diff --git a/.gitattributes b/.gitattributes index bed0738c7eeb449bca98b5d2f33c89a1ee56349a..b41216be5464f6b6b13e9f8c6bb03903eae72182 100644 --- a/.gitattributes +++ b/.gitattributes @@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text # Video files - compressed *.mp4 filter=lfs diff=lfs merge=lfs -text *.webm filter=lfs diff=lfs merge=lfs -text +data/splits/train.jsonl filter=lfs diff=lfs merge=lfs -text diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000000000000000000000000000000000000..91bf6438ed6a91bb8cf152f02e7b0030671d6856 --- /dev/null +++ b/.gitignore @@ -0,0 +1,61 @@ +# Python +__pycache__/ +*.py[cod] +*$py.class +*.egg-info/ +build/ +dist/ +.eggs/ +.venv/ +venv/ +env/ + +# Packaging +*.egg +*.wheel +pip-log.txt +pip-delete-this-directory.txt + +# uv lockfile — not committed; runtime deps are spec'd in pyproject.toml + +# requirements.txt. Anyone running the generator only needs PyYAML, which is +# stable enough that a pinned lockfile would add noise without value. +uv.lock + +# Testing +.pytest_cache/ +.coverage +htmlcov/ +.tox/ + +# Type checking +.mypy_cache/ +.ruff_cache/ + +# Editors +.vscode/ +.idea/ +*.swp +*~ +.DS_Store + +# Jupyter +.ipynb_checkpoints/ +*.ipynb + +# Environment +.env +.env.local +.env.*.local + +# Working artifacts — these are intentionally not committed. +# `data/splits/` IS committed (the release payload). +# But if anyone regenerates mid-review, they produce these: +data/splits/curation_pool.jsonl +data/splits/curation_pool_original.jsonl +data/splits/dtt.jsonl +data/splits/net.jsonl +data/splits/task_analysis.jsonl +data/splits/session_interpretation.jsonl + +# Curation review artifact (regenerated; not part of the release) +docs/curation/review.md diff --git a/CHANGELOG.md b/CHANGELOG.md new file mode 100644 index 0000000000000000000000000000000000000000..f2154b01a8968f6f3d6f64219b988709272429c7 --- /dev/null +++ b/CHANGELOG.md @@ -0,0 +1,30 @@ +# CHANGELOG + +All notable changes to TRACE are recorded here, starting with v1.0.0. +Future releases will follow semantic versioning: + +- **Patch** (v1.0.x) — typo fixes, metadata tweaks, no change to the JSONL splits. +- **Minor** (v1.x.0) — additions to the taxonomy or new teaching methods; existing examples preserved; new examples appended to splits. +- **Major** (v2.0.0) — schema changes or re-generations that invalidate prior splits; prior versions remain accessible via git tags. + +--- + +## v1.0.0 — 2026-04-25 + +Initial public release. 2,999 examples split across train (2,549) / valid (149) / test (281) / sanity (20). Paired with the dataset card, datasheet (Gebru et al. 2021), and data statement (Bender & Friedman 2018). + +### What's covered + +- **Teaching programs** across three methods — DTT (800), NET (500), Task Analysis (500, including toleration programs). +- **Session interpretations** — 1,200 multi-session behavioral logs across 12 trajectory patterns with 13 target behaviors. +- **Provenance** — every example carries full `meta.provenance.taxonomy_cells` for auditable traceback. +- **Reproducibility** — the corpus regenerates byte-identically from `(configs, seed)` on any platform; every `example_id` is verifiable as `sha256(user_content + assistant_content)[:16]` directly from the published JSONL row. + +### Known limitations + +See the dataset card section 6 and the data statement section H for the full list. Most notable: + +- English-only, US clinical register. +- Pattern frequencies are uniform for learnability rather than epidemiologically weighted. +- VB-MAPP + AFLS only. +- Single-reviewer clinical validation. diff --git a/CITATION.cff b/CITATION.cff new file mode 100644 index 0000000000000000000000000000000000000000..b9f42c3873ae619b093e1b49ebaf166a260c8e16 --- /dev/null +++ b/CITATION.cff @@ -0,0 +1,43 @@ +cff-version: 1.2.0 +message: "If you use TRACE in your research, please cite it as below." +title: "TRACE: Taxonomy-Referenced ABA Clinical Examples" +abstract: > + A 2,999-example synthetic instruction-tuning dataset for two clinical tasks + in Applied Behavior Analysis — teaching-program generation (DTT, NET, + Task Analysis) and multi-session behavioral interpretation (12 trajectory + patterns, 13 target behaviors). Every example is generated from a taxonomy + grounded in the canonical ABA literature and carries full sampling + provenance. +authors: + - family-names: Kahunla + given-names: Festus + affiliation: "Drexel University; Pombo Labs" +date-released: 2026-04-25 +version: "1.0.0" +license: CC-BY-NC-4.0 +repository-code: "https://github.com/Pombo-Labs/TRACE" +url: "https://huggingface.co/datasets/PomboLabs/TRACE" +type: dataset +keywords: + - applied-behavior-analysis + - ABA + - autism + - clinical-NLP + - synthetic-data + - instruction-tuning + - small-language-model + - taxonomy + - provenance +preferred-citation: + type: generic + title: > + TRACE: Taxonomy-Grounded Synthetic Data for Teaching Program Generation + and Session Interpretation in Applied Behavior Analysis + authors: + - family-names: Kahunla + given-names: Festus + affiliation: "Drexel University; Pombo Labs" + year: 2026 + publisher: + name: "Pombo Labs" + url: "https://github.com/Pombo-Labs/TRACE" diff --git a/LICENSE-CODE b/LICENSE-CODE new file mode 100644 index 0000000000000000000000000000000000000000..811535364375cef2db09513b95eb99f41a5b99ab --- /dev/null +++ b/LICENSE-CODE @@ -0,0 +1,26 @@ +MIT License + +Copyright (c) 2026 Pombo Labs + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + +This license applies to all source code, configuration files, and scripts in +this repository (the contents of `configs/`, `src/`, the top-level `pyproject.toml`, +and this LICENSE-CODE file itself). The dataset files in `data/splits/` are +licensed separately under CC BY-NC 4.0 (see LICENSE-DATA). diff --git a/LICENSE-DATA b/LICENSE-DATA new file mode 100644 index 0000000000000000000000000000000000000000..fb9553ec8def4ca709c3f4d82c0110e475215cc1 --- /dev/null +++ b/LICENSE-DATA @@ -0,0 +1,38 @@ +Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) + +Copyright (c) 2026 Pombo Labs + +The dataset files in `data/splits/` (train.jsonl, valid.jsonl, test.jsonl, +sanity.jsonl) — and any derivative dataset produced by running the generator +code in this repository — are licensed under the Creative Commons +Attribution-NonCommercial 4.0 International License. + +You are free to: + - Share — copy and redistribute the material in any medium or format + - Adapt — remix, transform, and build upon the material + +Under the following terms: + - Attribution — You must give appropriate credit, provide a link to the + license, and indicate if changes were made. + - NonCommercial — You may not use the material for commercial purposes. + +Attribution should include: + - The dataset name (TRACE) and version (v1) + - Author: Kahunla, F. + - Publisher: Pombo Labs + - The repository URL: https://github.com/Pombo-Labs/TRACE + +No warranty is given. TRACE is a research artifact and has not been +clinically validated. It is not a clinical tool. Any use of the dataset, +or any model derived from it, in a clinical setting is at the sole +responsibility of the user and their facility. The authors and Pombo Labs +make no representation of clinical suitability and accept no liability for +clinical outcomes. + +Full license text: +https://creativecommons.org/licenses/by-nc/4.0/legalcode + +Human-readable summary: +https://creativecommons.org/licenses/by-nc/4.0/ + +SPDX-License-Identifier: CC-BY-NC-4.0 diff --git a/README.md b/README.md new file mode 100644 index 0000000000000000000000000000000000000000..1a482ea657223253ee4f888d9425896499288721 --- /dev/null +++ b/README.md @@ -0,0 +1,149 @@ +# TRACE + +**T**axonomy-**R**eferenced **A**BA **C**linical **E**xamples + +A 2,999-example synthetic instruction-tuning dataset for two clinical tasks in Applied Behavior Analysis: **teaching-program generation** and **behavioral session interpretation**. Every example is generated from a taxonomy grounded in the canonical ABA literature (Cooper, Heron, & Heward 2020; VB-MAPP; AFLS; BACB Ethics Code 2020) and carries full sampling provenance — the exact taxonomy cells that produced it. + +| | | +|---|---| +| **Version** | v1.0.0 | +| **Examples** | 2,999 (2,549 train / 149 valid / 281 test / 20 sanity) | +| **Tasks** | Teaching-program generation (DTT, NET, Task Analysis); behavioral session interpretation (12 patterns, 13 target behaviors) | +| **Language** | English (US clinical register) | +| **Data license** | CC BY-NC 4.0 | +| **Code license** | MIT | +| **Maintainer** | Pombo Labs — [GitHub](https://github.com/Pombo-Labs) · [Hugging Face](https://huggingface.co/PomboLabs) | +| **Author** | Festus Kahunla | + +--- + +## What this is for + +ABA is a clinical discipline with high documentation workload. Board Certified Behavior Analysts (BCBAs) draft teaching programs and interpret multi-session behavioral logs constantly across their caseloads. TRACE is a **research** dataset for studying whether small language models can learn the structure of those documents well enough to produce useful first-pass drafts. It is a research artifact, not a clinical tool. + +Real session data is HIPAA-protected and gated by BACB confidentiality rules; it cannot be reliably de-identified without losing clinical detail. TRACE avoids the constraint by construction: the data never represented a real person. + +## What makes TRACE different + +1. **Two tasks in one dataset.** Structured teaching-program generation across DTT, NET, and Task Analysis; multi-session interpretation across 12 clinical trajectory patterns. +2. **Taxonomy-grounded.** Every category in the controlled vocabulary ties to a specific source — Cooper/Heron/Heward chapters, JABA papers, VB-MAPP/AFLS curricula, BACB Ethics Code 2020, ABAI 2010 Position Statement. +3. **Full provenance per example.** `meta.provenance.taxonomy_cells` records the exact values sampled from every taxonomy dimension that produced the example. This is the property that makes clinical auditing tractable. +4. **Practitioner-in-the-loop iteration.** Clinical accuracy was refined through targeted taxonomy edits rather than per-example rewrites — each flagged inaccuracy maps to a single cell whose fix then propagates across every example that sampled it. + +## Quick start + +```bash +# Load with Hugging Face datasets +from datasets import load_dataset +ds = load_dataset("PomboLabs/TRACE") +print(ds["train"][0]) +``` + +Or from the raw JSONL splits: +```python +import json +for line in open("data/splits/train.jsonl"): + example = json.loads(line) + system, user, assistant = example["messages"] + gold = example["meta"]["gold_labels"] + provenance = example["meta"]["provenance"]["taxonomy_cells"] +``` + +### Extend the generator + +The taxonomy YAMLs under `configs/` are the fork point — drop in a new teaching method, a new clinical area, or a new behavioral pattern, regenerate, and you have a corpus tailored to your scope. + +```bash +uv pip install -e . +uv run python src/generate.py --all +uv run python src/split_data.py +uv run python src/prepare_curation.py +uv run python src/compile_curation.py +``` + +(Without `uv`: `pip install -r requirements.txt` and drop the `uv run` prefix.) + +The pipeline is deterministic, so the same configs + seed produce the same corpus — but the more interesting use is to make your own. + +## Repository structure + +``` +. +├── configs/ # taxonomy YAMLs (controlled vocabulary) +│ ├── shared/ # cross-area: learner profiles, mastery states +│ ├── dtt/ # DTT taxonomy + template + compatibility +│ ├── net/ # NET area +│ ├── task_analysis/ # chaining (independence + toleration) +│ └── session_interpretation/ # patterns, behaviors, trajectories, recommendations +├── src/ +│ ├── generators/ # per-area generator code +│ ├── generate.py # orchestrator +│ ├── split_data.py # train / valid / curation_pool +│ ├── prepare_curation.py # browseable review.md +│ └── compile_curation.py # test + sanity splits +├── data/splits/ # the released JSONL splits +└── docs/ + ├── dataset-card.md # HF-style user-facing card + ├── datasheet.md # Gebru et al. (2021) template + ├── data-statement.md # Bender & Friedman (2018) template + ├── taxonomy-v1.md # operational definitions + citations + ├── schema-v1.md # wire format + slot specifications + ├── references.md # the focused citation shortlist (~30 papers) + └── curation/ + ├── README.md # review + compile workflow + └── LEGEND.md # session-log notation reference +``` + +## Documentation + +| | | +|---|---| +| **Dataset card** | [docs/dataset-card.md](docs/dataset-card.md) | +| **Datasheet** (Gebru et al. 2021) | [docs/datasheet.md](docs/datasheet.md) | +| **Data statement** (Bender & Friedman 2018) | [docs/data-statement.md](docs/data-statement.md) | +| **Taxonomy reference** | [docs/taxonomy-v1.md](docs/taxonomy-v1.md) | +| **Schema reference** | [docs/schema-v1.md](docs/schema-v1.md) | +| **Curation workflow** | [docs/curation/README.md](docs/curation/README.md) | +| **Session-log reading guide** | [docs/curation/LEGEND.md](docs/curation/LEGEND.md) | + +## Ethics and intended use + +- **For:** fine-tuning small models (recommended: Gemma 4 E2B, QLoRA, 4-bit) for on-device drafting assistants, evaluation baselines, research on clinical-NLP data pipelines. +- **Not for:** autonomous clinical decisions; writing final Behavior Intervention Plans without BCBA review; training models on real client data; medical diagnosis; insurance documentation. + +Crisis-plan content is deliberately conservative: physical-intervention procedures are not specified because they vary by jurisdiction, training certification, and learner-specific contraindications. See [docs/dataset-card.md section 6.3](docs/dataset-card.md) for details. + +**Responsibility.** TRACE is a research artifact. It is not a clinical tool, has not been clinically validated, and carries no clinical endorsement. Anyone who chooses to deploy TRACE — or any model derived from it — in a clinical setting does so entirely at their own responsibility and under their facility's own oversight. The authors and Pombo Labs make no representation of clinical suitability and accept no liability for clinical outcomes. + +## License + +- **Data** — JSONL splits under `data/splits/` — Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0). Research and non-commercial use permitted with attribution. See [LICENSE-DATA](LICENSE-DATA). +- **Code** — generator, scripts, configs — MIT. See [LICENSE-CODE](LICENSE-CODE). + +## Citation + +```bibtex +@dataset{trace_2026, + title = {TRACE: Taxonomy-Referenced ABA Clinical Examples}, + author = {Kahunla, Festus}, + year = {2026}, + publisher = {Pombo Labs}, + url = {https://github.com/Pombo-Labs/TRACE}, + note = {Taxonomy-Grounded Synthetic Data for Teaching Program Generation + and Session Interpretation in Applied Behavior Analysis} +} +``` + +Machine-readable metadata in [CITATION.cff](CITATION.cff). + +## Contributing + +The taxonomy YAMLs under `configs/` are the intended extension surface. Adding a new skill target, a new target behavior, a new mastery criterion, or a new teaching method follows a documented pattern — see `docs/taxonomy-v1.md` for the existing vocabulary and grounding conventions. Pull requests welcome. + +For clinical-accuracy corrections, open an issue with the flagged example's `example_id` and a description of the issue; we trace back to the offending taxonomy cell and apply a single-commit fix. + +## Scope of v1 + +2,999 examples covering DTT, NET, and Task Analysis (including toleration programs) for the teaching-program task, and 1,200 multi-session logs across 12 trajectory patterns for the session-interpretation task. What v1 does not cover is documented in the dataset card (section 6.5 Known limitations). + +Version history in [CHANGELOG.md](CHANGELOG.md). diff --git a/configs/dtt/compatibility.yaml b/configs/dtt/compatibility.yaml new file mode 100644 index 0000000000000000000000000000000000000000..6960f2296ec69378b59d552cbdcd691b266cad5f --- /dev/null +++ b/configs/dtt/compatibility.yaml @@ -0,0 +1,41 @@ +# DTT clinical-consistency rules — TRACE v1 +# Enforced by the DTT generator during sampling to avoid clinically-inconsistent combinations. + +# Errorless error-correction procedure pairs exclusively with most-to-least prompt hierarchy +# Citation: CHH Ch. 21 — errorless learning is the errorless most-to-least convention. +errorless_requires_hierarchy: most_to_least + +# Mastery state -> acceptable reinforcement schedule IDs +# Citation: CHH Ch. 13, 28 — reinforcement thinning follows acquisition -> mastery -> maintenance. +mastery_to_reinforcement: + emerging: [crf] + developing: [crf] + approaching: [crf, fr2] + near: [fr2, vr3] + mastered: [fr2, vr3, token_economy] + generalization: [vr3, token_economy] + +# VB-MAPP level -> compatible learner profile ids +# Citation: Sundberg 2008 (VB-MAPP manual) — level is anchored to developmental age. +level_to_learner_profiles: + L1: [early] + L2: [early, school_age] + L3: [school_age, adolescent] + +# Level -> array size for stimulus field (DTT-specific) +array_size_by_level: + L1: + n: 3 + text: "field of 3 (target + 2 distractors)" + L2: + n: 5 + text: "field of 5 (target + 4 distractors)" + L3: + n: 6 + text: "field of 6 (target + 5 distractors, mixed categories)" + +# Sampling weights for level selection (DTT biased toward L1/L2 discrete skills) +level_sampling_weights: + L1: 0.45 + L2: 0.40 + L3: 0.15 diff --git a/configs/dtt/taxonomy.yaml b/configs/dtt/taxonomy.yaml new file mode 100644 index 0000000000000000000000000000000000000000..30fbf6312c144226980b24232c6b1de6efc01607 --- /dev/null +++ b/configs/dtt/taxonomy.yaml @@ -0,0 +1,294 @@ +# DTT taxonomy — TRACE v1 +# Self-contained: everything the DTT generator needs beyond shared primitives. +# Citation: Lovaas 1987 JCCP; Smith 2001 Focus on Autism; CHH 2020 Chs. 17, 21. +# +# DTT fits array-based discrete-response skills. Only VB-MAPP domains where +# the target is discrete selection / labeling / matching appear here. + +method: + id: dtt + name: Discrete Trial Training + short: DTT + citation: "Lovaas, O. I. (1987). JCCP, 55(1), 3–9. https://doi.org/10.1037/0022-006X.55.1.3; Smith, T. (2001). Focus on Autism, 16(2), 86–92." + when_to_use: > + Acquisition of discrete, decontextualized skills (receptive labels, tacts, + matching, early imitation) — especially early in instruction when stimulus + control must be established. + +# VB-MAPP domains appropriate for DTT (array-based discrete-response skills). +# Other domains (Mand, Echoic, Motor Imitation, Intraverbal, Social, etc.) +# belong to NET, PRT, or other method taxonomies. +skill_domains: + vbmapp: + levels: + - id: L1 + name: Level 1 + developmental_age: "0–18 months" + - id: L2 + name: Level 2 + developmental_age: "18–30 months" + - id: L3 + name: Level 3 + developmental_age: "30–48 months" + + domains: + - id: tact + name: Tact + gloss: Labeling / naming + L1: + - tacts common objects (ball, cup, shoe) + - tacts familiar people by name + - tacts common actions (running, eating, sleeping) + - tacts body parts + - tacts common animals + L2: + - tacts colors of objects + - tacts shapes (circle, square, triangle) + - tacts adjectives (big/little, hot/cold) + - tacts prepositions (in, on, under) + - tacts emotions in self and others + L3: + - tacts community helpers and their roles + - tacts categories (fruits, vehicles, clothing) + - tacts features of objects (color, shape, function) + - tacts past-tense events + - tacts abstract concepts (same/different, first/last) + + - id: listener_responding + name: Listener Responding + gloss: Receptive language + L1: + - follows 1-step motor instructions (sit down, stand up) + - selects correct item from an array of 2 + - points to named body parts + - points to named common objects + - follows instructions involving objects (give me the cup) + L2: + - selects correct item from array of 4–6 + - follows 2-step instructions + - selects items by feature (find something red) + - selects items by function (find something you eat with) + - selects items by class (find an animal) + L3: + - selects items by multiple features simultaneously + - follows conditional instructions (if X, then Y) + - follows instructions involving temporal concepts (before, after) + + - id: vp_mts + name: Visual Perceptual Skills & Matching-to-Sample + gloss: VP-MTS + L1: + - matches identical objects + - matches identical pictures + - matches colors (identical chips) + - matches shapes (identical blocks) + - completes simple 3–4 piece puzzles + L2: + - matches non-identical items by category + - sorts items into 2–3 categories + - matches quantities (1–5) + - matches upper-case letters + - completes 8–12 piece puzzles + L3: + - matches associated items (sock-shoe, cup-plate) + - sorts by multiple attributes simultaneously + - matches upper to lower case letters + - sequences 3–4 step picture sequences + - reproduces block designs from model + + - id: lrffc + name: LRFFC + gloss: Listener Responding by Feature, Function, Class + L1: + - selects items by function (What do you drink from?) + - selects items by feature (What is round?) + - selects items by class (Find the animal) + - selects by single function from array of 3 + - selects by single visible feature from array of 3 + L2: + - selects items by function from array of 5–8 + - selects items by feature from array of 5–8 + - selects items by class from array of 5–8 + - selects items by multiple features (round and red) + - selects items by function when item not visible + L3: + - selects items given 2+ features/functions simultaneously + - selects items by class with exclusion (animal but not a pet) + - selects items by abstract features (something needed when cold) + - selects items by comparison (which is heavier) + - selects items by negative features (not round, not food) + + - id: reading + name: Reading + L1: + - matches letters to identical letters + - identifies own name in print + - identifies 5–10 upper-case letters + - matches words to identical words + - tracks print left to right + L2: + - identifies all upper-case letters + - identifies all lower-case letters + - reads 10–20 sight words + - sounds out CVC words (cat, dog, run) + - reads simple 2–3 word phrases + L3: + - reads sentences with comprehension + - reads short passages and answers questions + - phonetic decoding for novel words + + - id: writing + name: Writing + L1: + - traces horizontal and vertical lines + - traces basic shapes (circle, cross) + - copies basic shapes from model + - traces letters of own name + - writes own name from model + L2: + - writes own name independently + - copies all upper-case letters from model + - writes upper-case letters from dictation + - copies simple words from model + - writes numbers 1–10 + + - id: math + name: Math + L1: + - rote counts to 10 + - counts objects 1–5 with 1:1 correspondence + - identifies numerals 1–5 + - matches quantities to numerals 1–5 + - identifies basic shapes in math context + L2: + - counts objects 1–20 with 1:1 correspondence + - identifies numerals 1–20 + - compares quantities (more/less/same) + - solves single-digit addition with manipulatives + - identifies coins by name + L3: + - adds single-digit numbers without manipulatives + - subtracts single-digit numbers + - identifies place value (ones, tens) + - tells time to the hour and half-hour + - solves simple word problems + +# Prompt hierarchies that can be used in DTT +prompt_hierarchies: + - id: most_to_least + name: Most-to-Least (errorless) + sequence: [full_physical, partial_physical, gestural, positional, independent] + description: "Start with maximum assistance; fade systematically to independence." + best_for: "Acquisition, early learners, safety-critical skills." + + - id: least_to_most + name: Least-to-Most + sequence: [independent, gestural, positional, partial_physical, full_physical] + description: "Start with independent attempt; add assistance only as needed." + best_for: "Learner with partial repertoire; promotes independence." + + - id: time_delay + name: Time Delay (progressive) + sequence: [0s_delay, 2s_delay, 4s_delay, 6s_delay, independent] + description: "Increase latency between SD and prompt delivery across sessions." + best_for: "Prompt-dependency prevention; ideal for fading." + citation: "Touchette & Howard 1984 JABA." + + - id: graduated_guidance + name: Graduated Guidance + sequence: [hand_over_hand_full, hand_over_hand_light, shadow, independent] + description: "Hand-over-hand guidance with fading physical pressure." + best_for: "Motor skills, self-care chaining." + + - id: stimulus_fading + name: Stimulus Fading + sequence: [exaggerated_stimulus, moderate_fade, minimal_fade, natural_stimulus] + description: "Modify the stimulus itself; fade cue salience over sessions." + best_for: "Receptive discrimination, early reading." + + - id: stimulus_shaping + name: Stimulus Shaping + sequence: [modified_stimulus, intermediate_shape, target_stimulus] + description: "Change the stimulus topography to approximate the target." + best_for: "Complex visual discriminations." + +# Error correction procedures available for DTT +# Citation: Cooper, Heron, Heward (2020), Ch. 21. +error_corrections: + - id: transfer_trial + name: Transfer trial + steps: > + Re-present SD -> provide effective-level prompt -> reinforce prompted + response -> distractor trial -> re-present SD independently. + + - id: four_step + name: 4-step error correction + steps: > + Model correct response -> lead (do together) -> test (independent) -> + distractor -> retest. + + - id: backstep + name: Backstep + steps: > + Return to previous prompt level that produced success -> + successful response -> re-attempt at target prompt level. + + - id: simple_correction + name: Simple correction + steps: > + "No, watch me" -> model -> re-present SD -> differential reinforcement + for correct response. + + - id: errorless + name: Errorless (preventive) + steps: > + Prevent errors by using high prompts from the start; fade gradually. + Paired with most-to-least prompt hierarchy. + +# Reinforcement schedules relevant for DTT (acquisition and thinning) +# Citation: CHH (2020) Ch. 13, 22. +# Note: DRO/DRA/DRI are behavior-reduction schedules handled in FCT taxonomy. +reinforcement_schedules: + - id: crf + name: CRF (continuous reinforcement) + description: "Every correct response reinforced." + typical_phase: "Acquisition." + + - id: fr2 + name: FR-2 (fixed ratio 2) + description: "Every 2nd correct response reinforced." + typical_phase: "Thinning as accuracy stabilizes." + + - id: vr3 + name: VR-3 (variable ratio 3) + description: "Average of every 3rd correct response reinforced." + typical_phase: "Maintenance, resistance to extinction." + + - id: token_economy + name: Token economy + description: "Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item)." + typical_phase: "Group / classroom settings, mastered / generalization phase." + +# Mastery criteria common for DTT programs +# Citation: CHH (2020) Ch. 28. +mastery_criteria: + - id: mc_80_2sessions + text: "80% accuracy across 2 consecutive sessions with at least 10 trials each." + + - id: mc_90_3sessions + text: "90% accuracy across 3 consecutive sessions." + + - id: mc_80_generalization + text: "80% accuracy with 2 different therapists and 2 different settings." + + - id: mc_80_independent + text: "80% accuracy at independent prompt level across 3 sessions." + + - id: mc_90_first_trial + text: "90% first-trial-correct across 5 consecutive sessions." + + - id: mc_fluency + text: "Fluency criterion: correct responses within 3-second latency at 90% accuracy." + + - id: mc_natural_env + text: "Demonstrated use in natural environment across 3 independent instances." diff --git a/configs/dtt/template.yaml b/configs/dtt/template.yaml new file mode 100644 index 0000000000000000000000000000000000000000..063933fb65ff9e9d2ab0120e9c30e791b79dbefd --- /dev/null +++ b/configs/dtt/template.yaml @@ -0,0 +1,90 @@ +# DTT teaching program template — TRACE v1 +# Filled by src/generate.py with values sampled from taxonomy. + +template_id: task1_dtt_v1 +task_type: teaching_program +method_id: dtt + +# System prompt used for this task type +system_prompt: | + You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help + Board Certified Behavior Analysts and staff design teaching programs for + individuals with autism. Your responses are clinically accurate, individualized + to the learner profile, follow BACB ethical guidelines, and reference no real + client data. Select the appropriate teaching method (DTT, NET, Task Analysis, + FCT, BST, PRT) based on the skill target and learner profile. + +# User message template variants — one is sampled per example for phrasing diversity +user_variants: + - | + Generate a teaching program for the following target. + + Skill Target: {skill_target} + Curriculum Reference: {curriculum_ref} + Learner Profile: {learner_profile_name} + Current Mastery: {mastery_state_name} + Teaching Method: DTT + + Provide the full program structure appropriate to the selected method. + + - | + I need a DTT program for this skill. + + Target: {skill_target} + Curriculum: {curriculum_ref} + Learner: {learner_profile_name} + Mastery Status: {mastery_state_name} + + Please include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria. + + - | + Design a Discrete Trial Training program for the following. + + Skill: {skill_target} + Curriculum: {curriculum_ref} + Learner Level: {learner_profile_name} + Baseline: {mastery_state_name} + + Include the full DTT trial structure. + +# Assistant message template — filled by generator +# All slots in {braces} are computed by the generator from taxonomy values +assistant_template: | + ## Program Overview + - Target skill: {skill_target} + - Curriculum: {curriculum_ref} + - Learner: {learner_profile_name} + - Current mastery: {mastery_state_name} + + This DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure. + + ## Discriminative Stimulus (SD) + Primary SD: "{primary_sd}" + Variations: {sd_variations} + Presentation: {sd_presentation} + + ## Prompt Hierarchy + Strategy: {prompt_hierarchy_name} + Sequence: {prompt_sequence} + Current prompt level: At the {mastery_state_short} stage, {current_prompt_guidance} + + ## Stimulus Array + Array size: {array_size} + Target stimuli: {target_stimuli} + {distractor_block} + Rotation: Randomize position of target across trials to prevent position bias. + + ## Error Correction Procedure + {error_correction_steps} + + ## Reinforcement Schedule + {reinforcement_schedule_name} — {reinforcement_description}. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses. + + ## Mastery Criteria + {mastery_criterion_text} + + ## Data Collection + Trial-by-trial data recording prompt level and response accuracy. IOA scheduled every {ioa_frequency} sessions minimum; target agreement ≥ 80%. + + ## Generalization & Maintenance Plan + Post-mastery: probe with novel stimuli, across {n_generalization_therapists} additional therapists, and in at least {n_generalization_settings} natural contexts. Maintenance probes weekly for 4 weeks, then monthly. diff --git a/configs/generation.yaml b/configs/generation.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cdc1023c67bbdb37c7001291d58405c18beb4d67 --- /dev/null +++ b/configs/generation.yaml @@ -0,0 +1,44 @@ +# TRACE data generation config +# Controls per-area example counts, split ratios, and the global seed. + +seed: 42 +output_dir: data/splits + +# Per-area generation counts and enabled flags. +# Total target: ~3,500 examples (train + valid + curation-pool sums). +areas: + dtt: + enabled: true + n: 800 + description: "DTT teaching programs — array-based discrete-response skills." + net: + enabled: true + n: 500 + description: "NET teaching programs — motivation-based naturalistic teaching." + task_analysis: + enabled: true + n: 500 + description: "Task Analysis / Chaining — multi-step routines (AFLS)." + fct: + enabled: false + n: 300 + description: "FCT — replacement-behavior programs paired with deceleration targets." + bst: + enabled: false + n: 200 + description: "BST — staff-facing training programs." + prt: + enabled: false + n: 300 + description: "PRT — pivotal-behavior programs." + session_interpretation: + enabled: true + n: 1200 + description: "Task 2 — behavioral session interpretation (paper's innovation task)." + +# Split ratios applied after all enabled areas generate +# test + sanity are NOT generated here — they're curated by the user from curation_pool +splits: + train: 0.85 + valid: 0.05 + curation_pool: 0.10 # user reviews this pool and curates test.jsonl + sanity.jsonl diff --git a/configs/net/compatibility.yaml b/configs/net/compatibility.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3fe1c2b7af4471ddd1e93ab897d16fd010bbcb51 --- /dev/null +++ b/configs/net/compatibility.yaml @@ -0,0 +1,51 @@ +# NET clinical-consistency rules — TRACE v1 + +# VB-MAPP level -> compatible learner profiles +level_to_learner_profiles: + L1: [early] + L2: [early, school_age] + L3: [school_age, adolescent] + +# Sampling weights for level selection +# NET is used at all levels but slightly biased toward L2/L3 where mands grow +level_sampling_weights: + L1: 0.30 + L2: 0.40 + L3: 0.30 + +# Skill keyword -> MO category mapping +# Used to select the appropriate MO arrangement for a sampled skill. +skill_to_mo_category: + - keywords: [mand, item, preferred, 2-word] + category: mand_item + - keywords: [missing] + category: mand_missing + - keywords: [help] + category: mand_help + - keywords: [break] + category: mand_break + - keywords: ["all done", finished, completion] + category: mand_completion + - keywords: ["what"] + category: mand_info_what + - keywords: ["where"] + category: mand_info_where + - keywords: [attention, peers] + category: mand_attention + - keywords: [action] + category: mand_action + - keywords: [greet] + category: spontaneous_greeting + - keywords: [comment, describe, narrate, story] + category: spontaneous_comment + - keywords: [conversation, reciprocal, multi-turn] + category: intraverbal_multi_turn + - keywords: [turn-taking, turn] + category: social_turn_taking + - keywords: [initiate, peer] + category: social_initiation + - keywords: [routine] + category: intraverbal_routine + +# Default category when no keyword matches +default_mo_category: mand_item diff --git a/configs/net/taxonomy.yaml b/configs/net/taxonomy.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c3d0c149d2613b8f2206e796ab02856ba5b44802 --- /dev/null +++ b/configs/net/taxonomy.yaml @@ -0,0 +1,230 @@ +# NET (Natural Environment Teaching) taxonomy — TRACE v1 +# Self-contained: skills + MOs + natural contexts for naturalistic teaching. +# Citation: Hart & Risley 1975 JABA; McGee, Krantz, McClannahan 1985 JABA. + +method: + id: net + name: Natural Environment Teaching + short: NET + citation: "Hart, B. & Risley, T. R. (1975). JABA, 8(4), 411–420. https://doi.org/10.1901/jaba.1975.8-411; McGee, G. G., Krantz, P. J., & McClannahan, L. E. (1985). JABA, 18(1), 17–31." + when_to_use: > + Mand / requesting; generalization phase; when motivation is the limiting + factor; when the skill's natural occasion is ecologically predictable + (e.g., requesting food during snack time). + +# VB-MAPP domains appropriate for NET (motivation-driven skills). +skill_domains: + vbmapp: + levels: + - id: L1 + name: Level 1 + developmental_age: "0–18 months" + - id: L2 + name: Level 2 + developmental_age: "18–30 months" + - id: L3 + name: Level 3 + developmental_age: "30–48 months" + + domains: + - id: mand + name: Mand + gloss: Requesting + L1: + - single-word mands for preferred items + - mands for missing items needed to complete an activity + - mands for actions + - mands for help + - mands using 2-word phrases + - mands for a break (e.g., "break please", AAC "I need a break") + - mands for activity completion ("all done", "finished") + - mands to use the bathroom (e.g., "bathroom", "potty", "I need to go", AAC bathroom icon) + L2: + - mands for information using "what" + - mands for information using "where" + - mands using adjectives (big, little, more) + - mands for attention from peers + - mands for others to stop an action + - mands for a break during a non-preferred task ("break please, I need a break") + - mands to indicate completion of a step or activity ("all done with this one") + L3: + - mands for information using "why" + - mands for information using "when" + - mands using complete sentences with correct grammar + - mands for future events or items not present + - mands using polite social conventions + - mands for a break with reason ("I need a break, I'm frustrated") + - mands to indicate completion in context ("I'm all done with math, can I start reading?") + + - id: social + name: Social Behavior & Social Play + L1: + - makes eye contact during interactions + - responds to own name + - engages in social games (peek-a-boo, tickle) + - shows items to others spontaneously + - tolerates proximity of peers during parallel play + L2: + - initiates social interactions with peers + - takes turns during structured activities + - shares materials with peers when prompted + - engages in cooperative play with one peer + - responds to peer initiations appropriately + L3: + - maintains reciprocal conversations with peers + - demonstrates empathy and perspective-taking + - negotiates and compromises during group activities + - joins ongoing peer activities appropriately + - maintains friendships over time + + - id: spontaneous_vocal + name: Spontaneous Vocal Behavior + L1: + - spontaneously vocalizes during preferred activities + - spontaneously names items in the environment + - spontaneously requests desired items without prompts + - spontaneously greets familiar people + - spontaneously comments on events + L2: + - spontaneously describes ongoing activities + - spontaneously asks questions about the environment + - spontaneously reports past events + - spontaneously uses social phrases appropriately + - spontaneously initiates conversation with peers + L3: + - spontaneously tells stories or narratives + - spontaneously makes relevant comments in conversations + - spontaneously adjusts language to different listeners + - spontaneously uses humor appropriately + - spontaneously provides explanations and reasons + + - id: intraverbal + name: Intraverbal + gloss: Contextualized verbal-to-verbal responding + L2: + - answers WH-questions about familiar topics + - describes function of common objects + - names items in categories when given the category + - answers social questions (How are you?) + - describes recent events in sequence + L3: + - answers why- and how-questions + - engages in multi-turn conversations on a topic + - makes inferences from given information + - provides definitions of words + - answers hypothetical questions + +# MO (Motivating Operation) arrangement templates by skill-type category. +# The generator picks one based on the skill target keywords. +mo_arrangements: + - id: deprivation_preferred_item + applies_to: [mand_item, mand_action] + text: "Briefly withhold or restrict access to the target preferred item/activity for 30–60 seconds to establish motivation before the teaching opportunity." + context_examples: [snack time, free play, leisure time] + + - id: missing_item_completion + applies_to: [mand_help, mand_missing] + text: "Arrange an activity that requires a specific item to complete, then omit that item from the available materials. Wait for the learner to notice the gap." + context_examples: [art activity, snack preparation, puzzle] + + - id: break_opportunity + applies_to: [mand_break] + text: "Present a brief non-preferred task or demand at a level the learner can tolerate but may wish to pause. Wait for a request-for-break topography; honor the request immediately on first acquisition (CRF), thin later." + context_examples: [work time, tabletop task, academic period] + + - id: completion_opportunity + applies_to: [mand_completion] + text: "Set up a clearly-bounded activity with a visible endpoint (e.g., a worksheet, a puzzle, a sorting task). Wait for the learner to indicate completion verbally or via AAC; honor the indication immediately." + context_examples: [worksheet, puzzle, sorting task, meal] + + - id: bathroom_opportunity + applies_to: [mand_bathroom] + text: "Arrange a scheduled toileting window (typically 30–60 minutes after fluid intake) and observe for pre-toileting signs (body position, vocalization, squirming). Model the mand topography (vocal, sign, or AAC icon) at the first sign of need; reinforce any approximation with immediate escort to the bathroom. Do not prompt after an accident has started." + context_examples: [post-snack window, post-drink window, scheduled mid-morning check, end-of-activity transition] + + - id: unexpected_change + applies_to: [mand_info_what, mand_info_where, spontaneous_comment] + text: "Create a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question." + context_examples: [morning greeting, room transition, activity setup] + + - id: peer_presence + applies_to: [social_initiation, social_turn_taking, peer_mand] + text: "Arrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task)." + context_examples: [peer play, circle time, shared snack] + + - id: reciprocal_conversation + applies_to: [intraverbal_multi_turn, social_reciprocal, spontaneous_narrative] + text: "Engage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response." + context_examples: [transition conversation, post-activity debrief, leisure chat] + + - id: routine_lead_in + applies_to: [spontaneous_greeting, intraverbal_routine] + text: "Use the natural beginning/ending of an established routine (morning arrival, mealtime, departure) as the moment for the target response." + context_examples: [arrival, snack, departure] + +# Natural-context examples drawn into prompts for variation +natural_contexts: + - snack time + - free play with preferred materials + - transition between activities + - morning arrival routine + - peer group game + - art / craft activity + - leisure / sensory break + - outdoor play + - shared meal preparation + - departure routine + +# Prompt strategies used in NET (subset of full hierarchy — naturalistic) +prompt_strategies: + - id: time_delay + name: Time Delay + description: "Wait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt." + best_for: "Most NET mand and spontaneous vocal targets." + + - id: model_then_fade + name: Model + immediate fade + description: "Deliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions." + best_for: "Acquisition of new mand topographies." + + - id: expectant_look + name: Expectant look (pause) + description: "Deliver no prompt — maintain eye contact and hold the preferred item until the learner responds." + best_for: "Established mands; fading teacher-delivered prompts." + + - id: verbal_fill_in + name: Verbal fill-in + description: "Deliver a partial phrase and pause; learner completes the vocal response." + best_for: "Song, rhyme, or scripted social-phrase targets." + +# Natural reinforcers — always the functional outcome of the target response +natural_reinforcer_examples: + mand_item: "immediate access to the requested item for 30–60 seconds." + mand_action: "performance of the requested action." + mand_help: "adult delivers the requested help, task completed together." + mand_missing: "the missing item is provided, activity can continue." + mand_break: "immediate brief escape from the demand (30–60 seconds) to a neutral or preferred activity." + mand_completion: "the activity ends immediately and the learner transitions to the next preferred activity." + mand_bathroom: "immediate escort to the bathroom; the learner is not required to complete the current task before going. Pairing the mand with timely bathroom access is what teaches the learner that requesting works." + mand_info_what: "adult answers the question informatively." + mand_info_where: "adult answers the question or provides location." + mand_attention: "adult attends to the learner and engages briefly." + social_initiation: "peer engages back; activity continues with both." + social_turn_taking: "turn is delivered; game continues." + spontaneous_comment: "adult attends to the comment with matching affect." + spontaneous_greeting: "adult returns the greeting warmly." + intraverbal_multi_turn: "conversation continues with adult's naturally matched reply." + default: "natural social or tangible consequence matching the function of the response." + +# Mastery criteria for NET programs (slightly different emphasis than DTT) +mastery_criteria: + - id: mc_indep_natural_env + text: "Independent use of target response across 3 natural opportunities per day for 3 consecutive days." + - id: mc_generalize_multi_therapist + text: "Independent use across 3 different adults and 2 different settings within 2 weeks." + - id: mc_peer_initiation + text: "Independent use with a peer in unstructured free play across 2 consecutive sessions." + - id: mc_spontaneous_rate + text: "Average rate of ≥ 3 spontaneous responses per 30-minute session across 3 consecutive sessions." + - id: mc_fluent_exchange + text: "Fluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions." diff --git a/configs/net/template.yaml b/configs/net/template.yaml new file mode 100644 index 0000000000000000000000000000000000000000..759b1c0a64186fb127fb06e41da04ac24f28f21b --- /dev/null +++ b/configs/net/template.yaml @@ -0,0 +1,84 @@ +# NET teaching program template — TRACE v1 + +template_id: task1_net_v1 +task_type: teaching_program +method_id: net + +system_prompt: | + You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help + Board Certified Behavior Analysts and staff design teaching programs for + individuals with autism. Your responses are clinically accurate, individualized + to the learner profile, follow BACB ethical guidelines, and reference no real + client data. Select the appropriate teaching method (DTT, NET, Task Analysis, + FCT, BST, PRT) based on the skill target and learner profile. + +user_variants: + - | + Generate a teaching program for the following target. + + Skill Target: {skill_target} + Curriculum Reference: {curriculum_ref} + Learner Profile: {learner_profile_name} + Current Mastery: {mastery_state_name} + Teaching Method: NET (Natural Environment Teaching) + + Provide the full program structure appropriate to the selected method. + + - | + I need a NET program for this skill. + + Target: {skill_target} + Curriculum: {curriculum_ref} + Learner: {learner_profile_name} + Mastery Status: {mastery_state_name} + + Please include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria. + + - | + Design a Natural Environment Teaching program for the following. + + Skill: {skill_target} + Curriculum: {curriculum_ref} + Learner Level: {learner_profile_name} + Baseline: {mastery_state_name} + + Include the full naturalistic teaching structure. + +assistant_template: | + ## Program Overview + - Target skill: {skill_target} + - Curriculum: {curriculum_ref} + - Learner: {learner_profile_name} + - Current mastery: {mastery_state_name} + + This NET program teaches the target skill through naturalistic embedding — capturing motivated moments within the learner's routine rather than using contrived massed trials. + + ## Motivating Operation (MO) Arrangement + {mo_arrangement_text} + Primary natural context: {primary_natural_context}. + + ## Natural Opportunity + {natural_opportunity_text} + + ## Prompt Strategy + Strategy: {prompt_strategy_name} + {prompt_strategy_description} + Current prompt level: At the {mastery_state_short} stage, {current_prompt_guidance} + + ## Natural Reinforcer + {natural_reinforcer_text} + Avoid contrived consequences — the functional outcome IS the reinforcer. + + ## Generalization Tactics + - Program multiple exemplars across {n_exemplars} different stimuli or peers. + - Rotate between {n_settings} different natural settings. + - Rotate across {n_therapists} different adults during the teaching phase. + + ## Mastery Criteria + {mastery_criterion_text} + + ## Data Collection + Opportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every {ioa_frequency} sessions minimum; target agreement ≥ 80%. + + ## Generalization & Maintenance Plan + Post-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery. diff --git a/configs/session_interpretation/compatibility.yaml b/configs/session_interpretation/compatibility.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f71c62f41e45821deb72f7e4e9586e5528eec395 --- /dev/null +++ b/configs/session_interpretation/compatibility.yaml @@ -0,0 +1,90 @@ +# Session Interpretation compatibility rules — TRACE v1 + +# Pattern -> how many target behaviors are typically present in logs of this pattern +pattern_behavior_count_ranges: + mastery_progression: [0, 1] + regression: [1, 2] + plateau: [0, 1] + frustration_pattern: [1, 2] + variable_performance: [0, 2] + prompt_dependency: [0, 1] + rapid_acquisition: [0, 0] + generalization_failure: [0, 1] + extinction_burst: [1, 2] + skill_loss_after_break: [0, 1] + motivating_operation_shift: [0, 1] + setting_event_trigger: [0, 2] + +# Pattern -> likely behavior functions when behavior is present +pattern_function_bias: + frustration_pattern: [escape, escape, attention] # escape dominant + extinction_burst: [escape, attention, tangible] # depends on what's being extinguished + regression: [escape, attention, automatic] + variable_performance: [escape, attention, tangible, automatic] + plateau: [escape, automatic] + generalization_failure: [escape, attention] + setting_event_trigger: [escape, automatic, attention] + motivating_operation_shift: [tangible, automatic] + prompt_dependency: [attention, escape] + skill_loss_after_break: [escape, automatic] + mastery_progression: [automatic] # if any behavior present, likely automatic/low-stakes + rapid_acquisition: [] # no behaviors + +# Pattern -> behavioral indicator cluster to sample from when log includes indicators +pattern_behavioral_indicator_cluster: + frustration_pattern: frustration + regression: frustration + extinction_burst: frustration + variable_performance: disengagement + plateau: disengagement + prompt_dependency: disengagement + generalization_failure: disengagement + setting_event_trigger: disengagement + motivating_operation_shift: disengagement + mastery_progression: engagement + rapid_acquisition: engagement + skill_loss_after_break: disengagement + +# Escalation rules — raise above pattern default when severe behaviors present +# Behaviors with typical_severity "high" escalate one level (capped at 4). +escalation_rules: + severe_behavior_ids: [sib, aggression, pica, elopement, fecal_smearing] + severe_behavior_escalation_bump: 1 + safety_immediate_threshold_behaviors: [sib] # SIB at high frequency triggers 4 + safety_immediate_sib_rate_per_session: 8 # freq >= this in final 2 sessions -> escalation 4 + +# Confidence inference rules +confidence_rules: + min_sessions_for_moderate: 5 + min_sessions_for_high: 8 + ioa_required_for_high: true + variance_threshold_low: 0.25 # if accuracy SD > this, cap at moderate + +# Log length by pattern +log_length_by_pattern: + mastery_progression: [8, 12] + regression: [8, 12] + plateau: [8, 12] + frustration_pattern: [8, 12] + variable_performance: [8, 12] + prompt_dependency: [8, 12] + rapid_acquisition: [5, 8] + generalization_failure: [8, 12] + extinction_burst: [6, 10] + skill_loss_after_break: [6, 10] + motivating_operation_shift: [8, 12] + setting_event_trigger: [8, 12] + +# ABC and IOA presence (fractions are probabilities) +abc_inclusion_probability: 0.3 +ioa_inclusion_probability: 0.25 +ioa_agreement_range: [0.82, 0.97] +ioa_low_probability: 0.05 # small chance of below-80% IOA +ioa_low_range: [0.68, 0.79] + +# Sessions per log — max + min + typical duration +session_duration_minutes_range: [30, 60] +trials_per_program_per_session_range: [8, 15] + +# Sampling of programs per log +programs_per_log_range: [3, 6] diff --git a/configs/session_interpretation/recommendations.yaml b/configs/session_interpretation/recommendations.yaml new file mode 100644 index 0000000000000000000000000000000000000000..eafdf3b8e12272c5be155ecb54820504ff300e40 --- /dev/null +++ b/configs/session_interpretation/recommendations.yaml @@ -0,0 +1,155 @@ +# Per-pattern recommendation template pools — TRACE v1 +# Sampled by the generator to compose the "Programming Recommendations" section. +# Each pattern has bullets grouped under BIP-shaped sub-sections: +# antecedent / replacement / consequence / crisis +# Replacement and crisis sections only appear when target behaviors are in the log. +# +# Crisis-plan bullets are grounded in: +# - BACB Ethics Code for Behavior Analysts (2020), Section 3.05 (restrictive +# procedures) and Section 2.13 (referrals / coordination of services) +# - ABAI Position Statement on Restraint and Seclusion (2010) +# - Widely-used facility crisis-prevention training frameworks as examples only +# (Safety-Care, CPI, PMT, Therapeutic Crisis Intervention) +# Procedure specifics are intentionally left to the facility's training program +# because restraint / seclusion protocols vary by jurisdiction, training +# certification, and facility policy. + +patterns: + mastery_progression: + antecedent: + - "Continue current teaching procedure — it is effective for this learner." + - "Maintain current stimulus arrangement; begin planning next skill in sequence." + replacement: [] + consequence: + - "Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%." + - "Continue differential reinforcement for independent responses." + crisis: [] + + regression: + antecedent: + - "Conduct reinforcer assessment to identify current high-preference items." + - "Review recent environmental or schedule changes that may be affecting performance." + - "Return temporarily to a previously-mastered prompt level and re-fade systematically." + replacement: + - "If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function." + consequence: + - "Increase reinforcement magnitude and schedule temporarily (return to CRF if needed)." + - "Implement more frequent maintenance probes (at least weekly) after re-mastery." + crisis: + - "If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA." + - "Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent)." + - "Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures — in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention." + - "Document the incident per facility and BACB (Ethics Code 2020, Section 3.05) requirements within 24 hours; conduct post-incident debrief with all staff involved before resuming programming." + + plateau: + antecedent: + - "Modify teaching procedure — switch prompt type (e.g., from most-to-least to time delay)." + - "Conduct preference assessment and update reinforcer menu." + - "Break the current skill into smaller component steps if possible." + - "Increase practice opportunities by embedding targets in natural-environment training." + replacement: [] + consequence: + - "Check for inadvertent reinforcement of near-misses; tighten differential reinforcement." + crisis: [] + + frustration_pattern: + antecedent: + - "**Immediately** reduce task difficulty — drop back 1–2 prompt levels or present previously-mastered targets." + - "Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio." + - "Shorten session length and increase break frequency." + replacement: + - "Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior." + - "Pair the replacement response with short, predictable escape windows." + consequence: + - "Do NOT remove demand contingent on tantrum / escape behavior — this reinforces the escape function." + - "Provide brief break contingent on independent replacement response instead." + crisis: + - "If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards)." + - "If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol — only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact." + - "Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements." + - "Conduct a post-incident debrief with all staff involved to review what worked, what did not, and whether the BIP needs revision." + + variable_performance: + antecedent: + - "Check for environmental variables that differ across sessions (time of day, therapist, setting)." + - "Standardize session structure and reinforcement delivery across therapists." + replacement: [] + consequence: + - "Implement inter-observer agreement probes to rule out measurement inconsistency." + crisis: [] + + prompt_dependency: + antecedent: + - "Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding." + - "Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt." + replacement: [] + consequence: + - "Ensure differential reinforcement — independent responses receive higher-quality reinforcement than prompted responses." + - "Consider stimulus fading instead of response prompting for this skill." + crisis: [] + + rapid_acquisition: + antecedent: + - "Advance to next skill in the sequence immediately." + - "Begin generalization probes across people, settings, and materials." + replacement: [] + consequence: + - "Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets." + crisis: [] + + generalization_failure: + antecedent: + - "Implement structured generalization training: vary therapists, settings, and materials systematically." + - "Use multiple-exemplar training — teach with 3–5 different stimulus sets." + - "Program common stimuli across training and generalization settings." + replacement: [] + consequence: + - "Consider natural-environment training (NET) to supplement DTT for this skill." + crisis: [] + + extinction_burst: + antecedent: + - "Maintain current antecedent arrangements — do not add extra demand." + replacement: + - "Ensure replacement response is available and reinforced at high magnitude during the burst." + consequence: + - "**Maintain current protocol** — consistency during extinction bursts is critical." + - "Ensure all team members are trained on the procedure." + crisis: + - "If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent)." + - "De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP — contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only." + - "Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state." + - "Conduct a post-incident debrief with staff. If the burst exceeds the BIP's anticipated duration or intensity, the BIP should be re-reviewed before resuming." + + skill_loss_after_break: + antecedent: + - "Implement booster sessions at the most-recently-effective prompt level." + - "Increase session frequency temporarily until pre-break levels are recovered." + replacement: [] + consequence: + - "Review and strengthen the maintenance schedule to prevent future skill loss." + - "Consider sending practice materials home during extended breaks." + crisis: [] + + motivating_operation_shift: + antecedent: + - "Conduct preference assessment to identify a currently-valued reinforcer." + - "Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions." + - "Rotate reinforcers to prevent satiation." + replacement: [] + consequence: + - "Adjust reinforcement density to match current MO level." + crisis: [] + + setting_event_trigger: + antecedent: + - "Identify and document the setting event (sleep, illness, schedule change, medication)." + - "Modify programming demand on days when setting event is present." + - "Coordinate with caregivers / medical team on setting-event factors outside the program." + replacement: [] + consequence: + - "Track setting events alongside performance data to quantify their effect." + crisis: + - "If the setting event involves a medical concern (seizure-like activity, injury, illness): stop the program immediately, follow the facility's medical-incident protocol, and refer to the appropriate medical professional. Medical events are NOT behavioral targets and must not be treated as ABA programming issues." + - "If behavior escalates during the setting-event window: use verbal de-escalation and environmental safety first; do not add demand. Only staff currently certified in the facility's crisis-prevention framework may use physical intervention, and only if imminent safety risk warrants it AND the learner's BIP specifically authorizes it (contraindications and BIP restrictions may rule it out)." + - "Coordinate with caregivers, the medical team, and the supervising BCBA; document the setting event and the associated behavior for pattern analysis. Review the BIP in light of the setting-event data." diff --git a/configs/session_interpretation/taxonomy.yaml b/configs/session_interpretation/taxonomy.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d8b9bcf2ed4c769e242f28073c9c9128a4cae124 --- /dev/null +++ b/configs/session_interpretation/taxonomy.yaml @@ -0,0 +1,318 @@ +# Session Interpretation taxonomy — TRACE v1 +# The controlled vocabulary the Task 2 generator draws from. Self-contained. +# +# Citations for underlying concepts: see taxonomy-v1.md. + +task: + id: session_interpretation + name: Behavioral Session Interpretation + citation: > + Multi-session interpretation framework is TRACE's operationalization. + Underlying concepts grounded in CHH (2020) Ch. 6–7 (analyzing behavior change), + Iwata 1982/1994 (functional analysis), Hanley-Iwata-McCord 2003 (FBA review), + Bijou-Peterson-Ault 1968 (ABC recording), Michael 1993 (MOs). + +# 12 session patterns (each drives trajectory + interpretation) +patterns: + - id: mastery_progression + name: Mastery progression + description: "Steady improvement toward mastery criteria." + trend: ascending + concern_level: none + default_escalation: 1 + citation: "CHH Ch. 6." + + - id: regression + name: Regression + description: "Decline in performance after previous mastery or near-mastery." + trend: descending + concern_level: high + default_escalation: 3 + citation: "CHH Ch. 26 (maintenance)." + + - id: plateau + name: Plateau + description: "Accuracy stable but below mastery criteria for extended period." + trend: flat + concern_level: moderate + default_escalation: 2 + citation: "CHH Ch. 7." + + - id: frustration_pattern + name: Frustration pattern + description: "Declining accuracy accompanied by behavioral indicators of frustration and escape-maintained responding." + trend: descending + concern_level: high + default_escalation: 3 + citation: "Iwata 1994 (escape function); CHH Ch. 27." + + - id: variable_performance + name: Variable performance + description: "Inconsistent accuracy across sessions with no clear trend." + trend: variable + concern_level: moderate + default_escalation: 2 + citation: "CHH Ch. 5 (measurement reliability)." + + - id: prompt_dependency + name: Prompt dependency + description: "High accuracy with prompts but failure at independent level." + trend: flat_high_prompted + concern_level: moderate + default_escalation: 2 + citation: "Touchette & Howard 1984 JABA (time delay)." + + - id: rapid_acquisition + name: Rapid acquisition + description: "Quick skill acquisition exceeding expected timeline." + trend: sharply_ascending + concern_level: none + default_escalation: 1 + citation: "CHH Ch. 6." + + - id: generalization_failure + name: Generalization failure + description: "Strong performance in training but failure to generalize to novel conditions." + trend: context_dependent + concern_level: moderate + default_escalation: 2 + citation: "Stokes & Baer 1977 JABA; CHH Ch. 26." + + - id: extinction_burst + name: Extinction burst + description: "Temporary increase in problem behavior during behavior reduction." + trend: temporary_spike + concern_level: expected + default_escalation: 2 + citation: "CHH Ch. 24." + + - id: skill_loss_after_break + name: Skill loss after break + description: "Performance decline following extended absence or break, with recovery." + trend: drop_then_recovery + concern_level: moderate + default_escalation: 2 + citation: "CHH Ch. 26 (maintenance)." + + - id: motivating_operation_shift + name: Motivating operation shift + description: "Responding drops when MO changes (e.g., satiation); recovers when MO restored." + trend: mo_dip + concern_level: moderate + default_escalation: 2 + citation: "Michael 1993 JEAB; CHH Ch. 16." + + - id: setting_event_trigger + name: Setting event trigger + description: "Accuracy or behavior changes correlated with an external setting event (illness, sleep, schedule change)." + trend: setting_event + concern_level: moderate + default_escalation: 2 + citation: "Smith & Iwata 1997 JABA; Bijou & Baer 1961." + +# Behavior functions +functions: + - id: escape + name: escape + reinforcer: "Termination or avoidance of an aversive stimulus (demand, task, interaction)." + typical_antecedents: [demand presentation, difficult task, non-preferred activity, transition to work] + typical_consequences: [task removed, demand withdrawn, redirected to preferred activity] + - id: attention + name: attention + reinforcer: "Attention from others (adult or peer)." + typical_antecedents: [adult attention diverted, peer interaction, low-attention period] + typical_consequences: [adult attends, reprimand delivered, peer interaction initiated] + - id: tangible + name: tangible + reinforcer: "Access to an item or activity." + typical_antecedents: [preferred item removed, restricted access, another peer has item] + typical_consequences: [item returned, access granted, activity restored] + - id: automatic + name: automatic + reinforcer: "Self-produced reinforcement (sensory, proprioceptive)." + typical_antecedents: [low-stimulation period, alone time, waiting] + typical_consequences: [no social mediation; behavior self-terminates or continues] + - id: unknown + name: unknown + reinforcer: "Insufficient data to determine function." + +# Target behaviors (deceleration targets) +# For full operational definitions + citations see taxonomy-v1.md. +# Used when the generator samples 0–3 behaviors per log. +target_behaviors: + - id: tantrum + name: Tantrum + operational: "Co-occurring cluster of two or more of: crying/screaming, dropping to floor, kicking, hitting, throwing objects, lasting ≥ 3 s." + plausible_functions: [escape, attention, tangible] + typical_severity: moderate + - id: aggression + name: Aggression + operational: "Attempted or completed forceful contact directed toward another person (hit, kick, bite, scratch, pinch, throw-at)." + plausible_functions: [escape, tangible, attention] + typical_severity: high + - id: sib + name: Self-injurious behavior (SIB) + operational: "Any response that produces tissue damage or has potential to (head-hit, self-bite, face-slap, head-bang, skin-pick, self-pinch)." + plausible_functions: [automatic, attention, escape] + typical_severity: high + - id: elopement + name: Elopement + operational: "Full body crossing a designated boundary without adult approval." + plausible_functions: [escape, tangible, automatic] + typical_severity: high + - id: property_destruction + name: Property destruction + operational: "Hitting/kicking furniture or walls; throwing objects not meant to be thrown; tearing clothing, books, materials." + plausible_functions: [escape, attention, tangible] + typical_severity: moderate + - id: motor_stereotypy + name: Motor stereotypy + operational: "Repetitive non-functional motor movements (hand-flap, body-rock, finger-flick, spin)." + plausible_functions: [automatic] + typical_severity: low + - id: vocal_stereotypy + name: Vocal stereotypy + operational: "Non-contextual, non-communicative vocalizations outside appropriate conversational context." + plausible_functions: [automatic] + typical_severity: low + - id: non_compliance + name: Non-compliance + operational: "Failure to initiate a requested response within 5 s of an instructional prompt, OR active refusal." + plausible_functions: [escape] + typical_severity: low + - id: mouthing + name: Mouthing + operational: "Placing hand, fingers, or non-food object into the mouth outside scheduled meal/snack times." + plausible_functions: [automatic] + typical_severity: low + - id: pica + name: Pica + operational: "Placement of any inedible item past the plane of the lips, including ingestion." + plausible_functions: [automatic, attention] + typical_severity: high + - id: verbal_aggression + name: Verbal aggression + operational: "Yelling, cursing, name-calling, or verbal threats toward another person." + plausible_functions: [attention, escape] + typical_severity: moderate + - id: fecal_smearing + name: Fecal smearing (scatolia) + operational: "Reaching into diaper, pull-up, or pants to retrieve feces and/or spreading feces onto skin, clothing, walls, furniture, or other surfaces." + plausible_functions: [automatic, attention] + typical_severity: high + citation: "Piazza et al. 1996 JABA; Matson et al. 2008 (scatolia in ASD)." + - id: toileting_accident + name: Toileting accident (urine or bowel) + operational: "Session toileting data tracks both successful voids in the toilet (urine and BM) and accidents (voiding outside the toilet onto self, clothing, furniture, or floor). The deceleration target is accident frequency; in-toilet voids are recorded as context for toileting progress. Medical rule-out required before treating accidents behaviorally." + plausible_functions: [automatic, escape] + typical_severity: moderate + citation: "Azrin & Foxx 1971 JABA; Cicero & Pfadt 2002; LeBlanc et al. 2005." + +# Measurement types used in session logs +measurement_types: + - id: accuracy_pct + short: "accuracy" + description: "Trial-based percentage correct (X/N)." + used_for: [skill acquisition] + - id: frequency + short: "freq" + description: "Raw count per session." + used_for: [discrete behaviors, target behaviors] + - id: rate + short: "rate" + description: "Count per minute or hour." + used_for: [target behaviors] + - id: duration + short: "duration" + description: "Total elapsed time behavior occurred." + used_for: [tantrum, on-task engagement] + - id: latency + short: "latency" + description: "Time from SD to response initiation." + used_for: [skill responsivity] + - id: partial_interval + short: "PIR" + description: "Proportion of intervals in which behavior occurred at any point." + used_for: [stereotypy, low-intensity behavior] + - id: momentary_time_sampling + short: "MTS" + description: "Behavior sampled at end of each interval." + used_for: [group observation] + +# Escalation levels (safety-critical structured label) +escalation_levels: + - id: 1 + label: "Continue monitoring" + description: "Current programming is effective; no changes." + - id: 2 + label: "Adjust next session" + description: "Implement a specific programming change at the next session." + - id: 3 + label: "Supervisor review" + description: "Elevate to supervising BCBA within 24–48 hours." + - id: 4 + label: "Safety immediate" + description: "Stop program; contact supervisor / family; consider crisis-plan activation." + +# Confidence levels (calibration signal) +confidence_levels: + - id: high + label: high + description: "Pattern and recommendation well-supported by log data." + - id: moderate + label: moderate + description: "Data supports the hypothesis but alternatives cannot be fully ruled out." + - id: low + label: low + description: "Insufficient data to rule in a specific pattern; recommend data collection before programming changes." + +# Behavioral indicator phrases (pasted verbatim into session logs) +# Organized by cluster; Task 2 generator samples from here. +behavioral_indicators: + frustration: + - "Increased response latency (3–5× baseline)" + - "Pushing materials away from the work area" + - "Turning away from task or staff" + - "Vocal refusal ('no', 'I don't want to')" + - "Self-stimulatory behavior increase during demand periods" + - "Crying during demand presentation" + - "Elopement from the work area" + - "Attempts at property destruction (throwing materials, swiping items off the surface)" + - "Aggression attempt directed at staff (hit, kick, or scratch toward adult)" + - "Peer aggression (hitting or pushing peers during group activities)" + - "Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)" + engagement: + - "Eye contact maintained throughout trials" + - "Quick response latency (within 2–3 seconds)" + - "Reaching for materials proactively" + - "Smiling during reinforcement delivery" + - "Requesting more trials verbally" + - "Appropriate sitting posture maintained" + disengagement: + - "Looking away during SD presentation" + - "Delayed responding (>5 seconds consistently)" + - "Stereotypic behavior during inter-trial intervals" + - "Leaving seat without permission" + - "Playing with non-task materials" + - "Echolalic responding (repeating SD rather than answering)" + +# Teaching method metadata (for referencing methods in log programs) +teaching_methods: + - id: dtt + name: DTT + default_structure: "structured discrete trials, massed" + - id: net + name: NET + default_structure: "natural-environment embedded opportunities" + - id: task_analysis + name: Task Analysis + default_structure: "multi-step chaining, {chain_type}" + - id: fct + name: FCT + default_structure: "functional communication training, replacement response = {replacement}" + - id: prt + name: PRT + default_structure: "pivotal response training, child-choice, attempt reinforcement" + - id: bst + name: BST + default_structure: "staff-facing training program" diff --git a/configs/session_interpretation/template.yaml b/configs/session_interpretation/template.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c090085aacbdddfd96b30780f61eb123543bf34c --- /dev/null +++ b/configs/session_interpretation/template.yaml @@ -0,0 +1,68 @@ +# Session Interpretation template — TRACE v1 +# Task 2 — given a multi-session behavioral log, produce structured interpretation. + +template_id: task2_session_interp_v1 +task_type: session_interpretation + +system_prompt: | + You are an expert ABA (Applied Behavior Analysis) clinical assistant. You + analyze multi-session behavioral session logs for individuals with autism and + produce structured clinical interpretations that identify patterns, hypothesize + behavior functions when applicable, and recommend programming adjustments + structured along BIP lines (antecedent strategies, replacement behaviors, + consequence strategies, crisis plan). Your interpretation also includes an + escalation level and a confidence expression. Every recommendation is grounded + in the data provided. Follow BACB ethical guidelines and reference no real + client data. + +# User message variants — one is sampled per example to vary phrasing +user_variants: + - | + Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale. + + {session_log} + + - | + Review this session log and produce a structured clinical interpretation. + + {session_log} + + Please include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale. + + - | + Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format. + + {session_log} + +# Assistant template — filled section-by-section by the generator. +# Conditional sections (function hypothesis, replacement, crisis plan) are +# pre-computed by the generator as strings, possibly empty. +assistant_template: | + ## Clinical Concerns + {clinical_concerns_bullets} + + ## Pattern Classification + {pattern_class} + + {pattern_explanation} + {function_hypothesis_section} + ## Programming Recommendations + + ### Antecedent strategies + {antecedent_bullets} + {replacement_section} + ### Consequence strategies + {consequence_bullets} + {crisis_section} + ## Escalation Level + {escalation_level} — {escalation_label} + + {escalation_justification} + + ## Confidence + {confidence_level} + + {confidence_justification} + + ## Data-Supported Rationale + {rationale_bullets} diff --git a/configs/session_interpretation/trajectory_rules.yaml b/configs/session_interpretation/trajectory_rules.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9b39ed9cdb949b9df102233761ad41beab2e5b26 --- /dev/null +++ b/configs/session_interpretation/trajectory_rules.yaml @@ -0,0 +1,137 @@ +# Per-pattern trajectory rules — TRACE v1 +# Parameters that drive per-session accuracy and behavior-frequency generation. +# Each pattern ID matches configs/session_interpretation/taxonomy.yaml. +# +# Interpretation of fields: +# base_range: uniform-sample initial accuracy +# delta_per_session_range: per-session delta applied cumulatively +# noise: uniform ±noise added per session +# behavior_trajectory: how target-behavior frequencies evolve (optional) +# "stable_low" / "ascending" / "descending" / "spike" / etc. + +rules: + mastery_progression: + accuracy: + base_range: [0.30, 0.50] + delta_per_session_range: [0.04, 0.08] + noise: 0.05 + cap: 0.95 + behavior_trajectory: stable_low + + regression: + accuracy: + base_range: [0.70, 0.85] + delta_per_session_range: [-0.08, -0.04] + noise: 0.05 + floor: 0.10 + behavior_trajectory: ascending_moderate + + plateau: + accuracy: + base_range: [0.45, 0.65] + delta_per_session_range: [-0.02, 0.02] + noise: 0.08 + behavior_trajectory: stable_low + + frustration_pattern: + accuracy: + base_range: [0.70, 0.85] + delta_per_session_range: [-0.08, -0.04] + noise: 0.05 + floor: 0.10 + behavior_trajectory: ascending_strong + behavioral_indicators: frustration + + variable_performance: + accuracy: + base_range: [0.40, 0.70] + delta_per_session_range: [0.0, 0.0] + noise: 0.20 + behavior_trajectory: variable + + prompt_dependency: + accuracy: + base_range: [0.75, 0.95] # high but prompted + delta_per_session_range: [-0.02, 0.02] + noise: 0.04 + behavior_trajectory: stable_low + prompted_share: [0.70, 0.90] # share of trials that are prompted (not independent) + + rapid_acquisition: + accuracy: + base_range: [0.20, 0.40] + delta_per_session_range: [0.08, 0.12] + noise: 0.03 + cap: 0.97 + behavior_trajectory: stable_low + + generalization_failure: + accuracy: + alternating: + training_range: [0.75, 0.90] + novel_range: [0.25, 0.45] + noise: 0.05 + behavior_trajectory: stable_low + + extinction_burst: + accuracy: + base_range: [0.50, 0.70] + delta_per_session_range: [0.0, 0.03] + noise: 0.05 + behavior_trajectory: temporary_spike + burst_start_fraction: 0.33 # spike starts ~1/3 through log + burst_duration_sessions: 2 + burst_magnitude: [2.0, 3.5] # multiplier on baseline frequency + + skill_loss_after_break: + accuracy: + base_range: [0.70, 0.85] + initial_drop: [0.25, 0.40] + recovery_per_session_range: [0.05, 0.10] + noise: 0.03 + behavior_trajectory: stable_low + + motivating_operation_shift: + accuracy: + base_range: [0.70, 0.85] + mo_dip_sessions: [2, 4] # how many consecutive sessions show the dip + mo_dip_magnitude: [0.25, 0.40] # drop depth + recovery_per_session_range: [0.05, 0.08] + noise: 0.04 + behavior_trajectory: stable_low + + setting_event_trigger: + accuracy: + base_range: [0.55, 0.75] + setting_event_session_fraction: 0.40 # where the setting event hits + setting_event_magnitude: [0.20, 0.35] + recovery_per_session_range: [0.04, 0.08] + noise: 0.05 + behavior_trajectory: coincident_spike + +# Behavior-trajectory shorthand -> what it produces for frequency per session +behavior_trajectory_library: + stable_low: + base_freq_range: [0, 3] + delta_per_session: 0.0 + noise: 1.0 + ascending_moderate: + base_freq_range: [1, 3] + delta_per_session: 0.3 + noise: 1.0 + ascending_strong: + base_freq_range: [1, 3] + delta_per_session: 0.6 + noise: 1.0 + variable: + base_freq_range: [1, 5] + delta_per_session: 0.0 + noise: 3.0 + temporary_spike: + base_freq_range: [1, 3] + spike_multiplier_range: [2.5, 4.0] + noise: 1.0 + coincident_spike: + base_freq_range: [1, 3] + spike_multiplier_range: [2.0, 3.5] + noise: 1.0 diff --git a/configs/shared/learner_profiles.yaml b/configs/shared/learner_profiles.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9416241f28185a5da4b0f4ce74b3929b174ee1c3 --- /dev/null +++ b/configs/shared/learner_profiles.yaml @@ -0,0 +1,34 @@ +# Learner profile taxonomy — TRACE v1 +# Groups learners by developmental + chronological age + curriculum scope. + +profiles: + - id: early + name: Early Learner + developmental_age: "0–48 months" + curricula: [vbmapp] + typical_vbmapp_levels: [L1, L2] + note: "Foundational skills; discrete + naturalistic teaching appropriate." + + - id: school_age + name: School-Age Learner + developmental_age: "varies; often 24–60 months equivalent" + chronological_age: "6–12 years" + curricula: [vbmapp, afls] + typical_vbmapp_levels: [L2, L3] + typical_afls_modules: [school] + note: "Classroom skills + pre-academic + continuing language." + + - id: adolescent + name: Adolescent Learner + chronological_age: "13–17 years" + curricula: [afls, vbmapp] + typical_vbmapp_levels: [L3] + typical_afls_modules: [basic_living, home, community, school, vocational] + note: "Functional-living + vocational focus; residual language gaps addressed." + + - id: adult + name: Adult Learner + chronological_age: "18+ years" + curricula: [afls] + typical_afls_modules: [basic_living, home, community, vocational, independent_living] + note: "Independent-living + vocational focus; residential/day-program setting." diff --git a/configs/shared/mastery_states.yaml b/configs/shared/mastery_states.yaml new file mode 100644 index 0000000000000000000000000000000000000000..126aa896a0781ded3bb47d5e555a6ecfda4b3fd7 --- /dev/null +++ b/configs/shared/mastery_states.yaml @@ -0,0 +1,46 @@ +# Mastery states taxonomy — TRACE v1 +# Where a skill sits in its learning trajectory. +# Citation: Cooper, Heron, Heward (2020), Ch. 26, Ch. 28. + +states: + - id: emerging + name: Emerging + short: emerging + accuracy_range: [0.00, 0.30] + description: "Skill not yet acquired; high prompt levels needed." + prompt_level_guidance: "Begin at higher prompt levels (full physical / partial physical)." + + - id: developing + name: Developing + short: developing + accuracy_range: [0.30, 0.50] + description: "Responding established; stimulus control unstable." + prompt_level_guidance: "Partial physical or modeling prompt with systematic fading." + + - id: approaching + name: Approaching mastery + short: approaching mastery + accuracy_range: [0.50, 0.70] + description: "Moving toward independence; prompt fading in progress." + prompt_level_guidance: "Gestural or positional prompts; plan fade to independence." + + - id: near + name: Near mastery + short: near mastery + accuracy_range: [0.70, 0.85] + description: "Stimulus control solid; occasional errors on complex stimuli." + prompt_level_guidance: "Minimal prompts; increase expectations for independence." + + - id: mastered + name: Mastered at current level + short: mastered + accuracy_range: [0.85, 1.00] + description: "Meets mastery criterion for current step." + prompt_level_guidance: "Independent responding; move to generalization." + + - id: generalization + name: Generalization phase + short: generalization + accuracy_range: [0.85, 1.00] + description: "Mastered in training; extending to novel settings / materials / therapists." + prompt_level_guidance: "NET-style opportunities; multiple exemplar probing." diff --git a/configs/shared/prompt_types.yaml b/configs/shared/prompt_types.yaml new file mode 100644 index 0000000000000000000000000000000000000000..db278eca90d03d66936f2e5ae939b5d11600112b --- /dev/null +++ b/configs/shared/prompt_types.yaml @@ -0,0 +1,46 @@ +# Canonical prompt types — TRACE shared taxonomy +# Citation: Cooper, Heron, & Heward (2020), Ch. 21 ("Prompting and Fading"). +# +# Used across any teaching method that has a prompt hierarchy (DTT, Task Analysis, etc.). +# Each method's own taxonomy references these by id. + +prompt_types: + - id: full_physical + short: FP + name: Full physical + description: "Hand-over-hand / complete manual guidance through the movement." + + - id: partial_physical + short: PP + name: Partial physical + description: "Light guidance at a distal body part (elbow, shoulder)." + + - id: model + short: M + name: Modeling + description: "Demonstrate the response; learner copies." + + - id: gestural + short: G + name: Gestural + description: "Pointing or indicating without physical contact." + + - id: verbal + short: V + name: Verbal + description: "Spoken hint or instruction." + + - id: positional + short: Pos + name: Positional + description: "Target stimulus arranged closer to learner or more prominent." + + - id: visual + short: Vis + name: Visual + description: "Pictures, arrows, written cues placed for learner to see." + + - id: independent + short: I + name: Independent + description: "No prompt; learner responds without assistance." diff --git a/configs/task_analysis/compatibility.yaml b/configs/task_analysis/compatibility.yaml new file mode 100644 index 0000000000000000000000000000000000000000..edb850bd88b93a19194749a43237ba2512f43372 --- /dev/null +++ b/configs/task_analysis/compatibility.yaml @@ -0,0 +1,38 @@ +# Task Analysis clinical-consistency rules — TRACE v1 + +# Module -> typical learner profiles (some modules only fit older learners) +module_to_learner_profiles: + basic_living: [school_age, adolescent, adult] + home: [adolescent, adult] + community: [adolescent, adult] + vocational: [adolescent, adult] + independent_living: [adult] + +# Mastery state -> acceptable reinforcement schedules (chaining-specific) +mastery_to_reinforcement: + emerging: [crf_per_step] + developing: [crf_per_step] + approaching: [crf_per_step, differential_per_step] + near: [differential_per_step, token_economy] + mastered: [terminal_reinforcement, token_economy] + generalization: [terminal_reinforcement] + +# Chain type tends to follow skill preference; fallback is total_task +use_preferred_chain_type_probability: 0.75 + +# Sampling across modules (weighted toward typical clinical frequency) +module_sampling_weights: + basic_living: 0.35 + home: 0.20 + community: 0.15 + vocational: 0.15 + independent_living: 0.15 + +# Prompt strategy bias by mastery state +mastery_to_prompt_strategies: + emerging: [graduated_guidance, most_to_least] + developing: [graduated_guidance, most_to_least] + approaching: [most_to_least, least_to_most] + near: [least_to_most] + mastered: [least_to_most] + generalization: [least_to_most] diff --git a/configs/task_analysis/taxonomy.yaml b/configs/task_analysis/taxonomy.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ba7d0de7e5653cb9163722f84dc9316952c7e8d5 --- /dev/null +++ b/configs/task_analysis/taxonomy.yaml @@ -0,0 +1,413 @@ +# Task Analysis / Chaining taxonomy — TRACE v1 +# Self-contained: AFLS skills broken into step lists, chain types, prompting. +# Citation: CHH 2020 Ch. 23; Slocum & Tiger 2011 JABA; Partington & Mueller 2012. + +method: + id: task_analysis + name: Task Analysis / Chaining + short: Task Analysis + citation: "Cooper, Heron, Heward (2020) Applied Behavior Analysis (3rd ed.), Ch. 23; Slocum, S. K. & Tiger, J. H. (2011). JABA, 44(4), 793–805; Partington & Mueller 2012 (AFLS)." + when_to_use: > + Multi-step skills with a natural sequence — toileting, handwashing, + dressing, cooking, cleaning, classroom routines, community tasks, + vocational routines. + +# Chain types +chain_types: + - id: forward + name: Forward chaining + description: "Teach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps." + best_for: "Skills where early steps are simple and the chain grows in complexity." + - id: backward + name: Backward chaining + description: "Complete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner." + best_for: "Skills where the terminal reinforcer is salient (bathing, dressing, meal cleanup)." + - id: total_task + name: Total-task presentation + description: "Prompt the learner through the entire chain every trial; fade prompts across all steps simultaneously." + best_for: "Skills with few steps, or when all steps are similar in difficulty." + +# Prompt hierarchies commonly used in chaining +prompt_hierarchies: + - id: graduated_guidance + name: Graduated Guidance + description: "Hand-over-hand physical guidance with fading pressure; move to shadow, then independent." + best_for: "Self-care and motor-dominant chains." + - id: most_to_least + name: Most-to-Least + description: "Start with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent." + best_for: "Acquisition phase; new chains." + - id: least_to_most + name: Least-to-Most + description: "Allow independent attempt; add progressively more intrusive prompts only if learner does not initiate." + best_for: "Partial-repertoire chains; promoting independence." + +# Error correction procedures appropriate for chaining +error_corrections: + - id: back_to_previous_step + name: Back-step + re-approach + steps: "Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level." + - id: model_prompt_fade + name: Model -> prompt -> fade + steps: "Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials." + - id: task_reset + name: Task reset + steps: "Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step." + +# Reinforcement schedules for chaining +reinforcement_schedules: + - id: crf_per_step + name: CRF per step (acquisition) + description: "Continuous reinforcement for each correctly-completed step during acquisition." + - id: terminal_reinforcement + name: Terminal reinforcement (mastered) + description: "Reinforcement delivered only at chain completion; no per-step reinforcement." + - id: differential_per_step + name: Differential per-step + terminal + description: "Small per-step reinforcement for independent responses; larger reinforcer at chain completion." + - id: token_economy + name: Token economy + description: "One token per correctly-completed step; token exchange at chain completion." + +# Mastery criteria for chains (independence) vs toleration programs. +# Generator filters by applies_to when picking a criterion. +mastery_criteria: + - id: mc_all_steps_independent + applies_to: independence + text: "100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain." + - id: mc_fluent_chain + applies_to: independence + text: "Full chain completed within {fluency_seconds}s across 3 consecutive sessions with independent responding on all steps." + - id: mc_generalization + applies_to: independence + text: "Full chain completed independently with 2 different adults in 2 different settings across 2 consecutive days." + - id: mc_self_initiation + applies_to: independence + text: "Chain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities." + - id: mc_toleration_full_duration + applies_to: toleration + text: "End-goal duration tolerated across 3 consecutive sessions with no safety behaviors (refusal, crying, SIB, aggression, property destruction)." + - id: mc_toleration_caregivers + applies_to: toleration + text: "End-goal activity tolerated with 2 different caregivers (primary caregiver + one other) across 2 consecutive weeks." + - id: mc_toleration_settings + applies_to: toleration + text: "End-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors." + +# AFLS skill domains — learner profile + module + skill + step definitions +afls: + modules: + - id: basic_living + name: Basic Living Skills + learner_profiles: [school_age, adolescent, adult] + # Two kinds of programs in this module: + # - "steps" field = independence-focused chaining (learner performs the skill with + # graduated prompts) + # - "shaping_steps" field = toleration program (learner allows staff or caregiver + # to perform the skill; duration-based shaping of tolerance) + # Some real skills have BOTH variants depending on learner profile and history. + skills: + - name: washing hands + chain_type_preferred: total_task + steps: + - Walk to the sink. + - Turn on the water. + - Wet hands under the water. + - Apply soap to hands. + - Rub hands together for 15–20 seconds, covering fronts, backs, and between fingers. + - Rinse hands thoroughly under the water. + - Turn off the water. + - Dry hands with a towel. + - name: brushing teeth + chain_type_preferred: total_task + steps: + - Collect toothbrush and toothpaste. + - Wet the toothbrush under water. + - Apply a pea-sized amount of toothpaste. + - Brush upper-front teeth in small circles for 30 seconds. + - Brush lower-front teeth in small circles for 30 seconds. + - Brush upper-back teeth (left and right). + - Brush lower-back teeth (left and right). + - Spit into the sink and rinse mouth. + - Rinse the toothbrush and return it to the holder. + - name: putting on a shirt (pullover) + chain_type_preferred: backward + steps: + - Hold the shirt with the front facing away. + - Place both arms into the sleeves. + - Pull the opening over the head. + - Push arms through the sleeves fully. + - Pull the bottom of the shirt down to straighten it. + - name: putting on pants + chain_type_preferred: backward + steps: + - Sit or stand securely. + - Hold pants with the front facing away. + - Place one leg into the corresponding pant leg. + - Place the other leg into the other pant leg. + - Pull pants up to the waist. + - Fasten the button / zipper / velcro. + - name: showering + chain_type_preferred: backward + steps: + - Enter the shower and close the curtain. + - Turn on the water and adjust temperature. + - Wet body completely. + - Apply soap to a washcloth. + - Wash body systematically (face -> torso -> arms -> legs -> feet). + - Rinse all soap off. + - Apply shampoo and rub into hair. + - Rinse hair completely. + - Turn off the water. + - Exit the shower and dry off with a towel. + - name: toileting routine + chain_type_preferred: forward + steps: + - Recognize the need and walk to the bathroom. + - Close the door for privacy. + - Lower clothing appropriately. + - Sit on the toilet. + - Complete the elimination. + - Use toilet paper correctly. + - Stand and pull clothing up. + - Flush the toilet. + - Wash hands. + + # Toleration programs + # For learners who do not tolerate the activity being performed on them. + # Goal is NOT to do it independently — goal is to allow caregiver/staff to + # complete the activity without refusal, crying, SIB, or aggression, with + # duration shaped incrementally. + - name: tolerating tooth brushing by caregiver + program_type: toleration + target_activity: tooth brushing performed by caregiver + end_goal_description: "full 2-minute tooth-brushing routine completed by caregiver without refusal, crying, SIB, or aggression" + shaping_steps: + - "1-second toothbrush contact with teeth" + - "3 seconds of brushing" + - "5 seconds of brushing" + - "15 seconds of brushing" + - "30 seconds of brushing" + - "60 seconds of brushing" + - "Full 2-minute brushing routine" + + - name: tolerating hair washing by caregiver + program_type: toleration + target_activity: hair washing / shampooing performed by caregiver + end_goal_description: "full hair-washing routine completed by caregiver without refusal or safety behaviors" + shaping_steps: + - "5 seconds of water on hair" + - "15 seconds of water on hair" + - "Shampoo applied without lathering" + - "30 seconds of gentle lathering" + - "60 seconds of lathering" + - "Full rinse" + - "Full routine: wet -> shampoo -> lather -> rinse" + + - name: tolerating nail clipping + program_type: toleration + target_activity: nail clipping performed by caregiver + end_goal_description: "all 10 fingernails (and optionally toenails) clipped by caregiver without refusal, pulling hand away, or safety behaviors" + shaping_steps: + - "1 fingernail clipped" + - "2 fingernails clipped in one sitting" + - "5 fingernails clipped (one hand)" + - "All 10 fingernails clipped" + - "All 10 fingernails + 10 toenails clipped" + + - name: tolerating haircut + program_type: toleration + target_activity: haircut by caregiver or stylist + end_goal_description: "full haircut (10+ minutes) completed without refusal, leaving the chair, or safety behaviors" + shaping_steps: + - "Sit in the chair with cape on for 30 seconds" + - "Tolerate 1 minute of quiet sitting with cape" + - "Tolerate 2 minutes of brief clipping" + - "Tolerate 5 minutes of clipping" + - "Tolerate 10 minutes of clipping" + - "Tolerate full haircut routine" + + - name: tolerating showering assistance + program_type: toleration + target_activity: showering / bathing performed with caregiver assistance + end_goal_description: "full assisted shower routine completed without refusal, exiting the shower, or safety behaviors" + shaping_steps: + - "Stand in shower for 30 seconds with water off" + - "Tolerate 10 seconds of water on body" + - "Tolerate 30 seconds of water on body" + - "Tolerate soap application to torso" + - "Tolerate full body wash by caregiver" + - "Full assisted shower routine" + + - name: tolerating medical exam + program_type: toleration + target_activity: routine medical exam by clinician + end_goal_description: "full routine exam (stethoscope, blood pressure cuff, otoscope, temperature) completed without refusal or safety behaviors" + shaping_steps: + - "Enter and sit in the exam room" + - "Tolerate stethoscope contact for 5 seconds" + - "Tolerate blood pressure cuff on arm" + - "Tolerate otoscope in ear" + - "Tolerate temperature reading" + - "Tolerate full routine exam" + + - id: home + name: Home Skills + learner_profiles: [adolescent, adult] + skills: + - name: making the bed + chain_type_preferred: forward + steps: + - Strip any disheveled linens and lay them flat. + - Straighten the bottom sheet across the mattress. + - Place the top sheet evenly over the mattress. + - Tuck the top sheet in at the foot of the bed. + - Place the comforter over the top sheet. + - Place pillows at the head of the bed. + - Smooth the comforter surface. + - name: meal cleanup + chain_type_preferred: forward + steps: + - Stack used dishes on one side of the table. + - Carry dishes to the sink or dishwasher. + - Rinse food residue into the trash or disposal. + - Load dishes into the dishwasher (or wash by hand). + - Wipe the table with a damp cloth. + - Put away any leftover food into containers. + - Place containers in the refrigerator. + - name: dishwashing by hand + chain_type_preferred: forward + steps: + - Scrape food residue into the trash. + - Fill the sink with warm, soapy water. + - Wash each dish systematically with a sponge. + - Rinse each dish under clean water. + - Place each dish on the drying rack. + - Drain the sink and wipe it down. + - name: operating the dishwasher + chain_type_preferred: forward + steps: + - Collect dirty dishes from the table and sink area. + - Scrape any solid food residue into the trash. + - Rinse dishes briefly under water. + - Open the dishwasher door. + - Load plates and bowls into the bottom rack. + - Load cups, glasses, and small items into the top rack. + - Place utensils handle-down in the utensil basket. + - Add detergent to the detergent dispenser. + - Close the dishwasher door. + - Select the appropriate wash cycle and press Start. + - name: unloading the dishwasher + chain_type_preferred: forward + steps: + - Open the dishwasher once the cycle is complete. + - Unload the top rack first (cups, glasses, small items). + - Put cups and glasses away in the correct cabinet. + - Unload the utensil basket. + - Put utensils away in the drawer. + - Unload the bottom rack (plates, bowls, larger items). + - Put plates and bowls away in the correct cabinet. + - Close the dishwasher door. + - name: doing laundry + chain_type_preferred: total_task + steps: + - Sort laundry by color (lights / darks / whites). + - Load the washing machine with one color group. + - Add detergent according to instructions. + - Select the correct wash cycle and start the machine. + - When complete, move clothes to the dryer. + - Start the dryer on the correct cycle. + - When dry, fold clothes by category. + - Put folded clothes away in drawers or closet. + - name: simple meal preparation (sandwich) + chain_type_preferred: total_task + steps: + - Collect all ingredients (bread, filling, spread). + - Place two slices of bread on the plate. + - Apply spread to one slice. + - Place filling on the spread slice. + - Place the second slice on top. + - Cut the sandwich in half if desired. + - Place on a plate to serve. + + - id: community + name: Community Participation Skills + learner_profiles: [adolescent, adult] + skills: + - name: crossing a street safely + chain_type_preferred: total_task + steps: + - Approach the crosswalk and stop at the curb. + - Look left, look right, look left again. + - Identify the walk signal or absence of traffic. + - Begin to cross while continuing to scan for vehicles. + - Cross directly to the opposite curb. + - Step up onto the sidewalk. + - name: ordering in a restaurant + chain_type_preferred: total_task + steps: + - Approach the counter or wait to be seated. + - Review the menu. + - Decide on a choice. + - Place the order verbally (or with an AAC device). + - Respond to clarifying questions. + - Pay using money or a card. + - Wait for the order and collect it. + - name: using public transportation + chain_type_preferred: forward + steps: + - Confirm the correct stop or station. + - Purchase or have the fare ready. + - Wait at the designated boarding area. + - Board the vehicle when it arrives. + - Pay the fare or present the pass. + - Sit or stand safely. + - Monitor the route for the correct stop. + - Exit at the correct stop. + + - id: vocational + name: Vocational Skills + learner_profiles: [adolescent, adult] + skills: + - name: clocking in at start of shift + chain_type_preferred: total_task + steps: + - Arrive at the workplace by the scheduled time. + - Locate the time clock or digital system. + - Enter identification (ID card / badge / code). + - Confirm the clock-in timestamp. + - Notify the supervisor or check the day's task list. + - Proceed to the assigned workstation. + - name: completing assigned tasks in order + chain_type_preferred: forward + steps: + - Review the task list for the shift. + - Gather needed materials for the first task. + - Complete the first task to the specified standard. + - Mark the first task complete on the list. + - Return materials to storage. + - Move to the next task on the list and repeat. + + - id: independent_living + name: Independent Living Skills + learner_profiles: [adult] + skills: + - name: taking medication as prescribed + chain_type_preferred: total_task + steps: + - Check the medication schedule at the scheduled time. + - Collect the correct medication bottle. + - Confirm the dose against the label. + - Measure or count the correct dose. + - Take the medication with water. + - Log the dose in the medication tracker. + - Return the bottle to its storage location. + - name: scheduling a medical appointment + chain_type_preferred: forward + steps: + - Identify the reason for the appointment. + - Locate the provider's contact information. + - Call or use the online system. + - Provide name, date of birth, and insurance information. + - Request an available appointment time. + - Confirm the appointment and save it to the calendar. diff --git a/configs/task_analysis/template.yaml b/configs/task_analysis/template.yaml new file mode 100644 index 0000000000000000000000000000000000000000..51f2a73a50e7f614cced0c11d2355886e9be0c89 --- /dev/null +++ b/configs/task_analysis/template.yaml @@ -0,0 +1,82 @@ +# Task Analysis teaching program template — TRACE v1 + +template_id: task1_task_analysis_v1 +task_type: teaching_program +method_id: task_analysis + +system_prompt: | + You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help + Board Certified Behavior Analysts and staff design teaching programs for + individuals with autism. Your responses are clinically accurate, individualized + to the learner profile, follow BACB ethical guidelines, and reference no real + client data. Select the appropriate teaching method (DTT, NET, Task Analysis, + FCT, BST, PRT) based on the skill target and learner profile. + +user_variants: + - | + Generate a teaching program for the following target. + + Skill Target: {skill_target} + Curriculum Reference: {curriculum_ref} + Learner Profile: {learner_profile_name} + Current Mastery: {mastery_state_name} + Teaching Method: Task Analysis / Chaining + + Provide the full program structure appropriate to the selected method. + + - | + I need a Task Analysis program for this multi-step skill. + + Target: {skill_target} + Curriculum: {curriculum_ref} + Learner: {learner_profile_name} + Mastery Status: {mastery_state_name} + + Please include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria. + + - | + Design a chaining program for the following. + + Skill: {skill_target} + Curriculum: {curriculum_ref} + Learner Level: {learner_profile_name} + Baseline: {mastery_state_name} + + Include the full Task Analysis structure. + +assistant_template: | + ## Program Overview + - Target skill: {skill_target} + - Curriculum: {curriculum_ref} + - Learner: {learner_profile_name} + - Current mastery: {mastery_state_name} + + This Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently. + + ## Task Analysis + Chain type: {chain_type_name} + {chain_type_description} + + Steps: + {step_list} + + ## Prompt Strategy Per Step + Strategy: {prompt_strategy_name} + {prompt_strategy_description} + Current prompt level: At the {mastery_state_short} stage, {current_prompt_guidance} + + ## Error Correction Procedure + {error_correction_name}: {error_correction_steps} + + ## Reinforcement + {reinforcement_schedule_name} — {reinforcement_description}. + Pair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine). + + ## Mastery Criteria + {mastery_criterion_text} + + ## Data Collection + Per-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every {ioa_frequency} sessions minimum; target agreement ≥ 80%. + + ## Generalization & Maintenance Plan + Post-mastery: probe the full chain across {n_generalization_settings} different settings and with {n_generalization_therapists} different adults. Maintenance probes weekly for 4 weeks, then monthly. diff --git a/configs/task_analysis/template_toleration.yaml b/configs/task_analysis/template_toleration.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aa35fb72193eb5c5a5caa9f9c69bb7c8293e43d2 --- /dev/null +++ b/configs/task_analysis/template_toleration.yaml @@ -0,0 +1,91 @@ +# Toleration / Systematic-Desensitization program template — TRACE v1 +# Used for skills where the goal is NOT learner independence but rather the +# learner *allowing* caregiver/staff to perform the activity, with duration +# shaped incrementally across trials. + +template_id: task1_toleration_v1 +task_type: teaching_program +method_id: task_analysis # Administratively in the task_analysis area; clinically distinct approach + +system_prompt: | + You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help + Board Certified Behavior Analysts and staff design teaching programs for + individuals with autism. Your responses are clinically accurate, individualized + to the learner profile, follow BACB ethical guidelines, and reference no real + client data. Select the appropriate teaching method (DTT, NET, Task Analysis, + FCT, BST, PRT) based on the skill target and learner profile. + +user_variants: + - | + Generate a teaching program for the following target. + + Skill Target: {skill_target} + Curriculum Reference: {curriculum_ref} + Learner Profile: {learner_profile_name} + Current Mastery: {mastery_state_name} + Teaching Method: Toleration / Systematic Desensitization + + The goal is for the learner to allow the activity to be performed (not to perform it independently). Provide the full program structure including shaping steps and refusal/safety procedures. + + - | + I need a toleration program for this self-care / medical / caregiver-delivered skill. + + Target: {skill_target} + Curriculum: {curriculum_ref} + Learner: {learner_profile_name} + Mastery Status: {mastery_state_name} + + Please include the toleration end goal, shaping progression, antecedent strategies, reinforcement plan, refusal / safety procedures, and mastery criteria. + + - | + Design a systematic desensitization / toleration program for the following. + + Skill: {skill_target} + Curriculum: {curriculum_ref} + Learner Level: {learner_profile_name} + Baseline: {mastery_state_name} + + The learner currently refuses or exhibits safety behaviors during the activity; the program should build tolerance gradually. + +assistant_template: | + ## Program Overview + - Target skill: {skill_target} + - Curriculum: {curriculum_ref} + - Learner: {learner_profile_name} + - Current mastery: {mastery_state_name} + - Approach: Toleration / systematic desensitization (caregiver-delivered activity) + + This program builds tolerance for {target_activity} through systematic shaping. The learner is not being taught to perform the skill independently — the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression. + + ## Toleration End Goal + {end_goal_description} + + ## Shaping Progression + Teach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors. + + {shaping_step_list} + + Current shaping step: At the {mastery_state_short} stage, {current_step_guidance} + + ## Antecedent Strategies + - Pre-warn the learner before the activity begins (verbal or visual schedule). + - Offer a preferred item or sensory tool to hold during the activity. + - Where possible, offer a choice of position, location, or starting point. + - Keep the environment calm and predictable; avoid other demands immediately before or after. + + ## Reinforcement + Deliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step. + + ## Refusal and Safety Procedures + - If the learner refuses or turns away: pause briefly (5–10 seconds), re-present the activity at the current or previous shaping step. + - If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident. + - Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern — consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner. + + ## Mastery Criteria + {mastery_criterion_text} + + ## Data Collection + Per-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every {ioa_frequency} sessions minimum; target agreement ≥ 80%. + + ## Generalization & Maintenance Plan + Post-mastery: probe the full routine across {n_generalization_settings} different settings and with {n_generalization_therapists} different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly. diff --git a/data/splits/sanity.jsonl b/data/splits/sanity.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..852e79d3f35de43a10c64b1486d92050b05e15ae --- /dev/null +++ b/data/splits/sanity.jsonl @@ -0,0 +1,20 @@ +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: meal cleanup\nCurriculum Reference: AFLS Home Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Generalization phase\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: meal cleanup\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Stack used dishes on one side of the table.\n2. Carry dishes to the sink or dishwasher.\n3. Rinse food residue into the trash or disposal.\n4. Load dishes into the dishwasher (or wash by hand).\n5. Wipe the table with a damp cloth.\n6. Put away any leftover food into containers.\n7. Place containers in the refrigerator. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 120s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5d7083ebc56a4c74", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "meal cleanup", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "generalization", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "meal cleanup", "module": "home", "mastery_criterion": "mc_fluent_chain", "chain_type": "backward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "92010"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner: Adult Learner\nMastery Status: Approaching mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "61858acdbd184cf8", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adult", "mastery_state": "approaching", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_all_steps_independent", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "21081"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: washing hands\nCurriculum: AFLS Basic Living Skills\nLearner: Adolescent Learner\nMastery Status: Approaching mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: washing hands\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Walk to the sink.\n2. Turn on the water.\n3. Wet hands under the water.\n4. Apply soap to hands.\n5. Rub hands together for 15\u201320 seconds, covering fronts, backs, and between fingers.\n6. Rinse hands thoroughly under the water.\n7. Turn off the water.\n8. Dry hands with a towel.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nDifferential per-step + terminal \u2014 Small per-step reinforcement for independent responses; larger reinforcer at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 120s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a048233b2ba858c1", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "washing hands", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "approaching", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "washing hands", "module": "basic_living", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "differential_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "92010"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies 5\u201310 upper-case letters\nCurriculum Reference: Reading L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies 5\u201310 upper-case letters\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: J, A, D\nDistractor stimuli: E, F\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7ee5a348ee9d2761", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies 5\u201310 upper-case letters", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "token_economy", "error_correction": "four_step", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "69064"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: traces basic shapes (circle, cross)\nCurriculum: Writing L1\nLearner Level: Early Learner\nBaseline: Mastered at current level\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces basic shapes (circle, cross)\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: triangle, star, rectangle\nDistractor stimuli: circle, heart\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "65ca50da125ad263", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces basic shapes (circle, cross)", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "11906"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects items given 2+ features/functions simultaneously\nCurriculum Reference: LRFFC L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items given 2+ features/functions simultaneously\n- Curriculum: LRFFC L3\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: backpack, sock, pencil\nDistractor stimuli: key, car, cup, blanket, scissors\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "75278fd4d843e7b5", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L3", "learner_profile": "adolescent", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items given 2+ features/functions simultaneously", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "27367"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: sounds out CVC words (cat, dog, run)\nCurriculum Reference: Reading L2\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: sounds out CVC words (cat, dog, run)\n- Curriculum: Reading L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What word?\"\nVariations: \"Read this word\"; \"What does it say?\"\nPresentation: Present the sight-word card and deliver the SD.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: brush, blanket, towel\nDistractor stimuli: key, glue, apple, jacket\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "82ff6a74b2a01ba6", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "sounds out CVC words (cat, dog, run)", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "89319"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: compares quantities (more/less/same)\nCurriculum Reference: Math L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: compares quantities (more/less/same)\n- Curriculum: Math L2\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How many?\"\nVariations: \"Count these\"; \"What number is this?\"\nPresentation: Present the numeral card or a set of objects to count.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: key, block, blanket\nDistractor stimuli: ball, bottle, spoon, book\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "260624f458b79c7b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L2", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "compares quantities (more/less/same)", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "fr2", "error_correction": "four_step", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "97173"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9194\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Intraverbal + Mand + Writing\nDate range: Sessions 1\u201310 across 20 days (2026-05-06 to 2026-05-26)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n4. echoes 2-word phrases (Echoic)\n Method: DTT variant\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-06 \u2014 34 min \u2014 1 observer\n copies simple words from model: 5/11 correct (43%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n answers WH-questions about familiar topics: 5/14 correct (36%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n indicating completion ('all done', 'finished'): 4/11 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n echoes 2-word phrases: 4/9 correct (46%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n\nSession 2 \u2014 2026-05-08 \u2014 34 min \u2014 1 observer\n copies simple words from model: 6/13 correct (48%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d711\n answers WH-questions about familiar topics: 6/13 correct (44%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n indicating completion ('all done', 'finished'): 5/15 correct (34%); latency 4.8s; prompts FP\u00d73, PP\u00d70, G\u00d75, I\u00d77\n echoes 2-word phrases: 7/15 correct (48%); latency 3.6s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d79\n\nSession 3 \u2014 2026-05-10 \u2014 34 min \u2014 1 observer\n copies simple words from model: 9/14 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d76\n answers WH-questions about familiar topics: 6/12 correct (47%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d77\n indicating completion ('all done', 'finished'): 5/13 correct (41%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n echoes 2-word phrases: 7/14 correct (53%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n\nSession 4 \u2014 2026-05-12 \u2014 34 min \u2014 1 observer\n copies simple words from model: 10/14 correct (69%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n answers WH-questions about familiar topics: 6/11 correct (55%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n indicating completion ('all done', 'finished'): 6/12 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n echoes 2-word phrases: 7/11 correct (60%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n\nSession 5 \u2014 2026-05-14 \u2014 34 min \u2014 1 observer\n copies simple words from model: 10/13 correct (79%); latency 2.3s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n answers WH-questions about familiar topics: 8/12 correct (63%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n indicating completion ('all done', 'finished'): 7/12 correct (58%); latency 3.2s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d74\n echoes 2-word phrases: 8/12 correct (68%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n\nSession 6 \u2014 2026-05-16 \u2014 IOA SESSION \u2014 2 observers\n copies simple words from model: 88% (primary); IOA 93% agreement\n answers WH-questions about familiar topics: 70% (primary); IOA 93% agreement\n indicating completion ('all done', 'finished'): 63% (primary); IOA 93% agreement\n echoes 2-word phrases: 76% (primary); IOA 93% agreement\n\nSession 7 \u2014 2026-05-18 \u2014 34 min \u2014 1 observer\n copies simple words from model: 7/8 correct (90%); latency 1.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 11/13 correct (81%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n indicating completion ('all done', 'finished'): 10/15 correct (70%); latency 2.6s; prompts FP\u00d74, PP\u00d74, G\u00d75, I\u00d72\n echoes 2-word phrases: 9/10 correct (87%); latency 1.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n\nSession 8 \u2014 2026-05-20 \u2014 34 min \u2014 1 observer\n copies simple words from model: 10/10 correct (95%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n answers WH-questions about familiar topics: 7/8 correct (83%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 7/10 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n echoes 2-word phrases: 10/11 correct (89%); latency 1.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n\nSession 9 \u2014 2026-05-22 \u2014 34 min \u2014 1 observer\n copies simple words from model: 13/14 correct (95%); latency 1.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n answers WH-questions about familiar topics: 11/12 correct (92%); latency 2.1s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d76\n indicating completion ('all done', 'finished'): 8/10 correct (78%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n echoes 2-word phrases: 14/15 correct (95%); latency 1.5s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d78\n\nSession 10 \u2014 2026-05-24 \u2014 34 min \u2014 1 observer\n copies simple words from model: 9/9 correct (95%); latency 1.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n answers WH-questions about familiar topics: 11/12 correct (95%); latency 1.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n indicating completion ('all done', 'finished'): 10/12 correct (82%); latency 2.5s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n echoes 2-word phrases: 9/9 correct (95%); latency 1.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Reaching for materials proactively\n- Requesting more trials verbally\n- Quick response latency (within 2\u20133 seconds)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 43% to 95% across 10 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 76.8%.\n- Trend direction: upward (first session: 43%, last session: 95%).\n- IOA data: 93% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "79e0e2ac323f6c4d", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "early", "n_sessions": 10, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "55304"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7707\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-07-02 to 2026-07-22)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. requesting preferred items using AAC device (Mand)\n Method: NET\n3. identifying community safety signs (Community)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-02 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/12 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n requesting preferred items using AAC device: 8/15 correct (55%); latency 3.5s; prompts FP\u00d75, PP\u00d75, G\u00d75, I\u00d70\n identifying community safety signs: 5/9 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n\nSession 2 \u2014 2026-07-04 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/10 correct (44%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 8/14 correct (60%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d79\n identifying community safety signs: 4/8 correct (54%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n\nSession 3 \u2014 2026-07-06 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/14 correct (53%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d75\n requesting preferred items using AAC device: 6/13 correct (49%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n identifying community safety signs: 4/8 correct (54%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 4 \u2014 2026-07-08 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/15 correct (52%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d712\n requesting preferred items using AAC device: 4/8 correct (52%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n identifying community safety signs: 5/10 correct (49%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n\nSession 5 \u2014 2026-07-10 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/12 correct (47%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d74\n requesting preferred items using AAC device: 4/10 correct (43%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n identifying community safety signs: 8/15 correct (56%); latency 3.5s; prompts FP\u00d71, PP\u00d75, G\u00d73, I\u00d76\n\nSession 6 \u2014 2026-07-12 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/15 correct (43%); latency 4.0s; prompts FP\u00d71, PP\u00d75, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 8/14 correct (55%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n identifying community safety signs: 8/13 correct (62%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n\nSession 7 \u2014 2026-07-14 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/13 correct (44%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n requesting preferred items using AAC device: 7/14 correct (47%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n identifying community safety signs: 5/9 correct (58%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n\nSession 8 \u2014 2026-07-16 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/10 correct (37%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n requesting preferred items using AAC device: 5/13 correct (40%); latency 3.9s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n identifying community safety signs: 9/13 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n\nSession 9 \u2014 2026-07-18 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/14 correct (32%); latency 4.4s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n requesting preferred items using AAC device: 6/15 correct (39%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n identifying community safety signs: 10/15 correct (66%); latency 3.2s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n\nSession 10 \u2014 2026-07-20 \u2014 35 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/12 correct (39%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 3/8 correct (33%); latency 4.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n identifying community safety signs: 7/11 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 48% for 10 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 44.9%.\n- Trend direction: downward (first session: 57%, last session: 39%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a989003c3a53274e", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "70612"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands to indicate completion in context (\"I'm all done with math, can I start reading?\")\nCurriculum: Mand L3\nLearner: School-Age Learner\nMastery Status: Approaching mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands to indicate completion in context (\"I'm all done with math, can I start reading?\")\n- Curriculum: Mand L3\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nSet up a clearly-bounded activity with a visible endpoint (e.g., a worksheet, a puzzle, a sorting task). Wait for the learner to indicate completion verbally or via AAC; honor the indication immediately.\nPrimary natural context: worksheet.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's worksheet routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: the activity ends immediately and the learner transitions to the next preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "d3c962bf13d35ef1", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L3", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands to indicate completion in context (\"I'm all done with math, can I start reading?\")", "mo_category": "mand_completion", "mo_arrangement": "completion_opportunity", "prompt_strategy": "time_delay", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "76914"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: negotiates and compromises during group activities\nCurriculum: Social Behavior & Social Play L3\nLearner Level: Adolescent Learner\nBaseline: Emerging\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: negotiates and compromises during group activities\n- Curriculum: Social Behavior & Social Play L3\n- Learner: Adolescent Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: peer play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's peer play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "73dbfb550a0bf49a", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L3", "learner_profile": "adolescent", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "negotiates and compromises during group activities", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "44373"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: spontaneously provides explanations and reasons\nCurriculum: Spontaneous Vocal Behavior L3\nLearner Level: School-Age Learner\nBaseline: Generalization phase\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously provides explanations and reasons\n- Curriculum: Spontaneous Vocal Behavior L3\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: room transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's room transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "52367b6070b84c39", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L3", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously provides explanations and reasons", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "75835"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1411\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Self-Care\nDate range: Sessions 1\u201312 across 24 days (2026-07-24 to 2026-08-17)\n\nACCELERATION PROGRAMS\n1. identifying community safety signs (Community)\n Method: DTT\n2. making the bed (Home Skills)\n Method: Task Analysis\n3. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n4. washing hands independently (Self-Care)\n Method: Task Analysis\n5. requesting help when a task is difficult (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Motor stereotypy \u2014 function hypothesized: automatic\n2. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-24 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 12/15 correct (79%); latency 2.4s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n making the bed: 11/14 correct (76%); latency 2.4s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n tolerating tooth brushing by caregiver: 10/13 correct (81%); latency 2.3s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n washing hands independently: 12/15 correct (78%); latency 2.4s; prompts FP\u00d74, PP\u00d75, G\u00d74, I\u00d72\n requesting help when a task is difficult: 8/12 correct (70%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n Motor stereotypy: freq 1; PIR 7%\n Property destruction: freq 2\n\nSession 2 \u2014 2026-07-26 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 7/10 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n making the bed: 9/13 correct (72%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d76\n tolerating tooth brushing by caregiver: 12/15 correct (83%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n washing hands independently: 8/11 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n requesting help when a task is difficult: 8/13 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n Motor stereotypy: freq 2; PIR 18%\n Property destruction: freq 2\n\nSession 3 \u2014 2026-07-28 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 11/15 correct (70%); latency 2.8s; prompts FP\u00d75, PP\u00d74, G\u00d72, I\u00d74\n making the bed: 5/8 correct (65%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n tolerating tooth brushing by caregiver: 9/11 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n washing hands independently: 9/15 correct (60%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n requesting help when a task is difficult: 9/15 correct (61%); latency 3.2s; prompts FP\u00d71, PP\u00d75, G\u00d71, I\u00d78\n Motor stereotypy: freq 2; PIR 17%\n Property destruction: freq 3\n\nSession 4 \u2014 2026-07-30 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 8/13 correct (65%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n making the bed: 5/10 correct (49%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n tolerating tooth brushing by caregiver: 11/15 correct (73%); latency 2.7s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n washing hands independently: 7/12 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n requesting help when a task is difficult: 6/10 correct (57%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n Motor stereotypy: freq 3; PIR 14%\n Property destruction: freq 3\n\nSession 5 \u2014 2026-08-01 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 7/10 correct (67%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n making the bed: 7/13 correct (50%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n tolerating tooth brushing by caregiver: 7/11 correct (66%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n washing hands independently: 7/15 correct (47%); latency 3.8s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n requesting help when a task is difficult: 4/9 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n Motor stereotypy: freq 3; PIR 23%\n Property destruction: freq 4\n\nSession 6 \u2014 2026-08-03 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 8/14 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n making the bed: 3/9 correct (36%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n tolerating tooth brushing by caregiver: 9/14 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n washing hands independently: 7/15 correct (47%); latency 3.9s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n requesting help when a task is difficult: 6/12 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n Motor stereotypy: freq 3; PIR 19%\n Property destruction: freq 3\n\nSession 7 \u2014 2026-08-05 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 4/8 correct (50%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n making the bed: 2/8 correct (30%); latency 4.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n tolerating tooth brushing by caregiver: 7/12 correct (59%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n washing hands independently: 4/10 correct (35%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n requesting help when a task is difficult: 4/9 correct (43%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n Motor stereotypy: freq 4; PIR 22%\n Property destruction: freq 4\n\nSession 8 \u2014 2026-08-07 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 7/14 correct (52%); latency 3.8s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n making the bed: 2/11 correct (20%); latency 5.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n tolerating tooth brushing by caregiver: 5/9 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n washing hands independently: 3/11 correct (26%); latency 4.6s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n requesting help when a task is difficult: 4/14 correct (31%); latency 4.4s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n Motor stereotypy: freq 5; PIR 25%\n Property destruction: freq 5\n\nSession 9 \u2014 2026-08-09 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 4/9 correct (39%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n making the bed: 2/8 correct (20%); latency 5.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n tolerating tooth brushing by caregiver: 6/12 correct (50%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d75\n washing hands independently: 3/15 correct (22%); latency 5.0s; prompts FP\u00d75, PP\u00d71, G\u00d72, I\u00d77\n requesting help when a task is difficult: 4/14 correct (28%); latency 4.8s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n Motor stereotypy: freq 4; PIR 30%\n Property destruction: freq 5\n\nSession 10 \u2014 2026-08-11 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 6/15 correct (39%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d75, I\u00d76\n making the bed: 1/13 correct (10%); latency 5.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n tolerating tooth brushing by caregiver: 5/12 correct (44%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n washing hands independently: 2/9 correct (21%); latency 4.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n requesting help when a task is difficult: 2/9 correct (27%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n Motor stereotypy: freq 5; PIR 35%\n Property destruction: freq 5\n\nSession 11 \u2014 2026-08-13 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 4/14 correct (31%); latency 4.5s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n making the bed: 1/8 correct (10%); latency 5.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n tolerating tooth brushing by caregiver: 6/15 correct (41%); latency 4.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d75\n washing hands independently: 1/14 correct (10%); latency 5.6s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n requesting help when a task is difficult: 1/10 correct (14%); latency 5.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n Motor stereotypy: freq 5; PIR 23%\n Property destruction: freq 4\n\nSession 12 \u2014 2026-08-15 \u2014 44 min \u2014 1 observer\n identifying community safety signs: 3/11 correct (31%); latency 4.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n making the bed: 1/9 correct (10%); latency 5.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n tolerating tooth brushing by caregiver: 3/8 correct (35%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n washing hands independently: 1/12 correct (10%); latency 5.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n requesting help when a task is difficult: 2/14 correct (16%); latency 5.0s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n Motor stereotypy: freq 5; PIR 31%\n Property destruction: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Peer aggression (hitting or pushing peers during group activities)\n- Elopement from the work area\n- Vocal refusal ('no', 'I don't want to')\n- Increased response latency (3\u20135\u00d7 baseline)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 79% to 31% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\nProperty destruction: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 54.9%.\n- Trend direction: downward (first session: 79%, last session: 31%).\n- Motor stereotypy: mean 3.5/session (early-window mean 2.0 -> late-window mean 4.8).\n- Property destruction: mean 3.8/session (early-window mean 2.5 -> late-window mean 5.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a8fbbeda47ceb966", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Motor stereotypy": "automatic", "Property destruction": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["motor_stereotypy", "property_destruction"], "behavior_functions": ["automatic", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "19742"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9328\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u201310 across 20 days (2026-06-09 to 2026-06-29)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. brushing teeth independently (Self-Care)\n Method: Task Analysis\n3. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n4. doing laundry (Home Skills)\n Method: Task Analysis\n5. indicating 'all done' at task completion (Mand)\n Method: NET\n6. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n2. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-09 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 7/11 correct (64%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n brushing teeth independently: 3/8 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n scheduling medical appointments: 6/9 correct (64%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n doing laundry: 6/9 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n indicating 'all done' at task completion: 8/15 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d78\n wait calmly when a preferred activity is denied: 4/10 correct (42%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n Verbal aggression: freq 4\n Non-compliance: freq 3\n\nSession 2 \u2014 2026-06-11 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 8/9 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n brushing teeth independently: 7/15 correct (48%); latency 4.0s; prompts FP\u00d74, PP\u00d71, G\u00d75, I\u00d75\n scheduling medical appointments: 7/15 correct (48%); latency 3.9s; prompts FP\u00d75, PP\u00d75, G\u00d70, I\u00d75\n doing laundry: 6/9 correct (68%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n indicating 'all done' at task completion: 5/14 correct (39%); latency 4.0s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d74\n wait calmly when a preferred activity is denied: 6/13 correct (50%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d74\n Verbal aggression: freq 3\n Non-compliance: freq 3\n\nSession 3 \u2014 2026-06-13 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 4/8 correct (49%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n brushing teeth independently: 5/9 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n scheduling medical appointments: 7/11 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n doing laundry: 8/12 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n indicating 'all done' at task completion: 4/9 correct (49%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n wait calmly when a preferred activity is denied: 6/15 correct (37%); latency 4.1s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n Verbal aggression: freq 1\n Non-compliance: freq 4\n\nSession 4 \u2014 2026-06-15 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 6/11 correct (58%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n brushing teeth independently: 6/10 correct (55%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n scheduling medical appointments: 10/14 correct (74%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d78\n doing laundry: 8/12 correct (63%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n indicating 'all done' at task completion: 7/15 correct (49%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d75, I\u00d75\n wait calmly when a preferred activity is denied: 6/13 correct (50%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n Verbal aggression: freq 3\n Non-compliance: freq 1\n\nSession 5 \u2014 2026-06-17 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 8/10 correct (82%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n brushing teeth independently: 5/9 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n scheduling medical appointments: 8/11 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n doing laundry: 8/10 correct (81%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n indicating 'all done' at task completion: 6/15 correct (43%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d79\n wait calmly when a preferred activity is denied: 5/8 correct (61%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Verbal aggression: freq 2\n Non-compliance: freq 5\n\nSession 6 \u2014 2026-06-19 \u2014 IOA SESSION \u2014 2 observers\n completing assigned tasks in order: 51% (primary); IOA 96% agreement\n brushing teeth independently: 45% (primary); IOA 96% agreement\n scheduling medical appointments: 71% (primary); IOA 96% agreement\n doing laundry: 61% (primary); IOA 96% agreement\n indicating 'all done' at task completion: 46% (primary); IOA 96% agreement\n wait calmly when a preferred activity is denied: 51% (primary); IOA 96% agreement\n Verbal aggression: freq 3; IOA 96% agreement\n Non-compliance: freq 2; IOA 96% agreement\n\nSession 7 \u2014 2026-06-21 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 6/8 correct (80%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n brushing teeth independently: 5/8 correct (60%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n scheduling medical appointments: 5/8 correct (65%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n doing laundry: 5/9 correct (59%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n indicating 'all done' at task completion: 5/12 correct (39%); latency 4.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 4/13 correct (33%); latency 4.5s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d77\n Verbal aggression: freq 4\n Non-compliance: freq 6\n\nSession 8 \u2014 2026-06-23 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 8/13 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n brushing teeth independently: 4/9 correct (50%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n scheduling medical appointments: 5/10 correct (53%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n doing laundry: 6/11 correct (51%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n indicating 'all done' at task completion: 2/10 correct (24%); latency 4.9s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 5/10 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n Verbal aggression: freq 1\n Non-compliance: freq 2\n\nSession 9 \u2014 2026-06-25 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 10/13 correct (78%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n brushing teeth independently: 4/8 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n scheduling medical appointments: 11/15 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d712\n doing laundry: 5/8 correct (67%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n indicating 'all done' at task completion: 4/12 correct (30%); latency 4.9s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 6/11 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n Verbal aggression: freq 2\n Non-compliance: freq 3\n\nSession 10 \u2014 2026-06-27 \u2014 52 min \u2014 1 observer\n completing assigned tasks in order: 13/15 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d75, G\u00d72, I\u00d75\n brushing teeth independently: 6/9 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n scheduling medical appointments: 12/15 correct (77%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n doing laundry: 7/15 correct (50%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n indicating 'all done' at task completion: 6/10 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 6/9 correct (61%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n Verbal aggression: freq 4\n Non-compliance: freq 4\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 74% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 69.2%.\n- Trend direction: upward (first session: 64%, last session: 84%).\n- Verbal aggression: mean 2.7/session (early-window mean 2.7 -> late-window mean 2.3).\n- Non-compliance: mean 3.3/session (early-window mean 3.3 -> late-window mean 3.0).\n- IOA data: 96% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "1fb68e3f6908bb1b", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Verbal aggression": "escape", "Non-compliance": "escape"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adult", "n_sessions": 10, "n_programs": 6, "n_behaviors": 2, "behavior_ids": ["verbal_aggression", "non_compliance"], "behavior_functions": ["escape", "escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "346"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5412\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-01-14 to 2026-01-30)\n\nACCELERATION PROGRAMS\n1. indicating 'all done' at task completion (Mand)\n Method: NET\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-14 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 11/13 correct (82%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n self-advocacy statements: 13/15 correct (85%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n brushing teeth independently: 10/14 correct (69%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d76\n Verbal aggression: freq 2\n ABC (verbal aggression): A = demand to complete math problem; B = threw materials and vocalized 'no'; C = staff removed demand, redirected\n\nSession 2 \u2014 2026-01-16 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 6/8 correct (79%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n self-advocacy statements: 8/10 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n brushing teeth independently: 7/10 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n Verbal aggression: freq 4\n\nSession 3 \u2014 2026-01-18 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 9/12 correct (72%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n self-advocacy statements: 8/12 correct (69%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n brushing teeth independently: 9/14 correct (65%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n Verbal aggression: freq 4\n\nSession 4 \u2014 2026-01-20 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 5/8 correct (60%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n self-advocacy statements: 8/12 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n brushing teeth independently: 6/12 correct (54%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d75\n Verbal aggression: freq 5\n\nSession 5 \u2014 2026-01-22 \u2014 IOA SESSION \u2014 2 observers\n indicating 'all done' at task completion: 54% (primary); IOA 91% agreement\n self-advocacy statements: 61% (primary); IOA 91% agreement\n brushing teeth independently: 54% (primary); IOA 91% agreement\n Verbal aggression: freq 6; IOA 91% agreement\n\nSession 6 \u2014 2026-01-24 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 5/9 correct (55%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n self-advocacy statements: 5/10 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n brushing teeth independently: 5/12 correct (46%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d75\n Verbal aggression: freq 6\n ABC (verbal aggression): A = asked to finish dressing task; B = kicked at materials; C = task paused for 3 min\n\nSession 7 \u2014 2026-01-26 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 4/9 correct (42%); latency 4.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n self-advocacy statements: 6/12 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n brushing teeth independently: 4/9 correct (41%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n Verbal aggression: freq 7\n\nSession 8 \u2014 2026-01-28 \u2014 34 min \u2014 1 observer\n indicating 'all done' at task completion: 3/9 correct (36%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n self-advocacy statements: 4/9 correct (45%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n brushing teeth independently: 3/9 correct (38%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n Verbal aggression: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-stimulatory behavior increase during demand periods\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Elopement from the work area\n- Crying during demand presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (82% -> 36%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 60.1%.\n- Trend direction: downward (first session: 82%, last session: 36%).\n- Verbal aggression: mean 5.0/session (early-window mean 3.0 -> late-window mean 6.5).\n- IOA data: 91% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "5dcff7d20794a865", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Verbal aggression": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adult", "n_sessions": 8, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["verbal_aggression"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "97091"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7461\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201312 across 24 days (2026-04-27 to 2026-05-21)\n\nACCELERATION PROGRAMS\n1. requesting a break during non-preferred activities (Mand)\n Method: NET\n2. making the bed (Home Skills)\n Method: Task Analysis\n3. tolerating haircut (Self-Care)\n Method: Task Analysis\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-27 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/8 correct (70%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n making the bed: 8/11 correct (69%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n tolerating haircut: 9/12 correct (76%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 9/12 correct (71%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n\nSession 2 \u2014 2026-04-29 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 7/10 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n making the bed: 7/10 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n tolerating haircut: 10/13 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 7/10 correct (68%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n\nSession 3 \u2014 2026-05-01 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 8/12 correct (68%); latency 3.0s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n making the bed: 10/14 correct (69%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n tolerating haircut: 10/13 correct (75%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n wait calmly when access to preferred item is denied: 11/15 correct (73%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d79\n\nSession 4 \u2014 2026-05-03 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/15 correct (67%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d77\n making the bed: 8/12 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n tolerating haircut: 10/15 correct (70%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n wait calmly when access to preferred item is denied: 10/15 correct (67%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d75, I\u00d77\n\nSession 5 \u2014 2026-05-05 \u2014 IOA SESSION \u2014 2 observers\n requesting a break during non-preferred activities: 46% (primary); IOA 91% agreement\n making the bed: 46% (primary); IOA 91% agreement\n tolerating haircut: 47% (primary); IOA 91% agreement\n wait calmly when access to preferred item is denied: 40% (primary); IOA 91% agreement\n\nSession 6 \u2014 2026-05-07 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/11 correct (52%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n making the bed: 7/14 correct (53%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n tolerating haircut: 4/8 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n wait calmly when access to preferred item is denied: 3/8 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n\nSession 7 \u2014 2026-05-09 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 7/12 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n making the bed: 4/8 correct (56%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n tolerating haircut: 6/11 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 7/13 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d710\n\nSession 8 \u2014 2026-05-11 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 9/15 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d78\n making the bed: 7/12 correct (60%); latency 3.2s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d74\n tolerating haircut: 8/14 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 7/11 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n\nSession 9 \u2014 2026-05-13 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 9/14 correct (67%); latency 3.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n making the bed: 10/15 correct (66%); latency 3.1s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d73\n tolerating haircut: 9/15 correct (60%); latency 3.5s; prompts FP\u00d71, PP\u00d75, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 8/12 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n\nSession 10 \u2014 2026-05-15 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/9 correct (67%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n making the bed: 5/8 correct (66%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n tolerating haircut: 6/9 correct (69%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 9/14 correct (68%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n\nSession 11 \u2014 2026-05-17 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/9 correct (67%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n making the bed: 8/12 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n tolerating haircut: 11/15 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d710\n wait calmly when access to preferred item is denied: 9/14 correct (68%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n\nSession 12 \u2014 2026-05-19 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/15 correct (67%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d75\n making the bed: 10/15 correct (66%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d78\n tolerating haircut: 6/8 correct (73%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n wait calmly when access to preferred item is denied: 10/15 correct (68%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d710\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 12 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 63.4%.\n- Trend direction: downward (first session: 70%, last session: 67%).\n- IOA data: 91% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "a42e76febc72b1f1", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "33957"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3679\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Reading + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-05-22 to 2026-06-07)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. brushing teeth independently (Self-Care)\n Method: Task Analysis\n3. requesting a break during non-preferred activities (Mand)\n Method: NET\n4. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-22 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 10/15 correct (64%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d78\n brushing teeth independently: 7/13 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n requesting a break during non-preferred activities: 5/10 correct (53%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n tolerating tooth brushing by caregiver: 5/9 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n Non-compliance: freq 3\n\nSession 2 \u2014 2026-05-24 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 7/9 correct (73%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n brushing teeth independently: 5/8 correct (65%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n requesting a break during non-preferred activities: 7/14 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n tolerating tooth brushing by caregiver: 6/11 correct (54%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n Non-compliance: freq 3\n\nSession 3 \u2014 2026-05-26 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 8/12 correct (67%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n brushing teeth independently: 6/9 correct (61%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n requesting a break during non-preferred activities: 4/8 correct (53%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n tolerating tooth brushing by caregiver: 8/14 correct (60%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d77\n Non-compliance: freq 9\n\nSession 4 \u2014 2026-05-28 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 8/12 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n brushing teeth independently: 6/9 correct (67%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n requesting a break during non-preferred activities: 4/8 correct (55%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n tolerating tooth brushing by caregiver: 8/12 correct (67%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n Non-compliance: freq 9\n\nSession 5 \u2014 2026-05-30 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 6/8 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n brushing teeth independently: 6/9 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n requesting a break during non-preferred activities: 10/15 correct (64%); latency 3.0s; prompts FP\u00d72, PP\u00d75, G\u00d73, I\u00d75\n tolerating tooth brushing by caregiver: 7/11 correct (67%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n Non-compliance: freq 4\n\nSession 6 \u2014 2026-06-01 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 9/11 correct (81%); latency 2.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n brushing teeth independently: 9/13 correct (69%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d710\n requesting a break during non-preferred activities: 7/11 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n tolerating tooth brushing by caregiver: 9/14 correct (67%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n Non-compliance: freq 2\n\nSession 7 \u2014 2026-06-03 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 11/14 correct (79%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d78\n brushing teeth independently: 5/8 correct (65%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n requesting a break during non-preferred activities: 7/11 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n tolerating tooth brushing by caregiver: 8/12 correct (70%); latency 3.1s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n Non-compliance: freq 3\n\nSession 8 \u2014 2026-06-05 \u2014 43 min \u2014 1 observer\n sounds out CVC words: 11/14 correct (81%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n brushing teeth independently: 8/10 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n requesting a break during non-preferred activities: 7/11 correct (63%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n tolerating tooth brushing by caregiver: 9/12 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n Non-compliance: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Vocal refusal ('no', 'I don't want to')\n- Attempts at property destruction (throwing materials, swiping items off the surface)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 73.1%.\n- Trend direction: upward (first session: 64%, last session: 81%).\n- Non-compliance: mean 4.4/session (early-window mean 3.0 -> late-window mean 2.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e7dfdbd78593b422", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Non-compliance": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["non_compliance"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "78235"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1789\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Self-Care\nDate range: Sessions 1\u201310 across 20 days (2026-03-13 to 2026-04-02)\n\nACCELERATION PROGRAMS\n1. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n2. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n3. making the bed (Home Skills)\n Method: Task Analysis\n4. indicating 'all done' with a step or activity (Mand)\n Method: NET\n5. requesting help when a task is difficult (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-13 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 9/12 correct (76%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d76\n dressing (pants, shirt): 9/11 correct (79%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n making the bed: 13/15 correct (86%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n indicating 'all done' with a step or activity: 9/12 correct (74%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n requesting help when a task is difficult: 7/10 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n\nSession 2 \u2014 2026-03-15 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 8/10 correct (80%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n dressing (pants, shirt): 11/14 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n making the bed: 8/9 correct (84%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n indicating 'all done' with a step or activity: 9/12 correct (78%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n requesting help when a task is difficult: 8/12 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d76\n\nSession 3 \u2014 2026-03-17 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 8/11 correct (76%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n dressing (pants, shirt): 7/9 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n making the bed: 10/12 correct (86%); latency 2.3s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d72\n indicating 'all done' with a step or activity: 8/11 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n requesting help when a task is difficult: 7/9 correct (74%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n\nSession 4 \u2014 2026-03-19 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 6/15 correct (38%); latency 4.4s; prompts FP\u00d72, PP\u00d74, G\u00d75, I\u00d74\n dressing (pants, shirt): 6/14 correct (46%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n making the bed: 6/11 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n indicating 'all done' with a step or activity: 4/11 correct (37%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n requesting help when a task is difficult: 6/13 correct (47%); latency 4.2s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n\nSession 5 \u2014 2026-03-21 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 5/11 correct (41%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n dressing (pants, shirt): 3/8 correct (40%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n making the bed: 6/12 correct (48%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 5/12 correct (39%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d75\n requesting help when a task is difficult: 5/11 correct (48%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n\nSession 6 \u2014 2026-03-23 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 4/9 correct (39%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n dressing (pants, shirt): 5/10 correct (47%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n making the bed: 7/13 correct (52%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d77\n indicating 'all done' with a step or activity: 5/14 correct (36%); latency 4.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n requesting help when a task is difficult: 5/10 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n\nSession 7 \u2014 2026-03-25 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 5/13 correct (42%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n dressing (pants, shirt): 6/11 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n making the bed: 7/13 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d79\n indicating 'all done' with a step or activity: 5/11 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n requesting help when a task is difficult: 7/12 correct (59%); latency 3.5s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n\nSession 8 \u2014 2026-03-27 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 7/15 correct (47%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n dressing (pants, shirt): 8/14 correct (60%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n making the bed: 8/13 correct (60%); latency 3.4s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n indicating 'all done' with a step or activity: 4/10 correct (45%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n requesting help when a task is difficult: 6/9 correct (64%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n\nSession 9 \u2014 2026-03-29 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 7/13 correct (55%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n dressing (pants, shirt): 10/15 correct (65%); latency 2.8s; prompts FP\u00d75, PP\u00d72, G\u00d75, I\u00d73\n making the bed: 8/13 correct (65%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n indicating 'all done' with a step or activity: 7/13 correct (54%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n requesting help when a task is difficult: 7/10 correct (72%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n\nSession 10 \u2014 2026-03-31 \u2014 41 min \u2014 1 observer\n meal cleanup after eating: 7/13 correct (57%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n dressing (pants, shirt): 11/15 correct (73%); latency 2.8s; prompts FP\u00d75, PP\u00d70, G\u00d70, I\u00d710\n making the bed: 6/8 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n indicating 'all done' with a step or activity: 7/11 correct (62%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n requesting help when a task is difficult: 11/15 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d78\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Playing with non-task materials\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 55.1%.\n- Trend direction: downward (first session: 76%, last session: 57%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "da2fae2078c0e5b8", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "97069"}, "curation": {"target_split": "sanity"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3049\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u20136 across 12 days (2026-02-09 to 2026-02-21)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. requesting a break during non-preferred activities (Mand)\n Method: NET\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. ordering in a restaurant (Community)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-09 \u2014 41 min \u2014 1 observer\n sounds out CVC words: 3/10 correct (32%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n requesting a break during non-preferred activities: 5/10 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 4/11 correct (35%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n ordering in a restaurant: 6/14 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n\nSession 2 \u2014 2026-02-11 \u2014 41 min \u2014 1 observer\n sounds out CVC words: 3/8 correct (39%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n requesting a break during non-preferred activities: 6/9 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 4/10 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n ordering in a restaurant: 5/10 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n\nSession 3 \u2014 2026-02-13 \u2014 41 min \u2014 1 observer\n sounds out CVC words: 5/10 correct (45%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n requesting a break during non-preferred activities: 8/11 correct (69%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n wait calmly when access to preferred item is denied: 6/13 correct (46%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n ordering in a restaurant: 5/9 correct (57%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n\nSession 4 \u2014 2026-02-15 \u2014 41 min \u2014 1 observer\n sounds out CVC words: 6/13 correct (49%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n requesting a break during non-preferred activities: 8/10 correct (77%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 7/13 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n ordering in a restaurant: 5/8 correct (64%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 5 \u2014 2026-02-17 \u2014 41 min \u2014 1 observer\n sounds out CVC words: 6/12 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d77\n requesting a break during non-preferred activities: 12/15 correct (81%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d712\n wait calmly when access to preferred item is denied: 4/8 correct (56%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n ordering in a restaurant: 10/15 correct (70%); latency 3.1s; prompts FP\u00d75, PP\u00d72, G\u00d73, I\u00d75\n\nSession 6 \u2014 2026-02-19 \u2014 41 min \u2014 1 observer\n sounds out CVC words: 7/12 correct (55%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n requesting a break during non-preferred activities: 10/12 correct (81%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 5/9 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n ordering in a restaurant: 10/13 correct (73%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 32% at return, recovering to 55%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 6 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 45.4%.\n- Trend direction: upward (first session: 32%, last session: 55%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1a6e9bfc74bd401e", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adolescent", "n_sessions": 6, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "53323"}, "curation": {"target_split": "sanity"}}} diff --git a/data/splits/test.jsonl b/data/splits/test.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..fb86d342e40a5243270800cb4d235e3a180201e8 --- /dev/null +++ b/data/splits/test.jsonl @@ -0,0 +1,281 @@ +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a toleration program for this self-care / medical / caregiver-delivered skill.\n\nTarget: tolerating tooth brushing by caregiver\nCurriculum: AFLS Basic Living Skills\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the toleration end goal, shaping progression, antecedent strategies, reinforcement plan, refusal / safety procedures, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating tooth brushing by caregiver\n- Curriculum: AFLS Basic Living Skills\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for tooth brushing performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull 2-minute tooth-brushing routine completed by caregiver without refusal, crying, SIB, or aggression\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1-second toothbrush contact with teeth\n2. 3 seconds of brushing\n3. 5 seconds of brushing\n4. 15 seconds of brushing\n5. 30 seconds of brushing\n6. 60 seconds of brushing\n7. Full 2-minute brushing routine\n\nCurrent shaping step: At the generalization stage, begin at shaping step 6 (\"60 seconds of brushing\") and advance through step 7 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated with 2 different caregivers (primary caregiver + one other) across 2 consecutive weeks.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e3d789cfc2f4c909", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating tooth brushing by caregiver", "program_type": "toleration", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating tooth brushing by caregiver", "module": "basic_living", "mastery_criterion": "mc_toleration_caregivers", "shaping_n_steps": 7}, "teacher_model": null, "seed_tag": "39765"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects by single function from array of 3\nCurriculum Reference: LRFFC L1\nLearner Profile: Early Learner\nCurrent Mastery: Developing\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects by single function from array of 3\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: spoon, plate, jacket\nDistractor stimuli: shoe, car\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ca08f6959493ae2a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects by single function from array of 3", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "56565"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1860\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Writing\nDate range: Sessions 1\u20138 across 16 days (2026-06-10 to 2026-06-26)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. selects items by feature (Listener Responding)\n Method: DTT\n3. copies simple words from model (Writing)\n Method: DTT\n4. reads 10\u201320 sight words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Vocal stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-10 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/10 correct (38%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n selects items by feature: 5/12 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d77\n copies simple words from model: 4/13 correct (28%); latency 4.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n reads 10\u201320 sight words: 3/8 correct (36%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n Vocal stereotypy: freq 0; PIR 0%\n\nSession 2 \u2014 2026-06-12 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/14 correct (44%); latency 4.1s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n selects items by feature: 4/9 correct (47%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n copies simple words from model: 3/10 correct (33%); latency 4.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n reads 10\u201320 sight words: 4/8 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n Vocal stereotypy: freq 1; PIR 5%\n\nSession 3 \u2014 2026-06-14 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (53%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n selects items by feature: 8/15 correct (52%); latency 3.9s; prompts FP\u00d75, PP\u00d75, G\u00d70, I\u00d75\n copies simple words from model: 5/15 correct (36%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d713\n reads 10\u201320 sight words: 6/12 correct (53%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d75\n Vocal stereotypy: freq 0; PIR 10%\n\nSession 4 \u2014 2026-06-16 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (56%); latency 3.7s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n selects items by feature: 8/12 correct (64%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n copies simple words from model: 4/10 correct (43%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 4/8 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Vocal stereotypy: freq 0; PIR 0%\n\nSession 5 \u2014 2026-06-18 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/10 correct (56%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n selects items by feature: 10/15 correct (65%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d75, I\u00d75\n copies simple words from model: 4/8 correct (45%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n reads 10\u201320 sight words: 6/11 correct (58%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n Vocal stereotypy: freq 0; PIR 8%\n\nSession 6 \u2014 2026-06-20 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/11 correct (61%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n selects items by feature: 11/15 correct (73%); latency 2.7s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n copies simple words from model: 4/8 correct (50%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 10/15 correct (69%); latency 3.1s; prompts FP\u00d75, PP\u00d71, G\u00d70, I\u00d79\n Vocal stereotypy: freq 0; PIR 10%\n\nSession 7 \u2014 2026-06-22 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/12 correct (66%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n selects items by feature: 7/8 correct (88%); latency 1.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n copies simple words from model: 6/11 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 6/8 correct (71%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n Vocal stereotypy: freq 0; PIR 4%\n\nSession 8 \u2014 2026-06-24 \u2014 38 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/12 correct (72%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n selects items by feature: 14/15 correct (92%); latency 1.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n copies simple words from model: 9/13 correct (68%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 9/12 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n Vocal stereotypy: freq 0; PIR 3%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Requesting more trials verbally\n- Smiling during reinforcement delivery\n- Appropriate sitting posture maintained"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 38% to 72% across 8 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Behavior Function Hypothesis\nVocal stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 55.7%.\n- Trend direction: upward (first session: 38%, last session: 72%).\n- Vocal stereotypy: mean 0.1/session (early-window mean 0.5 -> late-window mean 0.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1e4bd278d3ad9077", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {"Vocal stereotypy": "automatic"}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "early", "n_sessions": 8, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["vocal_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "76532"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7364\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand + Tolerating Denied Access + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-07-25 to 2026-08-18)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n5. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n6. selects items by feature (Listener Responding)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-25 \u2014 49 min \u2014 1 observer\n copies simple words from model: 5/9 correct (57%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n wait calmly when access to preferred item is denied: 6/14 correct (41%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n answers WH-questions about familiar topics: 5/9 correct (59%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 9/14 correct (68%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n selects items by feature: 8/15 correct (53%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d711\n\nSession 2 \u2014 2026-07-27 \u2014 49 min \u2014 1 observer\n copies simple words from model: 7/10 correct (65%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/13 correct (51%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 4/10 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n answers WH-questions about familiar topics: 7/13 correct (55%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 5/8 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n selects items by feature: 7/13 correct (52%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n\nSession 3 \u2014 2026-07-29 \u2014 49 min \u2014 1 observer\n copies simple words from model: 10/14 correct (69%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (61%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 7/13 correct (54%); latency 3.3s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 8/14 correct (55%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 8/12 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n selects items by feature: 7/13 correct (54%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n\nSession 4 \u2014 2026-07-31 \u2014 49 min \u2014 1 observer\n copies simple words from model: 9/13 correct (72%); latency 2.6s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (56%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n wait calmly when access to preferred item is denied: 7/13 correct (55%); latency 3.3s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n answers WH-questions about familiar topics: 8/13 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 8/11 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n selects items by feature: 6/11 correct (54%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n\nSession 5 \u2014 2026-08-02 \u2014 49 min \u2014 1 observer\n copies simple words from model: 7/11 correct (65%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (55%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 6/10 correct (64%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n answers WH-questions about familiar topics: 4/8 correct (52%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 5/9 correct (58%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n selects items by feature: 5/12 correct (41%); latency 4.3s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n\nSession 6 \u2014 2026-08-04 \u2014 49 min \u2014 1 observer\n copies simple words from model: 11/15 correct (76%); latency 2.8s; prompts FP\u00d75, PP\u00d74, G\u00d74, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (63%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 4/8 correct (56%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n answers WH-questions about familiar topics: 6/10 correct (56%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 6/10 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n selects items by feature: 8/15 correct (52%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d75, I\u00d77\n\nSession 7 \u2014 2026-08-06 \u2014 49 min \u2014 1 observer\n copies simple words from model: 8/11 correct (73%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/11 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 6/10 correct (58%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n answers WH-questions about familiar topics: 9/15 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d75, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 8/13 correct (63%); latency 2.9s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n selects items by feature: 7/15 correct (46%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d75, I\u00d73\n\nSession 8 \u2014 2026-08-08 \u2014 49 min \u2014 1 observer\n copies simple words from model: 11/14 correct (75%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/14 correct (63%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 6/11 correct (55%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n answers WH-questions about familiar topics: 6/11 correct (59%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (67%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n selects items by feature: 4/10 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n\nSession 9 \u2014 2026-08-10 \u2014 49 min \u2014 1 observer\n copies simple words from model: 7/9 correct (74%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n wait calmly when access to preferred item is denied: 7/11 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n answers WH-questions about familiar topics: 5/8 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 7/10 correct (74%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n selects items by feature: 5/11 correct (48%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n\nSession 10 \u2014 2026-08-12 \u2014 49 min \u2014 1 observer\n copies simple words from model: 11/15 correct (74%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d711\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (63%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 7/12 correct (57%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d74\n answers WH-questions about familiar topics: 6/13 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 9/15 correct (59%); latency 3.5s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n selects items by feature: 5/15 correct (35%); latency 4.4s; prompts FP\u00d72, PP\u00d74, G\u00d75, I\u00d74\n\nSession 11 \u2014 2026-08-14 \u2014 49 min \u2014 1 observer\n copies simple words from model: 6/8 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (71%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 10/14 correct (73%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n answers WH-questions about familiar topics: 9/14 correct (65%); latency 3.1s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 8/13 correct (64%); latency 3.0s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n selects items by feature: 6/13 correct (44%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d79\n\nSession 12 \u2014 2026-08-16 \u2014 49 min \u2014 1 observer\n copies simple words from model: 7/9 correct (80%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/12 correct (66%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n wait calmly when access to preferred item is denied: 8/12 correct (69%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n answers WH-questions about familiar topics: 5/9 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 9/12 correct (71%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n selects items by feature: 7/15 correct (46%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d710\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 69% for 12 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 71.1%.\n- Trend direction: upward (first session: 57%, last session: 80%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e9ee174e3677e968", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "early", "n_sessions": 12, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "12904"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6041\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Motor Imitation + Reading + Tact + Writing\nDate range: Sessions 1\u20136 across 12 days (2026-08-02 to 2026-08-14)\n\nACCELERATION PROGRAMS\n1. tacts colors of objects (Tact)\n Method: DTT\n2. reads 10\u201320 sight words (Reading)\n Method: DTT\n3. copies simple words from model (Writing)\n Method: DTT\n4. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n6. selects items by feature (Listener Responding)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-02 \u2014 31 min \u2014 1 observer\n tacts colors of objects: 3/13 correct (23%); latency 4.7s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d77\n reads 10\u201320 sight words: 3/11 correct (23%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n copies simple words from model: 2/9 correct (28%); latency 4.6s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n imitates 3-step motor sequences: 3/12 correct (27%); latency 4.5s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/12 correct (18%); latency 5.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n selects items by feature: 2/10 correct (22%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n\nSession 2 \u2014 2026-08-04 \u2014 31 min \u2014 1 observer\n tacts colors of objects: 3/9 correct (34%); latency 4.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n reads 10\u201320 sight words: 4/13 correct (33%); latency 4.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n copies simple words from model: 5/13 correct (36%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d79\n imitates 3-step motor sequences: 5/14 correct (34%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (31%); latency 4.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n selects items by feature: 4/12 correct (32%); latency 4.4s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n\nSession 3 \u2014 2026-08-06 \u2014 31 min \u2014 1 observer\n tacts colors of objects: 4/12 correct (37%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 4/10 correct (42%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n copies simple words from model: 6/12 correct (46%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n imitates 3-step motor sequences: 4/10 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/15 correct (38%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d712\n selects items by feature: 5/13 correct (41%); latency 4.0s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n\nSession 4 \u2014 2026-08-08 \u2014 31 min \u2014 1 observer\n tacts colors of objects: 4/8 correct (49%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 5/10 correct (53%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n copies simple words from model: 6/11 correct (55%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n imitates 3-step motor sequences: 5/9 correct (53%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n selects items by feature: 5/9 correct (51%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n\nSession 5 \u2014 2026-08-10 \u2014 IOA SESSION \u2014 2 observers\n tacts colors of objects: 54% (primary); IOA 89% agreement\n reads 10\u201320 sight words: 57% (primary); IOA 89% agreement\n copies simple words from model: 62% (primary); IOA 89% agreement\n imitates 3-step motor sequences: 61% (primary); IOA 89% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 60% (primary); IOA 89% agreement\n selects items by feature: 60% (primary); IOA 89% agreement\n\nSession 6 \u2014 2026-08-12 \u2014 31 min \u2014 1 observer\n tacts colors of objects: 7/10 correct (65%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 8/11 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n copies simple words from model: 7/10 correct (70%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n imitates 3-step motor sequences: 10/14 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n selects items by feature: 10/15 correct (70%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d713\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Quick response latency (within 2\u20133 seconds)\n- Appropriate sitting posture maintained"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 23% to 65% across 6 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 6 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 43.5%.\n- Trend direction: upward (first session: 23%, last session: 65%).\n- IOA data: 89% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "522241129579b054", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "early", "n_sessions": 6, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "91206"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: spontaneously vocalizes during preferred activities\nCurriculum Reference: Spontaneous Vocal Behavior L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously vocalizes during preferred activities\n- Curriculum: Spontaneous Vocal Behavior L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: room transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's room transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "0fe5d4b85ec54243", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously vocalizes during preferred activities", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "54871"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: matches non-identical items by category\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L2\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches non-identical items by category\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: hat, spoon, phone\nDistractor stimuli: cup, blanket, backpack, scissors\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "514d19a4d252abe7", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches non-identical items by category", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "5667"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: crossing a street safely\nCurriculum: AFLS Community Participation Skills\nLearner: Adolescent Learner\nMastery Status: Mastered at current level\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: crossing a street safely\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Approach the crosswalk and stop at the curb.\n2. Look left, look right, look left again.\n3. Identify the walk signal or absence of traffic.\n4. Begin to cross while continuing to scan for vehicles.\n5. Cross directly to the opposite curb.\n6. Step up onto the sidewalk.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 180s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "1cfc4bb1fafba7bc", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "crossing a street safely", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "mastered", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "crossing a street safely", "module": "community", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "75542"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1781\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Social Behavior + Writing\nDate range: Sessions 1\u20138 across 16 days (2026-01-15 to 2026-01-31)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n4. copies simple words from model (Writing)\n Method: DTT\n5. requesting a break during non-preferred tasks (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Vocal stereotypy \u2014 function hypothesized: automatic\n2. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-15 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 4/9 correct (47%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n indicating 'all done' when an activity is complete: 6/9 correct (65%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n takes turns during structured activities: 6/14 correct (44%); latency 4.1s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n copies simple words from model: 11/15 correct (70%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d78\n requesting a break during non-preferred tasks: 9/15 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d711\n Vocal stereotypy: freq 2; PIR 18%\n Motor stereotypy: freq 1; PIR 9%\n ABC (vocal stereotypy): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 2 \u2014 2026-01-17 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 6/11 correct (51%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n indicating 'all done' when an activity is complete: 4/9 correct (42%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n takes turns during structured activities: 6/14 correct (40%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n copies simple words from model: 5/15 correct (32%); latency 4.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d711\n requesting a break during non-preferred tasks: 3/9 correct (32%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n Vocal stereotypy: freq 1; PIR 11%\n Motor stereotypy: freq 3; PIR 20%\n\nSession 3 \u2014 2026-01-19 \u2014 IOA SESSION \u2014 2 observers\n identifies numerals 1\u201320: 60% (primary); IOA 96% agreement\n indicating 'all done' when an activity is complete: 44% (primary); IOA 96% agreement\n takes turns during structured activities: 27% (primary); IOA 96% agreement\n copies simple words from model: 47% (primary); IOA 96% agreement\n requesting a break during non-preferred tasks: 27% (primary); IOA 96% agreement\n Vocal stereotypy: freq 3; PIR 17%; IOA 96% agreement\n Motor stereotypy: freq 1; PIR 14%; IOA 96% agreement\n\nSession 4 \u2014 2026-01-21 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 7/10 correct (66%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n indicating 'all done' when an activity is complete: 7/9 correct (81%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n takes turns during structured activities: 4/9 correct (49%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n copies simple words from model: 3/8 correct (43%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting a break during non-preferred tasks: 3/9 correct (34%); latency 4.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n Vocal stereotypy: freq 3; PIR 16%\n Motor stereotypy: freq 1; PIR 13%\n\nSession 5 \u2014 2026-01-23 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 6/11 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n indicating 'all done' when an activity is complete: 7/14 correct (49%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n takes turns during structured activities: 5/9 correct (61%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n copies simple words from model: 4/10 correct (41%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n requesting a break during non-preferred tasks: 5/10 correct (50%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n Vocal stereotypy: freq 2; PIR 12%\n Motor stereotypy: freq 2; PIR 16%\n\nSession 6 \u2014 2026-01-25 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 6/12 correct (48%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d76\n indicating 'all done' when an activity is complete: 4/8 correct (50%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n takes turns during structured activities: 4/15 correct (30%); latency 4.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n copies simple words from model: 3/9 correct (38%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n requesting a break during non-preferred tasks: 4/10 correct (35%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n Vocal stereotypy: freq 2; PIR 9%\n Motor stereotypy: freq 2; PIR 9%\n\nSession 7 \u2014 2026-01-27 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 3/9 correct (32%); latency 4.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n indicating 'all done' when an activity is complete: 5/8 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n takes turns during structured activities: 3/12 correct (23%); latency 4.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d77\n copies simple words from model: 4/12 correct (35%); latency 4.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n requesting a break during non-preferred tasks: 7/13 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n Vocal stereotypy: freq 2; PIR 9%\n Motor stereotypy: freq 5; PIR 16%\n\nSession 8 \u2014 2026-01-29 \u2014 36 min \u2014 1 observer\n identifies numerals 1\u201320: 7/13 correct (56%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d78\n indicating 'all done' when an activity is complete: 6/10 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n takes turns during structured activities: 3/8 correct (43%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n copies simple words from model: 7/13 correct (57%); latency 3.6s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n requesting a break during non-preferred tasks: 6/10 correct (58%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n Vocal stereotypy: freq 2; PIR 19%\n Motor stereotypy: freq 3; PIR 10%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Leaving seat without permission\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 52% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nVocal stereotypy: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\nMotor stereotypy: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 51.5%.\n- Trend direction: upward (first session: 47%, last session: 56%).\n- Vocal stereotypy: mean 2.1/session (early-window mean 1.5 -> late-window mean 2.0).\n- Motor stereotypy: mean 2.2/session (early-window mean 2.0 -> late-window mean 4.0).\n- IOA data: 96% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "11b0878c10d829bd", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Vocal stereotypy": "automatic", "Motor stereotypy": "automatic"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["vocal_stereotypy", "motor_stereotypy"], "behavior_functions": ["automatic", "automatic"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "41084"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies all lower-case letters\nCurriculum Reference: Reading L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Near mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies all lower-case letters\n- Curriculum: Reading L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: M, N, Z\nDistractor stimuli: K, C, V, E\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "23fa3255f35f0eff", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies all lower-case letters", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "vr3", "error_correction": "simple_correction", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "93657"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9955\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Self-Care\nDate range: Sessions 1\u201311 across 22 days (2026-02-02 to 2026-02-24)\n\nACCELERATION PROGRAMS\n1. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n2. requesting a break during non-preferred activities (Mand)\n Method: NET\n3. identifying community safety signs (Community)\n Method: DTT\n4. requesting preferred items using AAC device (Mand)\n Method: NET\n5. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-02 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 7/8 correct (84%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n requesting a break during non-preferred activities: 8/10 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n identifying community safety signs: 7/10 correct (72%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n requesting preferred items using AAC device: 7/10 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n brushing teeth independently: 7/8 correct (87%); latency 1.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n Non-compliance: freq 2\n\nSession 2 \u2014 2026-02-04 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 6/9 correct (72%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n requesting a break during non-preferred activities: 11/15 correct (70%); latency 2.9s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d77\n identifying community safety signs: 9/14 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n requesting preferred items using AAC device: 10/15 correct (65%); latency 3.2s; prompts FP\u00d75, PP\u00d74, G\u00d70, I\u00d76\n brushing teeth independently: 10/12 correct (80%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n Non-compliance: freq 3\n\nSession 3 \u2014 2026-02-06 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 6/9 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n requesting a break during non-preferred activities: 5/9 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n identifying community safety signs: 5/8 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n requesting preferred items using AAC device: 9/15 correct (62%); latency 3.2s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d74\n brushing teeth independently: 9/14 correct (64%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n Non-compliance: freq 4\n ABC (non-compliance): A = asked to finish dressing task; B = kicked at materials; C = task paused for 3 min\n\nSession 4 \u2014 2026-02-08 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 8/12 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n requesting a break during non-preferred activities: 8/15 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d75, G\u00d71, I\u00d77\n identifying community safety signs: 5/11 correct (49%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n requesting preferred items using AAC device: 8/14 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n brushing teeth independently: 5/8 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n Non-compliance: freq 5\n\nSession 5 \u2014 2026-02-10 \u2014 IOA SESSION \u2014 2 observers\n meal cleanup after eating: 63% (primary); IOA 86% agreement\n requesting a break during non-preferred activities: 53% (primary); IOA 86% agreement\n identifying community safety signs: 42% (primary); IOA 86% agreement\n requesting preferred items using AAC device: 50% (primary); IOA 86% agreement\n brushing teeth independently: 50% (primary); IOA 86% agreement\n Non-compliance: freq 5; IOA 86% agreement\n\nSession 6 \u2014 2026-02-12 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 6/12 correct (49%); latency 4.0s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n requesting a break during non-preferred activities: 6/13 correct (47%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d76\n identifying community safety signs: 6/15 correct (39%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d75, I\u00d77\n requesting preferred items using AAC device: 8/15 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d714\n brushing teeth independently: 6/12 correct (49%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d74\n Non-compliance: freq 6\n\nSession 7 \u2014 2026-02-14 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 5/11 correct (44%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n requesting a break during non-preferred activities: 5/14 correct (36%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n identifying community safety signs: 3/11 correct (23%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n requesting preferred items using AAC device: 5/11 correct (45%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n brushing teeth independently: 5/12 correct (40%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n Non-compliance: freq 6\n\nSession 8 \u2014 2026-02-16 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 4/10 correct (41%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n requesting a break during non-preferred activities: 4/12 correct (32%); latency 4.8s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n identifying community safety signs: 3/13 correct (23%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d79\n requesting preferred items using AAC device: 3/8 correct (38%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n brushing teeth independently: 4/15 correct (26%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n Non-compliance: freq 7\n ABC (non-compliance): A = transition to non-preferred activity; B = dropped to floor, cried; C = transition delayed, preferred item offered\n\nSession 9 \u2014 2026-02-18 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 5/13 correct (36%); latency 4.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n requesting a break during non-preferred activities: 3/12 correct (24%); latency 4.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n identifying community safety signs: 1/11 correct (12%); latency 5.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n requesting preferred items using AAC device: 3/10 correct (30%); latency 4.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n brushing teeth independently: 2/9 correct (17%); latency 5.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n Non-compliance: freq 8\n\nSession 10 \u2014 2026-02-20 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 3/12 correct (28%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n requesting a break during non-preferred activities: 3/15 correct (19%); latency 5.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n identifying community safety signs: 1/12 correct (10%); latency 5.6s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n requesting preferred items using AAC device: 3/9 correct (32%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n brushing teeth independently: 1/11 correct (13%); latency 5.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n Non-compliance: freq 8\n\nSession 11 \u2014 2026-02-22 \u2014 48 min \u2014 1 observer\n meal cleanup after eating: 4/15 correct (27%); latency 4.5s; prompts FP\u00d75, PP\u00d71, G\u00d74, I\u00d75\n requesting a break during non-preferred activities: 2/11 correct (21%); latency 5.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n identifying community safety signs: 1/11 correct (10%); latency 5.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n requesting preferred items using AAC device: 3/11 correct (24%); latency 5.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n brushing teeth independently: 1/12 correct (10%); latency 5.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d79\n Non-compliance: freq 9\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Peer aggression (hitting or pushing peers during group activities)\n- Turning away from task or staff\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (84% -> 27%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 52.6%.\n- Trend direction: downward (first session: 84%, last session: 27%).\n- Non-compliance: mean 5.7/session (early-window mean 3.0 -> late-window mean 8.3).\n- IOA data: 86% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "8c52dd7133f4d1fe", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Non-compliance": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adolescent", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["non_compliance"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "5662"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: mands for others to stop an action\nCurriculum: Mand L2\nLearner Level: Early Learner\nBaseline: Developing\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for others to stop an action\n- Curriculum: Mand L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: snack time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's snack time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: performance of the requested action.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "aaa9e8903eb01d6d", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for others to stop an action", "mo_category": "mand_action", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "70185"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: washing hands\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: washing hands\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Walk to the sink.\n2. Turn on the water.\n3. Wet hands under the water.\n4. Apply soap to hands.\n5. Rub hands together for 15\u201320 seconds, covering fronts, backs, and between fingers.\n6. Rinse hands thoroughly under the water.\n7. Turn off the water.\n8. Dry hands with a towel.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "168c6707572c4655", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "washing hands", "program_type": "independence", "learner_profile": "adult", "mastery_state": "approaching", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "washing hands", "module": "basic_living", "mastery_criterion": "mc_all_steps_independent", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "97806"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a toleration program for this self-care / medical / caregiver-delivered skill.\n\nTarget: tolerating nail clipping\nCurriculum: AFLS Basic Living Skills\nLearner: Adolescent Learner\nMastery Status: Emerging\n\nPlease include the toleration end goal, shaping progression, antecedent strategies, reinforcement plan, refusal / safety procedures, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating nail clipping\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Emerging\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for nail clipping performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nall 10 fingernails (and optionally toenails) clipped by caregiver without refusal, pulling hand away, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1 fingernail clipped\n2. 2 fingernails clipped in one sitting\n3. 5 fingernails clipped (one hand)\n4. All 10 fingernails clipped\n5. All 10 fingernails + 10 toenails clipped\n\nCurrent shaping step: At the emerging stage, begin at shaping step 1 (\"1 fingernail clipped\") and advance through step 2 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated with 2 different caregivers (primary caregiver + one other) across 2 consecutive weeks.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "dadc5d8b36a11468", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating nail clipping", "program_type": "toleration", "learner_profile": "adolescent", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating nail clipping", "module": "basic_living", "mastery_criterion": "mc_toleration_caregivers", "shaping_n_steps": 5}, "teacher_model": null, "seed_tag": "68917"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9164\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand + Reading + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u20138 across 16 days (2026-08-02 to 2026-08-18)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n3. requesting help when a task is difficult (Mand)\n Method: NET\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n5. ordering in a restaurant (Community)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-02 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 8/9 correct (89%); latency 1.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n dressing (pants, shirt): 8/9 correct (91%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n requesting help when a task is difficult: 13/15 correct (88%); latency 1.9s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n wait calmly when access to preferred item is denied: 7/9 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n ordering in a restaurant: 11/13 correct (81%); latency 2.4s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n Tantrum: freq 2, duration 10m total\n ABC (tantrum): A = transition to non-preferred activity; B = dropped to floor, cried; C = transition delayed, preferred item offered\n\nSession 2 \u2014 2026-08-04 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 13/14 correct (95%); latency 1.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n dressing (pants, shirt): 8/9 correct (90%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n requesting help when a task is difficult: 13/15 correct (89%); latency 1.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 9/12 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n ordering in a restaurant: 13/15 correct (86%); latency 2.2s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n Tantrum: freq 3, duration 6m total\n\nSession 3 \u2014 2026-08-06 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 10/11 correct (95%); latency 1.4s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n dressing (pants, shirt): 12/13 correct (89%); latency 2.1s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n requesting help when a task is difficult: 11/13 correct (87%); latency 2.1s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 8/11 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n ordering in a restaurant: 7/8 correct (89%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n Tantrum: freq 2, duration 8m total\n\nSession 4 \u2014 2026-08-08 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 13/14 correct (94%); latency 1.9s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n dressing (pants, shirt): 8/8 correct (96%); latency 1.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n requesting help when a task is difficult: 12/14 correct (83%); latency 1.9s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n wait calmly when access to preferred item is denied: 6/8 correct (76%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n ordering in a restaurant: 13/15 correct (87%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d713\n Tantrum: freq 4, duration 12m total\n ABC (tantrum): A = worksheet task presented; B = refused + pushed materials away; C = task removed for 2 min\n\nSession 5 \u2014 2026-08-10 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 11/11 correct (97%); latency 1.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n dressing (pants, shirt): 11/11 correct (96%); latency 1.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n requesting help when a task is difficult: 7/8 correct (85%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 10/15 correct (70%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d78\n ordering in a restaurant: 9/10 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n Tantrum: freq 2, duration 6m total\n\nSession 6 \u2014 2026-08-12 \u2014 IOA SESSION \u2014 2 observers\n sounds out CVC words: 95% (primary); IOA 89% agreement\n dressing (pants, shirt): 91% (primary); IOA 89% agreement\n requesting help when a task is difficult: 88% (primary); IOA 89% agreement\n wait calmly when access to preferred item is denied: 72% (primary); IOA 89% agreement\n ordering in a restaurant: 88% (primary); IOA 89% agreement\n Tantrum: freq 2, duration 4m total; IOA 89% agreement\n\nSession 7 \u2014 2026-08-14 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 8/8 correct (98%); latency 1.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n dressing (pants, shirt): 12/13 correct (90%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n requesting help when a task is difficult: 8/10 correct (82%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 9/13 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d710\n ordering in a restaurant: 12/13 correct (93%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n Tantrum: freq 3, duration 6m total\n\nSession 8 \u2014 2026-08-16 \u2014 52 min \u2014 1 observer\n sounds out CVC words: 8/8 correct (98%); latency 1.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n dressing (pants, shirt): 9/10 correct (94%); latency 1.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n requesting help when a task is difficult: 8/10 correct (83%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 8/12 correct (67%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n ordering in a restaurant: 13/14 correct (94%); latency 1.7s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n Tantrum: freq 3, duration 9m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (89%\u201398%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 95.2%.\n- Trend direction: upward (first session: 89%, last session: 98%).\n- Tantrum: mean 2.6/session (early-window mean 2.5 -> late-window mean 3.0).\n- IOA data: 89% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "34677ea7e560603b", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Tantrum": "escape"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "52515"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a systematic desensitization / toleration program for the following.\n\nSkill: tolerating showering assistance\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adult Learner\nBaseline: Mastered at current level\n\nThe learner currently refuses or exhibits safety behaviors during the activity; the program should build tolerance gradually."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating showering assistance\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for showering / bathing performed with caregiver assistance through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull assisted shower routine completed without refusal, exiting the shower, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. Stand in shower for 30 seconds with water off\n2. Tolerate 10 seconds of water on body\n3. Tolerate 30 seconds of water on body\n4. Tolerate soap application to torso\n5. Tolerate full body wash by caregiver\n6. Full assisted shower routine\n\nCurrent shaping step: At the mastered stage, begin at shaping step 5 (\"Tolerate full body wash by caregiver\") and advance through step 6 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "31946abf0054cd39", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating showering assistance", "program_type": "toleration", "learner_profile": "adult", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating showering assistance", "module": "basic_living", "mastery_criterion": "mc_toleration_settings", "shaping_n_steps": 6}, "teacher_model": null, "seed_tag": "80658"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9497\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Social Behavior\nDate range: Sessions 1\u201310 across 20 days (2026-08-10 to 2026-08-30)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. identifies numerals 1\u201320 (Math)\n Method: DTT\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. takes turns during structured activities (Social Behavior)\n Method: PRT\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-10 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 12/14 correct (83%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d710\n identifies numerals 1\u201320: 9/12 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/12 correct (83%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n takes turns during structured activities: 13/15 correct (87%); latency 2.3s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n\nSession 2 \u2014 2026-08-12 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 5/13 correct (35%); latency 4.5s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n identifies numerals 1\u201320: 4/8 correct (46%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (41%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n takes turns during structured activities: 4/13 correct (33%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n\nSession 3 \u2014 2026-08-14 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 7/9 correct (79%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n identifies numerals 1\u201320: 11/14 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/9 correct (78%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n takes turns during structured activities: 12/15 correct (82%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d712\n\nSession 4 \u2014 2026-08-16 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 3/8 correct (41%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n identifies numerals 1\u201320: 6/15 correct (42%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/14 correct (44%); latency 3.7s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n takes turns during structured activities: 5/13 correct (36%); latency 4.4s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n\nSession 5 \u2014 2026-08-18 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 8/11 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n identifies numerals 1\u201320: 11/15 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d713\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/12 correct (81%); latency 2.2s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d71\n takes turns during structured activities: 9/11 correct (82%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n\nSession 6 \u2014 2026-08-20 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 3/9 correct (35%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n identifies numerals 1\u201320: 3/8 correct (37%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (40%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n takes turns during structured activities: 4/11 correct (35%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n\nSession 7 \u2014 2026-08-22 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 10/12 correct (84%); latency 1.9s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n identifies numerals 1\u201320: 9/12 correct (73%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (77%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n takes turns during structured activities: 7/9 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n\nSession 8 \u2014 2026-08-24 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 5/13 correct (41%); latency 3.9s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n identifies numerals 1\u201320: 5/13 correct (41%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/11 correct (47%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n takes turns during structured activities: 5/14 correct (34%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n\nSession 9 \u2014 2026-08-26 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 11/15 correct (76%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d711\n identifies numerals 1\u201320: 6/8 correct (80%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 13/15 correct (84%); latency 2.1s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n takes turns during structured activities: 10/12 correct (79%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n\nSession 10 \u2014 2026-08-28 \u2014 41 min \u2014 1 observer\n reads 10\u201320 sight words: 6/14 correct (41%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d710\n identifies numerals 1\u201320: 4/9 correct (43%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/11 correct (39%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n takes turns during structured activities: 3/8 correct (41%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 59.1%.\n- Trend direction: downward (first session: 83%, last session: 41%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "2897dfede0e9df01", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "14320"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7060\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Tact + Writing\nDate range: Sessions 1\u20136 across 12 days (2026-06-23 to 2026-07-05)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n3. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n4. tacts colors of objects (Tact)\n Method: DTT\n5. tacts common objects (Tact)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-23 \u2014 42 min \u2014 1 observer\n copies simple words from model: 3/10 correct (30%); latency 4.5s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (24%); latency 4.7s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n indicating completion ('all done', 'finished'): 2/8 correct (29%); latency 4.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n tacts colors of objects: 3/9 correct (31%); latency 4.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n tacts common objects: 2/10 correct (25%); latency 4.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n\nSession 2 \u2014 2026-06-25 \u2014 42 min \u2014 1 observer\n copies simple words from model: 3/9 correct (37%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (31%); latency 4.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 3/8 correct (39%); latency 4.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n tacts colors of objects: 6/14 correct (40%); latency 4.1s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n tacts common objects: 3/9 correct (37%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-06-27 \u2014 42 min \u2014 1 observer\n copies simple words from model: 6/13 correct (44%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/14 correct (47%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n indicating completion ('all done', 'finished'): 4/9 correct (46%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n tacts colors of objects: 7/15 correct (49%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n tacts common objects: 4/8 correct (45%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n\nSession 4 \u2014 2026-06-29 \u2014 42 min \u2014 1 observer\n copies simple words from model: 8/15 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d713\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/9 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 9/15 correct (58%); latency 3.1s; prompts FP\u00d75, PP\u00d74, G\u00d72, I\u00d74\n tacts colors of objects: 4/8 correct (55%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n tacts common objects: 5/9 correct (54%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n\nSession 5 \u2014 2026-07-01 \u2014 42 min \u2014 1 observer\n copies simple words from model: 9/14 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/14 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n indicating completion ('all done', 'finished'): 9/14 correct (68%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n tacts colors of objects: 6/9 correct (64%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n tacts common objects: 6/10 correct (58%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n\nSession 6 \u2014 2026-07-03 \u2014 42 min \u2014 1 observer\n copies simple words from model: 6/8 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 11/13 correct (82%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n indicating completion ('all done', 'finished'): 7/9 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n tacts colors of objects: 7/10 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n tacts common objects: 6/9 correct (71%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Smiling during reinforcement delivery\n- Eye contact maintained throughout trials\n- Requesting more trials verbally"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 30% to 73% across 6 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nlow\n\nInsufficient data: only 6 sessions, high variance, and/or no IOA evidence. Recommend additional data collection before programming changes.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 50.8%.\n- Trend direction: upward (first session: 30%, last session: 73%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "c4bfb12d30859ea1", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "low", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "early", "n_sessions": 6, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "71683"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6315\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-02-01 to 2026-02-23)\n\nACCELERATION PROGRAMS\n1. counts objects 1\u201320 (Math)\n Method: DTT\n2. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n5. reads 10\u201320 sight words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Fecal smearing (scatolia) \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-01 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 10/14 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/12 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n wait calmly when access to preferred item is denied: 9/11 correct (78%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 12/14 correct (86%); latency 1.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n reads 10\u201320 sight words: 8/10 correct (81%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n Fecal smearing (scatolia): attempts 0\n\nSession 2 \u2014 2026-02-03 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 11/15 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/13 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 10/14 correct (72%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/9 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n reads 10\u201320 sight words: 9/11 correct (81%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 3 \u2014 2026-02-05 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 10/15 correct (65%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/10 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 8/12 correct (70%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (69%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 5/8 correct (66%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n Fecal smearing (scatolia): attempts 2 (0 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 4 \u2014 2026-02-07 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 5/8 correct (58%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/15 correct (51%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d712\n wait calmly when access to preferred item is denied: 8/13 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (65%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n reads 10\u201320 sight words: 9/15 correct (61%); latency 3.1s; prompts FP\u00d75, PP\u00d70, G\u00d72, I\u00d78\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 5 \u2014 2026-02-09 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 8/14 correct (55%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/10 correct (52%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 5/10 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/9 correct (60%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 8/14 correct (56%); latency 3.7s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n Fecal smearing (scatolia): attempts 3 (3 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 6 \u2014 2026-02-11 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 5/10 correct (50%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/11 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 3/8 correct (43%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (47%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 5/11 correct (43%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n Fecal smearing (scatolia): attempts 4 (4 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 7 \u2014 2026-02-13 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 5/11 correct (41%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/13 correct (34%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 6/13 correct (44%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 5/15 correct (36%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n Fecal smearing (scatolia): attempts 4 (3 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 8 \u2014 2026-02-15 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 6/15 correct (43%); latency 4.0s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/9 correct (32%); latency 4.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 3/12 correct (29%); latency 4.7s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/12 correct (39%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n reads 10\u201320 sight words: 4/14 correct (27%); latency 4.6s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n Fecal smearing (scatolia): attempts 6 (2 intercepted \u2014 staff redirected before smearing; 4 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 9 \u2014 2026-02-17 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 3/9 correct (32%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/11 correct (23%); latency 5.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 3/9 correct (30%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/13 correct (31%); latency 4.4s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 3/10 correct (28%); latency 4.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n Fecal smearing (scatolia): attempts 6 (2 intercepted \u2014 staff redirected before smearing; 4 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 10 \u2014 2026-02-19 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 3/11 correct (30%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 2/14 correct (12%); latency 5.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 2/8 correct (21%); latency 5.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/9 correct (31%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 1/8 correct (14%); latency 5.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n Fecal smearing (scatolia): attempts 6 (4 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 11 \u2014 2026-02-21 \u2014 47 min \u2014 1 observer\n counts objects 1\u201320: 3/10 correct (29%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 1/13 correct (10%); latency 5.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d710\n wait calmly when access to preferred item is denied: 1/12 correct (10%); latency 5.4s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/9 correct (17%); latency 5.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 1/9 correct (13%); latency 5.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n Fecal smearing (scatolia): attempts 8 (5 intercepted \u2014 staff redirected before smearing; 3 completed \u2014 feces transferred to skin, clothing, or surface)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-stimulatory behavior increase during demand periods\n- Crying during demand presentation\n- Peer aggression (hitting or pushing peers during group activities)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (73% -> 29%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nFecal smearing (scatolia): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 50.0%.\n- Trend direction: downward (first session: 73%, last session: 29%).\n- Fecal smearing (scatolia): mean 4.0/session (early-window mean 1.3 -> late-window mean 6.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "cb45405ddd8fc072", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Fecal smearing (scatolia)": "attention"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "school_age", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["fecal_smearing"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "7707"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6399\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math\nDate range: Sessions 1\u201310 across 20 days (2026-05-14 to 2026-06-03)\n\nACCELERATION PROGRAMS\n1. single-word requests for preferred items (Mand)\n Method: NET\n2. identifies numerals 1\u201320 (Math)\n Method: DTT\n3. mands for missing items needed to complete an activity (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: attention\n2. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-14 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 4/8 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n identifies numerals 1\u201320: 10/15 correct (64%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d76\n mands for missing items needed to complete an activity: 9/13 correct (68%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n Verbal aggression: freq 2\n Property destruction: freq 2\n\nSession 2 \u2014 2026-05-16 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 6/10 correct (56%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n identifies numerals 1\u201320: 10/15 correct (67%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n mands for missing items needed to complete an activity: 8/10 correct (77%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n Verbal aggression: freq 3\n Property destruction: freq 1\n ABC (property destruction): A = staff engaged in paperwork nearby; B = yelled name + tapped staff; C = staff attended and redirected\n\nSession 3 \u2014 2026-05-18 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 7/13 correct (58%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d76\n identifies numerals 1\u201320: 9/15 correct (63%); latency 2.9s; prompts FP\u00d74, PP\u00d75, G\u00d71, I\u00d75\n mands for missing items needed to complete an activity: 7/10 correct (74%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n Verbal aggression: freq 3\n Property destruction: freq 1\n\nSession 4 \u2014 2026-05-20 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 5/8 correct (60%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n identifies numerals 1\u201320: 9/13 correct (69%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n mands for missing items needed to complete an activity: 10/13 correct (74%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n Verbal aggression: freq 2\n Property destruction: freq 1\n\nSession 5 \u2014 2026-05-22 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 3/12 correct (24%); latency 5.2s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d77\n identifies numerals 1\u201320: 3/10 correct (29%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n mands for missing items needed to complete an activity: 4/10 correct (43%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n Verbal aggression: freq 7\n Property destruction: freq 3\n\nSession 6 \u2014 2026-05-24 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 4/11 correct (34%); latency 4.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n identifies numerals 1\u201320: 3/9 correct (38%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n mands for missing items needed to complete an activity: 5/10 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n Verbal aggression: freq 7\n Property destruction: freq 3\n\nSession 7 \u2014 2026-05-26 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 3/10 correct (34%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n identifies numerals 1\u201320: 5/11 correct (46%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n mands for missing items needed to complete an activity: 8/15 correct (54%); latency 3.7s; prompts FP\u00d75, PP\u00d71, G\u00d70, I\u00d79\n Verbal aggression: freq 1\n Property destruction: freq 1\n\nSession 8 \u2014 2026-05-28 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 5/14 correct (35%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n identifies numerals 1\u201320: 6/12 correct (54%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d79\n mands for missing items needed to complete an activity: 6/9 correct (64%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n Verbal aggression: freq 2\n Property destruction: freq 2\n\nSession 9 \u2014 2026-05-30 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 3/9 correct (36%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n identifies numerals 1\u201320: 6/10 correct (57%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 9/13 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n Verbal aggression: freq 2\n Property destruction: freq 1\n\nSession 10 \u2014 2026-06-01 \u2014 45 min \u2014 1 observer\n single-word requests for preferred items: 5/13 correct (38%); latency 4.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n identifies numerals 1\u201320: 8/13 correct (65%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n mands for missing items needed to complete an activity: 9/13 correct (71%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n Verbal aggression: freq 3\n Property destruction: freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nVerbal aggression: attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\nProperty destruction: attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 43.0%.\n- Trend direction: downward (first session: 55%, last session: 38%).\n- Verbal aggression: mean 3.2/session (early-window mean 2.7 -> late-window mean 2.3).\n- Property destruction: mean 1.6/session (early-window mean 1.3 -> late-window mean 1.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "809984a13faeceb2", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Verbal aggression": "attention", "Property destruction": "attention"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 3, "n_behaviors": 2, "behavior_ids": ["verbal_aggression", "property_destruction"], "behavior_functions": ["attention", "attention"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "70412"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by function from array of 5\u20138\nCurriculum: LRFFC L2\nLearner: Early Learner\nMastery Status: Approaching mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function from array of 5\u20138\n- Curriculum: LRFFC L2\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: pencil, toothbrush, sock\nDistractor stimuli: car, hat, blanket, phone\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ce083a6f263a2b97", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L2", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function from array of 5\u20138", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "68189"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: using public transportation\nCurriculum Reference: AFLS Community Participation Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Emerging\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: using public transportation\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adolescent Learner\n- Current mastery: Emerging\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Confirm the correct stop or station. <- Learner-taught step (forward chaining starts here)\n2. Purchase or have the fare ready.\n3. Wait at the designated boarding area.\n4. Board the vehicle when it arrives.\n5. Pay the fare or present the pass.\n6. Sit or stand safely.\n7. Monitor the route for the correct stop.\n8. Exit at the correct stop.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the emerging stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c43b29c3d4774850", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "using public transportation", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "emerging", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "using public transportation", "module": "community", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "44637"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner Level: Adolescent Learner\nBaseline: Developing\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the developing stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3b6c16bcd5e4e7ea", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "98892"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: answers social questions (How are you?)\nCurriculum: Intraverbal L2\nLearner: School-Age Learner\nMastery Status: Near mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers social questions (How are you?)\n- Curriculum: Intraverbal L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: transition conversation.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's transition conversation routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "7d7a50e3c6fd4134", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers social questions (How are you?)", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "83467"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tracks print left to right\nCurriculum: Reading L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tracks print left to right\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter or word on a flashcard.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: spoon, apple, pencil\nDistractor stimuli: ball, block\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "421660d3a275008b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tracks print left to right", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "36112"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5742\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Self-Care + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-01-01 to 2026-01-25)\n\nACCELERATION PROGRAMS\n1. brushing teeth independently (Self-Care)\n Method: Task Analysis\n2. washing hands independently (Self-Care)\n Method: Task Analysis\n3. requesting preferred items using AAC device (Mand)\n Method: NET\n4. making the bed (Home Skills)\n Method: Task Analysis\n5. copies simple words from model (Writing)\n Method: DTT\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n2. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-01 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 7/13 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d79\n washing hands independently: 4/9 correct (49%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n requesting preferred items using AAC device: 6/9 correct (65%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n making the bed: 4/12 correct (37%); latency 4.5s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d77\n copies simple words from model: 4/10 correct (45%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n Self-injurious behavior (SIB): freq 4\n Motor stereotypy: freq 5; PIR 34%\n\nSession 2 \u2014 2026-01-03 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 4/9 correct (48%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n washing hands independently: 7/9 correct (77%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n requesting preferred items using AAC device: 5/10 correct (53%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n making the bed: 5/8 correct (68%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n copies simple words from model: 10/14 correct (73%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n Self-injurious behavior (SIB): freq 5\n Motor stereotypy: freq 4; PIR 27%\n\nSession 3 \u2014 2026-01-05 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 6/11 correct (52%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n washing hands independently: 4/11 correct (40%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n requesting preferred items using AAC device: 10/13 correct (74%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n making the bed: 3/8 correct (42%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n copies simple words from model: 5/8 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n Self-injurious behavior (SIB): freq 3\n Motor stereotypy: freq 4; PIR 20%\n\nSession 4 \u2014 2026-01-07 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 5/8 correct (66%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n washing hands independently: 7/9 correct (75%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n requesting preferred items using AAC device: 5/8 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n making the bed: 8/13 correct (62%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n copies simple words from model: 7/12 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n Self-injurious behavior (SIB): freq 6\n Motor stereotypy: freq 6; PIR 36%\n\nSession 5 \u2014 2026-01-09 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 5/11 correct (45%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n washing hands independently: 10/15 correct (69%); latency 3.1s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d78\n requesting preferred items using AAC device: 12/15 correct (78%); latency 2.7s; prompts FP\u00d75, PP\u00d74, G\u00d71, I\u00d75\n making the bed: 6/8 correct (76%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n copies simple words from model: 5/9 correct (60%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n Self-injurious behavior (SIB): freq 3\n Motor stereotypy: freq 2; PIR 15%\n\nSession 6 \u2014 2026-01-11 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 9/15 correct (63%); latency 3.0s; prompts FP\u00d74, PP\u00d75, G\u00d75, I\u00d71\n washing hands independently: 11/14 correct (79%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n requesting preferred items using AAC device: 6/12 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d77\n making the bed: 4/9 correct (49%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n copies simple words from model: 6/15 correct (41%); latency 4.3s; prompts FP\u00d75, PP\u00d74, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 4\n Motor stereotypy: freq 5; PIR 29%\n\nSession 7 \u2014 2026-01-13 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 6/11 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n washing hands independently: 6/11 correct (56%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n requesting preferred items using AAC device: 7/15 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n making the bed: 5/8 correct (57%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n copies simple words from model: 6/9 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n Self-injurious behavior (SIB): freq 7\n Motor stereotypy: freq 1; PIR 10%\n\nSession 8 \u2014 2026-01-15 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 6/8 correct (74%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n washing hands independently: 4/8 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting preferred items using AAC device: 10/13 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d711\n making the bed: 10/13 correct (74%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n copies simple words from model: 5/14 correct (36%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d79\n Self-injurious behavior (SIB): freq 7\n Motor stereotypy: freq 3; PIR 15%\n\nSession 9 \u2014 2026-01-17 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 8/10 correct (77%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n washing hands independently: 6/9 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n requesting preferred items using AAC device: 12/15 correct (81%); latency 2.3s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d78\n making the bed: 7/11 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n copies simple words from model: 9/13 correct (68%); latency 3.1s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n Self-injurious behavior (SIB): freq 4\n Motor stereotypy: freq 3; PIR 21%\n\nSession 10 \u2014 2026-01-19 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 8/15 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d75, G\u00d72, I\u00d78\n washing hands independently: 6/14 correct (40%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n requesting preferred items using AAC device: 8/12 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n making the bed: 8/12 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n copies simple words from model: 6/10 correct (64%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n Self-injurious behavior (SIB): freq 6\n Motor stereotypy: freq 1; PIR 12%\n\nSession 11 \u2014 2026-01-21 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 8/14 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d78\n washing hands independently: 8/13 correct (60%); latency 3.5s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 9/15 correct (57%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d75\n making the bed: 5/11 correct (48%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n copies simple words from model: 7/11 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n Self-injurious behavior (SIB): freq 4\n Motor stereotypy: freq 6; PIR 32%\n\nSession 12 \u2014 2026-01-23 \u2014 36 min \u2014 1 observer\n brushing teeth independently: 9/12 correct (76%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n washing hands independently: 8/13 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d77\n requesting preferred items using AAC device: 6/11 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n making the bed: 7/14 correct (51%); latency 4.0s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n copies simple words from model: 8/13 correct (60%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n Self-injurious behavior (SIB): freq 6\n Motor stereotypy: freq 7; PIR 23%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 67% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 60.1%.\n- Trend direction: upward (first session: 57%, last session: 76%).\n- Self-injurious behavior (SIB): mean 4.9/session (early-window mean 4.5 -> late-window mean 5.0).\n- Motor stereotypy: mean 3.9/session (early-window mean 4.8 -> late-window mean 4.2).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "2355d4a3a0e38e8e", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Self-injurious behavior (SIB)": "escape", "Motor stereotypy": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["sib", "motor_stereotypy"], "behavior_functions": ["escape", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "90253"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner: Adolescent Learner\nMastery Status: Approaching mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5495eb32df5b6db1", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "approaching", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "80658"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a systematic desensitization / toleration program for the following.\n\nSkill: tolerating nail clipping\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adolescent Learner\nBaseline: Emerging\n\nThe learner currently refuses or exhibits safety behaviors during the activity; the program should build tolerance gradually."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating nail clipping\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Emerging\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for nail clipping performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nall 10 fingernails (and optionally toenails) clipped by caregiver without refusal, pulling hand away, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1 fingernail clipped\n2. 2 fingernails clipped in one sitting\n3. 5 fingernails clipped (one hand)\n4. All 10 fingernails clipped\n5. All 10 fingernails + 10 toenails clipped\n\nCurrent shaping step: At the emerging stage, begin at shaping step 1 (\"1 fingernail clipped\") and advance through step 2 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 2 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7172ec34fc2688b2", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating nail clipping", "program_type": "toleration", "learner_profile": "adolescent", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating nail clipping", "module": "basic_living", "mastery_criterion": "mc_toleration_settings", "shaping_n_steps": 5}, "teacher_model": null, "seed_tag": "68917"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9345\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Mand + Self-Care + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-10-20 to 2026-11-05)\n\nACCELERATION PROGRAMS\n1. brushing teeth independently (Self-Care)\n Method: Task Analysis\n2. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n3. doing laundry (Home Skills)\n Method: Task Analysis\n4. requesting assistance using AAC device (Mand)\n Method: NET\n5. washing hands independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-20 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 7/15 correct (50%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d710\n completing assigned tasks in order: 9/15 correct (61%); latency 3.3s; prompts FP\u00d75, PP\u00d74, G\u00d75, I\u00d71\n doing laundry: 7/15 correct (49%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n requesting assistance using AAC device: 6/9 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n washing hands independently: 6/9 correct (72%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n Non-compliance: freq 2\n\nSession 2 \u2014 2026-10-22 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 4/8 correct (53%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n completing assigned tasks in order: 9/13 correct (70%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d74\n doing laundry: 7/13 correct (51%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n requesting assistance using AAC device: 7/10 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n washing hands independently: 7/10 correct (72%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n Non-compliance: freq 2\n\nSession 3 \u2014 2026-10-24 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 8/14 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n completing assigned tasks in order: 10/13 correct (74%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d73\n doing laundry: 7/12 correct (58%); latency 3.2s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d74\n requesting assistance using AAC device: 9/13 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n washing hands independently: 9/13 correct (71%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n Non-compliance: freq 11\n\nSession 4 \u2014 2026-10-26 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 8/15 correct (55%); latency 3.8s; prompts FP\u00d75, PP\u00d75, G\u00d72, I\u00d73\n completing assigned tasks in order: 10/13 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n doing laundry: 9/15 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n requesting assistance using AAC device: 6/10 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n washing hands independently: 9/13 correct (69%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d710\n Non-compliance: freq 11\n\nSession 5 \u2014 2026-10-28 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 8/13 correct (64%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n completing assigned tasks in order: 7/9 correct (74%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n doing laundry: 6/9 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n requesting assistance using AAC device: 10/15 correct (66%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d710\n washing hands independently: 10/14 correct (74%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d710\n Non-compliance: freq 3\n\nSession 6 \u2014 2026-10-30 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 7/11 correct (64%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n completing assigned tasks in order: 10/14 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d77\n doing laundry: 5/8 correct (60%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n requesting assistance using AAC device: 10/14 correct (73%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n washing hands independently: 9/14 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n Non-compliance: freq 3\n\nSession 7 \u2014 2026-11-01 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 7/11 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n completing assigned tasks in order: 11/13 correct (82%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n doing laundry: 8/11 correct (70%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n requesting assistance using AAC device: 10/14 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n washing hands independently: 8/11 correct (68%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n Non-compliance: freq 2\n\nSession 8 \u2014 2026-11-03 \u2014 55 min \u2014 1 observer\n brushing teeth independently: 10/15 correct (68%); latency 2.8s; prompts FP\u00d75, PP\u00d73, G\u00d74, I\u00d73\n completing assigned tasks in order: 9/11 correct (79%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n doing laundry: 8/11 correct (74%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n requesting assistance using AAC device: 10/14 correct (69%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n washing hands independently: 10/14 correct (68%); latency 2.7s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d75\n Non-compliance: freq 4\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Turning away from task or staff\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Peer aggression (hitting or pushing peers during group activities)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 59.7%.\n- Trend direction: upward (first session: 50%, last session: 68%).\n- Non-compliance: mean 4.8/session (early-window mean 2.0 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1dd2723d41c62b51", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Non-compliance": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adult", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["non_compliance"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "44994"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: follows conditional instructions (if X, then Y)\nCurriculum Reference: Listener Responding L3\nLearner Profile: School-Age Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: follows conditional instructions (if X, then Y)\n- Curriculum: Listener Responding L3\n- Learner: School-Age Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: plate, brush, glue\nDistractor stimuli: cup, sock, pencil, block, apple\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fbbd06b71b7c164e", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L3", "learner_profile": "school_age", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "follows conditional instructions (if X, then Y)", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "token_economy", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "92219"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously requests desired items without prompts\nCurriculum: Spontaneous Vocal Behavior L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously requests desired items without prompts\n- Curriculum: Spontaneous Vocal Behavior L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: room transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's room transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "7483c51436980328", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously requests desired items without prompts", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "46838"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: matches colors (identical chips)\nCurriculum Reference: Visual Perceptual Skills & Matching-to-Sample L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches colors (identical chips)\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: red, blue, orange\nDistractor stimuli: green, white\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "44e8d61847443e95", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches colors (identical chips)", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "35263"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6948\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-02-13 to 2026-03-07)\n\nACCELERATION PROGRAMS\n1. selects items by feature (Listener Responding)\n Method: DTT\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n5. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-13 \u2014 35 min \u2014 1 observer\n selects items by feature: 8/10 correct (82%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (91%); latency 1.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n answers WH-questions about familiar topics: 10/12 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 8/11 correct (74%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n indicating completion ('all done', 'finished'): 7/9 correct (82%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n\nSession 2 \u2014 2026-02-15 \u2014 35 min \u2014 1 observer\n selects items by feature: 5/14 correct (36%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (42%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 6/15 correct (42%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d77\n wait calmly when access to preferred item is denied: 4/11 correct (37%); latency 4.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n indicating completion ('all done', 'finished'): 5/15 correct (31%); latency 4.6s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d79\n\nSession 3 \u2014 2026-02-17 \u2014 35 min \u2014 1 observer\n selects items by feature: 9/11 correct (81%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/9 correct (87%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n answers WH-questions about familiar topics: 8/10 correct (78%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 11/15 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d712\n indicating completion ('all done', 'finished'): 8/10 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n\nSession 4 \u2014 2026-02-19 \u2014 35 min \u2014 1 observer\n selects items by feature: 6/13 correct (45%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/12 correct (36%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n answers WH-questions about familiar topics: 4/11 correct (40%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n wait calmly when access to preferred item is denied: 6/14 correct (41%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n indicating completion ('all done', 'finished'): 3/11 correct (31%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n\nSession 5 \u2014 2026-02-21 \u2014 IOA SESSION \u2014 2 observers\n selects items by feature: 83% (primary); IOA 89% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 88% (primary); IOA 89% agreement\n answers WH-questions about familiar topics: 82% (primary); IOA 89% agreement\n wait calmly when access to preferred item is denied: 81% (primary); IOA 89% agreement\n indicating completion ('all done', 'finished'): 78% (primary); IOA 89% agreement\n\nSession 6 \u2014 2026-02-23 \u2014 35 min \u2014 1 observer\n selects items by feature: 4/9 correct (44%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/15 correct (34%); latency 4.3s; prompts FP\u00d70, PP\u00d73, G\u00d75, I\u00d77\n answers WH-questions about familiar topics: 4/9 correct (42%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 4/9 correct (44%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n indicating completion ('all done', 'finished'): 3/10 correct (27%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n\nSession 7 \u2014 2026-02-25 \u2014 35 min \u2014 1 observer\n selects items by feature: 6/8 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (82%); latency 2.6s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d78\n answers WH-questions about familiar topics: 10/13 correct (75%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d75\n wait calmly when access to preferred item is denied: 9/12 correct (79%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n indicating completion ('all done', 'finished'): 8/11 correct (74%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n\nSession 8 \u2014 2026-02-27 \u2014 35 min \u2014 1 observer\n selects items by feature: 5/13 correct (37%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/15 correct (38%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n answers WH-questions about familiar topics: 5/10 correct (49%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n wait calmly when access to preferred item is denied: 5/14 correct (35%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n indicating completion ('all done', 'finished'): 3/10 correct (29%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n\nSession 9 \u2014 2026-03-01 \u2014 35 min \u2014 1 observer\n selects items by feature: 13/15 correct (85%); latency 1.9s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (91%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n answers WH-questions about familiar topics: 7/9 correct (78%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 9/11 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n indicating completion ('all done', 'finished'): 7/10 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n\nSession 10 \u2014 2026-03-03 \u2014 35 min \u2014 1 observer\n selects items by feature: 4/10 correct (41%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (37%); latency 4.5s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n answers WH-questions about familiar topics: 7/14 correct (47%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 6/15 correct (43%); latency 4.2s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d79\n indicating completion ('all done', 'finished'): 3/9 correct (31%); latency 4.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n\nSession 11 \u2014 2026-03-05 \u2014 35 min \u2014 1 observer\n selects items by feature: 7/9 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/12 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n answers WH-questions about familiar topics: 10/13 correct (77%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 7/9 correct (73%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n indicating completion ('all done', 'finished'): 10/13 correct (79%); latency 2.3s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 63.2%.\n- Trend direction: downward (first session: 82%, last session: 82%).\n- IOA data: 89% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "c1877f5521ca22e8", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "early", "n_sessions": 11, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "82365"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3217\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Tact + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-04-12 to 2026-05-02)\n\nACCELERATION PROGRAMS\n1. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. tacts common objects (Tact)\n Method: DTT\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-12 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/8 correct (60%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n tacts common objects: 8/12 correct (66%); latency 3.1s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n wait calmly when access to preferred item is denied: 9/14 correct (63%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n\nSession 2 \u2014 2026-04-14 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/12 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (60%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n tacts common objects: 9/13 correct (68%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 5/8 correct (58%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n\nSession 3 \u2014 2026-04-16 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/11 correct (55%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n tacts common objects: 9/13 correct (69%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n wait calmly when access to preferred item is denied: 5/8 correct (60%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 4 \u2014 2026-04-18 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 7/12 correct (58%); latency 3.3s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (58%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n tacts common objects: 6/10 correct (65%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 7/11 correct (67%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n\nSession 5 \u2014 2026-04-20 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 4/14 correct (30%); latency 4.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (30%); latency 4.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d74\n tacts common objects: 5/14 correct (33%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d710\n wait calmly when access to preferred item is denied: 4/15 correct (27%); latency 4.8s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n\nSession 6 \u2014 2026-04-22 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/15 correct (34%); latency 4.7s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (32%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n tacts common objects: 4/11 correct (37%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 5/15 correct (32%); latency 4.6s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n\nSession 7 \u2014 2026-04-24 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/10 correct (46%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (39%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n tacts common objects: 5/13 correct (40%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 5/15 correct (35%); latency 4.2s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n\nSession 8 \u2014 2026-04-26 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/11 correct (49%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (47%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n tacts common objects: 4/8 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 6/15 correct (37%); latency 4.6s; prompts FP\u00d75, PP\u00d70, G\u00d70, I\u00d710\n\nSession 9 \u2014 2026-04-28 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/12 correct (54%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (54%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n tacts common objects: 6/11 correct (57%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 5/11 correct (42%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n\nSession 10 \u2014 2026-04-30 \u2014 55 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/9 correct (58%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (56%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d710\n tacts common objects: 8/12 correct (65%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n wait calmly when access to preferred item is denied: 8/15 correct (51%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 49.6%.\n- Trend direction: downward (first session: 60%, last session: 58%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "18156c9ea7eced13", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "early", "n_sessions": 10, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "74200"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8979\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Reading + Self-Care\nDate range: Sessions 1\u201312 across 24 days (2026-06-02 to 2026-06-26)\n\nACCELERATION PROGRAMS\n1. brushing teeth independently (Self-Care)\n Method: Task Analysis\n2. tolerating haircut (Self-Care)\n Method: Task Analysis\n3. requesting preferred items using AAC device (Mand)\n Method: NET\n4. reads 10\u201320 sight words (Reading)\n Method: DTT\n5. requesting help when a task is difficult (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n2. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-02 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 5/9 correct (60%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n tolerating haircut: 6/8 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n requesting preferred items using AAC device: 5/8 correct (60%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 9/14 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n requesting help when a task is difficult: 9/15 correct (62%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d78\n Non-compliance: freq 1\n Motor stereotypy: freq 2; PIR 9%\n\nSession 2 \u2014 2026-06-04 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 9/15 correct (60%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d712\n tolerating haircut: 10/15 correct (66%); latency 2.7s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d75\n requesting preferred items using AAC device: 6/10 correct (57%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n reads 10\u201320 sight words: 7/11 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n requesting help when a task is difficult: 7/12 correct (58%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n Non-compliance: freq 1\n Motor stereotypy: freq 1; PIR 8%\n\nSession 3 \u2014 2026-06-06 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 8/12 correct (64%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n tolerating haircut: 8/11 correct (69%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n requesting preferred items using AAC device: 8/15 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d75, G\u00d70, I\u00d78\n reads 10\u201320 sight words: 9/13 correct (69%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n requesting help when a task is difficult: 5/8 correct (59%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n Non-compliance: freq 2\n Motor stereotypy: freq 2; PIR 19%\n\nSession 4 \u2014 2026-06-08 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 9/13 correct (69%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n tolerating haircut: 10/14 correct (71%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d76\n requesting preferred items using AAC device: 6/10 correct (62%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 5/8 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n requesting help when a task is difficult: 5/8 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Non-compliance: freq 0\n Motor stereotypy: freq 1; PIR 14%\n\nSession 5 \u2014 2026-06-10 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 4/12 correct (34%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n tolerating haircut: 4/11 correct (40%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n requesting preferred items using AAC device: 3/9 correct (28%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 4/12 correct (32%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n requesting help when a task is difficult: 2/9 correct (25%); latency 5.2s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n Non-compliance: freq 3\n Motor stereotypy: freq 4; PIR 22%\n\nSession 6 \u2014 2026-06-12 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 3/9 correct (38%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n tolerating haircut: 7/15 correct (48%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 5/14 correct (32%); latency 4.8s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n reads 10\u201320 sight words: 5/14 correct (39%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n requesting help when a task is difficult: 4/12 correct (34%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n Non-compliance: freq 4\n Motor stereotypy: freq 3; PIR 9%\n\nSession 7 \u2014 2026-06-14 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 5/11 correct (44%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n tolerating haircut: 7/12 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d76\n requesting preferred items using AAC device: 4/10 correct (36%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 6/13 correct (47%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d77\n requesting help when a task is difficult: 4/11 correct (37%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n Non-compliance: freq 1\n Motor stereotypy: freq 1; PIR 6%\n\nSession 8 \u2014 2026-06-16 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 6/12 correct (50%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d75\n tolerating haircut: 7/11 correct (60%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n requesting preferred items using AAC device: 6/14 correct (45%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d79\n reads 10\u201320 sight words: 5/8 correct (58%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n requesting help when a task is difficult: 4/10 correct (40%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n Non-compliance: freq 0\n Motor stereotypy: freq 1; PIR 6%\n\nSession 9 \u2014 2026-06-18 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 5/8 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n tolerating haircut: 8/12 correct (67%); latency 3.2s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n requesting preferred items using AAC device: 6/10 correct (55%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n reads 10\u201320 sight words: 9/14 correct (66%); latency 3.1s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d76\n requesting help when a task is difficult: 4/9 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n Non-compliance: freq 1\n Motor stereotypy: freq 2; PIR 19%\n\nSession 10 \u2014 2026-06-20 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 5/8 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n tolerating haircut: 8/12 correct (67%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d74\n requesting preferred items using AAC device: 5/8 correct (60%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 9/14 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d710\n requesting help when a task is difficult: 7/15 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n Non-compliance: freq 1\n Motor stereotypy: freq 1; PIR 14%\n\nSession 11 \u2014 2026-06-22 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 6/10 correct (64%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n tolerating haircut: 9/13 correct (67%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 8/13 correct (60%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n reads 10\u201320 sight words: 6/9 correct (66%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n requesting help when a task is difficult: 8/14 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d711\n Non-compliance: freq 2\n Motor stereotypy: freq 2; PIR 11%\n\nSession 12 \u2014 2026-06-24 \u2014 31 min \u2014 1 observer\n brushing teeth independently: 10/15 correct (64%); latency 2.8s; prompts FP\u00d71, PP\u00d75, G\u00d75, I\u00d74\n tolerating haircut: 5/8 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n requesting preferred items using AAC device: 5/9 correct (60%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 9/14 correct (66%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n requesting help when a task is difficult: 7/12 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d79\n Non-compliance: freq 1\n Motor stereotypy: freq 2; PIR 11%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals\n- Looking away during SD presentation\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 56.2%.\n- Trend direction: upward (first session: 60%, last session: 64%).\n- Non-compliance: mean 1.4/session (early-window mean 1.0 -> late-window mean 1.2).\n- Motor stereotypy: mean 1.8/session (early-window mean 1.5 -> late-window mean 1.8).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "c8444262e057668a", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Non-compliance": "escape", "Motor stereotypy": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["non_compliance", "motor_stereotypy"], "behavior_functions": ["escape", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "76456"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7908\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Self-Care + Vocational\nDate range: Sessions 1\u20139 across 18 days (2026-06-16 to 2026-07-04)\n\nACCELERATION PROGRAMS\n1. brushing teeth independently (Self-Care)\n Method: Task Analysis\n2. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n3. managing a budget (Independent Living)\n Method: Task Analysis\n4. clocking in at start of shift (Vocational)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-16 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 11/15 correct (74%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d78\n dressing (pants, shirt): 10/13 correct (79%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n managing a budget: 9/11 correct (81%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n clocking in at start of shift: 12/15 correct (82%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n\nSession 2 \u2014 2026-06-18 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 8/11 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n dressing (pants, shirt): 8/10 correct (79%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n managing a budget: 10/12 correct (82%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n clocking in at start of shift: 9/11 correct (86%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n\nSession 3 \u2014 2026-06-20 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 11/15 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d75, G\u00d72, I\u00d76\n dressing (pants, shirt): 6/8 correct (75%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n managing a budget: 11/13 correct (81%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n clocking in at start of shift: 13/15 correct (86%); latency 1.8s; prompts FP\u00d72, PP\u00d74, G\u00d75, I\u00d74\n\nSession 4 \u2014 2026-06-22 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 3/9 correct (35%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n dressing (pants, shirt): 4/9 correct (44%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n managing a budget: 7/13 correct (55%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n clocking in at start of shift: 6/15 correct (41%); latency 4.0s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d73\n\nSession 5 \u2014 2026-06-24 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 3/8 correct (34%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n dressing (pants, shirt): 6/15 correct (41%); latency 4.4s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n managing a budget: 4/8 correct (54%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n clocking in at start of shift: 6/15 correct (41%); latency 4.3s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n\nSession 6 \u2014 2026-06-26 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 5/12 correct (38%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n dressing (pants, shirt): 4/8 correct (45%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n managing a budget: 8/14 correct (58%); latency 3.4s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d73\n clocking in at start of shift: 5/12 correct (43%); latency 3.9s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n\nSession 7 \u2014 2026-06-28 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 5/10 correct (47%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n dressing (pants, shirt): 5/11 correct (48%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n managing a budget: 7/14 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d710\n clocking in at start of shift: 6/13 correct (48%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n\nSession 8 \u2014 2026-06-30 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 7/13 correct (57%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n dressing (pants, shirt): 6/11 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n managing a budget: 5/9 correct (58%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n clocking in at start of shift: 8/14 correct (56%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n\nSession 9 \u2014 2026-07-02 \u2014 52 min \u2014 1 observer\n brushing teeth independently: 8/12 correct (63%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n dressing (pants, shirt): 9/15 correct (62%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n managing a budget: 9/14 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n clocking in at start of shift: 7/12 correct (62%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 54.7%.\n- Trend direction: downward (first session: 74%, last session: 63%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e8880784e0be88e1", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adult", "n_sessions": 9, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "39857"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: identifies 5\u201310 upper-case letters\nCurriculum: Reading L1\nLearner: Early Learner\nMastery Status: Approaching mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies 5\u201310 upper-case letters\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: C, F, D\nDistractor stimuli: I, B\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7026d711a01dbc3b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies 5\u201310 upper-case letters", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "46080"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8319\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20139 across 18 days (2026-08-19 to 2026-09-06)\n\nACCELERATION PROGRAMS\n1. requesting a break during extended tasks (Mand)\n Method: NET\n2. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n4. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n5. clocking in at start of shift (Vocational)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-19 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 7/10 correct (70%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n scheduling medical appointments: 4/8 correct (45%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n brushing teeth independently: 5/10 correct (47%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 10/15 correct (69%); latency 2.9s; prompts FP\u00d75, PP\u00d72, G\u00d70, I\u00d78\n clocking in at start of shift: 4/8 correct (50%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n\nSession 2 \u2014 2026-08-21 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 6/9 correct (67%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n scheduling medical appointments: 5/10 correct (55%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n brushing teeth independently: 5/12 correct (46%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d79\n wait calmly when a preferred activity is denied: 6/9 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n clocking in at start of shift: 6/12 correct (48%); latency 3.7s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d73\n\nSession 3 \u2014 2026-08-23 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 10/15 correct (65%); latency 3.3s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d75\n scheduling medical appointments: 4/8 correct (49%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n brushing teeth independently: 5/11 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n wait calmly when a preferred activity is denied: 8/12 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n clocking in at start of shift: 4/9 correct (46%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n\nSession 4 \u2014 2026-08-25 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 8/12 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d76\n scheduling medical appointments: 6/14 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n brushing teeth independently: 7/13 correct (51%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n wait calmly when a preferred activity is denied: 8/15 correct (56%); latency 3.2s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n clocking in at start of shift: 5/10 correct (50%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n\nSession 5 \u2014 2026-08-27 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 5/8 correct (66%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n scheduling medical appointments: 5/12 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n brushing teeth independently: 7/14 correct (51%); latency 3.6s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 6/9 correct (64%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n clocking in at start of shift: 4/8 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n\nSession 6 \u2014 2026-08-29 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 5/8 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n scheduling medical appointments: 4/8 correct (52%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n brushing teeth independently: 4/10 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 8/13 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n clocking in at start of shift: 4/9 correct (43%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n\nSession 7 \u2014 2026-08-31 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 7/9 correct (73%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n scheduling medical appointments: 4/9 correct (40%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n brushing teeth independently: 4/9 correct (42%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 6/9 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n clocking in at start of shift: 6/15 correct (42%); latency 4.3s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d710\n\nSession 8 \u2014 2026-09-02 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 9/12 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n scheduling medical appointments: 4/8 correct (53%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n brushing teeth independently: 4/10 correct (35%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n wait calmly when a preferred activity is denied: 8/14 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n clocking in at start of shift: 7/15 correct (43%); latency 4.0s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d74\n\nSession 9 \u2014 2026-09-04 \u2014 43 min \u2014 1 observer\n requesting a break during extended tasks: 9/14 correct (66%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n scheduling medical appointments: 4/11 correct (38%); latency 4.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n brushing teeth independently: 6/14 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n wait calmly when a preferred activity is denied: 7/10 correct (67%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n clocking in at start of shift: 5/13 correct (41%); latency 4.0s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 68% for 9 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 67.4%.\n- Trend direction: downward (first session: 70%, last session: 66%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "02c0ef3e43de42ad", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "adult", "n_sessions": 9, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "15263"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: spontaneously uses humor appropriately\nCurriculum Reference: Spontaneous Vocal Behavior L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Developing\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously uses humor appropriately\n- Curriculum: Spontaneous Vocal Behavior L3\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "dce5203c3082f7d9", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L3", "learner_profile": "adolescent", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously uses humor appropriately", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "76914"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by feature (What is round?)\nCurriculum: LRFFC L1\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by feature (What is round?)\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: towel, jacket, ball\nDistractor stimuli: book, blanket\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "665a1d982f3291a6", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by feature (What is round?)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "errorless", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "97173"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1664\nProfile: Adult Learner (18+ years)\nCurricula: Mand + Self-Care + Vocational\nDate range: Sessions 1\u201312 across 24 days (2026-04-02 to 2026-04-26)\n\nACCELERATION PROGRAMS\n1. indicating 'all done' at task completion (Mand)\n Method: NET\n2. requesting a break during extended tasks (Mand)\n Method: NET\n3. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n4. requesting assistance using AAC device (Mand)\n Method: NET\n5. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-02 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 7/12 correct (56%); latency 3.3s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n requesting a break during extended tasks: 5/8 correct (65%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n completing assigned tasks in order: 4/9 correct (44%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n requesting assistance using AAC device: 5/9 correct (60%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n brushing teeth independently: 4/8 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n\nSession 2 \u2014 2026-04-04 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 5/8 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n requesting a break during extended tasks: 7/10 correct (72%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n completing assigned tasks in order: 6/12 correct (47%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n requesting assistance using AAC device: 4/9 correct (49%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n brushing teeth independently: 6/12 correct (46%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n\nSession 3 \u2014 2026-04-06 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 5/8 correct (63%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n requesting a break during extended tasks: 9/15 correct (59%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d77\n completing assigned tasks in order: 7/15 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d78\n requesting assistance using AAC device: 4/8 correct (46%); latency 4.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n brushing teeth independently: 6/12 correct (49%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n\nSession 4 \u2014 2026-04-08 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 9/14 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n requesting a break during extended tasks: 9/15 correct (62%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d78\n completing assigned tasks in order: 5/10 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n requesting assistance using AAC device: 6/13 correct (45%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n brushing teeth independently: 4/8 correct (54%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n\nSession 5 \u2014 2026-04-10 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 4/8 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n requesting a break during extended tasks: 8/14 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d79\n completing assigned tasks in order: 3/8 correct (42%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n requesting assistance using AAC device: 7/13 correct (53%); latency 3.7s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n brushing teeth independently: 5/10 correct (49%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n\nSession 6 \u2014 2026-04-12 \u2014 IOA SESSION \u2014 2 observers\n indicating 'all done' at task completion: 49% (primary); IOA 93% agreement\n requesting a break during extended tasks: 67% (primary); IOA 93% agreement\n completing assigned tasks in order: 43% (primary); IOA 93% agreement\n requesting assistance using AAC device: 53% (primary); IOA 93% agreement\n brushing teeth independently: 51% (primary); IOA 93% agreement\n\nSession 7 \u2014 2026-04-14 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 4/8 correct (48%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n requesting a break during extended tasks: 9/13 correct (68%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n completing assigned tasks in order: 3/8 correct (34%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n requesting assistance using AAC device: 4/11 correct (40%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n brushing teeth independently: 8/15 correct (56%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n\nSession 8 \u2014 2026-04-16 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 6/11 correct (58%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n requesting a break during extended tasks: 6/10 correct (63%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n completing assigned tasks in order: 4/10 correct (37%); latency 4.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n requesting assistance using AAC device: 3/8 correct (39%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n brushing teeth independently: 4/10 correct (44%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n\nSession 9 \u2014 2026-04-18 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 6/14 correct (43%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n requesting a break during extended tasks: 7/11 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n completing assigned tasks in order: 5/13 correct (39%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n requesting assistance using AAC device: 6/11 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n brushing teeth independently: 4/8 correct (51%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n\nSession 10 \u2014 2026-04-20 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 4/9 correct (43%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n requesting a break during extended tasks: 5/8 correct (60%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n completing assigned tasks in order: 4/9 correct (39%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n requesting assistance using AAC device: 4/10 correct (42%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n brushing teeth independently: 5/10 correct (49%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n\nSession 11 \u2014 2026-04-22 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 5/10 correct (47%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n requesting a break during extended tasks: 6/10 correct (63%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n completing assigned tasks in order: 4/13 correct (31%); latency 4.5s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n requesting assistance using AAC device: 5/13 correct (35%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d711\n brushing teeth independently: 7/14 correct (51%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n\nSession 12 \u2014 2026-04-24 \u2014 54 min \u2014 1 observer\n indicating 'all done' at task completion: 6/12 correct (50%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n requesting a break during extended tasks: 6/9 correct (67%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n completing assigned tasks in order: 4/15 correct (28%); latency 4.6s; prompts FP\u00d71, PP\u00d75, G\u00d71, I\u00d78\n requesting assistance using AAC device: 3/8 correct (42%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n brushing teeth independently: 8/15 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Looking away during SD presentation\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 53% for 12 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 12 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 53.0%.\n- Trend direction: downward (first session: 56%, last session: 50%).\n- IOA data: 93% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "17b13ac4da2b3f04", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "adult", "n_sessions": 12, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "28229"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: compares quantities (more/less/same)\nCurriculum Reference: Math L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: compares quantities (more/less/same)\n- Curriculum: Math L2\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How many?\"\nVariations: \"Count these\"; \"What number is this?\"\nPresentation: Present the numeral card or a set of objects to count.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: napkin, scissors, apple\nDistractor stimuli: spoon, shoe, plate, cup\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7b9eacb92a8bc7db", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L2", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "compares quantities (more/less/same)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "errorless", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "27367"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: identifies 5\u201310 upper-case letters\nCurriculum: Reading L1\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies 5\u201310 upper-case letters\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: C, J, G\nDistractor stimuli: H, F\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "bb19268ca0ca039a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies 5\u201310 upper-case letters", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "93978"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: simple meal preparation (sandwich)\nCurriculum: AFLS Home Skills\nLearner: Adolescent Learner\nMastery Status: Mastered at current level\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: simple meal preparation (sandwich)\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Collect all ingredients (bread, filling, spread).\n2. Place two slices of bread on the plate.\n3. Apply spread to one slice.\n4. Place filling on the spread slice.\n5. Place the second slice on top.\n6. Cut the sandwich in half if desired.\n7. Place on a plate to serve.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7ebf1484022cb6be", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "simple meal preparation (sandwich)", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "mastered", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "simple meal preparation (sandwich)", "module": "home", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "42817"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: using public transportation\nCurriculum: AFLS Community Participation Skills\nLearner: Adolescent Learner\nMastery Status: Developing\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: using public transportation\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Confirm the correct stop or station. <- Learner-taught step (forward chaining starts here)\n2. Purchase or have the fare ready.\n3. Wait at the designated boarding area.\n4. Board the vehicle when it arrives.\n5. Pay the fare or present the pass.\n6. Sit or stand safely.\n7. Monitor the route for the correct stop.\n8. Exit at the correct stop.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "12eaa878425b5645", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "using public transportation", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "developing", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "using public transportation", "module": "community", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "92010"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a systematic desensitization / toleration program for the following.\n\nSkill: tolerating tooth brushing by caregiver\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adolescent Learner\nBaseline: Developing\n\nThe learner currently refuses or exhibits safety behaviors during the activity; the program should build tolerance gradually."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating tooth brushing by caregiver\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Developing\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for tooth brushing performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull 2-minute tooth-brushing routine completed by caregiver without refusal, crying, SIB, or aggression\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1-second toothbrush contact with teeth\n2. 3 seconds of brushing\n3. 5 seconds of brushing\n4. 15 seconds of brushing\n5. 30 seconds of brushing\n6. 60 seconds of brushing\n7. Full 2-minute brushing routine\n\nCurrent shaping step: At the developing stage, begin at shaping step 2 (\"3 seconds of brushing\") and advance through step 3 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal duration tolerated across 3 consecutive sessions with no safety behaviors (refusal, crying, SIB, aggression, property destruction).\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "21072f6b55e579d2", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating tooth brushing by caregiver", "program_type": "toleration", "learner_profile": "adolescent", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating tooth brushing by caregiver", "module": "basic_living", "mastery_criterion": "mc_toleration_full_duration", "shaping_n_steps": 7}, "teacher_model": null, "seed_tag": "98892"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts familiar people by name\nCurriculum: Tact L1\nLearner Level: Early Learner\nBaseline: Mastered at current level\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts familiar people by name\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: scissors, crayon, brush\nDistractor stimuli: car, apple\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "9c364ca007b1a2ad", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts familiar people by name", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "token_economy", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "69141"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies coins by name\nCurriculum Reference: Math L2\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies coins by name\n- Curriculum: Math L2\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: ball, towel, glue\nDistractor stimuli: book, blanket, apple, plate\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "bb30021e20265be7", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L2", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies coins by name", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "85854"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5737\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-08-07 to 2026-08-29)\n\nACCELERATION PROGRAMS\n1. ordering in a restaurant (Community)\n Method: Task Analysis\n2. washing hands independently (Self-Care)\n Method: Task Analysis\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n2. Property destruction \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-07 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 12/15 correct (77%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d79\n washing hands independently: 7/9 correct (81%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 12/15 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d75, I\u00d75\n Non-compliance: freq 2\n Property destruction: freq 3\n\nSession 2 \u2014 2026-08-09 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 7/10 correct (72%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n washing hands independently: 8/10 correct (76%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 6/8 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n Non-compliance: freq 2\n Property destruction: freq 4\n\nSession 3 \u2014 2026-08-11 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 6/10 correct (62%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n washing hands independently: 5/8 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 8/11 correct (70%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n Non-compliance: freq 2\n Property destruction: freq 3\n\nSession 4 \u2014 2026-08-13 \u2014 IOA SESSION \u2014 2 observers\n ordering in a restaurant: 58% (primary); IOA 97% agreement\n washing hands independently: 59% (primary); IOA 97% agreement\n wait calmly when access to preferred item is denied: 61% (primary); IOA 97% agreement\n Non-compliance: freq 2; IOA 97% agreement\n Property destruction: freq 5; IOA 97% agreement\n\nSession 5 \u2014 2026-08-15 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 6/13 correct (43%); latency 3.9s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d73\n washing hands independently: 7/15 correct (50%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n wait calmly when access to preferred item is denied: 5/8 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Non-compliance: freq 3\n Property destruction: freq 4\n\nSession 6 \u2014 2026-08-17 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 6/15 correct (39%); latency 4.2s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d77\n washing hands independently: 3/8 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 9/15 correct (57%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n Non-compliance: freq 3\n Property destruction: freq 5\n\nSession 7 \u2014 2026-08-19 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 3/10 correct (28%); latency 4.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n washing hands independently: 3/10 correct (34%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 4/9 correct (43%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n Non-compliance: freq 3\n Property destruction: freq 4\n\nSession 8 \u2014 2026-08-21 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 4/15 correct (23%); latency 5.2s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d74\n washing hands independently: 3/10 correct (30%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 3/8 correct (43%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n Non-compliance: freq 3\n Property destruction: freq 4\n\nSession 9 \u2014 2026-08-23 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 1/8 correct (15%); latency 5.6s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n washing hands independently: 4/13 correct (27%); latency 4.6s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 5/13 correct (40%); latency 4.2s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n Non-compliance: freq 4\n Property destruction: freq 6\n\nSession 10 \u2014 2026-08-25 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 2/15 correct (10%); latency 5.5s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d76\n washing hands independently: 2/12 correct (18%); latency 5.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 4/13 correct (32%); latency 4.3s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n Non-compliance: freq 4\n Property destruction: freq 7\n\nSession 11 \u2014 2026-08-27 \u2014 51 min \u2014 1 observer\n ordering in a restaurant: 1/9 correct (10%); latency 5.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n washing hands independently: 1/12 correct (10%); latency 5.7s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 3/11 correct (31%); latency 4.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n Non-compliance: freq 5\n Property destruction: freq 7\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Elopement from the work area\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 77% to 10% across 11 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nProperty destruction: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 39.7%.\n- Trend direction: downward (first session: 77%, last session: 10%).\n- Non-compliance: mean 3.0/session (early-window mean 2.0 -> late-window mean 4.3).\n- Property destruction: mean 4.7/session (early-window mean 3.3 -> late-window mean 6.7).\n- IOA data: 97% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "75afaba36c518978", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Non-compliance": "escape", "Property destruction": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adolescent", "n_sessions": 11, "n_programs": 3, "n_behaviors": 2, "behavior_ids": ["non_compliance", "property_destruction"], "behavior_functions": ["escape", "escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "46501"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: traces basic shapes (circle, cross)\nCurriculum Reference: Writing L1\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces basic shapes (circle, cross)\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: oval, rectangle, triangle\nDistractor stimuli: square, star\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3d5f7dbadda34951", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces basic shapes (circle, cross)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "68189"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: matches upper-case letters\nCurriculum Reference: Visual Perceptual Skills & Matching-to-Sample L2\nLearner Profile: Early Learner\nCurrent Mastery: Generalization phase\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches upper-case letters\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: Z, O, W\nDistractor stimuli: K, P, R, A\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "454315aea6f54820", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches upper-case letters", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "vr3", "error_correction": "simple_correction", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "97173"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: matches words to identical words\nCurriculum: Reading L1\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches words to identical words\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What word?\"\nVariations: \"Read this word\"; \"What does it say?\"\nPresentation: Present the sight-word card and deliver the SD.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: key, apple, car\nDistractor stimuli: book, brush\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "d84978d45bb8ebfa", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches words to identical words", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "93657"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously uses social phrases appropriately\nCurriculum: Spontaneous Vocal Behavior L2\nLearner: School-Age Learner\nMastery Status: Approaching mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously uses social phrases appropriately\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: activity setup.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's activity setup routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "3ce46eef40257ece", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously uses social phrases appropriately", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "time_delay", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "54871"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: describes recent events in sequence\nCurriculum Reference: Intraverbal L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: describes recent events in sequence\n- Curriculum: Intraverbal L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: transition conversation.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's transition conversation routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "ab4e819b342f5828", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "describes recent events in sequence", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "5962"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: engages in social games (peek-a-boo, tickle)\nCurriculum: Social Behavior & Social Play L1\nLearner Level: Early Learner\nBaseline: Developing\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: engages in social games (peek-a-boo, tickle)\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "c42d6dd38cae0351", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "engages in social games (peek-a-boo, tickle)", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "70185"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2769\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Reading + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-10-11 to 2026-10-27)\n\nACCELERATION PROGRAMS\n1. requesting preferred items using AAC device (Mand)\n Method: NET\n2. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n3. reads 10\u201320 sight words (Reading)\n Method: DTT\n4. indicating 'all done' with a step or activity (Mand)\n Method: NET\n5. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Aggression \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-11 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 11/15 correct (76%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d75, I\u00d75\n meal cleanup after eating: 10/15 correct (70%); latency 2.9s; prompts FP\u00d75, PP\u00d71, G\u00d75, I\u00d74\n reads 10\u201320 sight words: 6/8 correct (75%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n indicating 'all done' with a step or activity: 6/8 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n dressing (pants, shirt): 7/9 correct (77%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n Aggression: freq 2\n\nSession 2 \u2014 2026-10-13 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 9/13 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n meal cleanup after eating: 7/12 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 7/9 correct (76%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n indicating 'all done' with a step or activity: 8/11 correct (73%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n dressing (pants, shirt): 7/10 correct (69%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n Aggression: freq 3\n\nSession 3 \u2014 2026-10-15 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 9/13 correct (71%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n meal cleanup after eating: 5/10 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n reads 10\u201320 sight words: 7/10 correct (69%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n indicating 'all done' with a step or activity: 6/10 correct (60%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n dressing (pants, shirt): 6/10 correct (65%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n Aggression: freq 4\n\nSession 4 \u2014 2026-10-17 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 6/11 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n meal cleanup after eating: 5/10 correct (45%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n reads 10\u201320 sight words: 6/8 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n indicating 'all done' with a step or activity: 4/8 correct (53%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n dressing (pants, shirt): 7/11 correct (62%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n Aggression: freq 3\n\nSession 5 \u2014 2026-10-19 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 8/15 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n meal cleanup after eating: 5/15 correct (34%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d711\n reads 10\u201320 sight words: 5/8 correct (56%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n indicating 'all done' with a step or activity: 4/8 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n dressing (pants, shirt): 5/9 correct (58%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n Aggression: freq 4\n\nSession 6 \u2014 2026-10-21 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 4/10 correct (43%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n meal cleanup after eating: 4/15 correct (26%); latency 4.6s; prompts FP\u00d74, PP\u00d75, G\u00d73, I\u00d73\n reads 10\u201320 sight words: 4/8 correct (54%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n indicating 'all done' with a step or activity: 3/8 correct (38%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n dressing (pants, shirt): 5/10 correct (55%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n Aggression: freq 5\n\nSession 7 \u2014 2026-10-23 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 6/13 correct (42%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n meal cleanup after eating: 3/13 correct (27%); latency 5.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n reads 10\u201320 sight words: 4/8 correct (51%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n indicating 'all done' with a step or activity: 4/11 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n dressing (pants, shirt): 6/13 correct (49%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n Aggression: freq 5\n\nSession 8 \u2014 2026-10-25 \u2014 60 min \u2014 1 observer\n requesting preferred items using AAC device: 4/13 correct (30%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d79\n meal cleanup after eating: 2/14 correct (11%); latency 5.6s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n reads 10\u201320 sight words: 5/9 correct (51%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n indicating 'all done' with a step or activity: 4/13 correct (33%); latency 4.3s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n dressing (pants, shirt): 7/14 correct (47%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n Aggression: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-stimulatory behavior increase during demand periods\n- Vocal refusal ('no', 'I don't want to')\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Attempts at property destruction (throwing materials, swiping items off the surface)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (76% -> 30%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nAggression: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 55.6%.\n- Trend direction: downward (first session: 76%, last session: 30%).\n- Aggression: mean 4.0/session (early-window mean 2.5 -> late-window mean 5.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "ebee1bea6bf1e832", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Aggression": "attention"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["aggression"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "61520"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tracks print left to right\nCurriculum Reference: Reading L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tracks print left to right\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter or word on a flashcard.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: towel, shoe, toothbrush\nDistractor stimuli: hat, phone\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "da2ef8be0d1b5601", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tracks print left to right", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "fr2", "error_correction": "four_step", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "47307"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1818\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Writing\nDate range: Sessions 1\u20137 across 14 days (2026-03-17 to 2026-03-31)\n\nACCELERATION PROGRAMS\n1. ordering in a restaurant (Community)\n Method: Task Analysis\n2. making the bed (Home Skills)\n Method: Task Analysis\n3. requesting preferred items using AAC device (Mand)\n Method: NET\n4. copies simple words from model (Writing)\n Method: DTT\n5. identifying community safety signs (Community)\n Method: DTT\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: tangible\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-17 \u2014 36 min \u2014 1 observer\n ordering in a restaurant: 8/15 correct (52%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d77\n making the bed: 5/8 correct (60%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n requesting preferred items using AAC device: 8/13 correct (62%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n copies simple words from model: 6/12 correct (53%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n identifying community safety signs: 10/15 correct (64%); latency 3.0s; prompts FP\u00d73, PP\u00d75, G\u00d71, I\u00d76\n Tantrum: freq 2, duration 8m total\n\nSession 2 \u2014 2026-03-19 \u2014 36 min \u2014 1 observer\n ordering in a restaurant: 8/14 correct (56%); latency 3.5s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d74\n making the bed: 8/15 correct (55%); latency 3.7s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d78\n requesting preferred items using AAC device: 7/11 correct (63%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n copies simple words from model: 7/13 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d75\n identifying community safety signs: 7/12 correct (62%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n Tantrum: freq 2, duration 4m total\n\nSession 3 \u2014 2026-03-21 \u2014 IOA SESSION \u2014 2 observers\n ordering in a restaurant: 56% (primary); IOA 97% agreement\n making the bed: 55% (primary); IOA 97% agreement\n requesting preferred items using AAC device: 69% (primary); IOA 97% agreement\n copies simple words from model: 48% (primary); IOA 97% agreement\n identifying community safety signs: 65% (primary); IOA 97% agreement\n Tantrum: freq 7, duration 14m total; IOA 97% agreement\n\nSession 4 \u2014 2026-03-23 \u2014 36 min \u2014 1 observer\n ordering in a restaurant: 6/10 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n making the bed: 7/11 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n requesting preferred items using AAC device: 8/11 correct (73%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n copies simple words from model: 8/13 correct (58%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n identifying community safety signs: 7/10 correct (73%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n Tantrum: freq 7, duration 28m total\n\nSession 5 \u2014 2026-03-25 \u2014 36 min \u2014 1 observer\n ordering in a restaurant: 8/12 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d74\n making the bed: 9/15 correct (61%); latency 3.0s; prompts FP\u00d74, PP\u00d75, G\u00d73, I\u00d73\n requesting preferred items using AAC device: 11/15 correct (73%); latency 2.6s; prompts FP\u00d72, PP\u00d75, G\u00d73, I\u00d75\n copies simple words from model: 5/9 correct (54%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n identifying community safety signs: 5/8 correct (69%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Tantrum: freq 2, duration 6m total\n\nSession 6 \u2014 2026-03-27 \u2014 36 min \u2014 1 observer\n ordering in a restaurant: 8/13 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d78\n making the bed: 9/15 correct (60%); latency 3.1s; prompts FP\u00d75, PP\u00d74, G\u00d74, I\u00d72\n requesting preferred items using AAC device: 8/11 correct (74%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n copies simple words from model: 7/13 correct (53%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n identifying community safety signs: 6/9 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n Tantrum: freq 2, duration 8m total\n\nSession 7 \u2014 2026-03-29 \u2014 36 min \u2014 1 observer\n ordering in a restaurant: 6/9 correct (67%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n making the bed: 7/10 correct (66%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n requesting preferred items using AAC device: 11/14 correct (78%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n copies simple words from model: 6/12 correct (51%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d77\n identifying community safety signs: 7/10 correct (75%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n Tantrum: freq 2, duration 6m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Attempts at property destruction (throwing materials, swiping items off the surface)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Peer aggression (hitting or pushing peers during group activities)\n- Pushing materials away from the work area\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nTantrum: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 60.1%.\n- Trend direction: upward (first session: 52%, last session: 67%).\n- Tantrum: mean 3.4/session (early-window mean 2.0 -> late-window mean 2.0).\n- IOA data: 97% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "e438e7118ca6b489", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Tantrum": "tangible"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adolescent", "n_sessions": 7, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["tangible"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "41987"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9045\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Tact\nDate range: Sessions 1\u201312 across 24 days (2026-06-11 to 2026-07-05)\n\nACCELERATION PROGRAMS\n1. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n2. selects items by feature (Listener Responding)\n Method: DTT\n3. tacts common objects (Tact)\n Method: DTT\n4. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: automatic\n2. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-11 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 6/8 correct (73%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n selects items by feature: 9/13 correct (66%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n tacts common objects: 8/12 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 12/15 correct (77%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n Elopement: freq 2\n Tantrum: freq 2, duration 10m total\n\nSession 2 \u2014 2026-06-13 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 9/13 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n selects items by feature: 8/14 correct (59%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n tacts common objects: 10/14 correct (71%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/13 correct (73%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d78\n Elopement: freq 1\n Tantrum: freq 2, duration 6m total\n\nSession 3 \u2014 2026-06-15 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 5/9 correct (60%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n selects items by feature: 7/13 correct (55%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n tacts common objects: 6/9 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d73\n Elopement: freq 1\n Tantrum: freq 4, duration 8m total\n\nSession 4 \u2014 2026-06-17 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 6/14 correct (46%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d77\n selects items by feature: 4/8 correct (48%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n tacts common objects: 5/8 correct (58%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/15 correct (55%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n Elopement: freq 2\n Tantrum: freq 4, duration 16m total\n\nSession 5 \u2014 2026-06-19 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 5/10 correct (48%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n selects items by feature: 6/14 correct (43%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n tacts common objects: 5/9 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/15 correct (47%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d75, I\u00d79\n Elopement: freq 3\n Tantrum: freq 3, duration 15m total\n\nSession 6 \u2014 2026-06-21 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 5/15 correct (34%); latency 4.7s; prompts FP\u00d72, PP\u00d75, G\u00d71, I\u00d77\n selects items by feature: 3/11 correct (31%); latency 4.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n tacts common objects: 6/14 correct (44%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/8 correct (47%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n Elopement: freq 3\n Tantrum: freq 4, duration 20m total\n\nSession 7 \u2014 2026-06-23 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 4/14 correct (29%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n selects items by feature: 3/14 correct (23%); latency 4.8s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n tacts common objects: 4/9 correct (47%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/11 correct (37%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n Elopement: freq 3\n Tantrum: freq 3, duration 6m total\n\nSession 8 \u2014 2026-06-25 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 3/15 correct (23%); latency 4.7s; prompts FP\u00d75, PP\u00d70, G\u00d72, I\u00d78\n selects items by feature: 3/15 correct (19%); latency 5.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d712\n tacts common objects: 5/11 correct (42%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/12 correct (34%); latency 4.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n Elopement: freq 3\n Tantrum: freq 5, duration 15m total\n\nSession 9 \u2014 2026-06-27 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 1/11 correct (12%); latency 5.2s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n selects items by feature: 2/15 correct (14%); latency 5.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d711\n tacts common objects: 4/13 correct (32%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/15 correct (28%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n Elopement: freq 4\n Tantrum: freq 4, duration 16m total\n\nSession 10 \u2014 2026-06-29 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 1/14 correct (10%); latency 5.5s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n selects items by feature: 1/14 correct (10%); latency 5.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n tacts common objects: 4/10 correct (36%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/14 correct (21%); latency 4.8s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n Elopement: freq 3\n Tantrum: freq 5, duration 15m total\n\nSession 11 \u2014 2026-07-01 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 1/11 correct (10%); latency 5.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n selects items by feature: 1/9 correct (10%); latency 5.7s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n tacts common objects: 2/8 correct (24%); latency 5.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/11 correct (14%); latency 5.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n Elopement: freq 4\n Tantrum: freq 6, duration 24m total\n\nSession 12 \u2014 2026-07-03 \u2014 33 min \u2014 1 observer\n follows 1-step motor instructions: 2/15 correct (10%); latency 5.7s; prompts FP\u00d75, PP\u00d72, G\u00d72, I\u00d76\n selects items by feature: 1/8 correct (10%); latency 5.3s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n tacts common objects: 3/12 correct (24%); latency 5.1s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 1/14 correct (10%); latency 5.6s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d77\n Elopement: freq 4\n Tantrum: freq 5, duration 25m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Turning away from task or staff\n- Increased response latency (3\u20135\u00d7 baseline)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 73% to 10% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nElopement: automatic\n Evidence: behavior persists across contexts without clear social mediation.\nTantrum: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 35.4%.\n- Trend direction: downward (first session: 73%, last session: 10%).\n- Elopement: mean 2.8/session (early-window mean 1.5 -> late-window mean 3.8).\n- Tantrum: mean 3.9/session (early-window mean 3.0 -> late-window mean 5.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "86b21efbc1764ec5", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Elopement": "automatic", "Tantrum": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 12, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["elopement", "tantrum"], "behavior_functions": ["automatic", "escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "57327"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: putting on a shirt (pullover)\nCurriculum: AFLS Basic Living Skills\nLearner: Adult Learner\nMastery Status: Developing\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: putting on a shirt (pullover)\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Hold the shirt with the front facing away.\n2. Place both arms into the sleeves.\n3. Pull the opening over the head.\n4. Push arms through the sleeves fully.\n5. Pull the bottom of the shirt down to straighten it.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5c576fdd565d9397", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "putting on a shirt (pullover)", "program_type": "independence", "learner_profile": "adult", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "putting on a shirt (pullover)", "module": "basic_living", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "75542"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: engages in social games (peek-a-boo, tickle)\nCurriculum: Social Behavior & Social Play L1\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: engages in social games (peek-a-boo, tickle)\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "3544c885f5b9153e", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "engages in social games (peek-a-boo, tickle)", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "time_delay", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "16366"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: crossing a street safely\nCurriculum: AFLS Community Participation Skills\nLearner: Adult Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: crossing a street safely\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Approach the crosswalk and stop at the curb.\n2. Look left, look right, look left again.\n3. Identify the walk signal or absence of traffic.\n4. Begin to cross while continuing to scan for vehicles.\n5. Cross directly to the opposite curb.\n6. Step up onto the sidewalk.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 90s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "748e3971ed7eb394", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "crossing a street safely", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "crossing a street safely", "module": "community", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "98892"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6427\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Mand + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-06-07 to 2026-06-25)\n\nACCELERATION PROGRAMS\n1. doing laundry (Home Skills)\n Method: Task Analysis\n2. requesting a break during extended tasks (Mand)\n Method: NET\n3. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n4. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: tangible\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-07 \u2014 39 min \u2014 1 observer\n doing laundry: 7/9 correct (79%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n requesting a break during extended tasks: 6/8 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n wait calmly when a preferred activity is denied: 7/10 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n simple meal preparation: 9/11 correct (86%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n Tantrum: freq 1, duration 5m total\n\nSession 2 \u2014 2026-06-09 \u2014 39 min \u2014 1 observer\n doing laundry: 9/11 correct (80%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n requesting a break during extended tasks: 10/12 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 11/15 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d77\n simple meal preparation: 11/14 correct (82%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n Tantrum: freq 1, duration 4m total\n\nSession 3 \u2014 2026-06-11 \u2014 39 min \u2014 1 observer\n doing laundry: 6/8 correct (80%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting a break during extended tasks: 6/8 correct (80%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 6/8 correct (72%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n simple meal preparation: 7/9 correct (82%); latency 2.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n Tantrum: freq 1, duration 3m total\n\nSession 4 \u2014 2026-06-13 \u2014 39 min \u2014 1 observer\n doing laundry: 4/9 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n requesting a break during extended tasks: 4/8 correct (44%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 4/9 correct (43%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n simple meal preparation: 5/9 correct (51%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n Tantrum: freq 0\n\nSession 5 \u2014 2026-06-15 \u2014 39 min \u2014 1 observer\n doing laundry: 5/11 correct (46%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n requesting a break during extended tasks: 5/11 correct (47%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 5/11 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n simple meal preparation: 6/11 correct (53%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n Tantrum: freq 0\n\nSession 6 \u2014 2026-06-17 \u2014 39 min \u2014 1 observer\n doing laundry: 7/15 correct (46%); latency 4.1s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n requesting a break during extended tasks: 5/11 correct (49%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when a preferred activity is denied: 6/13 correct (43%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n simple meal preparation: 7/12 correct (55%); latency 3.7s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n Tantrum: freq 1, duration 4m total\n\nSession 7 \u2014 2026-06-19 \u2014 IOA SESSION \u2014 2 observers\n doing laundry: 43% (primary); IOA 91% agreement\n requesting a break during extended tasks: 47% (primary); IOA 91% agreement\n wait calmly when a preferred activity is denied: 50% (primary); IOA 91% agreement\n simple meal preparation: 51% (primary); IOA 91% agreement\n Tantrum: freq 1, duration 4m total; IOA 91% agreement\n\nSession 8 \u2014 2026-06-21 \u2014 39 min \u2014 1 observer\n doing laundry: 5/11 correct (47%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n requesting a break during extended tasks: 8/15 correct (50%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d75, I\u00d76\n wait calmly when a preferred activity is denied: 4/8 correct (54%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n simple meal preparation: 5/9 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n Tantrum: freq 0\n\nSession 9 \u2014 2026-06-23 \u2014 39 min \u2014 1 observer\n doing laundry: 7/14 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n requesting a break during extended tasks: 7/13 correct (57%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d79\n wait calmly when a preferred activity is denied: 8/13 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n simple meal preparation: 8/12 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d74\n Tantrum: freq 2, duration 4m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nTantrum: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 57.6%.\n- Trend direction: downward (first session: 79%, last session: 52%).\n- Tantrum: mean 0.8/session (early-window mean 1.0 -> late-window mean 1.0).\n- IOA data: 91% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "9fbe31b7612cabb9", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Tantrum": "tangible"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adult", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["tangible"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "37908"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8223\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-08-11 to 2026-08-29)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n4. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n5. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-11 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 6/9 correct (66%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n wait calmly when access to preferred item is denied: 9/14 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n indicating completion ('all done', 'finished'): 6/10 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n follows 1-step motor instructions: 8/12 correct (69%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n Property destruction: freq 2\n\nSession 2 \u2014 2026-08-13 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 5/8 correct (59%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 5/9 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n indicating completion ('all done', 'finished'): 7/12 correct (59%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n follows 1-step motor instructions: 9/13 correct (72%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d75\n Property destruction: freq 1\n\nSession 3 \u2014 2026-08-15 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 7/11 correct (67%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 6/9 correct (62%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n indicating completion ('all done', 'finished'): 8/11 correct (69%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (56%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d75\n follows 1-step motor instructions: 9/13 correct (71%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d711\n Property destruction: freq 4\n\nSession 4 \u2014 2026-08-17 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 6/8 correct (70%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 6/10 correct (64%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n indicating completion ('all done', 'finished'): 8/12 correct (68%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (51%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n follows 1-step motor instructions: 7/11 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n Property destruction: freq 5\n ABC (property destruction): A = staff engaged in paperwork nearby; B = yelled name + tapped staff; C = staff attended and redirected\n\nSession 5 \u2014 2026-08-19 \u2014 IOA SESSION \u2014 2 observers\n reads 10\u201320 sight words: 71% (primary); IOA 96% agreement\n wait calmly when access to preferred item is denied: 69% (primary); IOA 96% agreement\n indicating completion ('all done', 'finished'): 64% (primary); IOA 96% agreement\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 54% (primary); IOA 96% agreement\n follows 1-step motor instructions: 79% (primary); IOA 96% agreement\n Property destruction: freq 1; IOA 96% agreement\n\nSession 6 \u2014 2026-08-21 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 6/10 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 8/12 correct (66%); latency 3.0s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n indicating completion ('all done', 'finished'): 10/14 correct (74%); latency 2.5s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/9 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n follows 1-step motor instructions: 7/10 correct (74%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n Property destruction: freq 1\n\nSession 7 \u2014 2026-08-23 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 9/14 correct (65%); latency 3.4s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 11/14 correct (76%); latency 2.5s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n indicating completion ('all done', 'finished'): 6/9 correct (71%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/13 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d79\n follows 1-step motor instructions: 10/12 correct (80%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n Property destruction: freq 2\n\nSession 8 \u2014 2026-08-25 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 6/9 correct (71%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 10/14 correct (74%); latency 2.4s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n indicating completion ('all done', 'finished'): 6/8 correct (78%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (54%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n follows 1-step motor instructions: 11/12 correct (88%); latency 1.9s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n Property destruction: freq 1\n\nSession 9 \u2014 2026-08-27 \u2014 38 min \u2014 1 observer\n reads 10\u201320 sight words: 7/10 correct (65%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 11/15 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n indicating completion ('all done', 'finished'): 10/13 correct (75%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (54%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n follows 1-step motor instructions: 9/11 correct (81%); latency 2.0s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n Property destruction: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Pushing materials away from the work area\n- Self-stimulatory behavior increase during demand periods\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nProperty destruction: attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 66.5%.\n- Trend direction: downward (first session: 66%, last session: 65%).\n- Property destruction: mean 2.1/session (early-window mean 2.3 -> late-window mean 1.7).\n- IOA data: 96% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "7c191178acb6612f", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Property destruction": "attention"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "early", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["attention"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "23299"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner: Adult Learner\nMastery Status: Emerging\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the emerging stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3f1ba02b0880c98d", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adult", "mastery_state": "emerging", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_generalization", "chain_type": "total_task", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "39765"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: writes numbers 1\u201310\nCurriculum: Writing L2\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: writes numbers 1\u201310\n- Curriculum: Writing L2\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: 6, 10, 3\nDistractor stimuli: 7, 2, 1, 8\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5979b1067f9cb981", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L2", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "writes numbers 1\u201310", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "token_economy", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "15942"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a toleration program for this self-care / medical / caregiver-delivered skill.\n\nTarget: tolerating tooth brushing by caregiver\nCurriculum: AFLS Basic Living Skills\nLearner: Adolescent Learner\nMastery Status: Mastered at current level\n\nPlease include the toleration end goal, shaping progression, antecedent strategies, reinforcement plan, refusal / safety procedures, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating tooth brushing by caregiver\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for tooth brushing performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull 2-minute tooth-brushing routine completed by caregiver without refusal, crying, SIB, or aggression\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1-second toothbrush contact with teeth\n2. 3 seconds of brushing\n3. 5 seconds of brushing\n4. 15 seconds of brushing\n5. 30 seconds of brushing\n6. 60 seconds of brushing\n7. Full 2-minute brushing routine\n\nCurrent shaping step: At the mastered stage, begin at shaping step 5 (\"30 seconds of brushing\") and advance through step 6 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal duration tolerated across 3 consecutive sessions with no safety behaviors (refusal, crying, SIB, aggression, property destruction).\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "b26eab424c93c320", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating tooth brushing by caregiver", "program_type": "toleration", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating tooth brushing by caregiver", "module": "basic_living", "mastery_criterion": "mc_toleration_full_duration", "shaping_n_steps": 7}, "teacher_model": null, "seed_tag": "42817"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: completes simple 3\u20134 piece puzzles\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L1\nLearner Level: Early Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completes simple 3\u20134 piece puzzles\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: napkin, sock, backpack\nDistractor stimuli: cup, bottle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "09bdbd5a14b3a9fa", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "completes simple 3\u20134 piece puzzles", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "5667"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: completing assigned tasks in order\nCurriculum: AFLS Vocational Skills\nLearner Level: Adult Learner\nBaseline: Near mastery\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Review the task list for the shift. <- Learner-taught step (forward chaining starts here)\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "efc84cccae744192", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adult", "mastery_state": "near", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_all_steps_independent", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "57808"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: describes function of common objects\nCurriculum: Intraverbal L2\nLearner Level: School-Age Learner\nBaseline: Emerging\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: describes function of common objects\n- Curriculum: Intraverbal L2\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: transition conversation.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's transition conversation routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "d5fdbe0e45328959", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "describes function of common objects", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "70185"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8285\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u201312 across 24 days (2026-07-18 to 2026-08-11)\n\nACCELERATION PROGRAMS\n1. requesting a break during extended tasks (Mand)\n Method: NET\n2. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n3. indicating 'all done' at task completion (Mand)\n Method: NET\n4. managing a budget (Independent Living)\n Method: Task Analysis\n5. self-advocacy statements (Independent Living)\n Method: BST\n6. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Pica \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-18 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 6/13 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n completing assigned tasks in order: 4/11 correct (40%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n indicating 'all done' at task completion: 5/11 correct (43%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n managing a budget: 8/15 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d76\n self-advocacy statements: 5/15 correct (37%); latency 4.4s; prompts FP\u00d75, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 4/10 correct (43%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n Pica: attempts 0\n\nSession 2 \u2014 2026-07-20 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 5/11 correct (46%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n completing assigned tasks in order: 3/9 correct (37%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n indicating 'all done' at task completion: 6/11 correct (53%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n managing a budget: 4/8 correct (50%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n self-advocacy statements: 6/13 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n wait calmly when a preferred activity is denied: 5/11 correct (47%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-07-22 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 6/13 correct (50%); latency 3.8s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n completing assigned tasks in order: 4/9 correct (48%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n indicating 'all done' at task completion: 6/10 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n managing a budget: 6/9 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n self-advocacy statements: 6/11 correct (52%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 7/14 correct (48%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-07-24 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 6/10 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n completing assigned tasks in order: 7/13 correct (51%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n indicating 'all done' at task completion: 10/14 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n managing a budget: 6/9 correct (62%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n self-advocacy statements: 5/9 correct (57%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n wait calmly when a preferred activity is denied: 9/15 correct (60%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d76\n Pica: attempts 1 (0 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-07-26 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 8/13 correct (64%); latency 3.4s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n completing assigned tasks in order: 6/11 correct (59%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n indicating 'all done' at task completion: 6/8 correct (76%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n managing a budget: 10/14 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n self-advocacy statements: 9/14 correct (65%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 5/8 correct (65%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Pica: attempts 1 (0 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-07-28 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 6/9 correct (69%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n completing assigned tasks in order: 9/15 correct (58%); latency 3.5s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n indicating 'all done' at task completion: 9/12 correct (78%); latency 2.6s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n managing a budget: 12/15 correct (79%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d75, I\u00d78\n self-advocacy statements: 6/8 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n wait calmly when a preferred activity is denied: 5/8 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-07-30 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 7/9 correct (74%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n completing assigned tasks in order: 7/10 correct (67%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n indicating 'all done' at task completion: 9/10 correct (91%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n managing a budget: 8/10 correct (80%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n self-advocacy statements: 10/13 correct (81%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 9/14 correct (67%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-08-01 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 7/9 correct (78%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n completing assigned tasks in order: 10/13 correct (75%); latency 2.6s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n indicating 'all done' at task completion: 12/13 correct (94%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n managing a budget: 12/14 correct (89%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d79\n self-advocacy statements: 8/9 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 9/12 correct (75%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n Pica: attempts 0\n\nSession 9 \u2014 2026-08-03 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 10/13 correct (76%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n completing assigned tasks in order: 7/9 correct (83%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n indicating 'all done' at task completion: 10/10 correct (95%); latency 1.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n managing a budget: 14/15 correct (93%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d710\n self-advocacy statements: 11/12 correct (95%); latency 1.9s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d74\n wait calmly when a preferred activity is denied: 9/11 correct (85%); latency 2.0s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 10 \u2014 2026-08-05 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 11/14 correct (82%); latency 2.1s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n completing assigned tasks in order: 9/11 correct (82%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n indicating 'all done' at task completion: 10/10 correct (95%); latency 1.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n managing a budget: 8/8 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n self-advocacy statements: 10/10 correct (95%); latency 1.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 9/10 correct (86%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n Pica: attempts 0\n\nSession 11 \u2014 2026-08-07 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 10/12 correct (87%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n completing assigned tasks in order: 11/12 correct (92%); latency 1.6s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n indicating 'all done' at task completion: 9/9 correct (95%); latency 1.9s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n managing a budget: 8/8 correct (95%); latency 1.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n self-advocacy statements: 10/10 correct (95%); latency 1.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 10/11 correct (90%); latency 1.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 12 \u2014 2026-08-09 \u2014 42 min \u2014 1 observer\n requesting a break during extended tasks: 11/13 correct (87%); latency 2.0s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d76\n completing assigned tasks in order: 12/13 correct (95%); latency 2.0s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n indicating 'all done' at task completion: 10/11 correct (95%); latency 1.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n managing a budget: 9/9 correct (95%); latency 2.0s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n self-advocacy statements: 13/14 correct (95%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d710\n wait calmly when a preferred activity is denied: 9/10 correct (93%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Quick response latency (within 2\u20133 seconds)\n- Appropriate sitting posture maintained\n- Reaching for materials proactively\n- Smiling during reinforcement delivery\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 43% to 87% across 12 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Behavior Function Hypothesis\nPica: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 67.8%.\n- Trend direction: upward (first session: 43%, last session: 87%).\n- Pica: mean 0.9/session (early-window mean 1.0 -> late-window mean 1.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "16ed6f0f450e545e", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {"Pica": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "adult", "n_sessions": 12, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["pica"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "20691"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: scheduling a medical appointment\nCurriculum: AFLS Independent Living Skills\nLearner: Adult Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: scheduling a medical appointment\n- Curriculum: AFLS Independent Living Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Identify the reason for the appointment. <- Learner-taught step (forward chaining starts here)\n2. Locate the provider's contact information.\n3. Call or use the online system.\n4. Provide name, date of birth, and insurance information.\n5. Request an available appointment time.\n6. Confirm the appointment and save it to the calendar.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "498c53a8430cff22", "gold_labels": {"method": "task_analysis", "domain": "AFLS.independent_living", "skill": "scheduling a medical appointment", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "scheduling a medical appointment", "module": "independent_living", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "50550"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts common animals\nCurriculum: Tact L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common animals\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What animal is this?\"\nVariations: \"What is this?\"; \"Name this animal\"\nPresentation: Present an animal picture card and ask while pointing.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: frog, lion, horse\nDistractor stimuli: elephant, rabbit\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "33dfb6b1b25a4cbf", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common animals", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "69141"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3420\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-06-01 to 2026-06-17)\n\nACCELERATION PROGRAMS\n1. tolerating haircut (Self-Care)\n Method: Task Analysis\n2. ordering in a restaurant (Community)\n Method: Task Analysis\n3. identifying community safety signs (Community)\n Method: DTT\n4. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-01 \u2014 42 min \u2014 1 observer\n tolerating haircut: 10/14 correct (71%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n ordering in a restaurant: 7/10 correct (67%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n identifying community safety signs: 11/13 correct (85%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n tolerating tooth brushing by caregiver: 12/15 correct (80%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d710\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 2 \u2014 2026-06-03 \u2014 42 min \u2014 1 observer\n tolerating haircut: 9/12 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d77\n ordering in a restaurant: 6/9 correct (64%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n identifying community safety signs: 8/11 correct (72%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n tolerating tooth brushing by caregiver: 10/14 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d75\n Toileting accident (urine or bowel): urine: 3 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-06-05 \u2014 42 min \u2014 1 observer\n tolerating haircut: 7/11 correct (63%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n ordering in a restaurant: 7/11 correct (62%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n identifying community safety signs: 8/13 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n tolerating tooth brushing by caregiver: 9/15 correct (63%); latency 3.2s; prompts FP\u00d75, PP\u00d72, G\u00d71, I\u00d77\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = transition from preferred activity to non-preferred task; B = voided (urine) en route to the work area; C = staff redirected to bathroom, changed clothing\n\nSession 4 \u2014 2026-06-07 \u2014 42 min \u2014 1 observer\n tolerating haircut: 7/13 correct (54%); latency 3.3s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n ordering in a restaurant: 7/12 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d77\n identifying community safety signs: 9/15 correct (62%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d76\n tolerating tooth brushing by caregiver: 7/11 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 5 \u2014 2026-06-09 \u2014 42 min \u2014 1 observer\n tolerating haircut: 4/10 correct (44%); latency 4.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n ordering in a restaurant: 6/14 correct (46%); latency 4.1s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n identifying community safety signs: 6/11 correct (52%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n tolerating tooth brushing by caregiver: 7/13 correct (52%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d77\n Toileting accident (urine or bowel): urine: 0 in-toilet / 5 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = transition from preferred activity to non-preferred task; B = voided (urine) en route to the work area; C = staff redirected to bathroom, changed clothing\n\nSession 6 \u2014 2026-06-11 \u2014 42 min \u2014 1 observer\n tolerating haircut: 5/12 correct (40%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n ordering in a restaurant: 4/10 correct (42%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n identifying community safety signs: 4/9 correct (47%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n tolerating tooth brushing by caregiver: 7/13 correct (51%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d76\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 7 \u2014 2026-06-13 \u2014 42 min \u2014 1 observer\n tolerating haircut: 5/15 correct (32%); latency 4.5s; prompts FP\u00d75, PP\u00d70, G\u00d75, I\u00d75\n ordering in a restaurant: 5/14 correct (34%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d711\n identifying community safety signs: 4/10 correct (36%); latency 4.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n tolerating tooth brushing by caregiver: 7/15 correct (45%); latency 4.2s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n Toileting accident (urine or bowel): urine: 2 in-toilet / 4 accidents; BM: 0 in-toilet / 2 accidents\n\nSession 8 \u2014 2026-06-15 \u2014 42 min \u2014 1 observer\n tolerating haircut: 3/9 correct (28%); latency 4.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n ordering in a restaurant: 5/14 correct (35%); latency 4.4s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d78\n identifying community safety signs: 2/9 correct (27%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n tolerating tooth brushing by caregiver: 4/11 correct (32%); latency 4.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n Toileting accident (urine or bowel): urine: 0 in-toilet / 5 accidents; BM: 1 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Crying during demand presentation\n- Turning away from task or staff"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (71% -> 28%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 50.5%.\n- Trend direction: downward (first session: 71%, last session: 28%).\n- Toileting accident (urine or bowel): mean 4.0/session (early-window mean 2.5 -> late-window mean 5.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "47144ed42308d4ac", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Toileting accident (urine or bowel)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "94010"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8754\nProfile: Adult Learner (18+ years)\nCurricula: Community + Home Skills + Mand + Self-Care + Vocational\nDate range: Sessions 1\u201311 across 22 days (2026-02-17 to 2026-03-11)\n\nACCELERATION PROGRAMS\n1. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n2. requesting assistance using AAC device (Mand)\n Method: NET\n3. using public transportation (Community)\n Method: Task Analysis\n4. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n5. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Mouthing \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-17 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 9/13 correct (71%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n requesting assistance using AAC device: 8/11 correct (77%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n using public transportation: 7/8 correct (84%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n completing assigned tasks in order: 9/12 correct (76%); latency 2.6s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d75\n simple meal preparation: 8/11 correct (76%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n Mouthing: freq 2; PIR 12%\n\nSession 2 \u2014 2026-02-19 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 11/15 correct (74%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d712\n requesting assistance using AAC device: 8/11 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n using public transportation: 9/10 correct (85%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n completing assigned tasks in order: 6/8 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n simple meal preparation: 10/13 correct (76%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d79\n Mouthing: freq 0; PIR 4%\n\nSession 3 \u2014 2026-02-21 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 8/11 correct (76%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n requesting assistance using AAC device: 7/10 correct (72%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n using public transportation: 6/8 correct (78%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n completing assigned tasks in order: 8/11 correct (73%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n simple meal preparation: 6/8 correct (78%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n Mouthing: freq 2; PIR 13%\n\nSession 4 \u2014 2026-02-23 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 4/12 correct (33%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n requesting assistance using AAC device: 4/9 correct (39%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n using public transportation: 5/12 correct (45%); latency 4.2s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d74\n completing assigned tasks in order: 4/9 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n simple meal preparation: 4/8 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Mouthing: freq 1; PIR 10%\n\nSession 5 \u2014 2026-02-25 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 3/8 correct (36%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n requesting assistance using AAC device: 4/12 correct (36%); latency 4.2s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n using public transportation: 4/9 correct (47%); latency 3.9s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n completing assigned tasks in order: 6/13 correct (47%); latency 3.8s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n simple meal preparation: 6/12 correct (48%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n Mouthing: freq 1; PIR 6%\n\nSession 6 \u2014 2026-02-27 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 4/11 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n requesting assistance using AAC device: 5/11 correct (44%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n using public transportation: 4/8 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n completing assigned tasks in order: 6/11 correct (54%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n simple meal preparation: 7/14 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n Mouthing: freq 1; PIR 12%\n\nSession 7 \u2014 2026-03-01 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 4/11 correct (35%); latency 4.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n requesting assistance using AAC device: 4/9 correct (47%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n using public transportation: 6/10 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n completing assigned tasks in order: 7/11 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n simple meal preparation: 8/14 correct (60%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d77\n Mouthing: freq 1; PIR 13%\n\nSession 8 \u2014 2026-03-03 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 5/12 correct (40%); latency 4.2s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n requesting assistance using AAC device: 8/15 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n using public transportation: 8/12 correct (67%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n completing assigned tasks in order: 9/13 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n simple meal preparation: 10/15 correct (66%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d79\n Mouthing: freq 0; PIR 2%\n\nSession 9 \u2014 2026-03-05 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 4/9 correct (50%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n requesting assistance using AAC device: 8/14 correct (59%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d714\n using public transportation: 10/14 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d77\n completing assigned tasks in order: 7/10 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n simple meal preparation: 7/9 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n Mouthing: freq 0; PIR 6%\n\nSession 10 \u2014 2026-03-07 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 8/14 correct (61%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n requesting assistance using AAC device: 7/10 correct (69%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n using public transportation: 11/15 correct (76%); latency 2.5s; prompts FP\u00d75, PP\u00d71, G\u00d72, I\u00d77\n completing assigned tasks in order: 7/9 correct (73%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n simple meal preparation: 9/12 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n Mouthing: freq 2; PIR 14%\n\nSession 11 \u2014 2026-03-09 \u2014 53 min \u2014 1 observer\n dressing (pants, shirt): 7/10 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n requesting assistance using AAC device: 9/12 correct (75%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d710\n using public transportation: 6/8 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n completing assigned tasks in order: 9/13 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n simple meal preparation: 12/15 correct (79%); latency 2.7s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n Mouthing: freq 1; PIR 9%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n- Playing with non-task materials\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nMouthing: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 53.1%.\n- Trend direction: upward (first session: 71%, last session: 72%).\n- Mouthing: mean 1.0/session (early-window mean 1.3 -> late-window mean 1.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "414037a2432bb275", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Mouthing": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adult", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["mouthing"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "2880"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: provides definitions of words\nCurriculum: Intraverbal L3\nLearner: School-Age Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: provides definitions of words\n- Curriculum: Intraverbal L3\n- Learner: School-Age Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: post-activity debrief.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's post-activity debrief routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "b5331fa15b5b19d4", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "school_age", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "provides definitions of words", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "76914"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9260\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Motor Imitation + Tact + Writing\nDate range: Sessions 1\u20138 across 16 days (2026-06-04 to 2026-06-20)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. tacts common objects (Tact)\n Method: DTT\n3. selects items by feature (Listener Responding)\n Method: DTT\n4. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n2. Toileting accident (urine or bowel) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-04 \u2014 47 min \u2014 1 observer\n copies simple words from model: 7/10 correct (73%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n tacts common objects: 11/14 correct (80%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n selects items by feature: 6/8 correct (70%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n imitates 3-step motor sequences: 10/12 correct (81%); latency 2.4s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n Tantrum: freq 1, duration 4m total\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 2 \u2014 2026-06-06 \u2014 47 min \u2014 1 observer\n copies simple words from model: 6/10 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n tacts common objects: 10/15 correct (67%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n selects items by feature: 9/13 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n imitates 3-step motor sequences: 10/13 correct (73%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n Tantrum: freq 2, duration 6m total\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-06-08 \u2014 47 min \u2014 1 observer\n copies simple words from model: 6/10 correct (63%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n tacts common objects: 9/13 correct (68%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n selects items by feature: 6/10 correct (61%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n imitates 3-step motor sequences: 9/12 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n Tantrum: freq 2, duration 6m total\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 4 \u2014 2026-06-10 \u2014 47 min \u2014 1 observer\n copies simple words from model: 6/10 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n tacts common objects: 6/10 correct (57%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n selects items by feature: 7/13 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n imitates 3-step motor sequences: 7/11 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n Tantrum: freq 3, duration 9m total\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 5 \u2014 2026-06-12 \u2014 47 min \u2014 1 observer\n copies simple words from model: 6/14 correct (45%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n tacts common objects: 5/9 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n selects items by feature: 7/15 correct (48%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n imitates 3-step motor sequences: 6/9 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n Tantrum: freq 2, duration 6m total\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n ABC (tantrum): A = demand to complete math problem; B = threw materials and vocalized 'no'; C = staff removed demand, redirected\n\nSession 6 \u2014 2026-06-14 \u2014 47 min \u2014 1 observer\n copies simple words from model: 5/12 correct (42%); latency 4.3s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d75\n tacts common objects: 7/13 correct (54%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n selects items by feature: 5/10 correct (49%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n imitates 3-step motor sequences: 6/11 correct (52%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n Tantrum: freq 4, duration 16m total\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n ABC (tantrum): A = asked to finish dressing task; B = kicked at materials; C = task paused for 3 min\n\nSession 7 \u2014 2026-06-16 \u2014 47 min \u2014 1 observer\n copies simple words from model: 3/9 correct (28%); latency 4.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n tacts common objects: 7/14 correct (49%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n selects items by feature: 5/12 correct (38%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n imitates 3-step motor sequences: 6/12 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d76\n Tantrum: freq 4, duration 12m total\n Toileting accident (urine or bowel): urine: 1 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 8 \u2014 2026-06-18 \u2014 47 min \u2014 1 observer\n copies simple words from model: 3/11 correct (26%); latency 4.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n tacts common objects: 6/12 correct (46%); latency 4.0s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n selects items by feature: 4/10 correct (40%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n imitates 3-step motor sequences: 3/8 correct (39%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n Tantrum: freq 4, duration 20m total\n Toileting accident (urine or bowel): urine: 2 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Crying during demand presentation\n- Increased response latency (3\u20135\u00d7 baseline)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 73% to 26% across 8 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\nToileting accident (urine or bowel): automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 49.9%.\n- Trend direction: downward (first session: 73%, last session: 26%).\n- Tantrum: mean 2.8/session (early-window mean 1.5 -> late-window mean 4.0).\n- Toileting accident (urine or bowel): mean 3.0/session (early-window mean 2.5 -> late-window mean 3.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "177d6d6940e9ae56", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Tantrum": "escape", "Toileting accident (urine or bowel)": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 8, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["tantrum", "toileting_accident"], "behavior_functions": ["escape", "automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "93700"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously uses humor appropriately\nCurriculum: Spontaneous Vocal Behavior L3\nLearner: School-Age Learner\nMastery Status: Developing\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously uses humor appropriately\n- Curriculum: Spontaneous Vocal Behavior L3\n- Learner: School-Age Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: activity setup.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's activity setup routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "4f72329f66d84ca7", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L3", "learner_profile": "school_age", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously uses humor appropriately", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "33330"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: showering\nCurriculum: AFLS Basic Living Skills\nLearner: Adult Learner\nMastery Status: Approaching mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: showering\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Enter the shower and close the curtain.\n2. Turn on the water and adjust temperature.\n3. Wet body completely.\n4. Apply soap to a washcloth.\n5. Wash body systematically (face -> torso -> arms -> legs -> feet).\n6. Rinse all soap off.\n7. Apply shampoo and rub into hair.\n8. Rinse hair completely.\n9. Turn off the water.\n10. Exit the shower and dry off with a towel. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nDifferential per-step + terminal \u2014 Small per-step reinforcement for independent responses; larger reinforcer at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 180s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0738a04d982f38bb", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "showering", "program_type": "independence", "learner_profile": "adult", "mastery_state": "approaching", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "showering", "module": "basic_living", "mastery_criterion": "mc_fluent_chain", "chain_type": "backward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "differential_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "42817"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands to use the bathroom (e.g., \"bathroom\", \"potty\", \"I need to go\", AAC bathroom icon)\nCurriculum: Mand L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands to use the bathroom (e.g., \"bathroom\", \"potty\", \"I need to go\", AAC bathroom icon)\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a scheduled toileting window (typically 30\u201360 minutes after fluid intake) and observe for pre-toileting signs (body position, vocalization, squirming). Model the mand topography (vocal, sign, or AAC icon) at the first sign of need; reinforce any approximation with immediate escort to the bathroom. Do not prompt after an accident has started.\nPrimary natural context: end-of-activity transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's end-of-activity transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: immediate escort to the bathroom; the learner is not required to complete the current task before going. Pairing the mand with timely bathroom access is what teaches the learner that requesting works.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "93db21fda89a9ac4", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands to use the bathroom (e.g., \"bathroom\", \"potty\", \"I need to go\", AAC bathroom icon)", "mo_category": "mand_bathroom", "mo_arrangement": "bathroom_opportunity", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "44373"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a toleration program for this self-care / medical / caregiver-delivered skill.\n\nTarget: tolerating tooth brushing by caregiver\nCurriculum: AFLS Basic Living Skills\nLearner: School-Age Learner\nMastery Status: Developing\n\nPlease include the toleration end goal, shaping progression, antecedent strategies, reinforcement plan, refusal / safety procedures, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating tooth brushing by caregiver\n- Curriculum: AFLS Basic Living Skills\n- Learner: School-Age Learner\n- Current mastery: Developing\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for tooth brushing performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull 2-minute tooth-brushing routine completed by caregiver without refusal, crying, SIB, or aggression\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1-second toothbrush contact with teeth\n2. 3 seconds of brushing\n3. 5 seconds of brushing\n4. 15 seconds of brushing\n5. 30 seconds of brushing\n6. 60 seconds of brushing\n7. Full 2-minute brushing routine\n\nCurrent shaping step: At the developing stage, begin at shaping step 2 (\"3 seconds of brushing\") and advance through step 3 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated with 2 different caregivers (primary caregiver + one other) across 2 consecutive weeks.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f74496eb00c5aa6a", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating tooth brushing by caregiver", "program_type": "toleration", "learner_profile": "school_age", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating tooth brushing by caregiver", "module": "basic_living", "mastery_criterion": "mc_toleration_caregivers", "shaping_n_steps": 7}, "teacher_model": null, "seed_tag": "92010"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5476\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Tact\nDate range: Sessions 1\u20138 across 16 days (2026-10-20 to 2026-11-05)\n\nACCELERATION PROGRAMS\n1. tacts common objects (Tact)\n Method: DTT\n2. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n3. reads 10\u201320 sight words (Reading)\n Method: DTT\n4. tacts colors of objects (Tact)\n Method: DTT\n5. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n6. sounds out CVC words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n2. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-20 \u2014 58 min \u2014 1 observer\n tacts common objects: 10/14 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n follows 1-step motor instructions: 7/9 correct (76%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n reads 10\u201320 sight words: 9/13 correct (71%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n tacts colors of objects: 9/11 correct (82%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 5/8 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n sounds out CVC words: 10/12 correct (79%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n Verbal aggression: freq 1\n Motor stereotypy: freq 3; PIR 25%\n\nSession 2 \u2014 2026-10-22 \u2014 58 min \u2014 1 observer\n tacts common objects: 9/13 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d710\n follows 1-step motor instructions: 8/11 correct (76%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 7/11 correct (65%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n tacts colors of objects: 10/13 correct (78%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 9/14 correct (62%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n sounds out CVC words: 8/10 correct (77%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n Verbal aggression: freq 2\n Motor stereotypy: freq 2; PIR 13%\n\nSession 3 \u2014 2026-10-24 \u2014 58 min \u2014 1 observer\n tacts common objects: 7/11 correct (60%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n follows 1-step motor instructions: 10/15 correct (69%); latency 3.1s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d77\n reads 10\u201320 sight words: 7/10 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n tacts colors of objects: 6/9 correct (68%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 9/15 correct (57%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d79\n sounds out CVC words: 9/14 correct (67%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n Verbal aggression: freq 2\n Motor stereotypy: freq 4; PIR 17%\n\nSession 4 \u2014 2026-10-26 \u2014 58 min \u2014 1 observer\n tacts common objects: 4/8 correct (46%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n follows 1-step motor instructions: 8/13 correct (59%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d710\n reads 10\u201320 sight words: 8/14 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n tacts colors of objects: 8/13 correct (63%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 7/13 correct (50%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n sounds out CVC words: 9/13 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n Verbal aggression: freq 3\n Motor stereotypy: freq 3; PIR 13%\n\nSession 5 \u2014 2026-10-28 \u2014 58 min \u2014 1 observer\n tacts common objects: 7/15 correct (46%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n follows 1-step motor instructions: 6/12 correct (51%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d711\n reads 10\u201320 sight words: 5/11 correct (45%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n tacts colors of objects: 8/14 correct (59%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n requesting a break ('break please' or AAC 'I need a break'): 6/12 correct (51%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n sounds out CVC words: 7/12 correct (59%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n Verbal aggression: freq 2\n Motor stereotypy: freq 4; PIR 16%\n\nSession 6 \u2014 2026-10-30 \u2014 58 min \u2014 1 observer\n tacts common objects: 3/9 correct (34%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n follows 1-step motor instructions: 5/13 correct (42%); latency 4.1s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 4/9 correct (45%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n tacts colors of objects: 5/10 correct (48%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 6/13 correct (47%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n sounds out CVC words: 5/9 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n Verbal aggression: freq 3\n Motor stereotypy: freq 4; PIR 18%\n\nSession 7 \u2014 2026-11-01 \u2014 58 min \u2014 1 observer\n tacts common objects: 4/15 correct (29%); latency 4.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n follows 1-step motor instructions: 4/10 correct (39%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n reads 10\u201320 sight words: 4/14 correct (31%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n tacts colors of objects: 5/12 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 6/14 correct (42%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n sounds out CVC words: 7/15 correct (46%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n Verbal aggression: freq 4\n Motor stereotypy: freq 5; PIR 28%\n\nSession 8 \u2014 2026-11-03 \u2014 58 min \u2014 1 observer\n tacts common objects: 3/14 correct (19%); latency 5.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n follows 1-step motor instructions: 3/9 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 4/15 correct (29%); latency 4.5s; prompts FP\u00d75, PP\u00d75, G\u00d75, I\u00d70\n tacts colors of objects: 2/8 correct (30%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 4/15 correct (29%); latency 4.7s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d78\n sounds out CVC words: 4/10 correct (38%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n Verbal aggression: freq 4\n Motor stereotypy: freq 5; PIR 34%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Crying during demand presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 71% to 19% across 8 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 46.6%.\n- Trend direction: downward (first session: 71%, last session: 19%).\n- Verbal aggression: mean 2.6/session (early-window mean 1.5 -> late-window mean 4.0).\n- Motor stereotypy: mean 3.8/session (early-window mean 2.5 -> late-window mean 5.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a2270cd3bba824bd", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Verbal aggression": "escape", "Motor stereotypy": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 8, "n_programs": 6, "n_behaviors": 2, "behavior_ids": ["verbal_aggression", "motor_stereotypy"], "behavior_functions": ["escape", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "80693"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: writes own name from model\nCurriculum: Writing L1\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: writes own name from model\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: plate, cup, key\nDistractor stimuli: brush, sock\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0b4e2e44fb081590", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "writes own name from model", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "97173"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: copies simple words from model\nCurriculum: Writing L2\nLearner Level: School-Age Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies simple words from model\n- Curriculum: Writing L2\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: ball, napkin, book\nDistractor stimuli: scissors, bottle, crayon, shoe\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fc9f902f28333340", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L2", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies simple words from model", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "errorless", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "93978"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: washing hands\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adult Learner\nBaseline: Developing\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: washing hands\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Walk to the sink.\n2. Turn on the water.\n3. Wet hands under the water.\n4. Apply soap to hands.\n5. Rub hands together for 15\u201320 seconds, covering fronts, backs, and between fingers.\n6. Rinse hands thoroughly under the water.\n7. Turn off the water.\n8. Dry hands with a towel.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "b20b699b0e4041db", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "washing hands", "program_type": "independence", "learner_profile": "adult", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "washing hands", "module": "basic_living", "mastery_criterion": "mc_generalization", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "39765"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies coins by name\nCurriculum Reference: Math L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies coins by name\n- Curriculum: Math L2\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: book, bottle, blanket\nDistractor stimuli: key, toothbrush, block, crayon\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "dee5a7c95f01526a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L2", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies coins by name", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "46080"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: sequences 3\u20134 step picture sequences\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L3\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: sequences 3\u20134 step picture sequences\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L3\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: pencil, car, napkin\nDistractor stimuli: cup, glue, scissors, bottle, apple\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6037fda7fc41aae9", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L3", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "sequences 3\u20134 step picture sequences", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "vr3", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "87673"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: points to named common objects\nCurriculum: Listener Responding L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: points to named common objects\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: block, sock, car\nDistractor stimuli: toothbrush, blanket\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "66289c580f527dd4", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "points to named common objects", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "78310"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4550\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Reading\nDate range: Sessions 1\u20138 across 16 days (2026-03-16 to 2026-04-01)\n\nACCELERATION PROGRAMS\n1. requesting help when a task is difficult (Mand)\n Method: NET\n2. making the bed (Home Skills)\n Method: Task Analysis\n3. identifying community safety signs (Community)\n Method: DTT\n4. sounds out CVC words (Reading)\n Method: DTT\n5. requesting preferred items using AAC device (Mand)\n Method: NET\n6. indicating 'all done' with a step or activity (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-16 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 2/8 correct (23%); latency 4.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n making the bed: 3/14 correct (20%); latency 5.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d710\n identifying community safety signs: 3/8 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n sounds out CVC words: 2/9 correct (27%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n requesting preferred items using AAC device: 5/15 correct (34%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d79\n indicating 'all done' with a step or activity: 3/10 correct (35%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n\nSession 2 \u2014 2026-03-18 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 4/11 correct (34%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n making the bed: 3/11 correct (30%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n identifying community safety signs: 8/15 correct (50%); latency 3.9s; prompts FP\u00d73, PP\u00d75, G\u00d71, I\u00d76\n sounds out CVC words: 6/15 correct (37%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d75, I\u00d76\n requesting preferred items using AAC device: 4/8 correct (46%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n indicating 'all done' with a step or activity: 4/9 correct (43%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n\nSession 3 \u2014 2026-03-20 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 5/12 correct (41%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n making the bed: 4/10 correct (37%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n identifying community safety signs: 7/11 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n sounds out CVC words: 5/11 correct (45%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n requesting preferred items using AAC device: 4/8 correct (53%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n indicating 'all done' with a step or activity: 5/10 correct (52%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n\nSession 4 \u2014 2026-03-22 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 7/13 correct (51%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n making the bed: 7/15 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n identifying community safety signs: 10/15 correct (70%); latency 2.6s; prompts FP\u00d72, PP\u00d75, G\u00d71, I\u00d77\n sounds out CVC words: 7/13 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n requesting preferred items using AAC device: 6/9 correct (64%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 6/10 correct (64%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n\nSession 5 \u2014 2026-03-24 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 6/9 correct (63%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n making the bed: 7/13 correct (57%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n identifying community safety signs: 11/13 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n sounds out CVC words: 10/15 correct (67%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d714\n requesting preferred items using AAC device: 7/10 correct (73%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n indicating 'all done' with a step or activity: 7/9 correct (75%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n\nSession 6 \u2014 2026-03-26 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 7/10 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n making the bed: 8/14 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n identifying community safety signs: 13/14 correct (91%); latency 2.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n sounds out CVC words: 8/10 correct (75%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n requesting preferred items using AAC device: 7/9 correct (83%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n indicating 'all done' with a step or activity: 9/11 correct (82%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n\nSession 7 \u2014 2026-03-28 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 7/9 correct (83%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n making the bed: 10/14 correct (70%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n identifying community safety signs: 13/13 correct (97%); latency 1.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n sounds out CVC words: 10/12 correct (85%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n requesting preferred items using AAC device: 8/9 correct (89%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n indicating 'all done' with a step or activity: 11/12 correct (93%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n\nSession 8 \u2014 2026-03-30 \u2014 53 min \u2014 1 observer\n requesting help when a task is difficult: 11/12 correct (94%); latency 1.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n making the bed: 10/13 correct (79%); latency 2.3s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d74\n identifying community safety signs: 13/13 correct (97%); latency 1.6s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n sounds out CVC words: 14/15 correct (93%); latency 1.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d713\n requesting preferred items using AAC device: 9/9 correct (97%); latency 1.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n indicating 'all done' with a step or activity: 9/9 correct (97%); latency 1.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Quick response latency (within 2\u20133 seconds)\n- Requesting more trials verbally"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 23% to 94% across 8 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 57.4%.\n- Trend direction: upward (first session: 23%, last session: 94%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "c0aea6ab20025cac", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "31476"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously comments on events\nCurriculum: Spontaneous Vocal Behavior L1\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously comments on events\n- Curriculum: Spontaneous Vocal Behavior L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "9d734f49246a1442", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously comments on events", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "time_delay", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "70185"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts prepositions (in, on, under)\nCurriculum: Tact L2\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts prepositions (in, on, under)\n- Curriculum: Tact L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Where is it?\"\nVariations: \"Where is the {object}?\"; \"Tell me where\"\nPresentation: Arrange two objects in a target spatial relation and ask about the position.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: shoe, sock, plate\nDistractor stimuli: hat, car, backpack, napkin\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "2ad4df8308934d66", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts prepositions (in, on, under)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "69064"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4011\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Mand + Tact + Writing\nDate range: Sessions 1\u201310 across 20 days (2026-04-06 to 2026-04-26)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n3. echoes 2-word phrases (Echoic)\n Method: DTT variant\n4. copies simple words from model (Writing)\n Method: DTT\n5. tacts common objects (Tact)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-06 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (41%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 4/15 correct (26%); latency 4.7s; prompts FP\u00d75, PP\u00d73, G\u00d74, I\u00d73\n echoes 2-word phrases: 5/14 correct (33%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n copies simple words from model: 7/9 correct (73%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n tacts common objects: 6/10 correct (58%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n\nSession 2 \u2014 2026-04-08 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/14 correct (43%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 5/10 correct (46%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n echoes 2-word phrases: 5/13 correct (38%); latency 4.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n copies simple words from model: 7/10 correct (74%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n tacts common objects: 7/14 correct (50%); latency 4.0s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n\nSession 3 \u2014 2026-04-10 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/14 correct (48%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d711\n requesting a break ('break please' or AAC 'I need a break'): 6/12 correct (53%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n echoes 2-word phrases: 4/14 correct (30%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n copies simple words from model: 9/11 correct (79%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n tacts common objects: 3/9 correct (36%); latency 4.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n\nSession 4 \u2014 2026-04-12 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (64%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d710\n requesting a break ('break please' or AAC 'I need a break'): 6/11 correct (51%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n echoes 2-word phrases: 3/10 correct (27%); latency 4.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n copies simple words from model: 8/10 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n tacts common objects: 5/10 correct (50%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n\nSession 5 \u2014 2026-04-14 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (34%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 7/14 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n echoes 2-word phrases: 6/10 correct (63%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n copies simple words from model: 8/10 correct (80%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n tacts common objects: 5/14 correct (32%); latency 4.4s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n\nSession 6 \u2014 2026-04-16 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/12 correct (63%); latency 3.1s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 3/11 correct (29%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n echoes 2-word phrases: 3/11 correct (30%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n copies simple words from model: 6/8 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n tacts common objects: 8/14 correct (55%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n\nSession 7 \u2014 2026-04-18 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (61%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 3/12 correct (28%); latency 5.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d78\n echoes 2-word phrases: 7/15 correct (48%); latency 3.6s; prompts FP\u00d75, PP\u00d72, G\u00d75, I\u00d73\n copies simple words from model: 6/8 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n tacts common objects: 8/15 correct (54%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d79\n\nSession 8 \u2014 2026-04-20 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (57%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 3/9 correct (30%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n echoes 2-word phrases: 3/11 correct (30%); latency 4.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n copies simple words from model: 8/15 correct (54%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n tacts common objects: 6/15 correct (37%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d79\n\nSession 9 \u2014 2026-04-22 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/9 correct (40%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n requesting a break ('break please' or AAC 'I need a break'): 4/9 correct (49%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n echoes 2-word phrases: 4/8 correct (45%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n copies simple words from model: 6/10 correct (55%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n tacts common objects: 2/9 correct (22%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n\nSession 10 \u2014 2026-04-24 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (67%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 4/8 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n echoes 2-word phrases: 4/10 correct (42%); latency 4.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n copies simple words from model: 8/12 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n tacts common objects: 6/13 correct (47%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 54% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 51.7%.\n- Trend direction: upward (first session: 41%, last session: 67%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "2190c7a8f2f78d43", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "early", "n_sessions": 10, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "9737"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1654\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-04-19 to 2026-05-07)\n\nACCELERATION PROGRAMS\n1. simple meal preparation (Home Skills)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. indicating 'all done' at task completion (Mand)\n Method: NET\n4. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n5. self-advocacy statements (Independent Living)\n Method: BST\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-19 \u2014 55 min \u2014 1 observer\n simple meal preparation: 6/8 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n wait calmly when a preferred activity is denied: 12/14 correct (85%); latency 2.0s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d76\n indicating 'all done' at task completion: 10/13 correct (76%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n dressing (pants, shirt): 7/9 correct (80%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n self-advocacy statements: 9/11 correct (78%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n\nSession 2 \u2014 2026-04-21 \u2014 55 min \u2014 1 observer\n simple meal preparation: 11/15 correct (74%); latency 2.9s; prompts FP\u00d75, PP\u00d74, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 10/12 correct (83%); latency 2.3s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n indicating 'all done' at task completion: 11/15 correct (74%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n dressing (pants, shirt): 11/15 correct (76%); latency 2.5s; prompts FP\u00d74, PP\u00d70, G\u00d75, I\u00d76\n self-advocacy statements: 12/14 correct (83%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n\nSession 3 \u2014 2026-04-23 \u2014 55 min \u2014 1 observer\n simple meal preparation: 10/13 correct (77%); latency 2.4s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 9/11 correct (82%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n indicating 'all done' at task completion: 7/10 correct (72%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n dressing (pants, shirt): 6/8 correct (76%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n self-advocacy statements: 11/13 correct (85%); latency 1.9s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n\nSession 4 \u2014 2026-04-25 \u2014 55 min \u2014 1 observer\n simple meal preparation: 3/8 correct (37%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 4/9 correct (42%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n indicating 'all done' at task completion: 4/8 correct (44%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n dressing (pants, shirt): 4/9 correct (47%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n self-advocacy statements: 6/12 correct (47%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n\nSession 5 \u2014 2026-04-27 \u2014 55 min \u2014 1 observer\n simple meal preparation: 6/15 correct (41%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 5/10 correct (47%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n indicating 'all done' at task completion: 6/13 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n dressing (pants, shirt): 7/14 correct (50%); latency 4.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n self-advocacy statements: 4/9 correct (47%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n\nSession 6 \u2014 2026-04-29 \u2014 55 min \u2014 1 observer\n simple meal preparation: 5/13 correct (38%); latency 4.5s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 5/11 correct (43%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n indicating 'all done' at task completion: 6/11 correct (53%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n dressing (pants, shirt): 4/8 correct (46%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n self-advocacy statements: 5/9 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n\nSession 7 \u2014 2026-05-01 \u2014 55 min \u2014 1 observer\n simple meal preparation: 5/13 correct (37%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n wait calmly when a preferred activity is denied: 5/12 correct (45%); latency 3.9s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n indicating 'all done' at task completion: 5/9 correct (61%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n dressing (pants, shirt): 7/14 correct (47%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n self-advocacy statements: 5/9 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n\nSession 8 \u2014 2026-05-03 \u2014 55 min \u2014 1 observer\n simple meal preparation: 6/12 correct (47%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 5/10 correct (54%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n indicating 'all done' at task completion: 5/8 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n dressing (pants, shirt): 4/8 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n self-advocacy statements: 5/8 correct (61%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n\nSession 9 \u2014 2026-05-05 \u2014 55 min \u2014 1 observer\n simple meal preparation: 6/11 correct (58%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 8/13 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n indicating 'all done' at task completion: 10/15 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n dressing (pants, shirt): 6/12 correct (52%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d74\n self-advocacy statements: 8/12 correct (70%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 53.6%.\n- Trend direction: downward (first session: 75%, last session: 58%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1a87f0703e9bbe08", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adult", "n_sessions": 9, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "19304"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands for a break (e.g., \"break please\", AAC \"I need a break\")\nCurriculum: Mand L1\nLearner: Early Learner\nMastery Status: Approaching mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for a break (e.g., \"break please\", AAC \"I need a break\")\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nPresent a brief non-preferred task or demand at a level the learner can tolerate but may wish to pause. Wait for a request-for-break topography; honor the request immediately on first acquisition (CRF), thin later.\nPrimary natural context: academic period.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's academic period routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: immediate brief escape from the demand (30\u201360 seconds) to a neutral or preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "5e7ef552782ae9cc", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for a break (e.g., \"break please\", AAC \"I need a break\")", "mo_category": "mand_break", "mo_arrangement": "break_opportunity", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "16366"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: washing hands\nCurriculum: AFLS Basic Living Skills\nLearner: Adult Learner\nMastery Status: Developing\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: washing hands\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Walk to the sink.\n2. Turn on the water.\n3. Wet hands under the water.\n4. Apply soap to hands.\n5. Rub hands together for 15\u201320 seconds, covering fronts, backs, and between fingers.\n6. Rinse hands thoroughly under the water.\n7. Turn off the water.\n8. Dry hands with a towel.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e5d6245b70edb285", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "washing hands", "program_type": "independence", "learner_profile": "adult", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "washing hands", "module": "basic_living", "mastery_criterion": "mc_generalization", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "80658"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5210\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Self-Care + Writing\nDate range: Sessions 1\u201310 across 20 days (2026-03-17 to 2026-04-06)\n\nACCELERATION PROGRAMS\n1. requesting a break during non-preferred activities (Mand)\n Method: NET\n2. brushing teeth independently (Self-Care)\n Method: Task Analysis\n3. indicating 'all done' with a step or activity (Mand)\n Method: NET\n4. ordering in a restaurant (Community)\n Method: Task Analysis\n5. copies simple words from model (Writing)\n Method: DTT\n6. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-17 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/14 correct (73%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n brushing teeth independently: 6/8 correct (81%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n indicating 'all done' with a step or activity: 9/10 correct (91%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n ordering in a restaurant: 7/8 correct (93%); latency 1.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n copies simple words from model: 10/11 correct (87%); latency 2.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n meal cleanup after eating: 10/12 correct (79%); latency 2.6s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n\nSession 2 \u2014 2026-03-19 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/8 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n brushing teeth independently: 13/15 correct (87%); latency 1.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d77\n indicating 'all done' with a step or activity: 8/9 correct (89%); latency 1.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n ordering in a restaurant: 8/8 correct (97%); latency 1.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n copies simple words from model: 8/10 correct (81%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n meal cleanup after eating: 11/13 correct (83%); latency 2.2s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n\nSession 3 \u2014 2026-03-21 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 11/15 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n brushing teeth independently: 7/9 correct (83%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n indicating 'all done' with a step or activity: 11/12 correct (94%); latency 1.6s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n ordering in a restaurant: 8/8 correct (95%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n copies simple words from model: 10/12 correct (83%); latency 2.1s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n meal cleanup after eating: 11/14 correct (77%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d77\n\nSession 4 \u2014 2026-03-23 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/13 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d77\n brushing teeth independently: 7/8 correct (85%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n indicating 'all done' with a step or activity: 7/8 correct (94%); latency 1.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n ordering in a restaurant: 13/14 correct (95%); latency 1.9s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d74\n copies simple words from model: 11/13 correct (82%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n meal cleanup after eating: 10/13 correct (77%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n\nSession 5 \u2014 2026-03-25 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/14 correct (75%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d711\n brushing teeth independently: 7/8 correct (83%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n indicating 'all done' with a step or activity: 10/11 correct (90%); latency 1.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n ordering in a restaurant: 10/11 correct (92%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n copies simple words from model: 9/11 correct (80%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n meal cleanup after eating: 11/13 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n\nSession 6 \u2014 2026-03-27 \u2014 IOA SESSION \u2014 2 observers\n requesting a break during non-preferred activities: 72% (primary); IOA 84% agreement\n brushing teeth independently: 82% (primary); IOA 84% agreement\n indicating 'all done' with a step or activity: 93% (primary); IOA 84% agreement\n ordering in a restaurant: 96% (primary); IOA 84% agreement\n copies simple words from model: 78% (primary); IOA 84% agreement\n meal cleanup after eating: 80% (primary); IOA 84% agreement\n\nSession 7 \u2014 2026-03-29 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 11/15 correct (72%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d78\n brushing teeth independently: 9/11 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n indicating 'all done' with a step or activity: 8/9 correct (93%); latency 1.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n ordering in a restaurant: 8/8 correct (98%); latency 1.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n copies simple words from model: 10/13 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d79\n meal cleanup after eating: 8/9 correct (84%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n\nSession 8 \u2014 2026-03-31 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 5/8 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n brushing teeth independently: 10/13 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n indicating 'all done' with a step or activity: 10/11 correct (93%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n ordering in a restaurant: 11/11 correct (98%); latency 1.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n copies simple words from model: 6/8 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n meal cleanup after eating: 7/9 correct (82%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n\nSession 9 \u2014 2026-04-02 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/15 correct (68%); latency 3.0s; prompts FP\u00d75, PP\u00d71, G\u00d75, I\u00d74\n brushing teeth independently: 12/15 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d78\n indicating 'all done' with a step or activity: 9/10 correct (92%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n ordering in a restaurant: 8/8 correct (98%); latency 1.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n copies simple words from model: 11/15 correct (73%); latency 2.5s; prompts FP\u00d74, PP\u00d75, G\u00d71, I\u00d75\n meal cleanup after eating: 7/8 correct (82%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n\nSession 10 \u2014 2026-04-04 \u2014 54 min \u2014 1 observer\n requesting a break during non-preferred activities: 8/12 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n brushing teeth independently: 9/11 correct (77%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n indicating 'all done' with a step or activity: 12/12 correct (96%); latency 1.8s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n ordering in a restaurant: 10/10 correct (97%); latency 1.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n copies simple words from model: 11/15 correct (70%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n meal cleanup after eating: 8/10 correct (85%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (73%\u201368%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 71.6%.\n- Trend direction: downward (first session: 73%, last session: 68%).\n- IOA data: 84% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "f6eb4c97ec65c003", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "68487"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2587\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-05-20 to 2026-06-09)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. sounds out CVC words (Reading)\n Method: DTT\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n5. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n6. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-20 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 7/13 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n sounds out CVC words: 5/12 correct (38%); latency 4.1s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n wait calmly when access to preferred item is denied: 5/10 correct (51%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/12 correct (29%); latency 4.7s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n indicating 'all done' when an activity is complete: 6/14 correct (43%); latency 3.8s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/15 correct (51%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d77\n\nSession 2 \u2014 2026-05-22 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 5/13 correct (35%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d78\n sounds out CVC words: 10/14 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 9/15 correct (62%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/15 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d75, G\u00d73, I\u00d75\n indicating 'all done' when an activity is complete: 9/13 correct (73%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/8 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-05-24 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 6/10 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n sounds out CVC words: 5/10 correct (47%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 12/15 correct (82%); latency 2.2s; prompts FP\u00d74, PP\u00d75, G\u00d71, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/15 correct (39%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d77\n indicating 'all done' when an activity is complete: 6/13 correct (49%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/9 correct (30%); latency 4.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n\nSession 4 \u2014 2026-05-26 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 5/11 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n sounds out CVC words: 3/9 correct (36%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 6/9 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/9 correct (34%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n indicating 'all done' when an activity is complete: 8/11 correct (73%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/12 correct (49%); latency 3.5s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n\nSession 5 \u2014 2026-05-28 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 5/9 correct (53%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n sounds out CVC words: 6/10 correct (56%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 8/13 correct (63%); latency 3.1s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/15 correct (54%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d712\n indicating 'all done' when an activity is complete: 4/9 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/8 correct (59%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n\nSession 6 \u2014 2026-05-30 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 7/15 correct (46%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n sounds out CVC words: 5/9 correct (59%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 8/10 correct (79%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/15 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n indicating 'all done' when an activity is complete: 4/11 correct (35%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/11 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n\nSession 7 \u2014 2026-06-01 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 8/12 correct (69%); latency 3.0s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n sounds out CVC words: 7/12 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 7/13 correct (51%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/9 correct (47%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 5/13 correct (39%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/11 correct (47%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n\nSession 8 \u2014 2026-06-03 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 4/12 correct (37%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n sounds out CVC words: 9/15 correct (57%); latency 3.5s; prompts FP\u00d74, PP\u00d75, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 12/14 correct (83%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 2/8 correct (28%); latency 4.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n indicating 'all done' when an activity is complete: 8/14 correct (55%); latency 3.3s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n\nSession 9 \u2014 2026-06-05 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 4/13 correct (32%); latency 4.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n sounds out CVC words: 9/15 correct (63%); latency 3.4s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n wait calmly when access to preferred item is denied: 6/9 correct (70%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/14 correct (58%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d79\n indicating 'all done' when an activity is complete: 9/14 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/11 correct (29%); latency 4.8s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n\nSession 10 \u2014 2026-06-07 \u2014 51 min \u2014 1 observer\n identifies numerals 1\u201320: 9/13 correct (66%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n sounds out CVC words: 6/13 correct (44%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 11/14 correct (77%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/11 correct (31%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n indicating 'all done' when an activity is complete: 7/12 correct (57%); latency 3.5s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/9 correct (52%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 58% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 50.2%.\n- Trend direction: upward (first session: 51%, last session: 66%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "6256c5310000c637", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "91712"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: engages in social games (peek-a-boo, tickle)\nCurriculum Reference: Social Behavior & Social Play L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: engages in social games (peek-a-boo, tickle)\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "7f9858a2c702a1d9", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "engages in social games (peek-a-boo, tickle)", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "54871"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4343\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand + Self-Care + Vocational\nDate range: Sessions 1\u201311 across 22 days (2026-01-16 to 2026-02-07)\n\nACCELERATION PROGRAMS\n1. simple meal preparation (Home Skills)\n Method: Task Analysis\n2. requesting a break during extended tasks (Mand)\n Method: NET\n3. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. managing a budget (Independent Living)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-16 \u2014 59 min \u2014 1 observer\n simple meal preparation: 5/11 correct (49%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n requesting a break during extended tasks: 7/14 correct (53%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n completing assigned tasks in order: 9/14 correct (66%); latency 2.7s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n brushing teeth independently: 6/11 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n managing a budget: 5/10 correct (49%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n Self-injurious behavior (SIB): freq 3\n\nSession 2 \u2014 2026-01-18 \u2014 59 min \u2014 1 observer\n simple meal preparation: 8/15 correct (54%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d78\n requesting a break during extended tasks: 8/15 correct (51%); latency 3.8s; prompts FP\u00d72, PP\u00d75, G\u00d72, I\u00d76\n completing assigned tasks in order: 7/8 correct (85%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n brushing teeth independently: 6/10 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n managing a budget: 9/14 correct (67%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d74\n Self-injurious behavior (SIB): freq 7\n\nSession 3 \u2014 2026-01-20 \u2014 59 min \u2014 1 observer\n simple meal preparation: 5/13 correct (38%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n requesting a break during extended tasks: 7/12 correct (57%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d710\n completing assigned tasks in order: 10/15 correct (70%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n brushing teeth independently: 4/10 correct (41%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n managing a budget: 6/8 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Self-injurious behavior (SIB): freq 5\n\nSession 4 \u2014 2026-01-22 \u2014 59 min \u2014 1 observer\n simple meal preparation: 8/12 correct (66%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d77\n requesting a break during extended tasks: 4/8 correct (54%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n completing assigned tasks in order: 5/8 correct (61%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n brushing teeth independently: 7/15 correct (45%); latency 4.0s; prompts FP\u00d74, PP\u00d74, G\u00d75, I\u00d72\n managing a budget: 7/15 correct (49%); latency 3.9s; prompts FP\u00d71, PP\u00d75, G\u00d74, I\u00d75\n Self-injurious behavior (SIB): freq 2\n\nSession 5 \u2014 2026-01-24 \u2014 59 min \u2014 1 observer\n simple meal preparation: 10/15 correct (69%); latency 2.8s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n requesting a break during extended tasks: 3/10 correct (30%); latency 4.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n completing assigned tasks in order: 12/15 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d711\n brushing teeth independently: 5/10 correct (51%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n managing a budget: 5/10 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n Self-injurious behavior (SIB): freq 2\n\nSession 6 \u2014 2026-01-26 \u2014 59 min \u2014 1 observer\n simple meal preparation: 6/9 correct (65%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n requesting a break during extended tasks: 5/8 correct (57%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n completing assigned tasks in order: 8/10 correct (82%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n brushing teeth independently: 3/11 correct (25%); latency 4.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n managing a budget: 6/11 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-01-28 \u2014 59 min \u2014 1 observer\n simple meal preparation: 5/15 correct (36%); latency 4.5s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d78\n requesting a break during extended tasks: 5/11 correct (49%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n completing assigned tasks in order: 10/14 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n brushing teeth independently: 6/11 correct (51%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n managing a budget: 5/10 correct (48%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n Self-injurious behavior (SIB): freq 6\n\nSession 8 \u2014 2026-01-30 \u2014 59 min \u2014 1 observer\n simple meal preparation: 6/14 correct (45%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n requesting a break during extended tasks: 5/13 correct (41%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n completing assigned tasks in order: 7/9 correct (78%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n brushing teeth independently: 5/14 correct (35%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n managing a budget: 5/8 correct (67%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 7\n\nSession 9 \u2014 2026-02-01 \u2014 59 min \u2014 1 observer\n simple meal preparation: 6/8 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n requesting a break during extended tasks: 7/10 correct (67%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n completing assigned tasks in order: 11/14 correct (80%); latency 2.1s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n brushing teeth independently: 9/15 correct (59%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d79\n managing a budget: 5/13 correct (36%); latency 4.5s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n Self-injurious behavior (SIB): freq 5\n\nSession 10 \u2014 2026-02-03 \u2014 59 min \u2014 1 observer\n simple meal preparation: 5/12 correct (40%); latency 4.3s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n requesting a break during extended tasks: 6/12 correct (53%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n completing assigned tasks in order: 9/13 correct (71%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n brushing teeth independently: 8/14 correct (54%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n managing a budget: 10/14 correct (72%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d710\n Self-injurious behavior (SIB): freq 3\n\nSession 11 \u2014 2026-02-05 \u2014 59 min \u2014 1 observer\n simple meal preparation: 5/12 correct (45%); latency 3.7s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d71\n requesting a break during extended tasks: 3/8 correct (39%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n completing assigned tasks in order: 6/8 correct (75%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n brushing teeth independently: 4/12 correct (34%); latency 4.4s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n managing a budget: 6/10 correct (63%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 47% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 52.6%.\n- Trend direction: downward (first session: 49%, last session: 45%).\n- Self-injurious behavior (SIB): mean 4.1/session (early-window mean 5.0 -> late-window mean 3.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e62a7c80808f6e5a", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Self-injurious behavior (SIB)": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adult", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "7377"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5337\nProfile: Adult Learner (18+ years)\nCurricula: Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-07-03 to 2026-07-19)\n\nACCELERATION PROGRAMS\n1. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n2. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n4. washing hands independently (Self-Care)\n Method: Task Analysis\n5. clocking in at start of shift (Vocational)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-03 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 7/15 correct (47%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d78\n dressing (pants, shirt): 4/9 correct (42%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n brushing teeth independently: 4/10 correct (40%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n washing hands independently: 5/11 correct (43%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n clocking in at start of shift: 4/9 correct (49%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 2 \u2014 2026-07-05 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 8/15 correct (56%); latency 3.5s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d79\n dressing (pants, shirt): 7/15 correct (45%); latency 4.0s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n brushing teeth independently: 5/12 correct (45%); latency 4.0s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n washing hands independently: 6/12 correct (50%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n clocking in at start of shift: 9/15 correct (58%); latency 3.3s; prompts FP\u00d75, PP\u00d71, G\u00d74, I\u00d75\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-07-07 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 8/12 correct (65%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d77\n dressing (pants, shirt): 5/9 correct (53%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n brushing teeth independently: 7/14 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n washing hands independently: 5/8 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n clocking in at start of shift: 6/9 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 4 \u2014 2026-07-09 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 6/8 correct (74%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n dressing (pants, shirt): 8/13 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d711\n brushing teeth independently: 6/11 correct (54%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n washing hands independently: 9/13 correct (68%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n clocking in at start of shift: 10/14 correct (74%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d74\n Toileting accident (urine or bowel): urine: 5 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 5 \u2014 2026-07-11 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 9/11 correct (79%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n dressing (pants, shirt): 10/15 correct (64%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d79\n brushing teeth independently: 6/10 correct (61%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n washing hands independently: 6/8 correct (72%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n clocking in at start of shift: 11/15 correct (74%); latency 2.6s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d79\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = end of 90-minute block without scheduled toileting; B = soiled (bowel) at the work table; C = staff initiated clean-up and hygiene routine; program paused\n\nSession 6 \u2014 2026-07-13 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 10/13 correct (80%); latency 2.2s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d77\n dressing (pants, shirt): 9/13 correct (68%); latency 2.8s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n brushing teeth independently: 10/15 correct (68%); latency 2.9s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n washing hands independently: 11/15 correct (75%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d711\n clocking in at start of shift: 6/8 correct (74%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 1 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = transition from preferred activity to non-preferred task; B = voided (urine) en route to the work area; C = staff redirected to bathroom, changed clothing\n\nSession 7 \u2014 2026-07-15 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 10/13 correct (80%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d78\n dressing (pants, shirt): 7/10 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n brushing teeth independently: 8/11 correct (70%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n washing hands independently: 10/14 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n clocking in at start of shift: 8/11 correct (74%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Toileting accident (urine or bowel): urine: 2 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 8 \u2014 2026-07-17 \u2014 30 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 6/8 correct (80%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n dressing (pants, shirt): 9/13 correct (73%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n brushing teeth independently: 8/11 correct (77%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n washing hands independently: 7/10 correct (75%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n clocking in at start of shift: 7/10 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 0 in-toilet / 1 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 47% at return, recovering to 80%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 70.3%.\n- Trend direction: upward (first session: 47%, last session: 80%).\n- Toileting accident (urine or bowel): mean 2.8/session (early-window mean 2.5 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "0573834e468df7a5", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Toileting accident (urine or bowel)": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adult", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "99666"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4624\nProfile: Adult Learner (18+ years)\nCurricula: Community + Home Skills + Mand + Self-Care + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-09-10 to 2026-09-26)\n\nACCELERATION PROGRAMS\n1. using public transportation (Community)\n Method: Task Analysis\n2. clocking in at start of shift (Vocational)\n Method: Task Analysis\n3. requesting a break during extended tasks (Mand)\n Method: NET\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. simple meal preparation (Home Skills)\n Method: Task Analysis\n6. washing hands independently (Self-Care)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-10 \u2014 55 min \u2014 1 observer\n using public transportation: 9/10 correct (91%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n clocking in at start of shift: 11/13 correct (85%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d79\n requesting a break during extended tasks: 8/9 correct (85%); latency 1.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n brushing teeth independently: 9/11 correct (84%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n simple meal preparation: 8/10 correct (83%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n washing hands independently: 9/12 correct (78%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n\nSession 2 \u2014 2026-09-12 \u2014 55 min \u2014 1 observer\n using public transportation: 4/10 correct (41%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n clocking in at start of shift: 4/8 correct (47%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting a break during extended tasks: 4/10 correct (37%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n brushing teeth independently: 5/13 correct (35%); latency 4.7s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n simple meal preparation: 4/15 correct (29%); latency 4.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d77\n washing hands independently: 3/8 correct (41%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n\nSession 3 \u2014 2026-09-14 \u2014 55 min \u2014 1 observer\n using public transportation: 10/11 correct (92%); latency 1.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n clocking in at start of shift: 9/10 correct (89%); latency 2.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n requesting a break during extended tasks: 10/11 correct (88%); latency 1.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n brushing teeth independently: 8/9 correct (84%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n simple meal preparation: 9/11 correct (83%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n washing hands independently: 9/13 correct (72%); latency 2.5s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n\nSession 4 \u2014 2026-09-16 \u2014 55 min \u2014 1 observer\n using public transportation: 5/15 correct (36%); latency 4.4s; prompts FP\u00d75, PP\u00d72, G\u00d75, I\u00d73\n clocking in at start of shift: 4/10 correct (44%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n requesting a break during extended tasks: 4/11 correct (34%); latency 4.3s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n brushing teeth independently: 4/11 correct (37%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n simple meal preparation: 4/13 correct (32%); latency 4.6s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n washing hands independently: 4/9 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n\nSession 5 \u2014 2026-09-18 \u2014 55 min \u2014 1 observer\n using public transportation: 8/9 correct (86%); latency 2.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n clocking in at start of shift: 12/14 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n requesting a break during extended tasks: 11/12 correct (89%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n brushing teeth independently: 11/12 correct (89%); latency 2.2s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n simple meal preparation: 7/8 correct (89%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n washing hands independently: 6/8 correct (72%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n\nSession 6 \u2014 2026-09-20 \u2014 55 min \u2014 1 observer\n using public transportation: 5/15 correct (34%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d712\n clocking in at start of shift: 6/15 correct (37%); latency 4.6s; prompts FP\u00d73, PP\u00d72, G\u00d75, I\u00d75\n requesting a break during extended tasks: 4/10 correct (39%); latency 4.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n brushing teeth independently: 4/13 correct (34%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n simple meal preparation: 3/10 correct (32%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n washing hands independently: 6/12 correct (46%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n\nSession 7 \u2014 2026-09-22 \u2014 55 min \u2014 1 observer\n using public transportation: 12/14 correct (89%); latency 1.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n clocking in at start of shift: 12/14 correct (83%); latency 2.1s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n requesting a break during extended tasks: 8/9 correct (93%); latency 1.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n brushing teeth independently: 12/14 correct (87%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n simple meal preparation: 10/12 correct (83%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d74\n washing hands independently: 7/9 correct (76%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n\nSession 8 \u2014 2026-09-24 \u2014 55 min \u2014 1 observer\n using public transportation: 5/15 correct (34%); latency 4.4s; prompts FP\u00d75, PP\u00d73, G\u00d70, I\u00d77\n clocking in at start of shift: 5/14 correct (38%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d712\n requesting a break during extended tasks: 4/13 correct (34%); latency 4.8s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n brushing teeth independently: 3/8 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n simple meal preparation: 3/10 correct (30%); latency 4.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n washing hands independently: 4/8 correct (45%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 62.9%.\n- Trend direction: downward (first session: 91%, last session: 34%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "5f9e9dd2ffcbe477", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 8, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "9961"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously tells stories or narratives\nCurriculum: Spontaneous Vocal Behavior L3\nLearner: School-Age Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously tells stories or narratives\n- Curriculum: Spontaneous Vocal Behavior L3\n- Learner: School-Age Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: activity setup.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's activity setup routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "bfbfb047539aef8d", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L3", "learner_profile": "school_age", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously tells stories or narratives", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "76914"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects by single function from array of 3\nCurriculum: LRFFC L1\nLearner Level: Early Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects by single function from array of 3\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: blanket, ball, glue\nDistractor stimuli: key, spoon\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "64efc35e51522eba", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects by single function from array of 3", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "15942"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5679\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Reading + Tact\nDate range: Sessions 1\u20136 across 12 days (2026-09-06 to 2026-09-18)\n\nACCELERATION PROGRAMS\n1. tacts common objects (Tact)\n Method: DTT\n2. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. sounds out CVC words (Reading)\n Method: DTT\n5. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-06 \u2014 40 min \u2014 1 observer\n tacts common objects: 3/8 correct (35%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 3/8 correct (38%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/15 correct (28%); latency 4.9s; prompts FP\u00d75, PP\u00d72, G\u00d75, I\u00d73\n sounds out CVC words: 5/14 correct (33%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/15 correct (26%); latency 5.0s; prompts FP\u00d74, PP\u00d73, G\u00d75, I\u00d73\n\nSession 2 \u2014 2026-09-08 \u2014 40 min \u2014 1 observer\n tacts common objects: 6/13 correct (45%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 4/8 correct (47%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (35%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n sounds out CVC words: 3/8 correct (41%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/15 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d79\n\nSession 3 \u2014 2026-09-10 \u2014 40 min \u2014 1 observer\n tacts common objects: 5/10 correct (53%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n requesting a break ('break please' or AAC 'I need a break'): 7/14 correct (52%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/13 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n sounds out CVC words: 5/10 correct (47%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/15 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d712\n\nSession 4 \u2014 2026-09-12 \u2014 40 min \u2014 1 observer\n tacts common objects: 10/15 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n requesting a break ('break please' or AAC 'I need a break'): 8/13 correct (60%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (51%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n sounds out CVC words: 5/9 correct (55%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (58%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d79\n\nSession 5 \u2014 2026-09-14 \u2014 40 min \u2014 1 observer\n tacts common objects: 6/8 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 8/11 correct (70%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (61%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 8/12 correct (67%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/12 correct (72%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n\nSession 6 \u2014 2026-09-16 \u2014 40 min \u2014 1 observer\n tacts common objects: 11/14 correct (82%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 6/8 correct (78%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (67%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n sounds out CVC words: 7/9 correct (73%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 12/15 correct (80%); latency 2.6s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Requesting more trials verbally\n- Reaching for materials proactively\n- Quick response latency (within 2\u20133 seconds)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 35% to 82% across 6 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nlow\n\nInsufficient data: only 6 sessions, high variance, and/or no IOA evidence. Recommend additional data collection before programming changes.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 58.8%.\n- Trend direction: upward (first session: 35%, last session: 82%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "4296ba4bf545019c", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "low", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "early", "n_sessions": 6, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "81519"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: traces basic shapes (circle, cross)\nCurriculum: Writing L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces basic shapes (circle, cross)\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: square, triangle, diamond\nDistractor stimuli: oval, circle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "55c107a6331e4d5d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces basic shapes (circle, cross)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "15942"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: meal cleanup\nCurriculum: AFLS Home Skills\nLearner: Adult Learner\nMastery Status: Emerging\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: meal cleanup\n- Curriculum: AFLS Home Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Stack used dishes on one side of the table. <- Learner-taught step (forward chaining starts here)\n2. Carry dishes to the sink or dishwasher.\n3. Rinse food residue into the trash or disposal.\n4. Load dishes into the dishwasher (or wash by hand).\n5. Wipe the table with a damp cloth.\n6. Put away any leftover food into containers.\n7. Place containers in the refrigerator.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the emerging stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8e1038fba436c97c", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "meal cleanup", "program_type": "independence", "learner_profile": "adult", "mastery_state": "emerging", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "meal cleanup", "module": "home", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "6035"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7884\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u20137 across 14 days (2026-10-19 to 2026-11-02)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. sounds out CVC words (Reading)\n Method: DTT\n4. mands for missing items needed to complete an activity (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-19 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/15 correct (45%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n answers WH-questions about familiar topics: 7/13 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d710\n sounds out CVC words: 4/10 correct (41%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n mands for missing items needed to complete an activity: 5/13 correct (40%); latency 4.4s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n Verbal aggression: freq 2\n ABC (verbal aggression): A = transition to non-preferred activity; B = dropped to floor, cried; C = transition delayed, preferred item offered\n\nSession 2 \u2014 2026-10-21 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/9 correct (57%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n answers WH-questions about familiar topics: 4/8 correct (54%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n sounds out CVC words: 6/11 correct (51%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n mands for missing items needed to complete an activity: 6/13 correct (47%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n Verbal aggression: freq 2\n ABC (verbal aggression): A = demand to complete math problem; B = threw materials and vocalized 'no'; C = staff removed demand, redirected\n\nSession 3 \u2014 2026-10-23 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/11 correct (69%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n answers WH-questions about familiar topics: 9/15 correct (58%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d75, I\u00d76\n sounds out CVC words: 6/9 correct (62%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n mands for missing items needed to complete an activity: 5/9 correct (56%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n Verbal aggression: freq 0\n\nSession 4 \u2014 2026-10-25 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/9 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n answers WH-questions about familiar topics: 7/10 correct (67%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n sounds out CVC words: 7/10 correct (70%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n mands for missing items needed to complete an activity: 10/15 correct (64%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d75\n Verbal aggression: freq 2\n\nSession 5 \u2014 2026-10-27 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/13 correct (79%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n answers WH-questions about familiar topics: 11/15 correct (71%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d79\n sounds out CVC words: 10/14 correct (75%); latency 2.8s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n mands for missing items needed to complete an activity: 7/9 correct (73%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n Verbal aggression: freq 1\n ABC (verbal aggression): A = demand to complete math problem; B = threw materials and vocalized 'no'; C = staff removed demand, redirected\n\nSession 6 \u2014 2026-10-29 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/14 correct (79%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n answers WH-questions about familiar topics: 10/13 correct (77%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n sounds out CVC words: 11/15 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d76\n mands for missing items needed to complete an activity: 8/11 correct (76%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n Verbal aggression: freq 2\n\nSession 7 \u2014 2026-10-31 \u2014 33 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/12 correct (79%); latency 2.2s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n answers WH-questions about familiar topics: 8/10 correct (83%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n sounds out CVC words: 8/11 correct (75%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 6/8 correct (76%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n Verbal aggression: freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Looking away during SD presentation\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 45% at return, recovering to 79%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 69.5%.\n- Trend direction: upward (first session: 45%, last session: 79%).\n- Verbal aggression: mean 1.4/session (early-window mean 2.0 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "cce0ec4a4aecddae", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Verbal aggression": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "early", "n_sessions": 7, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["verbal_aggression"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "39736"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts colors of objects\nCurriculum: Tact L2\nLearner Level: School-Age Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts colors of objects\n- Curriculum: Tact L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What color?\"\nVariations: \"What color is this?\"; \"Tell me the color\"\nPresentation: Present a single-color object and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: red, green, yellow\nDistractor stimuli: black, purple, blue, orange\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fdfb3659e4e50012", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts colors of objects", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "69141"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: names items in categories when given the category\nCurriculum: Intraverbal L2\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: names items in categories when given the category\n- Curriculum: Intraverbal L2\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "a9824a31a75a5a55", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "names items in categories when given the category", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "time_delay", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "70431"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: ordering in a restaurant\nCurriculum Reference: AFLS Community Participation Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Generalization phase\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: ordering in a restaurant\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Approach the counter or wait to be seated.\n2. Review the menu.\n3. Decide on a choice.\n4. Place the order verbally (or with an AAC device).\n5. Respond to clarifying questions.\n6. Pay using money or a card.\n7. Wait for the order and collect it.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0d5a9e3ea8e2b305", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "ordering in a restaurant", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "ordering in a restaurant", "module": "community", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "50630"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4806\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand\nDate range: Sessions 1\u201311 across 22 days (2026-08-27 to 2026-09-18)\n\nACCELERATION PROGRAMS\n1. requesting preferred items using a full sentence (Mand)\n Method: NET\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n4. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n5. requesting a break during non-preferred tasks (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-27 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 4/8 correct (52%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n single-word requests for preferred items: 8/15 correct (53%); latency 3.6s; prompts FP\u00d75, PP\u00d70, G\u00d73, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/8 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n indicating 'all done' when an activity is complete: 7/13 correct (51%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n requesting a break during non-preferred tasks: 8/14 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n\nSession 2 \u2014 2026-08-29 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 8/13 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n single-word requests for preferred items: 8/14 correct (59%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/9 correct (53%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 8/15 correct (51%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n requesting a break during non-preferred tasks: 9/15 correct (62%); latency 3.4s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d76\n\nSession 3 \u2014 2026-08-31 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 10/15 correct (66%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d79\n single-word requests for preferred items: 6/11 correct (52%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/12 correct (44%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n indicating 'all done' when an activity is complete: 5/8 correct (58%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n requesting a break during non-preferred tasks: 10/14 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d77\n\nSession 4 \u2014 2026-09-02 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 6/11 correct (56%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n single-word requests for preferred items: 7/14 correct (50%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d711\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/15 correct (54%); latency 3.5s; prompts FP\u00d75, PP\u00d75, G\u00d72, I\u00d73\n indicating 'all done' when an activity is complete: 5/8 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n requesting a break during non-preferred tasks: 10/15 correct (63%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d78\n\nSession 5 \u2014 2026-09-04 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 10/14 correct (69%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n single-word requests for preferred items: 7/14 correct (52%); latency 3.9s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/15 correct (48%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n indicating 'all done' when an activity is complete: 5/9 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n requesting a break during non-preferred tasks: 8/11 correct (69%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n\nSession 6 \u2014 2026-09-06 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 6/8 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n single-word requests for preferred items: 5/10 correct (46%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/14 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n indicating 'all done' when an activity is complete: 6/11 correct (58%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n requesting a break during non-preferred tasks: 7/11 correct (64%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n\nSession 7 \u2014 2026-09-08 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 8/11 correct (70%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n single-word requests for preferred items: 5/11 correct (45%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/9 correct (48%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n indicating 'all done' when an activity is complete: 8/14 correct (55%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n requesting a break during non-preferred tasks: 9/14 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n\nSession 8 \u2014 2026-09-10 \u2014 IOA SESSION \u2014 2 observers\n requesting preferred items using a full sentence: 59% (primary); IOA 82% agreement\n single-word requests for preferred items: 40% (primary); IOA 82% agreement\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 51% (primary); IOA 82% agreement\n indicating 'all done' when an activity is complete: 59% (primary); IOA 82% agreement\n requesting a break during non-preferred tasks: 68% (primary); IOA 82% agreement\n\nSession 9 \u2014 2026-09-12 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 8/13 correct (60%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d711\n single-word requests for preferred items: 7/15 correct (46%); latency 4.1s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/12 correct (51%); latency 3.5s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d75\n indicating 'all done' when an activity is complete: 7/14 correct (49%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d710\n requesting a break during non-preferred tasks: 11/14 correct (77%); latency 2.4s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d77\n\nSession 10 \u2014 2026-09-14 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 8/12 correct (67%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n single-word requests for preferred items: 3/8 correct (34%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/14 correct (58%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 5/9 correct (50%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n requesting a break during non-preferred tasks: 7/9 correct (75%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n\nSession 11 \u2014 2026-09-16 \u2014 44 min \u2014 1 observer\n requesting preferred items using a full sentence: 6/8 correct (77%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n single-word requests for preferred items: 6/13 correct (45%); latency 4.1s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/13 correct (51%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n indicating 'all done' when an activity is complete: 4/8 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n requesting a break during non-preferred tasks: 6/8 correct (79%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 64% for 11 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 64.3%.\n- Trend direction: upward (first session: 52%, last session: 77%).\n- IOA data: 82% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "19459347411eaca6", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "school_age", "n_sessions": 11, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "57608"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6052\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand + Reading + Self-Care + Writing\nDate range: Sessions 1\u20139 across 18 days (2026-05-11 to 2026-05-29)\n\nACCELERATION PROGRAMS\n1. identifying community safety signs (Community)\n Method: DTT\n2. requesting a break during non-preferred activities (Mand)\n Method: NET\n3. copies simple words from model (Writing)\n Method: DTT\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. reads 10\u201320 sight words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-11 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 12/12 correct (96%); latency 1.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d77\n requesting a break during non-preferred activities: 14/15 correct (94%); latency 2.0s; prompts FP\u00d72, PP\u00d75, G\u00d74, I\u00d74\n copies simple words from model: 10/11 correct (89%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n brushing teeth independently: 9/12 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n reads 10\u201320 sight words: 8/8 correct (98%); latency 1.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 1\n\nSession 2 \u2014 2026-05-13 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 10/10 correct (97%); latency 1.3s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n requesting a break during non-preferred activities: 9/9 correct (95%); latency 1.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n copies simple words from model: 7/8 correct (88%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n brushing teeth independently: 9/11 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 12/12 correct (97%); latency 1.8s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n Self-injurious behavior (SIB): freq 1\n\nSession 3 \u2014 2026-05-15 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 14/14 correct (98%); latency 1.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d77\n requesting a break during non-preferred activities: 9/10 correct (94%); latency 1.6s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n copies simple words from model: 11/13 correct (85%); latency 1.9s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n brushing teeth independently: 10/12 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n reads 10\u201320 sight words: 13/14 correct (96%); latency 1.6s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 0\n\nSession 4 \u2014 2026-05-17 \u2014 IOA SESSION \u2014 2 observers\n identifying community safety signs: 98% (primary); IOA 85% agreement\n requesting a break during non-preferred activities: 90% (primary); IOA 85% agreement\n copies simple words from model: 83% (primary); IOA 85% agreement\n brushing teeth independently: 85% (primary); IOA 85% agreement\n reads 10\u201320 sight words: 88% (primary); IOA 85% agreement\n Self-injurious behavior (SIB): freq 0; IOA 85% agreement\n\nSession 5 \u2014 2026-05-19 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 13/13 correct (98%); latency 1.3s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d76\n requesting a break during non-preferred activities: 9/10 correct (94%); latency 1.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n copies simple words from model: 8/9 correct (83%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n brushing teeth independently: 12/14 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d77\n reads 10\u201320 sight words: 13/14 correct (90%); latency 1.8s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n Self-injurious behavior (SIB): freq 0\n\nSession 6 \u2014 2026-05-21 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 9/9 correct (98%); latency 1.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n requesting a break during non-preferred activities: 7/8 correct (93%); latency 1.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n copies simple words from model: 11/14 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n brushing teeth independently: 10/12 correct (85%); latency 2.3s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 10/12 correct (84%); latency 2.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d77\n Self-injurious behavior (SIB): freq 1\n\nSession 7 \u2014 2026-05-23 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 9/9 correct (98%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n requesting a break during non-preferred activities: 10/11 correct (87%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n copies simple words from model: 10/12 correct (80%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d78\n brushing teeth independently: 10/12 correct (86%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 8/9 correct (87%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 0\n\nSession 8 \u2014 2026-05-25 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 11/11 correct (98%); latency 1.6s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n requesting a break during non-preferred activities: 9/10 correct (88%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n copies simple words from model: 11/15 correct (74%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d79\n brushing teeth independently: 10/11 correct (91%); latency 1.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n reads 10\u201320 sight words: 9/11 correct (86%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n Self-injurious behavior (SIB): freq 2\n\nSession 9 \u2014 2026-05-27 \u2014 60 min \u2014 1 observer\n identifying community safety signs: 12/12 correct (98%); latency 1.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n requesting a break during non-preferred activities: 9/10 correct (88%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n copies simple words from model: 6/8 correct (76%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n brushing teeth independently: 13/15 correct (89%); latency 2.0s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 7/8 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n Self-injurious behavior (SIB): freq 0\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n- Leaving seat without permission\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (96%\u201398%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 97.7%.\n- Trend direction: upward (first session: 96%, last session: 98%).\n- Self-injurious behavior (SIB): mean 0.6/session (early-window mean 0.7 -> late-window mean 0.7).\n- IOA data: 85% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "b7bc12fb0edf56c7", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adolescent", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "22832"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: mands to indicate completion in context (\"I'm all done with math, can I start reading?\")\nCurriculum: Mand L3\nLearner Level: Adolescent Learner\nBaseline: Mastered at current level\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands to indicate completion in context (\"I'm all done with math, can I start reading?\")\n- Curriculum: Mand L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nSet up a clearly-bounded activity with a visible endpoint (e.g., a worksheet, a puzzle, a sorting task). Wait for the learner to indicate completion verbally or via AAC; honor the indication immediately.\nPrimary natural context: sorting task.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's sorting task routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: the activity ends immediately and the learner transitions to the next preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "9bc27945514f32f3", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands to indicate completion in context (\"I'm all done with math, can I start reading?\")", "mo_category": "mand_completion", "mo_arrangement": "completion_opportunity", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "44373"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: answers hypothetical questions\nCurriculum: Intraverbal L3\nLearner: Adolescent Learner\nMastery Status: Developing\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers hypothetical questions\n- Curriculum: Intraverbal L3\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "9d02438086548861", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "adolescent", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers hypothetical questions", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "time_delay", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "77480"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8083\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Reading\nDate range: Sessions 1\u201310 across 20 days (2026-08-20 to 2026-09-09)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. selects items by feature (Listener Responding)\n Method: DTT\n3. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n4. reads 10\u201320 sight words (Reading)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-20 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 3/9 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n selects items by feature: 6/13 correct (45%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n follows 1-step motor instructions: 4/13 correct (34%); latency 4.7s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 6/15 correct (43%); latency 3.9s; prompts FP\u00d73, PP\u00d75, G\u00d74, I\u00d73\n\nSession 2 \u2014 2026-08-22 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 5/12 correct (40%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n selects items by feature: 8/15 correct (55%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d79\n follows 1-step motor instructions: 5/13 correct (38%); latency 4.5s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n reads 10\u201320 sight words: 5/11 correct (45%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n\nSession 3 \u2014 2026-08-24 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 8/15 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d710\n selects items by feature: 7/11 correct (64%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n follows 1-step motor instructions: 7/15 correct (48%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d79\n reads 10\u201320 sight words: 5/9 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n\nSession 4 \u2014 2026-08-26 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 8/14 correct (57%); latency 3.4s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d77\n selects items by feature: 6/8 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n follows 1-step motor instructions: 8/15 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d75\n reads 10\u201320 sight words: 8/14 correct (58%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n\nSession 5 \u2014 2026-08-28 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 8/14 correct (57%); latency 3.7s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n selects items by feature: 7/10 correct (74%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n follows 1-step motor instructions: 6/10 correct (57%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 5/9 correct (57%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n\nSession 6 \u2014 2026-08-30 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 9/14 correct (66%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n selects items by feature: 8/10 correct (79%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n follows 1-step motor instructions: 10/14 correct (70%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n reads 10\u201320 sight words: 7/10 correct (68%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n\nSession 7 \u2014 2026-09-01 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 5/8 correct (64%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n selects items by feature: 10/12 correct (82%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d77\n follows 1-step motor instructions: 10/15 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 9/12 correct (71%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n\nSession 8 \u2014 2026-09-03 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 6/9 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n selects items by feature: 11/12 correct (95%); latency 1.7s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d76\n follows 1-step motor instructions: 10/13 correct (79%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 6/8 correct (70%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n\nSession 9 \u2014 2026-09-05 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 8/11 correct (73%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n selects items by feature: 11/12 correct (95%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d79\n follows 1-step motor instructions: 9/11 correct (84%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 10/13 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n\nSession 10 \u2014 2026-09-07 \u2014 49 min \u2014 1 observer\n sounds out CVC words: 8/11 correct (75%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n selects items by feature: 13/14 correct (95%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n follows 1-step motor instructions: 9/9 correct (95%); latency 1.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n reads 10\u201320 sight words: 8/10 correct (84%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Requesting more trials verbally\n- Smiling during reinforcement delivery"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 36% to 75% across 10 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 58.5%.\n- Trend direction: upward (first session: 36%, last session: 75%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "172f70bc92f86142", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "early", "n_sessions": 10, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "40195"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: points to named common objects\nCurriculum Reference: Listener Responding L1\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: points to named common objects\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: scissors, plate, blanket\nDistractor stimuli: cup, crayon\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f9393e468b1c91c3", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "points to named common objects", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "60118"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: completing assigned tasks in order\nCurriculum: AFLS Vocational Skills\nLearner: Adolescent Learner\nMastery Status: Approaching mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Review the task list for the shift. <- Learner-taught step (forward chaining starts here)\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 90s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "16caf940ee524f4f", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "approaching", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "50550"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5625\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Self-Care + Vocational\nDate range: Sessions 1\u20139 across 18 days (2026-05-09 to 2026-05-27)\n\nACCELERATION PROGRAMS\n1. clocking in at start of shift (Vocational)\n Method: Task Analysis\n2. brushing teeth independently (Self-Care)\n Method: Task Analysis\n3. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n4. managing a budget (Independent Living)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Vocal stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-09 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 8/14 correct (58%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n brushing teeth independently: 8/14 correct (61%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n dressing (pants, shirt): 7/9 correct (75%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n managing a budget: 9/14 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d75\n Vocal stereotypy: freq 2; PIR 15%\n\nSession 2 \u2014 2026-05-11 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 8/15 correct (57%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d710\n brushing teeth independently: 10/15 correct (68%); latency 2.7s; prompts FP\u00d75, PP\u00d75, G\u00d72, I\u00d73\n dressing (pants, shirt): 9/12 correct (79%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d75\n managing a budget: 6/12 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n Vocal stereotypy: freq 2; PIR 13%\n\nSession 3 \u2014 2026-05-13 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 8/13 correct (64%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n brushing teeth independently: 5/9 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n dressing (pants, shirt): 8/11 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n managing a budget: 8/14 correct (60%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d74\n Vocal stereotypy: freq 2; PIR 16%\n\nSession 4 \u2014 2026-05-15 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 3/15 correct (23%); latency 5.1s; prompts FP\u00d70, PP\u00d71, G\u00d75, I\u00d79\n brushing teeth independently: 3/8 correct (40%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n dressing (pants, shirt): 5/8 correct (59%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n managing a budget: 2/8 correct (27%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Vocal stereotypy: freq 4; PIR 20%\n\nSession 5 \u2014 2026-05-17 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 4/15 correct (29%); latency 5.0s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d710\n brushing teeth independently: 6/11 correct (51%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n dressing (pants, shirt): 8/12 correct (66%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n managing a budget: 5/15 correct (36%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n Vocal stereotypy: freq 5; PIR 27%\n\nSession 6 \u2014 2026-05-19 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 5/12 correct (41%); latency 4.3s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n brushing teeth independently: 8/13 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d74\n dressing (pants, shirt): 7/10 correct (74%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n managing a budget: 4/9 correct (45%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n Vocal stereotypy: freq 1; PIR 14%\n\nSession 7 \u2014 2026-05-21 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 5/10 correct (47%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n brushing teeth independently: 8/13 correct (64%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n dressing (pants, shirt): 10/14 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n managing a budget: 4/8 correct (48%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Vocal stereotypy: freq 1; PIR 7%\n\nSession 8 \u2014 2026-05-23 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 5/9 correct (55%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n brushing teeth independently: 7/11 correct (64%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n dressing (pants, shirt): 10/14 correct (74%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n managing a budget: 7/13 correct (54%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n Vocal stereotypy: freq 1; PIR 7%\n\nSession 9 \u2014 2026-05-25 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 8/13 correct (59%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n brushing teeth independently: 9/14 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n dressing (pants, shirt): 10/13 correct (74%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d710\n managing a budget: 7/13 correct (57%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n Vocal stereotypy: freq 2; PIR 14%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nVocal stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 47.9%.\n- Trend direction: upward (first session: 58%, last session: 59%).\n- Vocal stereotypy: mean 2.2/session (early-window mean 2.0 -> late-window mean 1.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "bbb60b2a6fe48cf6", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Vocal stereotypy": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adult", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["vocal_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "11442"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: completing assigned tasks in order\nCurriculum: AFLS Vocational Skills\nLearner: Adult Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Review the task list for the shift.\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "1b45ae57fb99c795", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_self_initiation", "chain_type": "backward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "6035"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tolerating nail clipping\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: Toleration / Systematic Desensitization\n\nThe goal is for the learner to allow the activity to be performed (not to perform it independently). Provide the full program structure including shaping steps and refusal/safety procedures."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating nail clipping\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for nail clipping performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nall 10 fingernails (and optionally toenails) clipped by caregiver without refusal, pulling hand away, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1 fingernail clipped\n2. 2 fingernails clipped in one sitting\n3. 5 fingernails clipped (one hand)\n4. All 10 fingernails clipped\n5. All 10 fingernails + 10 toenails clipped\n\nCurrent shaping step: At the mastered stage, begin at shaping step 5 (\"All 10 fingernails + 10 toenails clipped\") and advance through step 5 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "051b5e3d5b8c3bd9", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating nail clipping", "program_type": "toleration", "learner_profile": "adult", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating nail clipping", "module": "basic_living", "mastery_criterion": "mc_toleration_settings", "shaping_n_steps": 5}, "teacher_model": null, "seed_tag": "21081"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7206\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Self-Care + Writing\nDate range: Sessions 1\u20138 across 16 days (2026-05-02 to 2026-05-18)\n\nACCELERATION PROGRAMS\n1. identifying community safety signs (Community)\n Method: DTT\n2. copies simple words from model (Writing)\n Method: DTT\n3. tolerating haircut (Self-Care)\n Method: Task Analysis\n4. washing hands independently (Self-Care)\n Method: Task Analysis\n5. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: tangible\n2. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-02 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 7/11 correct (67%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n copies simple words from model: 6/9 correct (66%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n tolerating haircut: 9/15 correct (61%); latency 3.0s; prompts FP\u00d75, PP\u00d71, G\u00d74, I\u00d75\n washing hands independently: 5/9 correct (54%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n meal cleanup after eating: 6/9 correct (61%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n Elopement: freq 1\n Property destruction: freq 2\n\nSession 2 \u2014 2026-05-04 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 10/15 correct (68%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d76\n copies simple words from model: 6/9 correct (69%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n tolerating haircut: 6/10 correct (58%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n washing hands independently: 7/14 correct (52%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n meal cleanup after eating: 8/12 correct (67%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n Elopement: freq 0\n Property destruction: freq 1\n\nSession 3 \u2014 2026-05-06 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 8/12 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n copies simple words from model: 6/8 correct (70%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n tolerating haircut: 8/14 correct (60%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n washing hands independently: 5/8 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n meal cleanup after eating: 8/12 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n Elopement: freq 5\n Property destruction: freq 4\n\nSession 4 \u2014 2026-05-08 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 9/13 correct (69%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n copies simple words from model: 6/9 correct (69%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n tolerating haircut: 6/9 correct (67%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n washing hands independently: 5/10 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n meal cleanup after eating: 9/14 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d76\n Elopement: freq 4\n Property destruction: freq 4\n\nSession 5 \u2014 2026-05-10 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 8/12 correct (65%); latency 3.3s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n copies simple words from model: 7/10 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n tolerating haircut: 7/11 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n washing hands independently: 7/12 correct (57%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n meal cleanup after eating: 9/14 correct (67%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n Elopement: freq 1\n Property destruction: freq 2\n\nSession 6 \u2014 2026-05-12 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 8/12 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n copies simple words from model: 9/14 correct (67%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n tolerating haircut: 7/11 correct (66%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n washing hands independently: 8/12 correct (63%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n meal cleanup after eating: 10/14 correct (71%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d714\n Elopement: freq 1\n Property destruction: freq 0\n\nSession 7 \u2014 2026-05-14 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 6/8 correct (73%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n copies simple words from model: 10/15 correct (64%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n tolerating haircut: 9/13 correct (72%); latency 2.9s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n washing hands independently: 5/8 correct (63%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n meal cleanup after eating: 7/11 correct (67%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n Elopement: freq 1\n Property destruction: freq 1\n\nSession 8 \u2014 2026-05-16 \u2014 39 min \u2014 1 observer\n identifying community safety signs: 9/13 correct (72%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n copies simple words from model: 8/13 correct (64%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n tolerating haircut: 11/15 correct (71%); latency 3.0s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n washing hands independently: 6/9 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n meal cleanup after eating: 9/13 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d79\n Elopement: freq 1\n Property destruction: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Turning away from task or staff\n- Self-stimulatory behavior increase during demand periods\n- Peer aggression (hitting or pushing peers during group activities)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nElopement: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\nProperty destruction: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 68.3%.\n- Trend direction: upward (first session: 67%, last session: 72%).\n- Elopement: mean 1.8/session (early-window mean 0.5 -> late-window mean 1.0).\n- Property destruction: mean 2.0/session (early-window mean 1.5 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "4bd9407b04f41885", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Elopement": "tangible", "Property destruction": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["elopement", "property_destruction"], "behavior_functions": ["tangible", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "73910"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: operating the dishwasher\nCurriculum: AFLS Home Skills\nLearner Level: Adolescent Learner\nBaseline: Mastered at current level\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: operating the dishwasher\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Collect dirty dishes from the table and sink area. <- Learner-taught step (forward chaining starts here)\n2. Scrape any solid food residue into the trash.\n3. Rinse dishes briefly under water.\n4. Open the dishwasher door.\n5. Load plates and bowls into the bottom rack.\n6. Load cups, glasses, and small items into the top rack.\n7. Place utensils handle-down in the utensil basket.\n8. Add detergent to the detergent dispenser.\n9. Close the dishwasher door.\n10. Select the appropriate wash cycle and press Start.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "9a8ae982b3acf3aa", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "operating the dishwasher", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "operating the dishwasher", "module": "home", "mastery_criterion": "mc_all_steps_independent", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "39765"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5575\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-01-13 to 2026-01-29)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. doing laundry (Home Skills)\n Method: Task Analysis\n3. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n4. requesting assistance using AAC device (Mand)\n Method: NET\n5. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n6. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-13 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 8/13 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n doing laundry: 7/14 correct (49%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n dressing (pants, shirt): 4/8 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n requesting assistance using AAC device: 6/8 correct (70%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n wait calmly when a preferred activity is denied: 7/14 correct (52%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n simple meal preparation: 5/9 correct (56%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n\nSession 2 \u2014 2026-01-15 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 8/12 correct (66%); latency 3.0s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n doing laundry: 5/11 correct (48%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n dressing (pants, shirt): 7/13 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n requesting assistance using AAC device: 8/12 correct (63%); latency 3.1s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 7/11 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n simple meal preparation: 5/9 correct (57%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n\nSession 3 \u2014 2026-01-17 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 5/9 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n doing laundry: 6/13 correct (50%); latency 3.6s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n dressing (pants, shirt): 6/12 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d76\n requesting assistance using AAC device: 7/10 correct (67%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 7/11 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n simple meal preparation: 8/11 correct (69%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n\nSession 4 \u2014 2026-01-19 \u2014 IOA SESSION \u2014 2 observers\n completing assigned tasks in order: 66% (primary); IOA 94% agreement\n doing laundry: 44% (primary); IOA 94% agreement\n dressing (pants, shirt): 53% (primary); IOA 94% agreement\n requesting assistance using AAC device: 73% (primary); IOA 94% agreement\n wait calmly when a preferred activity is denied: 59% (primary); IOA 94% agreement\n simple meal preparation: 68% (primary); IOA 94% agreement\n\nSession 5 \u2014 2026-01-21 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 7/11 correct (62%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n doing laundry: 7/13 correct (52%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n dressing (pants, shirt): 4/8 correct (53%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n requesting assistance using AAC device: 6/10 correct (59%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 7/11 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n simple meal preparation: 7/11 correct (68%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n\nSession 6 \u2014 2026-01-23 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 6/9 correct (64%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n doing laundry: 6/15 correct (39%); latency 4.3s; prompts FP\u00d75, PP\u00d74, G\u00d73, I\u00d73\n dressing (pants, shirt): 6/10 correct (55%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n requesting assistance using AAC device: 6/8 correct (73%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 5/8 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n simple meal preparation: 8/15 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n\nSession 7 \u2014 2026-01-25 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 8/13 correct (62%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n doing laundry: 6/12 correct (48%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n dressing (pants, shirt): 5/10 correct (48%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n requesting assistance using AAC device: 5/8 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n wait calmly when a preferred activity is denied: 10/14 correct (72%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n simple meal preparation: 6/12 correct (51%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n\nSession 8 \u2014 2026-01-27 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 9/14 correct (67%); latency 3.0s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n doing laundry: 3/10 correct (32%); latency 4.5s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d74\n dressing (pants, shirt): 3/8 correct (40%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n requesting assistance using AAC device: 7/10 correct (73%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n wait calmly when a preferred activity is denied: 8/11 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n simple meal preparation: 7/13 correct (52%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 65% for 8 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 63.9%.\n- Trend direction: upward (first session: 63%, last session: 67%).\n- IOA data: 94% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "921c053f189072e5", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "adult", "n_sessions": 8, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "18407"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1294\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-07-16 to 2026-08-05)\n\nACCELERATION PROGRAMS\n1. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n2. requesting help when a task is difficult (Mand)\n Method: NET\n3. indicating 'all done' with a step or activity (Mand)\n Method: NET\n4. identifying community safety signs (Community)\n Method: DTT\n5. requesting a break during non-preferred activities (Mand)\n Method: NET\n6. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-16 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 8/15 correct (55%); latency 3.5s; prompts FP\u00d71, PP\u00d75, G\u00d70, I\u00d79\n requesting help when a task is difficult: 7/11 correct (62%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n indicating 'all done' with a step or activity: 6/9 correct (70%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n identifying community safety signs: 9/13 correct (66%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n requesting a break during non-preferred activities: 6/8 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n wait calmly when access to preferred item is denied: 7/12 correct (58%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n\nSession 2 \u2014 2026-07-18 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 7/11 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n requesting help when a task is difficult: 8/13 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n indicating 'all done' with a step or activity: 6/9 correct (64%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n identifying community safety signs: 8/12 correct (65%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d75\n requesting a break during non-preferred activities: 10/15 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 5/8 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-07-20 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 6/11 correct (57%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n requesting help when a task is difficult: 6/10 correct (62%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n indicating 'all done' with a step or activity: 8/11 correct (69%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n identifying community safety signs: 10/15 correct (68%); latency 3.1s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n requesting a break during non-preferred activities: 8/11 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 5/8 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n\nSession 4 \u2014 2026-07-22 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 9/15 correct (57%); latency 3.5s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n requesting help when a task is difficult: 6/11 correct (53%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n indicating 'all done' with a step or activity: 9/12 correct (71%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n identifying community safety signs: 6/9 correct (67%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n requesting a break during non-preferred activities: 10/13 correct (75%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d75\n wait calmly when access to preferred item is denied: 6/10 correct (56%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n\nSession 5 \u2014 2026-07-24 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 4/14 correct (29%); latency 4.9s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n requesting help when a task is difficult: 4/12 correct (29%); latency 4.8s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d73\n indicating 'all done' with a step or activity: 4/10 correct (40%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n identifying community safety signs: 6/14 correct (42%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d77\n requesting a break during non-preferred activities: 4/11 correct (35%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 3/9 correct (31%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n\nSession 6 \u2014 2026-07-26 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 4/11 correct (38%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n requesting help when a task is difficult: 5/14 correct (37%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d77\n indicating 'all done' with a step or activity: 5/10 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n identifying community safety signs: 6/13 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n requesting a break during non-preferred activities: 4/9 correct (42%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 4/12 correct (36%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n\nSession 7 \u2014 2026-07-28 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 5/10 correct (47%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n requesting help when a task is difficult: 7/14 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n indicating 'all done' with a step or activity: 7/12 correct (56%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n identifying community safety signs: 5/8 correct (57%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n requesting a break during non-preferred activities: 4/9 correct (47%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 6/12 correct (46%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n\nSession 8 \u2014 2026-07-30 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 7/13 correct (53%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d79\n requesting help when a task is difficult: 7/12 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 9/14 correct (66%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n identifying community safety signs: 7/12 correct (60%); latency 3.5s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n requesting a break during non-preferred activities: 7/15 correct (49%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d77\n wait calmly when access to preferred item is denied: 6/13 correct (46%); latency 4.1s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n\nSession 9 \u2014 2026-08-01 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 7/13 correct (53%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n requesting help when a task is difficult: 5/9 correct (58%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n indicating 'all done' with a step or activity: 7/10 correct (67%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n identifying community safety signs: 7/10 correct (69%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n requesting a break during non-preferred activities: 7/14 correct (53%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n wait calmly when access to preferred item is denied: 5/9 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n\nSession 10 \u2014 2026-08-03 \u2014 48 min \u2014 1 observer\n tolerating tooth brushing by caregiver: 5/8 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n requesting help when a task is difficult: 5/8 correct (58%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n indicating 'all done' with a step or activity: 8/12 correct (67%); latency 3.1s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d77\n identifying community safety signs: 6/9 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n requesting a break during non-preferred activities: 5/8 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n wait calmly when access to preferred item is denied: 5/8 correct (59%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Playing with non-task materials\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 51.2%.\n- Trend direction: upward (first session: 55%, last session: 59%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "ddcf715425482dc0", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "52659"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2923\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Writing\nDate range: Sessions 1\u201311 across 22 days (2026-03-24 to 2026-04-15)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. copies simple words from model (Writing)\n Method: DTT\n3. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-24 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 8/10 correct (85%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n copies simple words from model: 7/8 correct (82%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 7/8 correct (82%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n\nSession 2 \u2014 2026-03-26 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 11/14 correct (78%); latency 2.2s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d77\n copies simple words from model: 8/11 correct (76%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n indicating 'all done' when an activity is complete: 10/12 correct (83%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d79\n\nSession 3 \u2014 2026-03-28 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 9/11 correct (85%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n copies simple words from model: 11/14 correct (77%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n indicating 'all done' when an activity is complete: 11/13 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n\nSession 4 \u2014 2026-03-30 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 5/9 correct (55%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n copies simple words from model: 6/14 correct (40%); latency 4.0s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n indicating 'all done' when an activity is complete: 7/11 correct (64%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n\nSession 5 \u2014 2026-04-01 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 6/11 correct (50%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n copies simple words from model: 3/9 correct (37%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n indicating 'all done' when an activity is complete: 8/13 correct (62%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d77\n\nSession 6 \u2014 2026-04-03 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 5/9 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n copies simple words from model: 4/11 correct (37%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n indicating 'all done' when an activity is complete: 6/10 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n\nSession 7 \u2014 2026-04-05 \u2014 IOA SESSION \u2014 2 observers\n identifies numerals 1\u201320: 59% (primary); IOA 84% agreement\n copies simple words from model: 38% (primary); IOA 84% agreement\n indicating 'all done' when an activity is complete: 70% (primary); IOA 84% agreement\n\nSession 8 \u2014 2026-04-07 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 8/13 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n copies simple words from model: 4/10 correct (43%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n indicating 'all done' when an activity is complete: 11/15 correct (75%); latency 2.4s; prompts FP\u00d75, PP\u00d72, G\u00d70, I\u00d78\n\nSession 9 \u2014 2026-04-09 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 9/14 correct (68%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n copies simple words from model: 7/13 correct (51%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n indicating 'all done' when an activity is complete: 11/14 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n\nSession 10 \u2014 2026-04-11 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 6/9 correct (71%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n copies simple words from model: 7/13 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 13/15 correct (85%); latency 2.2s; prompts FP\u00d73, PP\u00d75, G\u00d72, I\u00d75\n\nSession 11 \u2014 2026-04-13 \u2014 59 min \u2014 1 observer\n identifies numerals 1\u201320: 6/8 correct (73%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n copies simple words from model: 6/10 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n indicating 'all done' when an activity is complete: 11/13 correct (85%); latency 2.2s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Stereotypic behavior during inter-trial intervals\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 67.3%.\n- Trend direction: downward (first session: 85%, last session: 73%).\n- IOA data: 84% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "6e2ad2ab8130afd0", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "school_age", "n_sessions": 11, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "54582"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3050\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Tolerating Denied Access\nDate range: Sessions 1\u201312 across 24 days (2026-09-11 to 2026-10-05)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n2. Tantrum \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-11 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/9 correct (85%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/12 correct (83%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 10/13 correct (76%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n Elopement: freq 2\n Tantrum: freq 3, duration 6m total\n\nSession 2 \u2014 2026-09-13 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (73%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 11/15 correct (77%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n Elopement: freq 2\n Tantrum: freq 2, duration 8m total\n\nSession 3 \u2014 2026-09-15 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/14 correct (73%); latency 2.5s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (63%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 6/9 correct (64%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n Elopement: freq 2\n Tantrum: freq 3, duration 12m total\n\nSession 4 \u2014 2026-09-17 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/9 correct (63%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (58%); latency 3.4s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 7/11 correct (64%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n Elopement: freq 3\n Tantrum: freq 3, duration 15m total\n\nSession 5 \u2014 2026-09-19 \u2014 IOA SESSION \u2014 2 observers\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 65% (primary); IOA 91% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 59% (primary); IOA 91% agreement\n wait calmly when access to preferred item is denied: 59% (primary); IOA 91% agreement\n Elopement: freq 5; IOA 91% agreement\n Tantrum: freq 4, duration 8m total; IOA 91% agreement\n\nSession 6 \u2014 2026-09-21 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (54%); latency 3.7s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/13 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 7/13 correct (57%); latency 3.4s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n Elopement: freq 4\n Tantrum: freq 5, duration 25m total\n\nSession 7 \u2014 2026-09-23 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (44%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 6/12 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n Elopement: freq 5\n Tantrum: freq 5, duration 10m total\n\nSession 8 \u2014 2026-09-25 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/12 correct (50%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/15 correct (34%); latency 4.6s; prompts FP\u00d75, PP\u00d71, G\u00d75, I\u00d74\n wait calmly when access to preferred item is denied: 3/8 correct (43%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n Elopement: freq 6\n Tantrum: freq 5, duration 15m total\n\nSession 9 \u2014 2026-09-27 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/12 correct (38%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/9 correct (26%); latency 4.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 5/12 correct (40%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d79\n Elopement: freq 7\n Tantrum: freq 6, duration 30m total\n\nSession 10 \u2014 2026-09-29 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/15 correct (37%); latency 4.5s; prompts FP\u00d75, PP\u00d73, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/13 correct (20%); latency 5.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 3/9 correct (33%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n Elopement: freq 7\n Tantrum: freq 8, duration 24m total\n\nSession 11 \u2014 2026-10-01 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/13 correct (35%); latency 4.3s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 1/11 correct (13%); latency 5.4s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 2/8 correct (29%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n Elopement: freq 7\n Tantrum: freq 8, duration 40m total\n\nSession 12 \u2014 2026-10-03 \u2014 33 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/10 correct (25%); latency 4.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 1/11 correct (11%); latency 5.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 2/11 correct (21%); latency 5.2s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n Elopement: freq 9\n Tantrum: freq 8, duration 40m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Vocal refusal ('no', 'I don't want to')\n- Crying during demand presentation\n- Pushing materials away from the work area\n- Attempts at property destruction (throwing materials, swiping items off the surface)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (85% -> 25%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nTantrum: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nhigh\n\nPattern is well-supported: 12 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 54.3%.\n- Trend direction: downward (first session: 85%, last session: 25%).\n- Elopement: mean 4.9/session (early-window mean 2.2 -> late-window mean 7.5).\n- Tantrum: mean 5.0/session (early-window mean 2.8 -> late-window mean 7.5).\n- IOA data: 91% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "d6705be2dbe0d3bb", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Elopement": "escape", "Tantrum": "attention"}, "escalation_level": 4, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "school_age", "n_sessions": 12, "n_programs": 3, "n_behaviors": 2, "behavior_ids": ["elopement", "tantrum"], "behavior_functions": ["escape", "attention"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "95315"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3442\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Social Behavior + Writing\nDate range: Sessions 1\u201310 across 20 days (2026-01-08 to 2026-01-28)\n\nACCELERATION PROGRAMS\n1. takes turns during structured activities (Social Behavior)\n Method: PRT\n2. requesting preferred items using a full sentence (Mand)\n Method: NET\n3. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n4. copies simple words from model (Writing)\n Method: DTT\n5. mands for missing items needed to complete an activity (Mand)\n Method: NET\n6. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-08 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 6/8 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n requesting preferred items using a full sentence: 11/14 correct (79%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 10/10 correct (95%); latency 2.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n copies simple words from model: 10/11 correct (91%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n mands for missing items needed to complete an activity: 13/15 correct (90%); latency 2.2s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (72%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n Verbal aggression: freq 1\n\nSession 2 \u2014 2026-01-10 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 8/11 correct (77%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n requesting preferred items using a full sentence: 8/10 correct (79%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 11/11 correct (98%); latency 1.3s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n copies simple words from model: 10/11 correct (87%); latency 2.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n mands for missing items needed to complete an activity: 7/8 correct (90%); latency 1.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/10 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n Verbal aggression: freq 1\n\nSession 3 \u2014 2026-01-12 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 8/10 correct (83%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n requesting preferred items using a full sentence: 11/14 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 12/12 correct (98%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d77\n copies simple words from model: 8/8 correct (95%); latency 1.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n mands for missing items needed to complete an activity: 11/13 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 11/15 correct (72%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d77\n Verbal aggression: freq 1\n\nSession 4 \u2014 2026-01-14 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 10/13 correct (77%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n requesting preferred items using a full sentence: 9/12 correct (77%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 10/10 correct (98%); latency 1.7s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n copies simple words from model: 8/8 correct (95%); latency 1.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n mands for missing items needed to complete an activity: 7/8 correct (89%); latency 1.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/11 correct (76%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n Verbal aggression: freq 1\n\nSession 5 \u2014 2026-01-16 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 12/15 correct (83%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d79\n requesting preferred items using a full sentence: 8/11 correct (74%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 12/12 correct (98%); latency 1.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n copies simple words from model: 13/14 correct (91%); latency 1.7s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d74\n mands for missing items needed to complete an activity: 9/10 correct (87%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/9 correct (77%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n Verbal aggression: freq 2\n\nSession 6 \u2014 2026-01-18 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 9/11 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n requesting preferred items using a full sentence: 10/14 correct (72%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 10/10 correct (98%); latency 1.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n copies simple words from model: 10/10 correct (97%); latency 1.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n mands for missing items needed to complete an activity: 9/10 correct (85%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (75%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Verbal aggression: freq 0\n\nSession 7 \u2014 2026-01-20 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 10/12 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n requesting preferred items using a full sentence: 9/13 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 14/14 correct (98%); latency 1.7s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n copies simple words from model: 11/11 correct (98%); latency 1.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n mands for missing items needed to complete an activity: 9/12 correct (77%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (73%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n Verbal aggression: freq 1\n\nSession 8 \u2014 2026-01-22 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 11/13 correct (88%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d78\n requesting preferred items using a full sentence: 8/12 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 15/15 correct (98%); latency 1.3s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d76\n copies simple words from model: 15/15 correct (98%); latency 1.8s; prompts FP\u00d74, PP\u00d75, G\u00d73, I\u00d73\n mands for missing items needed to complete an activity: 12/15 correct (80%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/9 correct (76%); latency 2.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n Verbal aggression: freq 0\n\nSession 9 \u2014 2026-01-24 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 10/12 correct (84%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n requesting preferred items using a full sentence: 7/10 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 14/14 correct (98%); latency 1.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n copies simple words from model: 8/8 correct (98%); latency 1.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n mands for missing items needed to complete an activity: 12/15 correct (78%); latency 2.3s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/10 correct (77%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n Verbal aggression: freq 1\n\nSession 10 \u2014 2026-01-26 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 11/13 correct (86%); latency 2.2s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n requesting preferred items using a full sentence: 8/12 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 14/14 correct (98%); latency 1.3s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n copies simple words from model: 8/8 correct (98%); latency 1.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 11/15 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 11/14 correct (76%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n Verbal aggression: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Delayed responding (>5 seconds consistently)\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (79%\u201386%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 81.8%.\n- Trend direction: upward (first session: 79%, last session: 86%).\n- Verbal aggression: mean 1.0/session (early-window mean 1.0 -> late-window mean 1.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "d0f7fca363ad9c8b", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Verbal aggression": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["verbal_aggression"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "71533"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts abstract concepts (same/different, first/last)\nCurriculum Reference: Tact L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts abstract concepts (same/different, first/last)\n- Curriculum: Tact L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: crayon, key, backpack\nDistractor stimuli: phone, hat, spoon, scissors, jacket\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "389196e0b33b3448", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts abstract concepts (same/different, first/last)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "20338"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: responds to peer initiations appropriately\nCurriculum Reference: Social Behavior & Social Play L2\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: responds to peer initiations appropriately\n- Curriculum: Social Behavior & Social Play L2\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "632d49434d10897a", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L2", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "responds to peer initiations appropriately", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "time_delay", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "76914"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6950\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand\nDate range: Sessions 1\u20139 across 18 days (2026-05-06 to 2026-05-24)\n\nACCELERATION PROGRAMS\n1. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. selects items by feature (Listener Responding)\n Method: DTT\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-06 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 9/10 correct (86%); latency 1.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d74\n answers WH-questions about familiar topics: 8/10 correct (78%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/10 correct (78%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n selects items by feature: 8/11 correct (76%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n Property destruction: freq 0\n\nSession 2 \u2014 2026-05-08 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 10/14 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n answers WH-questions about familiar topics: 10/13 correct (80%); latency 2.3s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n selects items by feature: 10/15 correct (70%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d75, I\u00d79\n Property destruction: freq 1\n\nSession 3 \u2014 2026-05-10 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 9/12 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n answers WH-questions about familiar topics: 10/14 correct (68%); latency 3.2s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n selects items by feature: 5/9 correct (60%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n Property destruction: freq 3\n\nSession 4 \u2014 2026-05-12 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 6/10 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n answers WH-questions about familiar topics: 6/8 correct (72%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (58%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n selects items by feature: 6/12 correct (53%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n Property destruction: freq 3\n\nSession 5 \u2014 2026-05-14 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 8/14 correct (59%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n answers WH-questions about familiar topics: 7/11 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (55%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n selects items by feature: 4/9 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n Property destruction: freq 4\n\nSession 6 \u2014 2026-05-16 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 7/14 correct (51%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n answers WH-questions about familiar topics: 5/8 correct (62%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (57%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n selects items by feature: 5/14 correct (37%); latency 4.1s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n Property destruction: freq 4\n\nSession 7 \u2014 2026-05-18 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 5/11 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n answers WH-questions about familiar topics: 5/9 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (46%); latency 3.7s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d75\n selects items by feature: 3/10 correct (34%); latency 4.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n Property destruction: freq 4\n\nSession 8 \u2014 2026-05-20 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 5/15 correct (31%); latency 4.4s; prompts FP\u00d75, PP\u00d74, G\u00d73, I\u00d73\n answers WH-questions about familiar topics: 6/11 correct (53%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (40%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n selects items by feature: 3/11 correct (29%); latency 4.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n Property destruction: freq 5\n\nSession 9 \u2014 2026-05-22 \u2014 38 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 3/12 correct (29%); latency 5.0s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n answers WH-questions about familiar topics: 7/15 correct (45%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/9 correct (40%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n selects items by feature: 2/11 correct (22%); latency 4.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n Property destruction: freq 5\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Turning away from task or staff\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (86% -> 29%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nProperty destruction: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 56.3%.\n- Trend direction: downward (first session: 86%, last session: 29%).\n- Property destruction: mean 3.2/session (early-window mean 1.3 -> late-window mean 4.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "ac7c2835f3befc58", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Property destruction": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "early", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "73913"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6120\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Self-Care\nDate range: Sessions 1\u201311 across 22 days (2026-01-07 to 2026-01-29)\n\nACCELERATION PROGRAMS\n1. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n2. washing hands independently (Self-Care)\n Method: Task Analysis\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: tangible\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-07 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 11/13 correct (81%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d710\n washing hands independently: 6/8 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n brushing teeth independently: 8/10 correct (82%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n Tantrum: freq 2, duration 10m total\n\nSession 2 \u2014 2026-01-09 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 11/13 correct (83%); latency 2.4s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n washing hands independently: 11/14 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d711\n brushing teeth independently: 9/11 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n Tantrum: freq 0\n\nSession 3 \u2014 2026-01-11 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 8/10 correct (79%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n washing hands independently: 7/9 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n brushing teeth independently: 10/12 correct (80%); latency 2.2s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n Tantrum: freq 2, duration 8m total\n\nSession 4 \u2014 2026-01-13 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 6/13 correct (48%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n washing hands independently: 6/12 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n brushing teeth independently: 8/15 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d77\n Tantrum: freq 1, duration 2m total\n\nSession 5 \u2014 2026-01-15 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 5/10 correct (47%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n washing hands independently: 7/14 correct (52%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n brushing teeth independently: 7/14 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n Tantrum: freq 1, duration 2m total\n\nSession 6 \u2014 2026-01-17 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 6/13 correct (49%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n washing hands independently: 8/15 correct (54%); latency 3.7s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d75\n brushing teeth independently: 5/9 correct (55%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n Tantrum: freq 2, duration 4m total\n\nSession 7 \u2014 2026-01-19 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 8/15 correct (52%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d77\n washing hands independently: 6/12 correct (48%); latency 4.0s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n brushing teeth independently: 7/11 correct (65%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n Tantrum: freq 1, duration 2m total\n\nSession 8 \u2014 2026-01-21 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 6/12 correct (54%); latency 3.6s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n washing hands independently: 6/10 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n brushing teeth independently: 11/15 correct (76%); latency 2.6s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d78\n Tantrum: freq 1, duration 3m total\n\nSession 9 \u2014 2026-01-23 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 6/10 correct (63%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n washing hands independently: 6/10 correct (62%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n brushing teeth independently: 7/9 correct (81%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n Tantrum: freq 2, duration 8m total\n\nSession 10 \u2014 2026-01-25 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 8/12 correct (70%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n washing hands independently: 7/10 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n brushing teeth independently: 10/12 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n Tantrum: freq 2, duration 8m total\n ABC (tantrum): A = peer took preferred toy; B = grabbed toy back + cried; C = staff mediated and returned item\n\nSession 11 \u2014 2026-01-27 \u2014 44 min \u2014 1 observer\n meal cleanup after eating: 12/15 correct (79%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n washing hands independently: 9/12 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n brushing teeth independently: 9/11 correct (81%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n Tantrum: freq 1, duration 4m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nTantrum: tangible\n Evidence: ABC events in the log support the tangible function; behavior occurs when preferred items are restricted and is followed by access to the item.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 64.1%.\n- Trend direction: downward (first session: 81%, last session: 79%).\n- Tantrum: mean 1.4/session (early-window mean 1.3 -> late-window mean 1.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "8567f65e7969a8e9", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Tantrum": "tangible"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adolescent", "n_sessions": 11, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["tangible"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "1699"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9274\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Reading + Writing\nDate range: Sessions 1\u20137 across 14 days (2026-07-07 to 2026-07-21)\n\nACCELERATION PROGRAMS\n1. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n2. copies simple words from model (Writing)\n Method: DTT\n3. reads 10\u201320 sight words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Mouthing \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-07 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/13 correct (46%); latency 4.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d79\n copies simple words from model: 5/11 correct (41%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n reads 10\u201320 sight words: 4/14 correct (32%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n Mouthing: freq 0; PIR 3%\n\nSession 2 \u2014 2026-07-09 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/13 correct (49%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n copies simple words from model: 6/14 correct (44%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d75\n reads 10\u201320 sight words: 6/15 correct (40%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d76\n Mouthing: freq 1; PIR 6%\n\nSession 3 \u2014 2026-07-11 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/11 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n copies simple words from model: 7/15 correct (47%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d712\n reads 10\u201320 sight words: 5/11 correct (49%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n Mouthing: freq 0; PIR 0%\n\nSession 4 \u2014 2026-07-13 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/9 correct (59%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n copies simple words from model: 8/14 correct (55%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d79\n reads 10\u201320 sight words: 7/12 correct (56%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n Mouthing: freq 2; PIR 9%\n ABC (mouthing): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 5 \u2014 2026-07-15 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 9/14 correct (63%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n copies simple words from model: 8/12 correct (63%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n reads 10\u201320 sight words: 8/13 correct (65%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n Mouthing: freq 1; PIR 9%\n\nSession 6 \u2014 2026-07-17 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 7/10 correct (68%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n copies simple words from model: 8/12 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d74\n reads 10\u201320 sight words: 8/12 correct (70%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n Mouthing: freq 2; PIR 11%\n\nSession 7 \u2014 2026-07-19 \u2014 39 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/9 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n copies simple words from model: 9/13 correct (72%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n reads 10\u201320 sight words: 11/15 correct (72%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d711\n Mouthing: freq 0; PIR 10%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Leaving seat without permission\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 46% at return, recovering to 71%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nMouthing: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 58.6%.\n- Trend direction: upward (first session: 46%, last session: 71%).\n- Mouthing: mean 0.9/session (early-window mean 0.5 -> late-window mean 1.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "033a32bfc33c1177", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Mouthing": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "early", "n_sessions": 7, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["mouthing"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "23188"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5515\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand + Reading + Tact\nDate range: Sessions 1\u20139 across 18 days (2026-04-18 to 2026-05-06)\n\nACCELERATION PROGRAMS\n1. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. reads 10\u201320 sight words (Reading)\n Method: DTT\n4. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n5. tacts colors of objects (Tact)\n Method: DTT\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-18 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 5/9 correct (50%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 6/13 correct (47%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n reads 10\u201320 sight words: 7/15 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d711\n requesting a break ('break please' or AAC 'I need a break'): 6/14 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n tacts colors of objects: 7/14 correct (51%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n Self-injurious behavior (SIB): freq 0\n\nSession 2 \u2014 2026-04-20 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 5/9 correct (53%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n answers WH-questions about familiar topics: 7/12 correct (58%); latency 3.5s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n reads 10\u201320 sight words: 7/13 correct (55%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 6/12 correct (52%); latency 3.6s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n tacts colors of objects: 5/9 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n Self-injurious behavior (SIB): freq 2\n\nSession 3 \u2014 2026-04-22 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 5/8 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n answers WH-questions about familiar topics: 7/10 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 6/10 correct (60%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 5/9 correct (59%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n tacts colors of objects: 7/10 correct (66%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n Self-injurious behavior (SIB): freq 1\n\nSession 4 \u2014 2026-04-24 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 9/15 correct (63%); latency 3.2s; prompts FP\u00d75, PP\u00d71, G\u00d75, I\u00d74\n answers WH-questions about familiar topics: 6/8 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 7/10 correct (69%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (66%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n tacts colors of objects: 10/13 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 0\n\nSession 5 \u2014 2026-04-26 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 8/12 correct (68%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d76\n answers WH-questions about familiar topics: 12/15 correct (83%); latency 2.0s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d77\n reads 10\u201320 sight words: 7/9 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 8/10 correct (76%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n tacts colors of objects: 8/10 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 1\n\nSession 6 \u2014 2026-04-28 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 7/9 correct (75%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n answers WH-questions about familiar topics: 10/12 correct (83%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 6/8 correct (79%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 8/10 correct (82%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n tacts colors of objects: 8/10 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n Self-injurious behavior (SIB): freq 1\n\nSession 7 \u2014 2026-04-30 \u2014 IOA SESSION \u2014 2 observers\n follows 1-step motor instructions: 75% (primary); IOA 83% agreement\n answers WH-questions about familiar topics: 83% (primary); IOA 83% agreement\n reads 10\u201320 sight words: 79% (primary); IOA 83% agreement\n requesting a break ('break please' or AAC 'I need a break'): 83% (primary); IOA 83% agreement\n tacts colors of objects: 80% (primary); IOA 83% agreement\n Self-injurious behavior (SIB): freq 2; IOA 83% agreement\n\nSession 8 \u2014 2026-05-02 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 8/10 correct (75%); latency 2.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 8/10 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 11/14 correct (79%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 8/10 correct (83%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n tacts colors of objects: 10/13 correct (80%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n Self-injurious behavior (SIB): freq 1\n\nSession 9 \u2014 2026-05-04 \u2014 44 min \u2014 1 observer\n follows 1-step motor instructions: 6/8 correct (75%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 12/14 correct (83%); latency 2.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n reads 10\u201320 sight words: 7/9 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 11/13 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d75\n tacts colors of objects: 7/9 correct (80%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 50% at return, recovering to 75%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 66.0%.\n- Trend direction: upward (first session: 50%, last session: 75%).\n- Self-injurious behavior (SIB): mean 1.0/session (early-window mean 1.0 -> late-window mean 1.3).\n- IOA data: 83% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "1088fded3c749272", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "early", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "60991"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: using public transportation\nCurriculum Reference: AFLS Community Participation Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: using public transportation\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Confirm the correct stop or station. <- Learner-taught step (forward chaining starts here)\n2. Purchase or have the fare ready.\n3. Wait at the designated boarding area.\n4. Board the vehicle when it arrives.\n5. Pay the fare or present the pass.\n6. Sit or stand safely.\n7. Monitor the route for the correct stop.\n8. Exit at the correct stop.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3e9f1e4489999688", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "using public transportation", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "using public transportation", "module": "community", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "92010"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9796\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Self-Care\nDate range: Sessions 1\u201310 across 20 days (2026-02-25 to 2026-03-17)\n\nACCELERATION PROGRAMS\n1. tolerating haircut (Self-Care)\n Method: Task Analysis\n2. indicating 'all done' with a step or activity (Mand)\n Method: NET\n3. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n4. requesting help when a task is difficult (Mand)\n Method: NET\n5. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-25 \u2014 54 min \u2014 1 observer\n tolerating haircut: 4/9 correct (44%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n indicating 'all done' with a step or activity: 5/11 correct (45%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n meal cleanup after eating: 7/12 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n requesting help when a task is difficult: 10/13 correct (75%); latency 2.7s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n brushing teeth independently: 5/8 correct (56%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n\nSession 2 \u2014 2026-02-27 \u2014 54 min \u2014 1 observer\n tolerating haircut: 8/11 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n indicating 'all done' with a step or activity: 6/10 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n meal cleanup after eating: 7/12 correct (57%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n requesting help when a task is difficult: 10/13 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n brushing teeth independently: 6/12 correct (50%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n\nSession 3 \u2014 2026-03-01 \u2014 54 min \u2014 1 observer\n tolerating haircut: 7/12 correct (56%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n indicating 'all done' with a step or activity: 6/9 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n meal cleanup after eating: 9/15 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d710\n requesting help when a task is difficult: 12/14 correct (85%); latency 2.2s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n brushing teeth independently: 5/8 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n\nSession 4 \u2014 2026-03-03 \u2014 54 min \u2014 1 observer\n tolerating haircut: 6/8 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n indicating 'all done' with a step or activity: 7/13 correct (56%); latency 3.7s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n meal cleanup after eating: 8/12 correct (68%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n requesting help when a task is difficult: 6/11 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n brushing teeth independently: 5/8 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n\nSession 5 \u2014 2026-03-05 \u2014 54 min \u2014 1 observer\n tolerating haircut: 9/14 correct (67%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d75\n indicating 'all done' with a step or activity: 8/13 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n meal cleanup after eating: 7/9 correct (81%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n requesting help when a task is difficult: 8/10 correct (79%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n brushing teeth independently: 4/8 correct (53%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n\nSession 6 \u2014 2026-03-07 \u2014 IOA SESSION \u2014 2 observers\n tolerating haircut: 56% (primary); IOA 87% agreement\n indicating 'all done' with a step or activity: 53% (primary); IOA 87% agreement\n meal cleanup after eating: 87% (primary); IOA 87% agreement\n requesting help when a task is difficult: 61% (primary); IOA 87% agreement\n brushing teeth independently: 65% (primary); IOA 87% agreement\n\nSession 7 \u2014 2026-03-09 \u2014 54 min \u2014 1 observer\n tolerating haircut: 6/12 correct (46%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n indicating 'all done' with a step or activity: 4/9 correct (44%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n meal cleanup after eating: 6/8 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n requesting help when a task is difficult: 10/12 correct (81%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n brushing teeth independently: 5/13 correct (42%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n\nSession 8 \u2014 2026-03-11 \u2014 54 min \u2014 1 observer\n tolerating haircut: 7/9 correct (76%); latency 2.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n indicating 'all done' with a step or activity: 8/13 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n meal cleanup after eating: 6/9 correct (72%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n requesting help when a task is difficult: 8/11 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n brushing teeth independently: 4/10 correct (39%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n\nSession 9 \u2014 2026-03-13 \u2014 54 min \u2014 1 observer\n tolerating haircut: 9/13 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n indicating 'all done' with a step or activity: 5/12 correct (45%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n meal cleanup after eating: 6/10 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n requesting help when a task is difficult: 5/8 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n brushing teeth independently: 8/13 correct (60%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n\nSession 10 \u2014 2026-03-15 \u2014 54 min \u2014 1 observer\n tolerating haircut: 7/10 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n indicating 'all done' with a step or activity: 6/10 correct (64%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n meal cleanup after eating: 11/13 correct (82%); latency 2.6s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n requesting help when a task is difficult: 5/8 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n brushing teeth independently: 3/9 correct (38%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 58% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 63.8%.\n- Trend direction: upward (first session: 44%, last session: 73%).\n- IOA data: 87% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "97f9122e04592919", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "81514"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7332\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-01-17 to 2026-02-10)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. sounds out CVC words (Reading)\n Method: DTT\n4. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n5. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-17 \u2014 41 min \u2014 1 observer\n copies simple words from model: 10/13 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n single-word requests for preferred items: 8/11 correct (72%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n sounds out CVC words: 10/14 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n indicating 'all done' when an activity is complete: 7/9 correct (81%); latency 2.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (81%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 1\n\nSession 2 \u2014 2026-01-19 \u2014 41 min \u2014 1 observer\n copies simple words from model: 6/8 correct (79%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n single-word requests for preferred items: 7/9 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n sounds out CVC words: 10/14 correct (74%); latency 2.4s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d75\n indicating 'all done' when an activity is complete: 7/8 correct (84%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/9 correct (82%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 1\n\nSession 3 \u2014 2026-01-21 \u2014 41 min \u2014 1 observer\n copies simple words from model: 8/11 correct (73%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n single-word requests for preferred items: 6/8 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n sounds out CVC words: 11/15 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n indicating 'all done' when an activity is complete: 12/15 correct (80%); latency 2.1s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/11 correct (79%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n Self-injurious behavior (SIB): freq 1\n\nSession 4 \u2014 2026-01-23 \u2014 41 min \u2014 1 observer\n copies simple words from model: 9/12 correct (75%); latency 2.7s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n single-word requests for preferred items: 6/8 correct (71%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n sounds out CVC words: 6/9 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n indicating 'all done' when an activity is complete: 11/13 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/9 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 0\n\nSession 5 \u2014 2026-01-25 \u2014 41 min \u2014 1 observer\n copies simple words from model: 4/9 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n single-word requests for preferred items: 5/12 correct (42%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n sounds out CVC words: 5/15 correct (34%); latency 4.5s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d78\n indicating 'all done' when an activity is complete: 5/13 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/11 correct (38%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n Self-injurious behavior (SIB): freq 0\n\nSession 6 \u2014 2026-01-27 \u2014 41 min \u2014 1 observer\n copies simple words from model: 6/13 correct (46%); latency 3.9s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n single-word requests for preferred items: 4/12 correct (35%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n sounds out CVC words: 5/14 correct (36%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n indicating 'all done' when an activity is complete: 5/10 correct (46%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/9 correct (38%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n Self-injurious behavior (SIB): freq 2\n\nSession 7 \u2014 2026-01-29 \u2014 41 min \u2014 1 observer\n copies simple words from model: 4/8 correct (52%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n single-word requests for preferred items: 5/12 correct (40%); latency 4.0s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n sounds out CVC words: 3/8 correct (38%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 6/14 correct (46%); latency 4.1s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/10 correct (40%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 1\n\nSession 8 \u2014 2026-01-31 \u2014 41 min \u2014 1 observer\n copies simple words from model: 5/8 correct (58%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n single-word requests for preferred items: 3/8 correct (39%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n sounds out CVC words: 5/14 correct (39%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n indicating 'all done' when an activity is complete: 6/13 correct (49%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/14 correct (44%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d710\n Self-injurious behavior (SIB): freq 2\n\nSession 9 \u2014 2026-02-02 \u2014 41 min \u2014 1 observer\n copies simple words from model: 8/13 correct (61%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d79\n single-word requests for preferred items: 5/12 correct (42%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n sounds out CVC words: 5/11 correct (44%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 6/11 correct (54%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/8 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 1\n\nSession 10 \u2014 2026-02-04 \u2014 41 min \u2014 1 observer\n copies simple words from model: 10/15 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n single-word requests for preferred items: 6/13 correct (46%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d710\n sounds out CVC words: 6/12 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d75\n indicating 'all done' when an activity is complete: 6/10 correct (57%); latency 3.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (56%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n Self-injurious behavior (SIB): freq 2\n\nSession 11 \u2014 2026-02-06 \u2014 41 min \u2014 1 observer\n copies simple words from model: 8/11 correct (73%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n single-word requests for preferred items: 5/10 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 7/12 correct (61%); latency 3.1s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n indicating 'all done' when an activity is complete: 5/8 correct (67%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/15 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d713\n Self-injurious behavior (SIB): freq 1\n\nSession 12 \u2014 2026-02-08 \u2014 41 min \u2014 1 observer\n copies simple words from model: 11/15 correct (76%); latency 2.5s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d79\n single-word requests for preferred items: 7/13 correct (56%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n sounds out CVC words: 10/15 correct (66%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n indicating 'all done' when an activity is complete: 6/8 correct (71%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/15 correct (66%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d711\n Self-injurious behavior (SIB): freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 65.7%.\n- Trend direction: downward (first session: 79%, last session: 76%).\n- Self-injurious behavior (SIB): mean 1.1/session (early-window mean 0.8 -> late-window mean 1.2).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "2b825786b519b265", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Self-injurious behavior (SIB)": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "school_age", "n_sessions": 12, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "74717"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1192\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Social Behavior + Writing\nDate range: Sessions 1\u201311 across 22 days (2026-04-12 to 2026-05-04)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n4. copies simple words from model (Writing)\n Method: DTT\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-12 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/14 correct (87%); latency 2.1s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 7/8 correct (86%); latency 1.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n takes turns during structured activities: 7/8 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n copies simple words from model: 8/9 correct (84%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n Verbal aggression: freq 1\n\nSession 2 \u2014 2026-04-14 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (39%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n indicating 'all done' when an activity is complete: 4/13 correct (32%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d711\n takes turns during structured activities: 3/9 correct (39%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n copies simple words from model: 5/10 correct (45%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n Verbal aggression: freq 1\n\nSession 3 \u2014 2026-04-16 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/13 correct (85%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n indicating 'all done' when an activity is complete: 11/15 correct (77%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d75, I\u00d79\n takes turns during structured activities: 8/9 correct (85%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n copies simple words from model: 12/15 correct (82%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n Verbal aggression: freq 0\n\nSession 4 \u2014 2026-04-18 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/15 correct (42%); latency 4.4s; prompts FP\u00d71, PP\u00d75, G\u00d72, I\u00d77\n indicating 'all done' when an activity is complete: 3/10 correct (33%); latency 4.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n takes turns during structured activities: 5/14 correct (38%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d77\n copies simple words from model: 5/12 correct (45%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n Verbal aggression: freq 0\n\nSession 5 \u2014 2026-04-20 \u2014 IOA SESSION \u2014 2 observers\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 92% (primary); IOA 89% agreement\n indicating 'all done' when an activity is complete: 83% (primary); IOA 89% agreement\n takes turns during structured activities: 91% (primary); IOA 89% agreement\n copies simple words from model: 78% (primary); IOA 89% agreement\n Verbal aggression: freq 0; IOA 89% agreement\n\nSession 6 \u2014 2026-04-22 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/13 correct (35%); latency 4.2s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d74\n indicating 'all done' when an activity is complete: 4/11 correct (33%); latency 4.7s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n takes turns during structured activities: 5/13 correct (36%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n copies simple words from model: 3/8 correct (43%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Verbal aggression: freq 1\n\nSession 7 \u2014 2026-04-24 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/14 correct (85%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n indicating 'all done' when an activity is complete: 10/13 correct (79%); latency 2.4s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n takes turns during structured activities: 11/13 correct (86%); latency 2.2s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n copies simple words from model: 6/8 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Verbal aggression: freq 0\n\nSession 8 \u2014 2026-04-26 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (43%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 3/9 correct (37%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n takes turns during structured activities: 5/11 correct (42%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n copies simple words from model: 6/12 correct (48%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n Verbal aggression: freq 0\n\nSession 9 \u2014 2026-04-28 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (87%); latency 1.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n indicating 'all done' when an activity is complete: 11/14 correct (82%); latency 2.4s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n takes turns during structured activities: 9/11 correct (82%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n copies simple words from model: 7/8 correct (86%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n Verbal aggression: freq 0\n\nSession 10 \u2014 2026-04-30 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (44%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n indicating 'all done' when an activity is complete: 5/11 correct (41%); latency 4.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n takes turns during structured activities: 6/13 correct (44%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n copies simple words from model: 6/14 correct (41%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n Verbal aggression: freq 1\n\nSession 11 \u2014 2026-05-02 \u2014 60 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/13 correct (88%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n indicating 'all done' when an activity is complete: 11/14 correct (78%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d76\n takes turns during structured activities: 13/15 correct (90%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n copies simple words from model: 10/13 correct (77%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n Verbal aggression: freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 66.2%.\n- Trend direction: upward (first session: 87%, last session: 88%).\n- Verbal aggression: mean 0.5/session (early-window mean 0.7 -> late-window mean 0.7).\n- IOA data: 89% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "a61de0f2c5761341", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {"Verbal aggression": "escape"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "school_age", "n_sessions": 11, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["verbal_aggression"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "4269"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: spontaneously greets familiar people\nCurriculum Reference: Spontaneous Vocal Behavior L1\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously greets familiar people\n- Curriculum: Spontaneous Vocal Behavior L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nUse the natural beginning/ending of an established routine (morning arrival, mealtime, departure) as the moment for the target response.\nPrimary natural context: departure.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's departure routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult returns the greeting warmly.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "7a8b6a280f60c544", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously greets familiar people", "mo_category": "spontaneous_greeting", "mo_arrangement": "routine_lead_in", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "44373"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: identifies basic shapes in math context\nCurriculum: Math L1\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies basic shapes in math context\n- Curriculum: Math L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: heart, star, circle\nDistractor stimuli: diamond, triangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "245293d827f46696", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies basic shapes in math context", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "35263"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4405\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Reading + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u20138 across 16 days (2026-05-06 to 2026-05-22)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. requesting preferred items using AAC device (Mand)\n Method: NET\n3. indicating 'all done' with a step or activity (Mand)\n Method: NET\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. sounds out CVC words (Reading)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-06 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/11 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n requesting preferred items using AAC device: 8/15 correct (54%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d712\n indicating 'all done' with a step or activity: 5/14 correct (36%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d710\n brushing teeth independently: 6/12 correct (47%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n sounds out CVC words: 3/9 correct (38%); latency 4.5s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n\nSession 2 \u2014 2026-05-08 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/15 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n requesting preferred items using AAC device: 4/8 correct (54%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n indicating 'all done' with a step or activity: 3/8 correct (33%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n brushing teeth independently: 8/15 correct (54%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n sounds out CVC words: 6/15 correct (39%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d79\n\nSession 3 \u2014 2026-05-10 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/11 correct (53%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n requesting preferred items using AAC device: 8/14 correct (59%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d78\n indicating 'all done' with a step or activity: 4/9 correct (44%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n brushing teeth independently: 5/9 correct (52%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n sounds out CVC words: 4/10 correct (45%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n\nSession 4 \u2014 2026-05-12 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/9 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n requesting preferred items using AAC device: 7/11 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n indicating 'all done' with a step or activity: 6/14 correct (44%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n brushing teeth independently: 5/9 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n sounds out CVC words: 6/11 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n\nSession 5 \u2014 2026-05-14 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/10 correct (62%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n requesting preferred items using AAC device: 10/14 correct (70%); latency 2.7s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d73\n indicating 'all done' with a step or activity: 8/15 correct (51%); latency 3.6s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d75\n brushing teeth independently: 8/12 correct (64%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n sounds out CVC words: 4/8 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n\nSession 6 \u2014 2026-05-16 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/15 correct (72%); latency 2.8s; prompts FP\u00d75, PP\u00d74, G\u00d75, I\u00d71\n requesting preferred items using AAC device: 7/11 correct (66%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n indicating 'all done' with a step or activity: 7/12 correct (56%); latency 3.2s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n brushing teeth independently: 6/8 correct (72%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n sounds out CVC words: 6/10 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n\nSession 7 \u2014 2026-05-18 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/9 correct (78%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n requesting preferred items using AAC device: 10/14 correct (74%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n indicating 'all done' with a step or activity: 9/15 correct (57%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d76\n brushing teeth independently: 11/14 correct (76%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n sounds out CVC words: 7/10 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n\nSession 8 \u2014 2026-05-20 \u2014 48 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/10 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n requesting preferred items using AAC device: 11/14 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n indicating 'all done' with a step or activity: 8/13 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n brushing teeth independently: 11/12 correct (88%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n sounds out CVC words: 11/15 correct (74%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d711\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Appropriate sitting posture maintained\n- Requesting more trials verbally\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 39% to 78% across 8 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 61.2%.\n- Trend direction: upward (first session: 39%, last session: 78%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "5315591797aed264", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "5794"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: engages in social games (peek-a-boo, tickle)\nCurriculum: Social Behavior & Social Play L1\nLearner Level: Early Learner\nBaseline: Generalization phase\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: engages in social games (peek-a-boo, tickle)\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "df0b036de2688cb5", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "engages in social games (peek-a-boo, tickle)", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "67055"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: solves simple word problems\nCurriculum: Math L3\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: solves simple word problems\n- Curriculum: Math L3\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How many?\"\nVariations: \"Count these\"; \"What number is this?\"\nPresentation: Present the numeral card or a set of objects to count.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: bottle, ball, apple\nDistractor stimuli: book, towel, spoon, glue, hat\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "963fc2aabc513c65", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L3", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "solves simple word problems", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "token_economy", "error_correction": "backstep", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "69141"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2052\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math\nDate range: Sessions 1\u20139 across 18 days (2026-09-24 to 2026-10-12)\n\nACCELERATION PROGRAMS\n1. single-word requests for preferred items (Mand)\n Method: NET\n2. requesting preferred items using a full sentence (Mand)\n Method: NET\n3. counts objects 1\u201320 (Math)\n Method: DTT\n4. identifies numerals 1\u201320 (Math)\n Method: DTT\n\nDECELERATION TARGETS\n1. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-24 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 4/9 correct (43%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n requesting preferred items using a full sentence: 4/9 correct (46%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n counts objects 1\u201320: 6/15 correct (42%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d77\n identifies numerals 1\u201320: 6/15 correct (41%); latency 3.9s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d76\n Motor stereotypy: freq 1; PIR 13%\n\nSession 2 \u2014 2026-09-26 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 6/14 correct (44%); latency 4.2s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n requesting preferred items using a full sentence: 7/14 correct (47%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n counts objects 1\u201320: 7/14 correct (48%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n identifies numerals 1\u201320: 5/10 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n Motor stereotypy: freq 0; PIR 7%\n\nSession 3 \u2014 2026-09-28 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 6/11 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n requesting preferred items using a full sentence: 4/8 correct (54%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n counts objects 1\u201320: 8/14 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n identifies numerals 1\u201320: 9/14 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n Motor stereotypy: freq 1; PIR 5%\n\nSession 4 \u2014 2026-09-30 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 5/9 correct (59%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n requesting preferred items using a full sentence: 5/8 correct (68%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n counts objects 1\u201320: 8/12 correct (66%); latency 3.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n identifies numerals 1\u201320: 9/13 correct (66%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n Motor stereotypy: freq 1; PIR 5%\n\nSession 5 \u2014 2026-10-02 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 8/12 correct (70%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d72\n requesting preferred items using a full sentence: 7/9 correct (72%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n counts objects 1\u201320: 8/11 correct (76%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n identifies numerals 1\u201320: 8/11 correct (74%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n Motor stereotypy: freq 0; PIR 10%\n\nSession 6 \u2014 2026-10-04 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 9/11 correct (78%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n requesting preferred items using a full sentence: 10/12 correct (84%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n counts objects 1\u201320: 12/15 correct (79%); latency 2.5s; prompts FP\u00d70, PP\u00d75, G\u00d75, I\u00d75\n identifies numerals 1\u201320: 6/8 correct (77%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n Motor stereotypy: freq 2; PIR 12%\n\nSession 7 \u2014 2026-10-06 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 12/15 correct (80%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d78\n requesting preferred items using a full sentence: 7/8 correct (90%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n counts objects 1\u201320: 8/9 correct (87%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n identifies numerals 1\u201320: 13/15 correct (84%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n Motor stereotypy: freq 1; PIR 11%\n\nSession 8 \u2014 2026-10-08 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 9/10 correct (92%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n requesting preferred items using a full sentence: 14/15 correct (95%); latency 1.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n counts objects 1\u201320: 14/15 correct (95%); latency 1.7s; prompts FP\u00d73, PP\u00d75, G\u00d72, I\u00d75\n identifies numerals 1\u201320: 8/9 correct (93%); latency 1.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n Motor stereotypy: freq 0; PIR 3%\n\nSession 9 \u2014 2026-10-10 \u2014 59 min \u2014 1 observer\n single-word requests for preferred items: 8/8 correct (95%); latency 1.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n requesting preferred items using a full sentence: 11/12 correct (95%); latency 1.5s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n counts objects 1\u201320: 13/14 correct (95%); latency 1.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n identifies numerals 1\u201320: 8/8 correct (95%); latency 1.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n Motor stereotypy: freq 0; PIR 2%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Appropriate sitting posture maintained\n- Smiling during reinforcement delivery\n- Eye contact maintained throughout trials\n- Reaching for materials proactively"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 43% to 95% across 9 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Behavior Function Hypothesis\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 68.7%.\n- Trend direction: upward (first session: 43%, last session: 95%).\n- Motor stereotypy: mean 0.7/session (early-window mean 0.7 -> late-window mean 0.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "8a4557cc64117938", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {"Motor stereotypy": "automatic"}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "school_age", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["motor_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "97400"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects correct item from an array of 2\nCurriculum Reference: Listener Responding L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from an array of 2\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: crayon, hat, napkin\nDistractor stimuli: block, key\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "543a505338f0e98c", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from an array of 2", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "93657"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: answers hypothetical questions\nCurriculum Reference: Intraverbal L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Generalization phase\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers hypothetical questions\n- Curriculum: Intraverbal L3\n- Learner: Adolescent Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: post-activity debrief.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's post-activity debrief routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "d9deb6595de84cb9", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "adolescent", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers hypothetical questions", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "time_delay", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "30191"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5878\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-06-03 to 2026-06-23)\n\nACCELERATION PROGRAMS\n1. doing laundry (Home Skills)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-03 \u2014 50 min \u2014 1 observer\n doing laundry: 8/10 correct (78%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 7/9 correct (81%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n simple meal preparation: 6/8 correct (79%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n Elopement: freq 1\n\nSession 2 \u2014 2026-06-05 \u2014 50 min \u2014 1 observer\n doing laundry: 12/15 correct (80%); latency 2.3s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n wait calmly when a preferred activity is denied: 6/9 correct (69%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n simple meal preparation: 6/8 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Elopement: freq 3\n\nSession 3 \u2014 2026-06-07 \u2014 50 min \u2014 1 observer\n doing laundry: 7/10 correct (69%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n wait calmly when a preferred activity is denied: 7/11 correct (68%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n simple meal preparation: 8/13 correct (64%); latency 3.4s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n Elopement: freq 2\n\nSession 4 \u2014 2026-06-09 \u2014 50 min \u2014 1 observer\n doing laundry: 7/11 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 6/10 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n simple meal preparation: 6/10 correct (63%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n Elopement: freq 3\n\nSession 5 \u2014 2026-06-11 \u2014 50 min \u2014 1 observer\n doing laundry: 8/12 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 5/8 correct (56%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n simple meal preparation: 5/10 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n Elopement: freq 4\n\nSession 6 \u2014 2026-06-13 \u2014 50 min \u2014 1 observer\n doing laundry: 9/15 correct (59%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d710\n wait calmly when a preferred activity is denied: 7/14 correct (50%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d76\n simple meal preparation: 4/8 correct (46%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n Elopement: freq 5\n\nSession 7 \u2014 2026-06-15 \u2014 50 min \u2014 1 observer\n doing laundry: 6/10 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n wait calmly when a preferred activity is denied: 4/10 correct (44%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n simple meal preparation: 3/8 correct (38%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Elopement: freq 4\n\nSession 8 \u2014 2026-06-17 \u2014 50 min \u2014 1 observer\n doing laundry: 7/14 correct (51%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n wait calmly when a preferred activity is denied: 3/8 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n simple meal preparation: 3/12 correct (27%); latency 5.1s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d75\n Elopement: freq 6\n\nSession 9 \u2014 2026-06-19 \u2014 50 min \u2014 1 observer\n doing laundry: 4/9 correct (42%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n wait calmly when a preferred activity is denied: 4/11 correct (38%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n simple meal preparation: 2/9 correct (19%); latency 5.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n Elopement: freq 6\n\nSession 10 \u2014 2026-06-21 \u2014 50 min \u2014 1 observer\n doing laundry: 5/13 correct (41%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n wait calmly when a preferred activity is denied: 4/12 correct (34%); latency 4.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n simple meal preparation: 2/15 correct (10%); latency 5.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n Elopement: freq 8\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Turning away from task or staff\n- Crying during demand presentation\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (78% -> 41%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 60.3%.\n- Trend direction: downward (first session: 78%, last session: 41%).\n- Elopement: mean 4.2/session (early-window mean 2.0 -> late-window mean 6.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "92ba46c1f5e394ad", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Elopement": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adult", "n_sessions": 10, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "8458"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: unloading the dishwasher\nCurriculum: AFLS Home Skills\nLearner Level: Adolescent Learner\nBaseline: Near mastery\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: unloading the dishwasher\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Open the dishwasher once the cycle is complete. <- Learner-taught step (forward chaining starts here)\n2. Unload the top rack first (cups, glasses, small items).\n3. Put cups and glasses away in the correct cabinet.\n4. Unload the utensil basket.\n5. Put utensils away in the drawer.\n6. Unload the bottom rack (plates, bowls, larger items).\n7. Put plates and bowls away in the correct cabinet.\n8. Close the dishwasher door.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 90s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ac15d726f8d8b745", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "unloading the dishwasher", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "near", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "unloading the dishwasher", "module": "home", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "44637"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands for information using \"what\"\nCurriculum: Mand L2\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for information using \"what\"\n- Curriculum: Mand L2\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult answers the question informatively.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "9bc65664e489d56c", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for information using \"what\"", "mo_category": "mand_info_what", "mo_arrangement": "unexpected_change", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "40151"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: spontaneously initiates conversation with peers\nCurriculum: Spontaneous Vocal Behavior L2\nLearner Level: School-Age Learner\nBaseline: Generalization phase\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously initiates conversation with peers\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: room transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's room transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "5fd6bd510eb6da06", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously initiates conversation with peers", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "40151"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: matches non-identical items by category\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L2\nLearner Level: School-Age Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches non-identical items by category\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: sock, shoe, hat\nDistractor stimuli: brush, phone, jacket, towel\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c9be181d3e3812d2", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches non-identical items by category", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "17197"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects correct item from an array of 2\nCurriculum Reference: Listener Responding L1\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from an array of 2\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: apple, bottle, brush\nDistractor stimuli: spoon, glue\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "38ef29638c95b0c1", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from an array of 2", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "75131"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tolerating tooth brushing by caregiver\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Emerging\nTeaching Method: Toleration / Systematic Desensitization\n\nThe goal is for the learner to allow the activity to be performed (not to perform it independently). Provide the full program structure including shaping steps and refusal/safety procedures."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating tooth brushing by caregiver\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for tooth brushing performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull 2-minute tooth-brushing routine completed by caregiver without refusal, crying, SIB, or aggression\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1-second toothbrush contact with teeth\n2. 3 seconds of brushing\n3. 5 seconds of brushing\n4. 15 seconds of brushing\n5. 30 seconds of brushing\n6. 60 seconds of brushing\n7. Full 2-minute brushing routine\n\nCurrent shaping step: At the emerging stage, begin at shaping step 1 (\"1-second toothbrush contact with teeth\") and advance through step 2 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 2 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fd02b7c9fec2c31c", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating tooth brushing by caregiver", "program_type": "toleration", "learner_profile": "adult", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating tooth brushing by caregiver", "module": "basic_living", "mastery_criterion": "mc_toleration_settings", "shaping_n_steps": 7}, "teacher_model": null, "seed_tag": "50550"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7450\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Social Behavior\nDate range: Sessions 1\u201310 across 20 days (2026-04-26 to 2026-05-16)\n\nACCELERATION PROGRAMS\n1. counts objects 1\u201320 (Math)\n Method: DTT\n2. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-26 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 7/12 correct (55%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 10/14 correct (69%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n takes turns during structured activities: 9/14 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n\nSession 2 \u2014 2026-04-28 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 8/13 correct (63%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d75\n indicating 'all done' when an activity is complete: 5/8 correct (68%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n takes turns during structured activities: 9/14 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n\nSession 3 \u2014 2026-04-30 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 9/14 correct (64%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n indicating 'all done' when an activity is complete: 8/12 correct (67%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n takes turns during structured activities: 5/9 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n\nSession 4 \u2014 2026-05-02 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 9/15 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n indicating 'all done' when an activity is complete: 5/8 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n takes turns during structured activities: 8/14 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n\nSession 5 \u2014 2026-05-04 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 3/12 correct (27%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 4/9 correct (42%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n takes turns during structured activities: 4/13 correct (27%); latency 4.6s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n\nSession 6 \u2014 2026-05-06 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 4/13 correct (32%); latency 4.7s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n indicating 'all done' when an activity is complete: 7/14 correct (50%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n takes turns during structured activities: 3/9 correct (36%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n\nSession 7 \u2014 2026-05-08 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 5/12 correct (41%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 5/9 correct (54%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n takes turns during structured activities: 3/8 correct (41%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n\nSession 8 \u2014 2026-05-10 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 7/15 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 9/14 correct (62%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n takes turns during structured activities: 7/15 correct (44%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n\nSession 9 \u2014 2026-05-12 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 7/13 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n indicating 'all done' when an activity is complete: 10/15 correct (65%); latency 2.8s; prompts FP\u00d75, PP\u00d72, G\u00d72, I\u00d76\n takes turns during structured activities: 8/14 correct (54%); latency 3.6s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n\nSession 10 \u2014 2026-05-14 \u2014 39 min \u2014 1 observer\n counts objects 1\u201320: 9/15 correct (59%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n indicating 'all done' when an activity is complete: 8/12 correct (65%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n takes turns during structured activities: 5/9 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 50.5%.\n- Trend direction: upward (first session: 55%, last session: 59%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "01a9a4d0e97c76a1", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "53478"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: clocking in at start of shift\nCurriculum Reference: AFLS Vocational Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f2ffda94885d3bc3", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_all_steps_independent", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "39765"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1849\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Social Behavior\nDate range: Sessions 1\u20138 across 16 days (2026-07-05 to 2026-07-21)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. requesting a break during non-preferred tasks (Mand)\n Method: NET\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n4. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n5. counts objects 1\u201320 (Math)\n Method: DTT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-05 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 7/9 correct (77%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n requesting a break during non-preferred tasks: 9/11 correct (82%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n takes turns during structured activities: 6/8 correct (70%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/11 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n counts objects 1\u201320: 9/11 correct (85%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n Elopement: freq 2\n\nSession 2 \u2014 2026-07-07 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 10/12 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n requesting a break during non-preferred tasks: 7/9 correct (75%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n takes turns during structured activities: 10/13 correct (74%); latency 2.6s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 13/15 correct (85%); latency 2.3s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d77\n counts objects 1\u201320: 8/9 correct (84%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n Elopement: freq 2\n ABC (elopement): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 3 \u2014 2026-07-09 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 4/10 correct (41%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n requesting a break during non-preferred tasks: 5/12 correct (44%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d77\n takes turns during structured activities: 6/15 correct (37%); latency 4.2s; prompts FP\u00d75, PP\u00d73, G\u00d70, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/10 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n counts objects 1\u201320: 5/11 correct (45%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n Elopement: freq 3\n\nSession 4 \u2014 2026-07-11 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 3/8 correct (39%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n requesting a break during non-preferred tasks: 5/11 correct (47%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n takes turns during structured activities: 4/10 correct (42%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/12 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n counts objects 1\u201320: 4/9 correct (45%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n Elopement: freq 3\n\nSession 5 \u2014 2026-07-13 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 4/10 correct (44%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n requesting a break during non-preferred tasks: 6/14 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d76\n takes turns during structured activities: 5/13 correct (36%); latency 4.2s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/9 correct (48%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n counts objects 1\u201320: 5/10 correct (51%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n Elopement: freq 2\n ABC (elopement): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 6 \u2014 2026-07-15 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 6/14 correct (39%); latency 4.2s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d73\n requesting a break during non-preferred tasks: 6/14 correct (45%); latency 3.9s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n takes turns during structured activities: 6/14 correct (41%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/8 correct (48%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n counts objects 1\u201320: 6/14 correct (43%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n Elopement: freq 3\n\nSession 7 \u2014 2026-07-17 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 6/14 correct (45%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n requesting a break during non-preferred tasks: 6/12 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n takes turns during structured activities: 4/9 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/14 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n counts objects 1\u201320: 8/15 correct (53%); latency 3.7s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n Elopement: freq 3\n\nSession 8 \u2014 2026-07-19 \u2014 55 min \u2014 1 observer\n reads 10\u201320 sight words: 5/10 correct (51%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n requesting a break during non-preferred tasks: 5/9 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n takes turns during structured activities: 5/9 correct (57%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/11 correct (59%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n counts objects 1\u201320: 5/9 correct (59%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n Elopement: freq 3\n ABC (elopement): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n- Looking away during SD presentation\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nElopement: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 52.2%.\n- Trend direction: downward (first session: 77%, last session: 51%).\n- Elopement: mean 2.6/session (early-window mean 2.0 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "7b64026b8d55d001", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Elopement": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "87536"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3922\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Mand + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-09-02 to 2026-09-22)\n\nACCELERATION PROGRAMS\n1. doing laundry (Home Skills)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. requesting a break during extended tasks (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Aggression \u2014 function hypothesized: escape\n2. Pica \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-02 \u2014 39 min \u2014 1 observer\n doing laundry: 10/14 correct (68%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n wait calmly when a preferred activity is denied: 8/14 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n requesting a break during extended tasks: 6/8 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n Aggression: freq 0\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 2 \u2014 2026-09-04 \u2014 39 min \u2014 1 observer\n doing laundry: 7/12 correct (62%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n wait calmly when a preferred activity is denied: 7/11 correct (59%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n requesting a break during extended tasks: 5/8 correct (68%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n Aggression: freq 2\n Pica: attempts 3 (0 unsuccessful \u2014 staff retrieved item before ingestion; 3 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-09-06 \u2014 39 min \u2014 1 observer\n doing laundry: 7/11 correct (64%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n wait calmly when a preferred activity is denied: 5/10 correct (52%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n requesting a break during extended tasks: 8/12 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d78\n Aggression: freq 0\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-09-08 \u2014 39 min \u2014 1 observer\n doing laundry: 6/9 correct (68%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d76\n wait calmly when a preferred activity is denied: 5/9 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n requesting a break during extended tasks: 8/12 correct (68%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n Aggression: freq 2\n Pica: attempts 3 (2 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-09-10 \u2014 39 min \u2014 1 observer\n doing laundry: 3/9 correct (33%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 4/14 correct (29%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d76\n requesting a break during extended tasks: 4/8 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Aggression: freq 4\n Pica: attempts 7 (7 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-09-12 \u2014 39 min \u2014 1 observer\n doing laundry: 4/9 correct (41%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n wait calmly when a preferred activity is denied: 3/10 correct (33%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n requesting a break during extended tasks: 7/12 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n Aggression: freq 4\n Pica: attempts 7 (6 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-09-14 \u2014 39 min \u2014 1 observer\n doing laundry: 5/13 correct (41%); latency 4.3s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n wait calmly when a preferred activity is denied: 3/8 correct (41%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n requesting a break during extended tasks: 7/10 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n Aggression: freq 0\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-09-16 \u2014 39 min \u2014 1 observer\n doing laundry: 6/13 correct (45%); latency 4.2s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 5/12 correct (44%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d77\n requesting a break during extended tasks: 7/10 correct (70%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n Aggression: freq 1\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 9 \u2014 2026-09-18 \u2014 39 min \u2014 1 observer\n doing laundry: 6/12 correct (53%); latency 3.6s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n wait calmly when a preferred activity is denied: 5/9 correct (50%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n requesting a break during extended tasks: 11/15 correct (71%); latency 3.0s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n Aggression: freq 2\n Pica: attempts 3 (1 unsuccessful \u2014 staff retrieved item before ingestion; 2 successful \u2014 item ingested)\n\nSession 10 \u2014 2026-09-20 \u2014 39 min \u2014 1 observer\n doing laundry: 8/13 correct (60%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 5/9 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n requesting a break during extended tasks: 10/14 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n Aggression: freq 0\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Leaving seat without permission\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nAggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nPica: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n### Crisis plan\n- If the setting event involves a medical concern (seizure-like activity, injury, illness): stop the program immediately, follow the facility's medical-incident protocol, and refer to the appropriate medical professional. Medical events are NOT behavioral targets and must not be treated as ABA programming issues.\n- If behavior escalates during the setting-event window: use verbal de-escalation and environmental safety first; do not add demand. Only staff currently certified in the facility's crisis-prevention framework may use physical intervention, and only if imminent safety risk warrants it AND the learner's BIP specifically authorizes it (contraindications and BIP restrictions may rule it out).\n- Coordinate with caregivers, the medical team, and the supervising BCBA; document the setting event and the associated behavior for pattern analysis. Review the BIP in light of the setting-event data.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 53.5%.\n- Trend direction: downward (first session: 68%, last session: 60%).\n- Aggression: mean 1.5/session (early-window mean 0.7 -> late-window mean 1.0).\n- Pica: mean 3.3/session (early-window mean 2.3 -> late-window mean 2.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1101c5df682a1f90", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Aggression": "escape", "Pica": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adult", "n_sessions": 10, "n_programs": 3, "n_behaviors": 2, "behavior_ids": ["aggression", "pica"], "behavior_functions": ["escape", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "86328"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: unloading the dishwasher\nCurriculum: AFLS Home Skills\nLearner: Adolescent Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: unloading the dishwasher\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Open the dishwasher once the cycle is complete. <- Learner-taught step (forward chaining starts here)\n2. Unload the top rack first (cups, glasses, small items).\n3. Put cups and glasses away in the correct cabinet.\n4. Unload the utensil basket.\n5. Put utensils away in the drawer.\n6. Unload the bottom rack (plates, bowls, larger items).\n7. Put plates and bowls away in the correct cabinet.\n8. Close the dishwasher door.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 180s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a9114ef5b73f2211", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "unloading the dishwasher", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "generalization", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "unloading the dishwasher", "module": "home", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "92010"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5711\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Reading + Self-Care\nDate range: Sessions 1\u201312 across 24 days (2026-08-17 to 2026-09-10)\n\nACCELERATION PROGRAMS\n1. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n2. ordering in a restaurant (Community)\n Method: Task Analysis\n3. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n4. reads 10\u201320 sight words (Reading)\n Method: DTT\n5. sounds out CVC words (Reading)\n Method: DTT\n6. requesting preferred items using AAC device (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-17 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 10/12 correct (81%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d75\n ordering in a restaurant: 11/14 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d711\n meal cleanup after eating: 8/10 correct (76%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n reads 10\u201320 sight words: 7/9 correct (77%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n sounds out CVC words: 8/10 correct (76%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n requesting preferred items using AAC device: 7/9 correct (77%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n Self-injurious behavior (SIB): freq 2\n\nSession 2 \u2014 2026-08-19 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 9/12 correct (77%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d78\n ordering in a restaurant: 8/10 correct (78%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n meal cleanup after eating: 5/8 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 7/10 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n sounds out CVC words: 7/10 correct (72%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n requesting preferred items using AAC device: 9/13 correct (72%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d74\n Self-injurious behavior (SIB): freq 2\n\nSession 3 \u2014 2026-08-21 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 8/12 correct (70%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n ordering in a restaurant: 7/9 correct (73%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n meal cleanup after eating: 8/14 correct (58%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n reads 10\u201320 sight words: 7/10 correct (67%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n sounds out CVC words: 5/9 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n requesting preferred items using AAC device: 7/10 correct (69%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 2\n\nSession 4 \u2014 2026-08-23 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 6/10 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n ordering in a restaurant: 5/8 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n meal cleanup after eating: 7/12 correct (56%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n reads 10\u201320 sight words: 5/8 correct (57%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n sounds out CVC words: 6/12 correct (53%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n requesting preferred items using AAC device: 6/9 correct (70%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n Self-injurious behavior (SIB): freq 2\n\nSession 5 \u2014 2026-08-25 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 9/15 correct (61%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d712\n ordering in a restaurant: 5/9 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n meal cleanup after eating: 7/15 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n reads 10\u201320 sight words: 4/8 correct (54%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n sounds out CVC words: 6/12 correct (53%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n requesting preferred items using AAC device: 8/13 correct (65%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n Self-injurious behavior (SIB): freq 4\n\nSession 6 \u2014 2026-08-27 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 6/12 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n ordering in a restaurant: 6/12 correct (53%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n meal cleanup after eating: 4/12 correct (37%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d79\n reads 10\u201320 sight words: 5/11 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n sounds out CVC words: 3/8 correct (44%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n requesting preferred items using AAC device: 6/11 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-08-29 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 6/14 correct (46%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n ordering in a restaurant: 5/12 correct (40%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n meal cleanup after eating: 4/10 correct (38%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 4/8 correct (46%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n sounds out CVC words: 4/11 correct (37%); latency 4.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n requesting preferred items using AAC device: 8/15 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n Self-injurious behavior (SIB): freq 4\n\nSession 8 \u2014 2026-08-31 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 4/12 correct (37%); latency 4.6s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d75\n ordering in a restaurant: 5/15 correct (35%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n meal cleanup after eating: 4/15 correct (30%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n reads 10\u201320 sight words: 3/8 correct (40%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n sounds out CVC words: 2/9 correct (22%); latency 4.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n requesting preferred items using AAC device: 7/13 correct (50%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d75\n Self-injurious behavior (SIB): freq 5\n\nSession 9 \u2014 2026-09-02 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 4/12 correct (36%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n ordering in a restaurant: 3/10 correct (33%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n meal cleanup after eating: 1/8 correct (17%); latency 5.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 3/11 correct (31%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n sounds out CVC words: 2/11 correct (14%); latency 5.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n requesting preferred items using AAC device: 4/8 correct (46%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n Self-injurious behavior (SIB): freq 4\n\nSession 10 \u2014 2026-09-04 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 4/13 correct (31%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n ordering in a restaurant: 3/12 correct (25%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n meal cleanup after eating: 1/12 correct (11%); latency 5.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 2/9 correct (25%); latency 5.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n sounds out CVC words: 1/9 correct (11%); latency 5.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n requesting preferred items using AAC device: 6/13 correct (43%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n Self-injurious behavior (SIB): freq 5\n\nSession 11 \u2014 2026-09-06 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 2/10 correct (20%); latency 5.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n ordering in a restaurant: 3/14 correct (19%); latency 5.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n meal cleanup after eating: 2/14 correct (12%); latency 5.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n reads 10\u201320 sight words: 2/14 correct (18%); latency 5.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d712\n sounds out CVC words: 1/8 correct (10%); latency 5.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n requesting preferred items using AAC device: 5/15 correct (36%); latency 4.2s; prompts FP\u00d72, PP\u00d74, G\u00d75, I\u00d74\n Self-injurious behavior (SIB): freq 5\n\nSession 12 \u2014 2026-09-08 \u2014 37 min \u2014 1 observer\n dressing (pants, shirt): 2/15 correct (13%); latency 5.3s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d78\n ordering in a restaurant: 2/14 correct (12%); latency 5.2s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n meal cleanup after eating: 1/11 correct (10%); latency 5.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n reads 10\u201320 sight words: 2/10 correct (18%); latency 5.3s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n sounds out CVC words: 1/14 correct (10%); latency 5.3s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n requesting preferred items using AAC device: 4/11 correct (34%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 5\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Crying during demand presentation\n- Peer aggression (hitting or pushing peers during group activities)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 81% to 13% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 48.6%.\n- Trend direction: downward (first session: 81%, last session: 13%).\n- Self-injurious behavior (SIB): mean 3.6/session (early-window mean 2.0 -> late-window mean 4.8).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "963373482b529468", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Self-injurious behavior (SIB)": "attention"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "44149"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: sorts by multiple attributes simultaneously\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L3\nLearner: School-Age Learner\nMastery Status: Approaching mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: sorts by multiple attributes simultaneously\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L3\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: scissors, napkin, block\nDistractor stimuli: sock, ball, backpack, glue, jacket\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "4e0130eb311b49fb", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L3", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "sorts by multiple attributes simultaneously", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "5667"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts colors of objects\nCurriculum: Tact L2\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts colors of objects\n- Curriculum: Tact L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What color?\"\nVariations: \"What color is this?\"; \"Tell me the color\"\nPresentation: Present a single-color object and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: pink, black, green\nDistractor stimuli: brown, purple, white, orange\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "28293e5b40be667e", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts colors of objects", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "87673"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: maintains friendships over time\nCurriculum: Social Behavior & Social Play L3\nLearner: School-Age Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: maintains friendships over time\n- Curriculum: Social Behavior & Social Play L3\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "70c3671e3d3a2db7", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L3", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "maintains friendships over time", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "46838"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1511\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Mand + Reading + Tact + Tolerating Denied Access\nDate range: Sessions 1\u20138 across 16 days (2026-08-14 to 2026-08-30)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. sounds out CVC words (Reading)\n Method: DTT\n3. tacts colors of objects (Tact)\n Method: DTT\n4. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n5. reads 10\u201320 sight words (Reading)\n Method: DTT\n6. echoes 2-word phrases (Echoic)\n Method: DTT variant\n\nDECELERATION TARGETS\n1. Fecal smearing (scatolia) \u2014 function hypothesized: attention\n2. Tantrum \u2014 function hypothesized: tangible\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-14 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/8 correct (65%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n sounds out CVC words: 7/11 correct (62%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n tacts colors of objects: 8/12 correct (68%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n reads 10\u201320 sight words: 6/9 correct (63%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n echoes 2-word phrases: 7/12 correct (62%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n Fecal smearing (scatolia): attempts 3 (1 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 2, duration 8m total\n\nSession 2 \u2014 2026-08-16 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/13 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n sounds out CVC words: 5/9 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n tacts colors of objects: 6/8 correct (70%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/9 correct (50%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n reads 10\u201320 sight words: 7/10 correct (66%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n echoes 2-word phrases: 7/12 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d78\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 2, duration 6m total\n\nSession 3 \u2014 2026-08-18 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/11 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n sounds out CVC words: 9/15 correct (59%); latency 3.5s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n tacts colors of objects: 10/14 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (55%); latency 3.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d79\n reads 10\u201320 sight words: 8/11 correct (69%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n echoes 2-word phrases: 5/9 correct (56%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n Fecal smearing (scatolia): attempts 8 (5 intercepted \u2014 staff redirected before smearing; 3 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 8, duration 24m total\n\nSession 4 \u2014 2026-08-20 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/11 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n sounds out CVC words: 6/10 correct (65%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n tacts colors of objects: 7/9 correct (74%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (56%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d710\n reads 10\u201320 sight words: 6/8 correct (70%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n echoes 2-word phrases: 6/11 correct (58%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n Fecal smearing (scatolia): attempts 8 (4 intercepted \u2014 staff redirected before smearing; 4 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 7, duration 35m total\n\nSession 5 \u2014 2026-08-22 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/10 correct (69%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n sounds out CVC words: 10/14 correct (68%); latency 3.2s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n tacts colors of objects: 9/12 correct (77%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/9 correct (59%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 10/15 correct (65%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d711\n echoes 2-word phrases: 5/8 correct (61%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Fecal smearing (scatolia): attempts 3 (1 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 3, duration 12m total\n\nSession 6 \u2014 2026-08-24 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/11 correct (66%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n sounds out CVC words: 10/15 correct (68%); latency 3.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d79\n tacts colors of objects: 10/14 correct (71%); latency 2.7s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/13 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d710\n reads 10\u201320 sight words: 6/8 correct (70%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n echoes 2-word phrases: 8/14 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n Fecal smearing (scatolia): attempts 3 (1 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 3, duration 6m total\n\nSession 7 \u2014 2026-08-26 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/14 correct (71%); latency 2.6s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n sounds out CVC words: 8/11 correct (73%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n tacts colors of objects: 11/14 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/11 correct (71%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 10/14 correct (69%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n echoes 2-word phrases: 8/14 correct (56%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 2, duration 10m total\n\nSession 8 \u2014 2026-08-28 \u2014 43 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/12 correct (65%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n sounds out CVC words: 12/15 correct (78%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d79\n tacts colors of objects: 12/14 correct (84%); latency 2.5s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/15 correct (70%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d79\n reads 10\u201320 sight words: 7/10 correct (73%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n echoes 2-word phrases: 6/11 correct (54%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 3, duration 6m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Crying during demand presentation\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nFecal smearing (scatolia): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\nTantrum: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 65.5%.\n- Trend direction: downward (first session: 65%, last session: 65%).\n- Fecal smearing (scatolia): mean 4.1/session (early-window mean 3.0 -> late-window mean 2.5).\n- Tantrum: mean 3.8/session (early-window mean 2.0 -> late-window mean 2.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "af653837865a040f", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Fecal smearing (scatolia)": "attention", "Tantrum": "tangible"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "early", "n_sessions": 8, "n_programs": 6, "n_behaviors": 2, "behavior_ids": ["fecal_smearing", "tantrum"], "behavior_functions": ["attention", "tangible"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "89331"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6823\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Motor Imitation + Reading + Tact + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-10-07 to 2026-10-27)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. reads 10\u201320 sight words (Reading)\n Method: DTT\n3. tacts common objects (Tact)\n Method: DTT\n4. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n5. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n6. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-07 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/15 correct (33%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 5/14 correct (39%); latency 4.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d74\n tacts common objects: 4/8 correct (45%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 5/13 correct (39%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n imitates 3-step motor sequences: 5/14 correct (38%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n follows 1-step motor instructions: 4/11 correct (37%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n\nSession 2 \u2014 2026-10-09 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/14 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n reads 10\u201320 sight words: 4/8 correct (48%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n tacts common objects: 7/13 correct (52%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 6/14 correct (46%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n imitates 3-step motor sequences: 6/15 correct (41%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d77\n follows 1-step motor instructions: 6/13 correct (43%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n\nSession 3 \u2014 2026-10-11 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/15 correct (43%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d75, I\u00d79\n reads 10\u201320 sight words: 6/11 correct (53%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n tacts common objects: 5/9 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 6/11 correct (56%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n imitates 3-step motor sequences: 5/13 correct (41%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d79\n follows 1-step motor instructions: 5/11 correct (41%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n\nSession 4 \u2014 2026-10-13 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/8 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 5/8 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n tacts common objects: 9/15 correct (61%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d713\n requesting a break ('break please' or AAC 'I need a break'): 6/10 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n imitates 3-step motor sequences: 4/8 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n follows 1-step motor instructions: 5/10 correct (51%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n\nSession 5 \u2014 2026-10-15 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/11 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 7/11 correct (59%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n tacts common objects: 7/10 correct (74%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 6/10 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n imitates 3-step motor sequences: 4/8 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n follows 1-step motor instructions: 8/14 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n\nSession 6 \u2014 2026-10-17 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/11 correct (65%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n reads 10\u201320 sight words: 9/14 correct (67%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d710\n tacts common objects: 7/9 correct (78%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n requesting a break ('break please' or AAC 'I need a break'): 8/12 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d78\n imitates 3-step motor sequences: 4/8 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n follows 1-step motor instructions: 6/10 correct (58%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n\nSession 7 \u2014 2026-10-19 \u2014 IOA SESSION \u2014 2 observers\n wait calmly when access to preferred item is denied: 82% (primary); IOA 72% agreement\n reads 10\u201320 sight words: 72% (primary); IOA 72% agreement\n tacts common objects: 87% (primary); IOA 72% agreement\n requesting a break ('break please' or AAC 'I need a break'): 70% (primary); IOA 72% agreement\n imitates 3-step motor sequences: 66% (primary); IOA 72% agreement\n follows 1-step motor instructions: 60% (primary); IOA 72% agreement\n\nSession 8 \u2014 2026-10-21 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/11 correct (87%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n reads 10\u201320 sight words: 8/12 correct (68%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n tacts common objects: 11/13 correct (88%); latency 2.1s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 7/9 correct (79%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n imitates 3-step motor sequences: 9/12 correct (73%); latency 2.5s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n follows 1-step motor instructions: 8/11 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n\nSession 9 \u2014 2026-10-23 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 13/14 correct (91%); latency 1.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n reads 10\u201320 sight words: 11/14 correct (75%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d77\n tacts common objects: 8/8 correct (95%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 10/13 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n imitates 3-step motor sequences: 8/11 correct (75%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n follows 1-step motor instructions: 6/8 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n\nSession 10 \u2014 2026-10-25 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/12 correct (95%); latency 1.6s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d74\n reads 10\u201320 sight words: 8/10 correct (77%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n tacts common objects: 13/14 correct (95%); latency 1.7s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 10/12 correct (85%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n imitates 3-step motor sequences: 7/9 correct (80%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n follows 1-step motor instructions: 9/12 correct (75%); latency 2.7s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Appropriate sitting posture maintained\n- Smiling during reinforcement delivery\n- Eye contact maintained throughout trials"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 33% to 95% across 10 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 65.2%.\n- Trend direction: upward (first session: 33%, last session: 95%).\n- IOA data: 72% agreement on IOA session \u2014 below threshold."}], "meta": {"task_type": "session_interpretation", "example_id": "681c4cb017f3a621", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "early", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "15583"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5982\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand + Reading\nDate range: Sessions 1\u20138 across 16 days (2026-10-05 to 2026-10-21)\n\nACCELERATION PROGRAMS\n1. selects items by feature (Listener Responding)\n Method: DTT\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n5. sounds out CVC words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-05 \u2014 37 min \u2014 1 observer\n selects items by feature: 7/10 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n answers WH-questions about familiar topics: 7/10 correct (72%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/15 correct (83%); latency 2.3s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d710\n requesting a break ('break please' or AAC 'I need a break'): 12/14 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d76\n sounds out CVC words: 10/13 correct (80%); latency 2.3s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n Elopement: freq 3\n\nSession 2 \u2014 2026-10-07 \u2014 37 min \u2014 1 observer\n selects items by feature: 9/13 correct (69%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n answers WH-questions about familiar topics: 9/14 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (73%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 7/10 correct (74%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n sounds out CVC words: 9/12 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n Elopement: freq 2\n\nSession 3 \u2014 2026-10-09 \u2014 37 min \u2014 1 observer\n selects items by feature: 7/11 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n answers WH-questions about familiar topics: 9/14 correct (64%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (67%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (70%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n sounds out CVC words: 6/8 correct (79%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n Elopement: freq 3\n\nSession 4 \u2014 2026-10-11 \u2014 37 min \u2014 1 observer\n selects items by feature: 8/13 correct (62%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n answers WH-questions about familiar topics: 6/9 correct (66%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/14 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 9/14 correct (62%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n sounds out CVC words: 8/11 correct (70%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n Elopement: freq 5\n\nSession 5 \u2014 2026-10-13 \u2014 37 min \u2014 1 observer\n selects items by feature: 4/9 correct (48%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n answers WH-questions about familiar topics: 5/8 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (60%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 5/10 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n sounds out CVC words: 9/14 correct (64%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n Elopement: freq 6\n\nSession 6 \u2014 2026-10-15 \u2014 37 min \u2014 1 observer\n selects items by feature: 7/14 correct (51%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d710\n answers WH-questions about familiar topics: 5/9 correct (52%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/15 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d75, G\u00d74, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 6/12 correct (54%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d76\n sounds out CVC words: 5/9 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n Elopement: freq 6\n\nSession 7 \u2014 2026-10-17 \u2014 37 min \u2014 1 observer\n selects items by feature: 3/8 correct (40%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n answers WH-questions about familiar topics: 5/10 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/15 correct (43%); latency 4.3s; prompts FP\u00d75, PP\u00d70, G\u00d75, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 5/12 correct (40%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n sounds out CVC words: 7/13 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n Elopement: freq 6\n\nSession 8 \u2014 2026-10-19 \u2014 37 min \u2014 1 observer\n selects items by feature: 5/11 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n answers WH-questions about familiar topics: 4/10 correct (43%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/13 correct (44%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 4/11 correct (32%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n sounds out CVC words: 5/10 correct (49%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n Elopement: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Crying during demand presentation\n- Vocal refusal ('no', 'I don't want to')\n- Pushing materials away from the work area\n- Increased response latency (3\u20135\u00d7 baseline)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (68% -> 43%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 55.8%.\n- Trend direction: downward (first session: 68%, last session: 43%).\n- Elopement: mean 4.6/session (early-window mean 2.5 -> late-window mean 6.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "28555857785d8a78", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Elopement": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "early", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "35802"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5338\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Social Behavior + Writing\nDate range: Sessions 1\u20139 across 18 days (2026-08-05 to 2026-08-23)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. identifies numerals 1\u201320 (Math)\n Method: DTT\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n4. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Aggression \u2014 function hypothesized: tangible\n2. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-05 \u2014 54 min \u2014 1 observer\n copies simple words from model: 5/8 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n identifies numerals 1\u201320: 5/8 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n takes turns during structured activities: 7/14 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/12 correct (61%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (60%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Aggression: freq 2\n Non-compliance: freq 2\n\nSession 2 \u2014 2026-08-07 \u2014 54 min \u2014 1 observer\n copies simple words from model: 5/9 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n identifies numerals 1\u201320: 10/14 correct (69%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n takes turns during structured activities: 6/11 correct (53%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/12 correct (59%); latency 3.2s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n Aggression: freq 1\n Non-compliance: freq 2\n\nSession 3 \u2014 2026-08-09 \u2014 54 min \u2014 1 observer\n copies simple words from model: 7/11 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n identifies numerals 1\u201320: 10/13 correct (75%); latency 2.7s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n takes turns during structured activities: 6/11 correct (51%); latency 4.0s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/10 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n Aggression: freq 4\n Non-compliance: freq 9\n\nSession 4 \u2014 2026-08-11 \u2014 54 min \u2014 1 observer\n copies simple words from model: 6/10 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n identifies numerals 1\u201320: 8/12 correct (70%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d76\n takes turns during structured activities: 6/10 correct (59%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/14 correct (63%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (60%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n Aggression: freq 4\n Non-compliance: freq 9\n\nSession 5 \u2014 2026-08-13 \u2014 54 min \u2014 1 observer\n copies simple words from model: 8/13 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n identifies numerals 1\u201320: 9/13 correct (72%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n takes turns during structured activities: 9/15 correct (57%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d710\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/11 correct (66%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (64%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n Aggression: freq 1\n Non-compliance: freq 4\n\nSession 6 \u2014 2026-08-15 \u2014 54 min \u2014 1 observer\n copies simple words from model: 10/15 correct (67%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d75, I\u00d75\n identifies numerals 1\u201320: 9/12 correct (78%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n takes turns during structured activities: 8/14 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/12 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (65%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d78\n Aggression: freq 2\n Non-compliance: freq 3\n\nSession 7 \u2014 2026-08-17 \u2014 54 min \u2014 1 observer\n copies simple words from model: 7/11 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n identifies numerals 1\u201320: 11/15 correct (71%); latency 3.0s; prompts FP\u00d75, PP\u00d72, G\u00d72, I\u00d76\n takes turns during structured activities: 7/13 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 10/14 correct (70%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/12 correct (69%); latency 2.7s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n Aggression: freq 1\n Non-compliance: freq 2\n\nSession 8 \u2014 2026-08-19 \u2014 54 min \u2014 1 observer\n copies simple words from model: 6/8 correct (72%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n identifies numerals 1\u201320: 6/9 correct (71%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n takes turns during structured activities: 9/14 correct (62%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d712\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 11/15 correct (72%); latency 2.9s; prompts FP\u00d75, PP\u00d70, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (68%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n Aggression: freq 2\n Non-compliance: freq 3\n\nSession 9 \u2014 2026-08-21 \u2014 54 min \u2014 1 observer\n copies simple words from model: 7/10 correct (69%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n identifies numerals 1\u201320: 9/12 correct (77%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d77\n takes turns during structured activities: 7/12 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/8 correct (70%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n Aggression: freq 1\n Non-compliance: freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Peer aggression (hitting or pushing peers during group activities)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nAggression: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 63.0%.\n- Trend direction: upward (first session: 58%, last session: 69%).\n- Aggression: mean 2.0/session (early-window mean 2.3 -> late-window mean 1.3).\n- Non-compliance: mean 4.1/session (early-window mean 4.3 -> late-window mean 2.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "9d3b3b903243d101", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Aggression": "tangible", "Non-compliance": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "school_age", "n_sessions": 9, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["aggression", "non_compliance"], "behavior_functions": ["tangible", "escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "89484"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: responds to own name\nCurriculum: Social Behavior & Social Play L1\nLearner Level: Early Learner\nBaseline: Mastered at current level\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: responds to own name\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "0ee3cb6bf7eb71a4", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "responds to own name", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "87353"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by function (What do you drink from?)\nCurriculum: LRFFC L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function (What do you drink from?)\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: ball, napkin, brush\nDistractor stimuli: glue, phone\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e75c0e20c67cd9a1", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function (What do you drink from?)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "60118"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously tells stories or narratives\nCurriculum: Spontaneous Vocal Behavior L3\nLearner: School-Age Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously tells stories or narratives\n- Curriculum: Spontaneous Vocal Behavior L3\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "9207dde2f9f1ddb0", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L3", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously tells stories or narratives", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "75835"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4656\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Reading + Self-Care + Writing\nDate range: Sessions 1\u201310 across 20 days (2026-06-10 to 2026-06-30)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. copies simple words from model (Writing)\n Method: DTT\n3. identifying community safety signs (Community)\n Method: DTT\n4. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-10 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 7/9 correct (80%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n copies simple words from model: 8/10 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n identifying community safety signs: 11/14 correct (79%); latency 2.4s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d75\n tolerating tooth brushing by caregiver: 9/13 correct (71%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n Self-injurious behavior (SIB): freq 2\n\nSession 2 \u2014 2026-06-12 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 9/12 correct (77%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n copies simple words from model: 9/11 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n identifying community safety signs: 9/11 correct (79%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n tolerating tooth brushing by caregiver: 10/15 correct (69%); latency 3.0s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d76\n Self-injurious behavior (SIB): freq 3\n\nSession 3 \u2014 2026-06-14 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 9/11 correct (80%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n copies simple words from model: 12/14 correct (84%); latency 2.2s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n identifying community safety signs: 10/12 correct (84%); latency 2.1s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d72\n tolerating tooth brushing by caregiver: 7/11 correct (68%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 2\n\nSession 4 \u2014 2026-06-16 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 6/13 correct (47%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n copies simple words from model: 5/11 correct (43%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n identifying community safety signs: 5/10 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n tolerating tooth brushing by caregiver: 5/14 correct (39%); latency 4.4s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 2\n\nSession 5 \u2014 2026-06-18 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 5/11 correct (49%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n copies simple words from model: 6/13 correct (44%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n identifying community safety signs: 5/13 correct (42%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n tolerating tooth brushing by caregiver: 5/12 correct (38%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n Self-injurious behavior (SIB): freq 2\n\nSession 6 \u2014 2026-06-20 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 6/11 correct (50%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n copies simple words from model: 6/14 correct (40%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n identifying community safety signs: 3/8 correct (42%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n tolerating tooth brushing by caregiver: 4/9 correct (39%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-06-22 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 7/13 correct (55%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n copies simple words from model: 7/15 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n identifying community safety signs: 4/8 correct (45%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n tolerating tooth brushing by caregiver: 4/9 correct (44%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n Self-injurious behavior (SIB): freq 3\n\nSession 8 \u2014 2026-06-24 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 9/15 correct (60%); latency 3.3s; prompts FP\u00d75, PP\u00d72, G\u00d72, I\u00d76\n copies simple words from model: 7/13 correct (51%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n identifying community safety signs: 6/13 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n tolerating tooth brushing by caregiver: 7/13 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 2\n ABC (self-injurious behavior (sib)): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 9 \u2014 2026-06-26 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 8/12 correct (68%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n copies simple words from model: 5/8 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n identifying community safety signs: 7/13 correct (57%); latency 3.7s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n tolerating tooth brushing by caregiver: 5/8 correct (64%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n Self-injurious behavior (SIB): freq 2\n ABC (self-injurious behavior (sib)): A = waiting in line transition; B = mouthed own hand; C = staff redirected to task materials\n\nSession 10 \u2014 2026-06-28 \u2014 32 min \u2014 1 observer\n sounds out CVC words: 9/11 correct (77%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n copies simple words from model: 7/11 correct (65%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n identifying community safety signs: 5/9 correct (61%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n tolerating tooth brushing by caregiver: 8/11 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n Self-injurious behavior (SIB): freq 2\n ABC (self-injurious behavior (sib)): A = waiting in line transition; B = mouthed own hand; C = staff redirected to task materials\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 64.3%.\n- Trend direction: downward (first session: 80%, last session: 77%).\n- Self-injurious behavior (SIB): mean 2.3/session (early-window mean 2.3 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "d4ab433eabbf7259", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Self-injurious behavior (SIB)": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "76861"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts common animals\nCurriculum: Tact L1\nLearner Level: Early Learner\nBaseline: Generalization phase\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common animals\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What animal is this?\"\nVariations: \"What is this?\"; \"Name this animal\"\nPresentation: Present an animal picture card and ask while pointing.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: bear, cow, frog\nDistractor stimuli: duck, dog\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ed95cdbf660ddf42", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common animals", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "four_step", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "15203"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5464\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math\nDate range: Sessions 1\u201310 across 20 days (2026-06-24 to 2026-07-14)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. counts objects 1\u201320 (Math)\n Method: DTT\n4. requesting preferred items using a full sentence (Mand)\n Method: NET\n5. identifies numerals 1\u201320 (Math)\n Method: DTT\n6. requesting a break during non-preferred tasks (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-24 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/10 correct (39%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n single-word requests for preferred items: 3/10 correct (33%); latency 4.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n counts objects 1\u201320: 3/8 correct (31%); latency 4.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n requesting preferred items using a full sentence: 5/11 correct (44%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n identifies numerals 1\u201320: 3/9 correct (30%); latency 4.6s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n requesting a break during non-preferred tasks: 4/11 correct (41%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 2 \u2014 2026-06-26 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/12 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n single-word requests for preferred items: 3/9 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n counts objects 1\u201320: 6/15 correct (42%); latency 4.1s; prompts FP\u00d75, PP\u00d75, G\u00d72, I\u00d73\n requesting preferred items using a full sentence: 6/12 correct (51%); latency 3.9s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d75\n identifies numerals 1\u201320: 4/9 correct (42%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n requesting a break during non-preferred tasks: 6/13 correct (47%); latency 3.7s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-06-28 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/12 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d76\n single-word requests for preferred items: 4/11 correct (38%); latency 4.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n counts objects 1\u201320: 5/12 correct (42%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n requesting preferred items using a full sentence: 5/8 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n identifies numerals 1\u201320: 6/13 correct (44%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d77\n requesting a break during non-preferred tasks: 5/11 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n Toileting accident (urine or bowel): urine: 5 in-toilet / 1 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 4 \u2014 2026-06-30 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (60%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n single-word requests for preferred items: 5/11 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n counts objects 1\u201320: 5/11 correct (50%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n requesting preferred items using a full sentence: 8/11 correct (68%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n identifies numerals 1\u201320: 5/11 correct (48%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n requesting a break during non-preferred tasks: 8/14 correct (57%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n Toileting accident (urine or bowel): urine: 5 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 5 \u2014 2026-07-02 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (58%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d79\n single-word requests for preferred items: 7/13 correct (57%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n counts objects 1\u201320: 6/13 correct (50%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n requesting preferred items using a full sentence: 10/14 correct (73%); latency 2.5s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d74\n identifies numerals 1\u201320: 9/15 correct (57%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n requesting a break during non-preferred tasks: 5/8 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 6 \u2014 2026-07-04 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/9 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n single-word requests for preferred items: 9/15 correct (59%); latency 3.5s; prompts FP\u00d70, PP\u00d74, G\u00d75, I\u00d76\n counts objects 1\u201320: 5/9 correct (51%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n requesting preferred items using a full sentence: 10/13 correct (77%); latency 2.4s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n identifies numerals 1\u201320: 9/14 correct (64%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d77\n requesting a break during non-preferred tasks: 8/13 correct (62%); latency 2.9s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 7 \u2014 2026-07-06 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/12 correct (75%); latency 2.5s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n single-word requests for preferred items: 9/15 correct (61%); latency 3.0s; prompts FP\u00d72, PP\u00d75, G\u00d71, I\u00d77\n counts objects 1\u201320: 8/13 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n requesting preferred items using a full sentence: 8/10 correct (85%); latency 1.9s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n identifies numerals 1\u201320: 6/9 correct (62%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n requesting a break during non-preferred tasks: 8/11 correct (77%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 8 \u2014 2026-07-08 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/13 correct (80%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d79\n single-word requests for preferred items: 9/14 correct (67%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n counts objects 1\u201320: 6/10 correct (62%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n requesting preferred items using a full sentence: 9/10 correct (92%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n identifies numerals 1\u201320: 6/8 correct (75%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n requesting a break during non-preferred tasks: 11/14 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n Toileting accident (urine or bowel): urine: 2 in-toilet / 1 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 9 \u2014 2026-07-10 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 13/14 correct (93%); latency 1.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n single-word requests for preferred items: 6/8 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n counts objects 1\u201320: 7/10 correct (69%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n requesting preferred items using a full sentence: 10/11 correct (95%); latency 1.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n identifies numerals 1\u201320: 6/8 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n requesting a break during non-preferred tasks: 11/13 correct (81%); latency 2.1s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 1 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 10 \u2014 2026-07-12 \u2014 54 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/10 correct (95%); latency 2.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n single-word requests for preferred items: 9/11 correct (82%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n counts objects 1\u201320: 7/9 correct (79%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n requesting preferred items using a full sentence: 11/12 correct (95%); latency 1.7s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n identifies numerals 1\u201320: 8/9 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n requesting a break during non-preferred tasks: 11/13 correct (86%); latency 2.1s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Quick response latency (within 2\u20133 seconds)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 39% to 95% across 10 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 65.8%.\n- Trend direction: upward (first session: 39%, last session: 95%).\n- Toileting accident (urine or bowel): mean 1.5/session (early-window mean 1.0 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a9bd3c2f58eee9d8", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {"Toileting accident (urine or bowel)": "automatic"}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "68883"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9052\nProfile: Adult Learner (18+ years)\nCurricula: Community + Home Skills + Independent Living + Mand + Vocational\nDate range: Sessions 1\u201311 across 22 days (2026-08-11 to 2026-09-02)\n\nACCELERATION PROGRAMS\n1. using public transportation (Community)\n Method: Task Analysis\n2. managing a budget (Independent Living)\n Method: Task Analysis\n3. simple meal preparation (Home Skills)\n Method: Task Analysis\n4. requesting assistance using AAC device (Mand)\n Method: NET\n5. clocking in at start of shift (Vocational)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-11 \u2014 31 min \u2014 1 observer\n using public transportation: 13/15 correct (88%); latency 2.1s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d75\n managing a budget: 11/15 correct (73%); latency 3.0s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d74\n simple meal preparation: 12/15 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d710\n requesting assistance using AAC device: 12/15 correct (81%); latency 2.4s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d710\n clocking in at start of shift: 8/9 correct (86%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n Tantrum: freq 3, duration 6m total\n\nSession 2 \u2014 2026-08-13 \u2014 31 min \u2014 1 observer\n using public transportation: 3/14 correct (21%); latency 5.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n managing a budget: 6/13 correct (43%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n simple meal preparation: 5/13 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n requesting assistance using AAC device: 6/15 correct (43%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d712\n clocking in at start of shift: 3/11 correct (30%); latency 4.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n Tantrum: freq 3, duration 15m total\n\nSession 3 \u2014 2026-08-15 \u2014 31 min \u2014 1 observer\n using public transportation: 7/8 correct (87%); latency 2.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n managing a budget: 8/10 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n simple meal preparation: 8/9 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n requesting assistance using AAC device: 12/14 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d76\n clocking in at start of shift: 10/11 correct (90%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n Tantrum: freq 3, duration 6m total\n\nSession 4 \u2014 2026-08-17 \u2014 31 min \u2014 1 observer\n using public transportation: 3/13 correct (23%); latency 4.9s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n managing a budget: 3/10 correct (35%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n simple meal preparation: 5/15 correct (33%); latency 4.4s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d77\n requesting assistance using AAC device: 4/9 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n clocking in at start of shift: 4/11 correct (37%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n Tantrum: freq 2, duration 4m total\n\nSession 5 \u2014 2026-08-19 \u2014 31 min \u2014 1 observer\n using public transportation: 12/14 correct (85%); latency 2.2s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n managing a budget: 11/15 correct (74%); latency 2.4s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d74\n simple meal preparation: 7/9 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n requesting assistance using AAC device: 7/9 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n clocking in at start of shift: 9/11 correct (83%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n Tantrum: freq 2, duration 6m total\n\nSession 6 \u2014 2026-08-21 \u2014 31 min \u2014 1 observer\n using public transportation: 2/9 correct (27%); latency 5.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n managing a budget: 6/15 correct (39%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d79\n simple meal preparation: 3/9 correct (34%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n requesting assistance using AAC device: 5/13 correct (40%); latency 4.1s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n clocking in at start of shift: 4/12 correct (35%); latency 4.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n Tantrum: freq 3, duration 6m total\n\nSession 7 \u2014 2026-08-23 \u2014 31 min \u2014 1 observer\n using public transportation: 7/8 correct (86%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n managing a budget: 8/10 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n simple meal preparation: 9/11 correct (83%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n requesting assistance using AAC device: 10/12 correct (80%); latency 2.5s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n clocking in at start of shift: 7/8 correct (89%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n Tantrum: freq 2, duration 4m total\n\nSession 8 \u2014 2026-08-25 \u2014 31 min \u2014 1 observer\n using public transportation: 2/9 correct (21%); latency 5.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n managing a budget: 6/14 correct (40%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d710\n simple meal preparation: 4/11 correct (33%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n requesting assistance using AAC device: 6/15 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d711\n clocking in at start of shift: 5/15 correct (35%); latency 4.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n Tantrum: freq 2, duration 8m total\n\nSession 9 \u2014 2026-08-27 \u2014 31 min \u2014 1 observer\n using public transportation: 13/14 correct (95%); latency 1.7s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d73\n managing a budget: 6/8 correct (74%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n simple meal preparation: 7/9 correct (77%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n requesting assistance using AAC device: 10/13 correct (78%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n clocking in at start of shift: 8/9 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n Tantrum: freq 2, duration 10m total\n\nSession 10 \u2014 2026-08-29 \u2014 31 min \u2014 1 observer\n using public transportation: 3/11 correct (24%); latency 4.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n managing a budget: 5/13 correct (35%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n simple meal preparation: 4/11 correct (32%); latency 4.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n requesting assistance using AAC device: 5/12 correct (42%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d76\n clocking in at start of shift: 4/11 correct (38%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n Tantrum: freq 3, duration 12m total\n\nSession 11 \u2014 2026-08-31 \u2014 31 min \u2014 1 observer\n using public transportation: 7/8 correct (91%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n managing a budget: 9/12 correct (73%); latency 3.0s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n simple meal preparation: 7/8 correct (83%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n requesting assistance using AAC device: 7/8 correct (82%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n clocking in at start of shift: 8/9 correct (89%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n Tantrum: freq 1, duration 3m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 58.9%.\n- Trend direction: upward (first session: 88%, last session: 91%).\n- Tantrum: mean 2.4/session (early-window mean 3.0 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e6ba83392341f144", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {"Tantrum": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "85192"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2389\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-05-06 to 2026-05-22)\n\nACCELERATION PROGRAMS\n1. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n2. requesting preferred items using AAC device (Mand)\n Method: NET\n3. making the bed (Home Skills)\n Method: Task Analysis\n4. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-06 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 5/13 correct (42%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n requesting preferred items using AAC device: 6/12 correct (48%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n making the bed: 7/12 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n tolerating tooth brushing by caregiver: 7/15 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d75\n Self-injurious behavior (SIB): freq 1\n\nSession 2 \u2014 2026-05-08 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 7/14 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d76\n requesting preferred items using AAC device: 4/9 correct (50%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n making the bed: 5/8 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n tolerating tooth brushing by caregiver: 6/12 correct (54%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 1\n\nSession 3 \u2014 2026-05-10 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 6/10 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n requesting preferred items using AAC device: 5/9 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n making the bed: 7/9 correct (76%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n tolerating tooth brushing by caregiver: 6/9 correct (62%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 1\n\nSession 4 \u2014 2026-05-12 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 10/14 correct (69%); latency 2.6s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n requesting preferred items using AAC device: 9/14 correct (65%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n making the bed: 11/13 correct (84%); latency 2.1s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n tolerating tooth brushing by caregiver: 11/15 correct (71%); latency 2.5s; prompts FP\u00d70, PP\u00d75, G\u00d70, I\u00d710\n Self-injurious behavior (SIB): freq 2\n\nSession 5 \u2014 2026-05-14 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 6/8 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n requesting preferred items using AAC device: 10/15 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n making the bed: 8/10 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n tolerating tooth brushing by caregiver: 10/13 correct (76%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n Self-injurious behavior (SIB): freq 3\n\nSession 6 \u2014 2026-05-16 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 7/10 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n requesting preferred items using AAC device: 9/13 correct (73%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n making the bed: 8/9 correct (84%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n tolerating tooth brushing by caregiver: 11/14 correct (80%); latency 2.4s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-05-18 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 8/11 correct (71%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n requesting preferred items using AAC device: 10/13 correct (79%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n making the bed: 10/12 correct (84%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n tolerating tooth brushing by caregiver: 10/13 correct (80%); latency 2.3s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n Self-injurious behavior (SIB): freq 1\n\nSession 8 \u2014 2026-05-20 \u2014 57 min \u2014 1 observer\n dressing (pants, shirt): 11/15 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d710\n requesting preferred items using AAC device: 9/11 correct (83%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n making the bed: 7/8 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n tolerating tooth brushing by caregiver: 11/14 correct (80%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d77\n Self-injurious behavior (SIB): freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Stereotypic behavior during inter-trial intervals\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 42% at return, recovering to 71%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 62.8%.\n- Trend direction: upward (first session: 42%, last session: 71%).\n- Self-injurious behavior (SIB): mean 1.8/session (early-window mean 1.0 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "86a1df0db19c4a24", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "15986"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts colors of objects\nCurriculum: Tact L2\nLearner Level: Early Learner\nBaseline: Generalization phase\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts colors of objects\n- Curriculum: Tact L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What color?\"\nVariations: \"What color is this?\"; \"Tell me the color\"\nPresentation: Present a single-color object and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: purple, yellow, blue\nDistractor stimuli: white, brown, green, black\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "cedb4db27b03394d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts colors of objects", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "38800"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: traces basic shapes (circle, cross)\nCurriculum Reference: Writing L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces basic shapes (circle, cross)\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: square, diamond, oval\nDistractor stimuli: heart, circle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8d7b775a0f2b2b88", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces basic shapes (circle, cross)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "50881"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: copies simple words from model\nCurriculum: Writing L2\nLearner Level: School-Age Learner\nBaseline: Developing\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies simple words from model\n- Curriculum: Writing L2\n- Learner: School-Age Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: backpack, crayon, ball\nDistractor stimuli: jacket, block, glue, apple\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0bcbe1298ea440d0", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L2", "learner_profile": "school_age", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies simple words from model", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "36112"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: completing assigned tasks in order\nCurriculum: AFLS Vocational Skills\nLearner: Adult Learner\nMastery Status: Mastered at current level\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Review the task list for the shift. <- Learner-taught step (forward chaining starts here)\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "22881db4ba09fdc7", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_generalization", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "51834"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands for a break (e.g., \"break please\", AAC \"I need a break\")\nCurriculum Reference: Mand L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for a break (e.g., \"break please\", AAC \"I need a break\")\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nPresent a brief non-preferred task or demand at a level the learner can tolerate but may wish to pause. Wait for a request-for-break topography; honor the request immediately on first acquisition (CRF), thin later.\nPrimary natural context: academic period.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's academic period routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: immediate brief escape from the demand (30\u201360 seconds) to a neutral or preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "4f67b13403a98c47", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for a break (e.g., \"break please\", AAC \"I need a break\")", "mo_category": "mand_break", "mo_arrangement": "break_opportunity", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "44373"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: copies all upper-case letters from model\nCurriculum: Writing L2\nLearner Level: Early Learner\nBaseline: Developing\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies all upper-case letters from model\n- Curriculum: Writing L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: U, Q, Y\nDistractor stimuli: K, A, T, N\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "46e4e7bed0b015a3", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies all upper-case letters from model", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "28081"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: scheduling a medical appointment\nCurriculum: AFLS Independent Living Skills\nLearner: Adult Learner\nMastery Status: Mastered at current level\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: scheduling a medical appointment\n- Curriculum: AFLS Independent Living Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Identify the reason for the appointment. <- Learner-taught step (forward chaining starts here)\n2. Locate the provider's contact information.\n3. Call or use the online system.\n4. Provide name, date of birth, and insurance information.\n5. Request an available appointment time.\n6. Confirm the appointment and save it to the calendar.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "73bab6d8fa853146", "gold_labels": {"method": "task_analysis", "domain": "AFLS.independent_living", "skill": "scheduling a medical appointment", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "scheduling a medical appointment", "module": "independent_living", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "68917"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7704\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Reading + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-09-23 to 2026-10-13)\n\nACCELERATION PROGRAMS\n1. requesting a break during non-preferred activities (Mand)\n Method: NET\n2. ordering in a restaurant (Community)\n Method: Task Analysis\n3. tolerating haircut (Self-Care)\n Method: Task Analysis\n4. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n5. reads 10\u201320 sight words (Reading)\n Method: DTT\n6. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-23 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 7/9 correct (74%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n ordering in a restaurant: 10/14 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n tolerating haircut: 10/15 correct (69%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d75, I\u00d72\n meal cleanup after eating: 8/10 correct (75%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n reads 10\u201320 sight words: 6/10 correct (62%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n wait calmly when access to preferred item is denied: 10/15 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d76\n\nSession 2 \u2014 2026-09-25 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/9 correct (67%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n ordering in a restaurant: 7/9 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n tolerating haircut: 9/13 correct (70%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n meal cleanup after eating: 7/10 correct (73%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n reads 10\u201320 sight words: 6/10 correct (57%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 9/13 correct (68%); latency 3.0s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n\nSession 3 \u2014 2026-09-27 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/14 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d710\n ordering in a restaurant: 9/12 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n tolerating haircut: 8/11 correct (75%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n meal cleanup after eating: 11/15 correct (74%); latency 2.6s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 5/8 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 10/14 correct (73%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n\nSession 4 \u2014 2026-09-29 \u2014 IOA SESSION \u2014 2 observers\n requesting a break during non-preferred activities: 72% (primary); IOA 88% agreement\n ordering in a restaurant: 69% (primary); IOA 88% agreement\n tolerating haircut: 74% (primary); IOA 88% agreement\n meal cleanup after eating: 70% (primary); IOA 88% agreement\n reads 10\u201320 sight words: 60% (primary); IOA 88% agreement\n wait calmly when access to preferred item is denied: 64% (primary); IOA 88% agreement\n\nSession 5 \u2014 2026-10-01 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 4/9 correct (48%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n ordering in a restaurant: 7/15 correct (44%); latency 4.1s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d76\n tolerating haircut: 4/8 correct (55%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n meal cleanup after eating: 5/8 correct (57%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 6/13 correct (43%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n wait calmly when access to preferred item is denied: 3/9 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n\nSession 6 \u2014 2026-10-03 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 4/9 correct (49%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n ordering in a restaurant: 5/8 correct (57%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n tolerating haircut: 5/9 correct (58%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n meal cleanup after eating: 7/12 correct (61%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 4/8 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 5/12 correct (39%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n\nSession 7 \u2014 2026-10-05 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/11 correct (56%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n ordering in a restaurant: 8/13 correct (60%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n tolerating haircut: 9/15 correct (63%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d75, I\u00d75\n meal cleanup after eating: 7/10 correct (71%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n reads 10\u201320 sight words: 6/9 correct (61%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n wait calmly when access to preferred item is denied: 6/14 correct (45%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d79\n\nSession 8 \u2014 2026-10-07 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 9/14 correct (62%); latency 3.0s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n ordering in a restaurant: 8/11 correct (71%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n tolerating haircut: 7/10 correct (69%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n meal cleanup after eating: 6/8 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n reads 10\u201320 sight words: 7/11 correct (61%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 6/12 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n\nSession 9 \u2014 2026-10-09 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 5/8 correct (62%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n ordering in a restaurant: 9/13 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d77\n tolerating haircut: 11/15 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n meal cleanup after eating: 6/8 correct (74%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 9/14 correct (61%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d711\n wait calmly when access to preferred item is denied: 8/14 correct (54%); latency 3.8s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n\nSession 10 \u2014 2026-10-11 \u2014 46 min \u2014 1 observer\n requesting a break during non-preferred activities: 8/12 correct (69%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d77\n ordering in a restaurant: 10/14 correct (72%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d75\n tolerating haircut: 8/12 correct (70%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d75\n meal cleanup after eating: 7/9 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 7/11 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 4/8 correct (55%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Stereotypic behavior during inter-trial intervals\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 63.0%.\n- Trend direction: downward (first session: 74%, last session: 69%).\n- IOA data: 88% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "e9a943d58300ff67", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "61246"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts common actions (running, eating, sleeping)\nCurriculum Reference: Tact L1\nLearner Profile: Early Learner\nCurrent Mastery: Generalization phase\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common actions (running, eating, sleeping)\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is he doing?\"\nVariations: \"What action?\"; \"What is she doing?\"; \"What is happening here?\"\nPresentation: Present an action picture, video clip, or live demonstration.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: reading, drawing, pouring\nDistractor stimuli: sleeping, eating\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a56463556b4b2305", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common actions (running, eating, sleeping)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "token_economy", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "78310"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: points to named body parts\nCurriculum: Listener Responding L1\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: points to named body parts\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: car, spoon, plate\nDistractor stimuli: jacket, ball\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "54d90740f4fe2e24", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "points to named body parts", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "87673"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: names items in categories when given the category\nCurriculum Reference: Intraverbal L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: names items in categories when given the category\n- Curriculum: Intraverbal L2\n- Learner: School-Age Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: transition conversation.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's transition conversation routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "e7527886bb53d118", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "school_age", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "names items in categories when given the category", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "time_delay", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "83467"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands for missing items needed to complete an activity\nCurriculum: Mand L1\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for missing items needed to complete an activity\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange an activity that requires a specific item to complete, then omit that item from the available materials. Wait for the learner to notice the gap.\nPrimary natural context: art activity.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's art activity routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: the missing item is provided, activity can continue.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "866b3fe3f8e154a9", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for missing items needed to complete an activity", "mo_category": "mand_missing", "mo_arrangement": "missing_item_completion", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "75835"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3262\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand + Tact + Tolerating Denied Access\nDate range: Sessions 1\u201312 across 24 days (2026-10-07 to 2026-10-31)\n\nACCELERATION PROGRAMS\n1. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n2. tacts common objects (Tact)\n Method: DTT\n3. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n4. tacts colors of objects (Tact)\n Method: DTT\n5. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n6. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-07 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 7/9 correct (82%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n tacts common objects: 11/13 correct (83%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n answers WH-questions about familiar topics: 6/8 correct (78%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n tacts colors of objects: 9/12 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d77\n indicating completion ('all done', 'finished'): 9/11 correct (84%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 7/8 correct (82%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n\nSession 2 \u2014 2026-10-09 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 7/8 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n tacts common objects: 9/11 correct (84%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n answers WH-questions about familiar topics: 11/14 correct (77%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n tacts colors of objects: 12/15 correct (78%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n indicating completion ('all done', 'finished'): 7/9 correct (78%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 7/9 correct (82%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n\nSession 3 \u2014 2026-10-11 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 12/14 correct (84%); latency 2.0s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n tacts common objects: 8/9 correct (87%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n answers WH-questions about familiar topics: 11/14 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d77\n tacts colors of objects: 11/15 correct (74%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d76\n indicating completion ('all done', 'finished'): 10/12 correct (84%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d75\n wait calmly when access to preferred item is denied: 10/12 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n\nSession 4 \u2014 2026-10-13 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 12/14 correct (86%); latency 2.2s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n tacts common objects: 9/11 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n answers WH-questions about familiar topics: 11/14 correct (79%); latency 2.5s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d78\n tacts colors of objects: 10/13 correct (80%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n indicating completion ('all done', 'finished'): 10/13 correct (80%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 9/11 correct (80%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n\nSession 5 \u2014 2026-10-15 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 6/12 correct (52%); latency 3.4s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n tacts common objects: 5/11 correct (44%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n answers WH-questions about familiar topics: 4/9 correct (45%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n tacts colors of objects: 4/9 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n indicating completion ('all done', 'finished'): 7/14 correct (50%); latency 3.7s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 6/11 correct (56%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n\nSession 6 \u2014 2026-10-17 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 8/15 correct (55%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d710\n tacts common objects: 6/13 correct (43%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n answers WH-questions about familiar topics: 4/9 correct (44%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d76\n tacts colors of objects: 6/13 correct (46%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n indicating completion ('all done', 'finished'): 5/9 correct (54%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 9/15 correct (57%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n\nSession 7 \u2014 2026-10-19 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 5/8 correct (63%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n tacts common objects: 6/13 correct (49%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n answers WH-questions about familiar topics: 5/11 correct (42%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n tacts colors of objects: 3/8 correct (41%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n indicating completion ('all done', 'finished'): 5/8 correct (57%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 9/15 correct (59%); latency 3.3s; prompts FP\u00d73, PP\u00d75, G\u00d71, I\u00d76\n\nSession 8 \u2014 2026-10-21 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 9/13 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d73\n tacts common objects: 5/8 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n answers WH-questions about familiar topics: 4/9 correct (43%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n tacts colors of objects: 3/8 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n indicating completion ('all done', 'finished'): 6/10 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 8/12 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n\nSession 9 \u2014 2026-10-23 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 8/11 correct (76%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n tacts common objects: 7/10 correct (67%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n answers WH-questions about familiar topics: 6/13 correct (50%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n tacts colors of objects: 4/8 correct (45%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n indicating completion ('all done', 'finished'): 5/8 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 9/12 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n\nSession 10 \u2014 2026-10-25 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 9/11 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n tacts common objects: 11/15 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n answers WH-questions about familiar topics: 5/8 correct (60%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n tacts colors of objects: 6/11 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n indicating completion ('all done', 'finished'): 10/14 correct (71%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 11/13 correct (81%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n\nSession 11 \u2014 2026-10-27 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 8/10 correct (85%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n tacts common objects: 11/14 correct (77%); latency 2.6s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n answers WH-questions about familiar topics: 9/13 correct (69%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n tacts colors of objects: 6/12 correct (54%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d77\n indicating completion ('all done', 'finished'): 9/12 correct (79%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 13/15 correct (84%); latency 2.0s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d77\n\nSession 12 \u2014 2026-10-29 \u2014 40 min \u2014 1 observer\n follows 1-step motor instructions: 13/15 correct (85%); latency 2.1s; prompts FP\u00d75, PP\u00d75, G\u00d71, I\u00d74\n tacts common objects: 9/11 correct (84%); latency 2.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n answers WH-questions about familiar topics: 10/13 correct (77%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n tacts colors of objects: 7/12 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 10/12 correct (80%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n wait calmly when access to preferred item is denied: 7/8 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Stereotypic behavior during inter-trial intervals\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 74.9%.\n- Trend direction: upward (first session: 82%, last session: 85%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1b5cc587f01f0c10", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "early", "n_sessions": 12, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "99371"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: washing hands\nCurriculum: AFLS Basic Living Skills\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: washing hands\n- Curriculum: AFLS Basic Living Skills\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Walk to the sink.\n2. Turn on the water.\n3. Wet hands under the water.\n4. Apply soap to hands.\n5. Rub hands together for 15\u201320 seconds, covering fronts, backs, and between fingers.\n6. Rinse hands thoroughly under the water.\n7. Turn off the water.\n8. Dry hands with a towel.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "40e25cab9194baab", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "washing hands", "program_type": "independence", "learner_profile": "school_age", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "washing hands", "module": "basic_living", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "98892"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2139\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math\nDate range: Sessions 1\u20135 across 10 days (2026-08-25 to 2026-09-04)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n3. single-word requests for preferred items (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-25 \u2014 39 min \u2014 1 observer\n identifies numerals 1\u201320: 4/12 correct (35%); latency 4.2s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/14 correct (27%); latency 5.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d79\n single-word requests for preferred items: 4/12 correct (36%); latency 4.5s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n\nSession 2 \u2014 2026-08-27 \u2014 39 min \u2014 1 observer\n identifies numerals 1\u201320: 6/14 correct (43%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/13 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n single-word requests for preferred items: 6/13 correct (50%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n\nSession 3 \u2014 2026-08-29 \u2014 39 min \u2014 1 observer\n identifies numerals 1\u201320: 4/8 correct (53%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/14 correct (47%); latency 4.0s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n single-word requests for preferred items: 5/9 correct (58%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n\nSession 4 \u2014 2026-08-31 \u2014 39 min \u2014 1 observer\n identifies numerals 1\u201320: 8/13 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (53%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n single-word requests for preferred items: 9/13 correct (70%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n\nSession 5 \u2014 2026-09-02 \u2014 39 min \u2014 1 observer\n identifies numerals 1\u201320: 8/10 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/14 correct (65%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n single-word requests for preferred items: 10/13 correct (77%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Quick response latency (within 2\u20133 seconds)\n- Smiling during reinforcement delivery"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 35% to 77% across 5 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 5 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 5 sessions: 54.4%.\n- Trend direction: upward (first session: 35%, last session: 77%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "c37d37801c39adf9", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "school_age", "n_sessions": 5, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "83942"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8371\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Reading + Self-Care\nDate range: Sessions 1\u201311 across 22 days (2026-02-13 to 2026-03-07)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. indicating 'all done' with a step or activity (Mand)\n Method: NET\n3. requesting preferred items using AAC device (Mand)\n Method: NET\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Mouthing \u2014 function hypothesized: automatic\n2. Property destruction \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-13 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 6/8 correct (77%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n indicating 'all done' with a step or activity: 11/14 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d79\n requesting preferred items using AAC device: 13/15 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n brushing teeth independently: 9/11 correct (84%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n Mouthing: freq 1; PIR 5%\n Property destruction: freq 2\n\nSession 2 \u2014 2026-02-15 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 9/12 correct (76%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n indicating 'all done' with a step or activity: 10/14 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d710\n requesting preferred items using AAC device: 7/9 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n brushing teeth independently: 9/12 correct (71%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n Mouthing: freq 3; PIR 11%\n Property destruction: freq 1\n\nSession 3 \u2014 2026-02-17 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 6/9 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n indicating 'all done' with a step or activity: 8/13 correct (62%); latency 2.9s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d74\n requesting preferred items using AAC device: 7/9 correct (74%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n brushing teeth independently: 7/10 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n Mouthing: freq 2; PIR 18%\n Property destruction: freq 2\n ABC (mouthing): A = waiting in line transition; B = mouthed own hand; C = staff redirected to task materials\n\nSession 4 \u2014 2026-02-19 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 6/10 correct (58%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n indicating 'all done' with a step or activity: 10/15 correct (64%); latency 3.4s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d77\n requesting preferred items using AAC device: 5/8 correct (67%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n brushing teeth independently: 9/14 correct (67%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n Mouthing: freq 2; PIR 13%\n Property destruction: freq 2\n ABC (property destruction): A = transition to non-preferred activity; B = dropped to floor, cried; C = transition delayed, preferred item offered\n\nSession 5 \u2014 2026-02-21 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 7/12 correct (60%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n indicating 'all done' with a step or activity: 9/15 correct (57%); latency 3.1s; prompts FP\u00d74, PP\u00d75, G\u00d71, I\u00d75\n requesting preferred items using AAC device: 6/10 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n brushing teeth independently: 6/12 correct (53%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n Mouthing: freq 3; PIR 21%\n Property destruction: freq 1\n\nSession 6 \u2014 2026-02-23 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 6/12 correct (51%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 5/10 correct (52%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n requesting preferred items using AAC device: 4/8 correct (53%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n brushing teeth independently: 6/11 correct (50%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n Mouthing: freq 4; PIR 21%\n Property destruction: freq 2\n\nSession 7 \u2014 2026-02-25 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 6/13 correct (45%); latency 4.2s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n indicating 'all done' with a step or activity: 5/12 correct (45%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d75\n requesting preferred items using AAC device: 5/13 correct (40%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d79\n brushing teeth independently: 7/15 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d78\n Mouthing: freq 5; PIR 21%\n Property destruction: freq 3\n\nSession 8 \u2014 2026-02-27 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 4/9 correct (45%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n indicating 'all done' with a step or activity: 5/12 correct (42%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n requesting preferred items using AAC device: 5/13 correct (39%); latency 4.0s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n brushing teeth independently: 3/8 correct (40%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Mouthing: freq 5; PIR 20%\n Property destruction: freq 4\n ABC (property destruction): A = worksheet task presented; B = refused + pushed materials away; C = task removed for 2 min\n\nSession 9 \u2014 2026-03-01 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 3/8 correct (32%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n indicating 'all done' with a step or activity: 5/13 correct (37%); latency 4.4s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n requesting preferred items using AAC device: 2/8 correct (26%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n brushing teeth independently: 3/9 correct (29%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n Mouthing: freq 5; PIR 22%\n Property destruction: freq 4\n\nSession 10 \u2014 2026-03-03 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 4/15 correct (28%); latency 4.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d710\n indicating 'all done' with a step or activity: 4/11 correct (35%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n requesting preferred items using AAC device: 2/10 correct (21%); latency 5.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n brushing teeth independently: 3/12 correct (29%); latency 4.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n Mouthing: freq 5; PIR 22%\n Property destruction: freq 3\n\nSession 11 \u2014 2026-03-05 \u2014 34 min \u2014 1 observer\n sounds out CVC words: 5/15 correct (32%); latency 4.4s; prompts FP\u00d70, PP\u00d75, G\u00d75, I\u00d75\n indicating 'all done' with a step or activity: 3/8 correct (34%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n requesting preferred items using AAC device: 2/13 correct (16%); latency 5.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d710\n brushing teeth independently: 1/9 correct (17%); latency 5.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n Mouthing: freq 5; PIR 33%\n Property destruction: freq 4\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Attempts at property destruction (throwing materials, swiping items off the surface)\n- Self-stimulatory behavior increase during demand periods"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 77% to 32% across 11 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nMouthing: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\nProperty destruction: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 52.1%.\n- Trend direction: downward (first session: 77%, last session: 32%).\n- Mouthing: mean 3.6/session (early-window mean 2.0 -> late-window mean 5.0).\n- Property destruction: mean 2.5/session (early-window mean 1.7 -> late-window mean 3.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "dfb65de9daa9d3b9", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Mouthing": "automatic", "Property destruction": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adolescent", "n_sessions": 11, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["mouthing", "property_destruction"], "behavior_functions": ["automatic", "escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "90680"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: traces letters of own name\nCurriculum: Writing L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces letters of own name\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: O, B, A\nDistractor stimuli: E, Y\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6310a06ef21744ae", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces letters of own name", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "60118"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9146\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Math + Reading + Writing\nDate range: Sessions 1\u20138 across 16 days (2026-09-09 to 2026-09-25)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. reads 10\u201320 sight words (Reading)\n Method: DTT\n3. counts objects 1\u201320 (Math)\n Method: DTT\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-09 \u2014 45 min \u2014 1 observer\n copies simple words from model: 9/11 correct (81%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 11/14 correct (82%); latency 2.5s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d74\n counts objects 1\u201320: 8/10 correct (80%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Property destruction: freq 2\n\nSession 2 \u2014 2026-09-11 \u2014 45 min \u2014 1 observer\n copies simple words from model: 7/9 correct (83%); latency 2.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 12/15 correct (83%); latency 2.3s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n counts objects 1\u201320: 12/14 correct (86%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d713\n Property destruction: freq 2\n\nSession 3 \u2014 2026-09-13 \u2014 45 min \u2014 1 observer\n copies simple words from model: 7/8 correct (84%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n reads 10\u201320 sight words: 9/10 correct (85%); latency 2.0s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n counts objects 1\u201320: 13/15 correct (86%); latency 2.3s; prompts FP\u00d75, PP\u00d72, G\u00d73, I\u00d75\n Property destruction: freq 3\n\nSession 4 \u2014 2026-09-15 \u2014 45 min \u2014 1 observer\n copies simple words from model: 9/11 correct (81%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n reads 10\u201320 sight words: 9/11 correct (83%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n counts objects 1\u201320: 7/8 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n Property destruction: freq 2\n\nSession 5 \u2014 2026-09-17 \u2014 45 min \u2014 1 observer\n copies simple words from model: 7/9 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n reads 10\u201320 sight words: 10/12 correct (83%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n counts objects 1\u201320: 7/8 correct (85%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Property destruction: freq 2\n\nSession 6 \u2014 2026-09-19 \u2014 IOA SESSION \u2014 2 observers\n copies simple words from model: 83% (primary); IOA 82% agreement\n reads 10\u201320 sight words: 86% (primary); IOA 82% agreement\n counts objects 1\u201320: 86% (primary); IOA 82% agreement\n Property destruction: freq 2; IOA 82% agreement\n\nSession 7 \u2014 2026-09-21 \u2014 45 min \u2014 1 observer\n copies simple words from model: 10/12 correct (80%); latency 2.2s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 9/11 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n counts objects 1\u201320: 12/13 correct (90%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d710\n Property destruction: freq 2\n\nSession 8 \u2014 2026-09-23 \u2014 45 min \u2014 1 observer\n copies simple words from model: 12/15 correct (78%); latency 2.7s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d74\n reads 10\u201320 sight words: 9/11 correct (82%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n counts objects 1\u201320: 9/10 correct (85%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n Property destruction: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (81%\u201378%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nProperty destruction: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 80.9%.\n- Trend direction: downward (first session: 81%, last session: 78%).\n- Property destruction: mean 2.1/session (early-window mean 2.0 -> late-window mean 2.0).\n- IOA data: 82% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "a103fd8f3d78e3c3", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Property destruction": "escape"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "8756"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6092\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u201310 across 20 days (2026-07-16 to 2026-08-05)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. simple meal preparation (Home Skills)\n Method: Task Analysis\n3. requesting assistance using AAC device (Mand)\n Method: NET\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n6. doing laundry (Home Skills)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-16 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 9/15 correct (57%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d712\n simple meal preparation: 8/12 correct (68%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n requesting assistance using AAC device: 5/11 correct (44%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n brushing teeth independently: 6/11 correct (57%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 5/12 correct (44%); latency 4.3s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n doing laundry: 5/8 correct (58%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 3\n\nSession 2 \u2014 2026-07-18 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 6/11 correct (50%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n simple meal preparation: 7/14 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n requesting assistance using AAC device: 4/9 correct (43%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n brushing teeth independently: 5/8 correct (66%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 4/8 correct (45%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n doing laundry: 9/15 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n Self-injurious behavior (SIB): freq 2\n\nSession 3 \u2014 2026-07-20 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 6/10 correct (63%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n simple meal preparation: 5/8 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n requesting assistance using AAC device: 4/9 correct (50%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n brushing teeth independently: 8/13 correct (59%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d711\n wait calmly when a preferred activity is denied: 3/8 correct (37%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n doing laundry: 7/12 correct (57%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d75\n Self-injurious behavior (SIB): freq 3\n\nSession 4 \u2014 2026-07-22 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 7/11 correct (64%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n simple meal preparation: 6/10 correct (56%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n requesting assistance using AAC device: 4/8 correct (46%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n brushing teeth independently: 7/10 correct (67%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n wait calmly when a preferred activity is denied: 3/8 correct (33%); latency 4.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n doing laundry: 5/8 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n Self-injurious behavior (SIB): freq 3\n\nSession 5 \u2014 2026-07-24 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 5/10 correct (54%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n simple meal preparation: 6/11 correct (59%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n requesting assistance using AAC device: 6/13 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n brushing teeth independently: 10/14 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n wait calmly when a preferred activity is denied: 6/15 correct (39%); latency 4.4s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n doing laundry: 5/8 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n Self-injurious behavior (SIB): freq 2\n\nSession 6 \u2014 2026-07-26 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 5/8 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n simple meal preparation: 9/13 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d74\n requesting assistance using AAC device: 7/12 correct (56%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d76\n brushing teeth independently: 9/14 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n wait calmly when a preferred activity is denied: 4/14 correct (31%); latency 4.8s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n doing laundry: 8/13 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-07-28 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 9/14 correct (66%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d76\n simple meal preparation: 9/15 correct (62%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d77\n requesting assistance using AAC device: 8/15 correct (54%); latency 3.5s; prompts FP\u00d73, PP\u00d75, G\u00d72, I\u00d75\n brushing teeth independently: 8/12 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 5/13 correct (40%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n doing laundry: 5/8 correct (57%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n Self-injurious behavior (SIB): freq 3\n\nSession 8 \u2014 2026-07-30 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 8/15 correct (56%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d713\n simple meal preparation: 6/10 correct (63%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n requesting assistance using AAC device: 6/10 correct (56%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n brushing teeth independently: 5/8 correct (67%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 4/10 correct (39%); latency 4.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d74\n doing laundry: 7/11 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n Self-injurious behavior (SIB): freq 3\n\nSession 9 \u2014 2026-08-01 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 6/11 correct (58%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n simple meal preparation: 9/14 correct (65%); latency 3.3s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n requesting assistance using AAC device: 7/12 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n brushing teeth independently: 8/11 correct (74%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 5/15 correct (35%); latency 4.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n doing laundry: 5/8 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 3\n\nSession 10 \u2014 2026-08-03 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 10/15 correct (69%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d74\n simple meal preparation: 8/14 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n requesting assistance using AAC device: 6/12 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n brushing teeth independently: 8/11 correct (74%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 5/15 correct (33%); latency 4.5s; prompts FP\u00d72, PP\u00d75, G\u00d70, I\u00d78\n doing laundry: 8/14 correct (54%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d77\n Self-injurious behavior (SIB): freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 63% for 10 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 60.5%.\n- Trend direction: upward (first session: 57%, last session: 69%).\n- Self-injurious behavior (SIB): mean 2.8/session (early-window mean 2.7 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "b8379739a966e4f3", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "adult", "n_sessions": 10, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "52665"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts common objects (ball, cup, shoe)\nCurriculum: Tact L1\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common objects (ball, cup, shoe)\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: phone, pencil, napkin\nDistractor stimuli: hat, backpack\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "84ce86cebd1f1cfe", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common objects (ball, cup, shoe)", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "token_economy", "error_correction": "four_step", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "93657"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: sorts items into 2\u20133 categories\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L2\nLearner: Early Learner\nMastery Status: Approaching mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: sorts items into 2\u20133 categories\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L2\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: towel, napkin, jacket\nDistractor stimuli: pencil, apple, cup, phone\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3913a0455c682fe8", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L2", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "sorts items into 2\u20133 categories", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "errorless", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "85854"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5252\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-03-10 to 2026-03-26)\n\nACCELERATION PROGRAMS\n1. doing laundry (Home Skills)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. clocking in at start of shift (Vocational)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-10 \u2014 35 min \u2014 1 observer\n doing laundry: 9/11 correct (85%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n wait calmly when a preferred activity is denied: 13/15 correct (87%); latency 2.1s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d74\n clocking in at start of shift: 8/9 correct (85%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n\nSession 2 \u2014 2026-03-12 \u2014 35 min \u2014 1 observer\n doing laundry: 2/8 correct (26%); latency 4.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n wait calmly when a preferred activity is denied: 3/9 correct (33%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n clocking in at start of shift: 2/8 correct (28%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n\nSession 3 \u2014 2026-03-14 \u2014 IOA SESSION \u2014 2 observers\n doing laundry: 90% (primary); IOA 91% agreement\n wait calmly when a preferred activity is denied: 80% (primary); IOA 91% agreement\n clocking in at start of shift: 81% (primary); IOA 91% agreement\n\nSession 4 \u2014 2026-03-16 \u2014 35 min \u2014 1 observer\n doing laundry: 3/11 correct (26%); latency 5.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 4/10 correct (37%); latency 4.0s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n clocking in at start of shift: 4/14 correct (30%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n\nSession 5 \u2014 2026-03-18 \u2014 35 min \u2014 1 observer\n doing laundry: 7/8 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 8/10 correct (80%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n clocking in at start of shift: 11/13 correct (87%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n\nSession 6 \u2014 2026-03-20 \u2014 35 min \u2014 1 observer\n doing laundry: 2/9 correct (28%); latency 5.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 3/10 correct (34%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n clocking in at start of shift: 4/14 correct (26%); latency 4.6s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d73\n\nSession 7 \u2014 2026-03-22 \u2014 35 min \u2014 1 observer\n doing laundry: 7/8 correct (85%); latency 1.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 8/10 correct (79%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n clocking in at start of shift: 13/15 correct (87%); latency 2.0s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n\nSession 8 \u2014 2026-03-24 \u2014 35 min \u2014 1 observer\n doing laundry: 3/8 correct (32%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 4/11 correct (34%); latency 4.7s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n clocking in at start of shift: 4/15 correct (28%); latency 4.8s; prompts FP\u00d75, PP\u00d74, G\u00d71, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Stereotypic behavior during inter-trial intervals\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 56.9%.\n- Trend direction: downward (first session: 85%, last session: 32%).\n- IOA data: 91% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "50f8a62f6f53880e", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 8, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "52979"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands for missing items needed to complete an activity\nCurriculum: Mand L1\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for missing items needed to complete an activity\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange an activity that requires a specific item to complete, then omit that item from the available materials. Wait for the learner to notice the gap.\nPrimary natural context: art activity.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's art activity routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: the missing item is provided, activity can continue.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "5e95c63a3261363d", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for missing items needed to complete an activity", "mo_category": "mand_missing", "mo_arrangement": "missing_item_completion", "prompt_strategy": "time_delay", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "83467"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects items by function when item not visible\nCurriculum: LRFFC L2\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function when item not visible\n- Curriculum: LRFFC L2\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: jacket, ball, backpack\nDistractor stimuli: apple, plate, block, key\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ce1403ef1332f36b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L2", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function when item not visible", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "17197"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2005\nProfile: Adult Learner (18+ years)\nCurricula: Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20137 across 14 days (2026-04-06 to 2026-04-20)\n\nACCELERATION PROGRAMS\n1. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n4. requesting a break during extended tasks (Mand)\n Method: NET\n5. washing hands independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-06 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 6/12 correct (54%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n wait calmly when a preferred activity is denied: 6/8 correct (73%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n completing assigned tasks in order: 7/13 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n requesting a break during extended tasks: 8/12 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d73\n washing hands independently: 9/15 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n Aggression: freq 1\n\nSession 2 \u2014 2026-04-08 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 5/10 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 8/11 correct (74%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n completing assigned tasks in order: 9/15 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d79\n requesting a break during extended tasks: 7/10 correct (74%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n washing hands independently: 5/9 correct (60%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n Aggression: freq 3\n\nSession 3 \u2014 2026-04-10 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 7/11 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 6/8 correct (73%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n completing assigned tasks in order: 8/15 correct (51%); latency 3.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d79\n requesting a break during extended tasks: 10/14 correct (71%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d79\n washing hands independently: 5/9 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n Aggression: freq 5\n\nSession 4 \u2014 2026-04-12 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 8/14 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n wait calmly when a preferred activity is denied: 8/11 correct (73%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n completing assigned tasks in order: 8/15 correct (53%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n requesting a break during extended tasks: 9/12 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n washing hands independently: 7/13 correct (56%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n Aggression: freq 5\n\nSession 5 \u2014 2026-04-14 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 6/11 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 6/8 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n completing assigned tasks in order: 5/10 correct (53%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n requesting a break during extended tasks: 10/14 correct (74%); latency 3.0s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n washing hands independently: 7/13 correct (53%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d79\n Aggression: freq 2\n\nSession 6 \u2014 2026-04-16 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 6/9 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n wait calmly when a preferred activity is denied: 11/15 correct (76%); latency 2.9s; prompts FP\u00d72, PP\u00d75, G\u00d75, I\u00d73\n completing assigned tasks in order: 7/11 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n requesting a break during extended tasks: 6/8 correct (74%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n washing hands independently: 6/11 correct (53%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n Aggression: freq 2\n\nSession 7 \u2014 2026-04-18 \u2014 50 min \u2014 1 observer\n dressing (pants, shirt): 7/11 correct (67%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n wait calmly when a preferred activity is denied: 6/8 correct (78%); latency 2.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n completing assigned tasks in order: 7/13 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n requesting a break during extended tasks: 12/15 correct (83%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d711\n washing hands independently: 6/11 correct (56%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n Aggression: freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Peer aggression (hitting or pushing peers during group activities)\n- Turning away from task or staff\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Crying during demand presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nAggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 58.9%.\n- Trend direction: upward (first session: 54%, last session: 67%).\n- Aggression: mean 2.7/session (early-window mean 2.0 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "6ad02e6384b6d574", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Aggression": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adult", "n_sessions": 7, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["aggression"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "54431"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9867\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-04-02 to 2026-04-18)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. washing hands independently (Self-Care)\n Method: Task Analysis\n4. self-advocacy statements (Independent Living)\n Method: BST\n5. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-02 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 12/14 correct (86%); latency 2.2s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n wait calmly when a preferred activity is denied: 10/11 correct (87%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n washing hands independently: 9/12 correct (79%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n self-advocacy statements: 9/10 correct (86%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n simple meal preparation: 7/8 correct (91%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n\nSession 2 \u2014 2026-04-04 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 3/9 correct (35%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n wait calmly when a preferred activity is denied: 5/14 correct (37%); latency 4.6s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n washing hands independently: 2/9 correct (25%); latency 5.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n self-advocacy statements: 5/11 correct (41%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n simple meal preparation: 5/11 correct (47%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n\nSession 3 \u2014 2026-04-06 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 13/15 correct (90%); latency 1.7s; prompts FP\u00d75, PP\u00d73, G\u00d74, I\u00d73\n wait calmly when a preferred activity is denied: 11/13 correct (84%); latency 2.4s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n washing hands independently: 9/11 correct (80%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n self-advocacy statements: 12/14 correct (86%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d710\n simple meal preparation: 12/14 correct (84%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n\nSession 4 \u2014 2026-04-08 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 3/10 correct (34%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 3/9 correct (32%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n washing hands independently: 3/13 correct (23%); latency 5.2s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n self-advocacy statements: 6/15 correct (37%); latency 4.5s; prompts FP\u00d72, PP\u00d74, G\u00d75, I\u00d74\n simple meal preparation: 5/12 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n\nSession 5 \u2014 2026-04-10 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 12/14 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d73\n wait calmly when a preferred activity is denied: 8/9 correct (89%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n washing hands independently: 11/13 correct (81%); latency 2.4s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n self-advocacy statements: 13/15 correct (89%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d715\n simple meal preparation: 11/13 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n\nSession 6 \u2014 2026-04-12 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 3/8 correct (34%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 5/11 correct (42%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n washing hands independently: 3/11 correct (27%); latency 5.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n self-advocacy statements: 5/13 correct (40%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n simple meal preparation: 3/8 correct (43%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n\nSession 7 \u2014 2026-04-14 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 9/10 correct (93%); latency 2.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 14/15 correct (92%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d712\n washing hands independently: 11/14 correct (76%); latency 2.7s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n self-advocacy statements: 7/8 correct (89%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n simple meal preparation: 12/14 correct (84%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n\nSession 8 \u2014 2026-04-16 \u2014 58 min \u2014 1 observer\n completing assigned tasks in order: 3/9 correct (39%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 4/11 correct (40%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n washing hands independently: 2/8 correct (28%); latency 4.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n self-advocacy statements: 5/12 correct (42%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n simple meal preparation: 4/11 correct (41%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Echolalic responding (repeating SD rather than answering)\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 61.9%.\n- Trend direction: downward (first session: 86%, last session: 39%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "2a430140599afd91", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 8, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "27450"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: traces letters of own name\nCurriculum: Writing L1\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces letters of own name\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: Z, R, J\nDistractor stimuli: C, S\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "97dacbc1748705d0", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces letters of own name", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "vr3", "error_correction": "simple_correction", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "69141"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts body parts\nCurriculum: Tact L1\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts body parts\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me\"; \"Name this body part\"\nPresentation: Point to the target body part (on self, learner, or doll) and ask.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: sock, cup, phone\nDistractor stimuli: block, glue\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "51df15c5d8b1e0d3", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts body parts", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "36112"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands to indicate completion of a step or activity (\"all done with this one\")\nCurriculum Reference: Mand L2\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands to indicate completion of a step or activity (\"all done with this one\")\n- Curriculum: Mand L2\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nSet up a clearly-bounded activity with a visible endpoint (e.g., a worksheet, a puzzle, a sorting task). Wait for the learner to indicate completion verbally or via AAC; honor the indication immediately.\nPrimary natural context: meal.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's meal routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: the activity ends immediately and the learner transitions to the next preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "1293c2195a8abcc4", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands to indicate completion of a step or activity (\"all done with this one\")", "mo_category": "mand_completion", "mo_arrangement": "completion_opportunity", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "36908"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: engages in social games (peek-a-boo, tickle)\nCurriculum: Social Behavior & Social Play L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: engages in social games (peek-a-boo, tickle)\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "bd75fe66d23005e4", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "engages in social games (peek-a-boo, tickle)", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "54871"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a systematic desensitization / toleration program for the following.\n\nSkill: tolerating nail clipping\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adult Learner\nBaseline: Generalization phase\n\nThe learner currently refuses or exhibits safety behaviors during the activity; the program should build tolerance gradually."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating nail clipping\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for nail clipping performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nall 10 fingernails (and optionally toenails) clipped by caregiver without refusal, pulling hand away, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1 fingernail clipped\n2. 2 fingernails clipped in one sitting\n3. 5 fingernails clipped (one hand)\n4. All 10 fingernails clipped\n5. All 10 fingernails + 10 toenails clipped\n\nCurrent shaping step: At the generalization stage, begin at shaping step 5 (\"All 10 fingernails + 10 toenails clipped\") and advance through step 5 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated with 2 different caregivers (primary caregiver + one other) across 2 consecutive weeks.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 2 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5ec50cf81011767a", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating nail clipping", "program_type": "toleration", "learner_profile": "adult", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating nail clipping", "module": "basic_living", "mastery_criterion": "mc_toleration_caregivers", "shaping_n_steps": 5}, "teacher_model": null, "seed_tag": "97806"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6485\nProfile: Adult Learner (18+ years)\nCurricula: Mand + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-02-10 to 2026-03-02)\n\nACCELERATION PROGRAMS\n1. requesting a break during extended tasks (Mand)\n Method: NET\n2. brushing teeth independently (Self-Care)\n Method: Task Analysis\n3. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-10 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 5/12 correct (44%); latency 4.1s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n brushing teeth independently: 3/8 correct (34%); latency 4.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n wait calmly when a preferred activity is denied: 3/8 correct (42%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n\nSession 2 \u2014 2026-02-12 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 6/11 correct (53%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n brushing teeth independently: 4/10 correct (43%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n wait calmly when a preferred activity is denied: 7/15 correct (50%); latency 3.5s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d76\n\nSession 3 \u2014 2026-02-14 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 8/13 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n brushing teeth independently: 8/15 correct (51%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d75\n wait calmly when a preferred activity is denied: 6/10 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n\nSession 4 \u2014 2026-02-16 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 7/10 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n brushing teeth independently: 6/10 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n wait calmly when a preferred activity is denied: 7/10 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n\nSession 5 \u2014 2026-02-18 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 12/15 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n brushing teeth independently: 5/8 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 9/13 correct (72%); latency 3.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n\nSession 6 \u2014 2026-02-20 \u2014 IOA SESSION \u2014 2 observers\n requesting a break during extended tasks: 81% (primary); IOA 70% agreement\n brushing teeth independently: 69% (primary); IOA 70% agreement\n wait calmly when a preferred activity is denied: 72% (primary); IOA 70% agreement\n\nSession 7 \u2014 2026-02-22 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 12/15 correct (81%); latency 2.1s; prompts FP\u00d75, PP\u00d73, G\u00d70, I\u00d77\n brushing teeth independently: 6/9 correct (71%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n wait calmly when a preferred activity is denied: 7/10 correct (72%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n\nSession 8 \u2014 2026-02-24 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 8/10 correct (81%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n brushing teeth independently: 9/13 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 6/8 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n\nSession 9 \u2014 2026-02-26 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 7/9 correct (81%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n brushing teeth independently: 6/9 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 9/13 correct (72%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n\nSession 10 \u2014 2026-02-28 \u2014 51 min \u2014 1 observer\n requesting a break during extended tasks: 8/10 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n brushing teeth independently: 6/8 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n wait calmly when a preferred activity is denied: 6/8 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 44% at return, recovering to 81%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 71.5%.\n- Trend direction: upward (first session: 44%, last session: 81%).\n- IOA data: 70% agreement on IOA session \u2014 below threshold."}], "meta": {"task_type": "session_interpretation", "example_id": "ceb8b81cd0e690c7", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adult", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "70409"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6673\nProfile: Adult Learner (18+ years)\nCurricula: Community + Mand + Self-Care + Vocational\nDate range: Sessions 1\u20139 across 18 days (2026-01-25 to 2026-02-12)\n\nACCELERATION PROGRAMS\n1. requesting assistance using AAC device (Mand)\n Method: NET\n2. washing hands independently (Self-Care)\n Method: Task Analysis\n3. clocking in at start of shift (Vocational)\n Method: Task Analysis\n4. requesting a break during extended tasks (Mand)\n Method: NET\n5. using public transportation (Community)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-25 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 11/14 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n washing hands independently: 7/9 correct (80%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n clocking in at start of shift: 8/9 correct (89%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n requesting a break during extended tasks: 14/15 correct (93%); latency 2.1s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d75\n using public transportation: 7/8 correct (82%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n Property destruction: freq 1\n\nSession 2 \u2014 2026-01-27 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 3/10 correct (27%); latency 4.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n washing hands independently: 6/14 correct (40%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n clocking in at start of shift: 5/15 correct (34%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d77\n requesting a break during extended tasks: 6/15 correct (37%); latency 4.3s; prompts FP\u00d75, PP\u00d70, G\u00d71, I\u00d79\n using public transportation: 3/10 correct (34%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n Property destruction: freq 1\n\nSession 3 \u2014 2026-01-29 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 7/9 correct (77%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n washing hands independently: 6/8 correct (71%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n clocking in at start of shift: 10/12 correct (87%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n requesting a break during extended tasks: 9/10 correct (93%); latency 2.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n using public transportation: 8/10 correct (83%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n Property destruction: freq 2\n\nSession 4 \u2014 2026-01-31 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 3/9 correct (33%); latency 4.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n washing hands independently: 3/9 correct (31%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n clocking in at start of shift: 4/13 correct (34%); latency 4.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n requesting a break during extended tasks: 5/13 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n using public transportation: 4/12 correct (37%); latency 4.1s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d71\n Property destruction: freq 3\n\nSession 5 \u2014 2026-02-02 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 9/11 correct (78%); latency 2.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n washing hands independently: 10/14 correct (71%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n clocking in at start of shift: 12/14 correct (85%); latency 2.0s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n requesting a break during extended tasks: 9/10 correct (86%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n using public transportation: 10/11 correct (89%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Property destruction: freq 1\n\nSession 6 \u2014 2026-02-04 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 3/9 correct (29%); latency 4.6s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n washing hands independently: 3/10 correct (34%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n clocking in at start of shift: 5/14 correct (38%); latency 4.1s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d75\n requesting a break during extended tasks: 5/12 correct (39%); latency 4.3s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n using public transportation: 3/11 correct (27%); latency 5.0s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n Property destruction: freq 2\n ABC (property destruction): A = transition to non-preferred activity; B = dropped to floor, cried; C = transition delayed, preferred item offered\n\nSession 7 \u2014 2026-02-06 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 6/8 correct (77%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n washing hands independently: 11/14 correct (76%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d711\n clocking in at start of shift: 12/13 correct (91%); latency 1.7s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n requesting a break during extended tasks: 11/12 correct (89%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d78\n using public transportation: 8/9 correct (90%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n Property destruction: freq 2\n\nSession 8 \u2014 2026-02-08 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 4/15 correct (24%); latency 5.2s; prompts FP\u00d71, PP\u00d75, G\u00d73, I\u00d76\n washing hands independently: 3/10 correct (34%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n clocking in at start of shift: 5/15 correct (33%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n requesting a break during extended tasks: 3/8 correct (43%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n using public transportation: 2/8 correct (30%); latency 4.6s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n Property destruction: freq 2\n\nSession 9 \u2014 2026-02-10 \u2014 49 min \u2014 1 observer\n requesting assistance using AAC device: 7/9 correct (81%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n washing hands independently: 9/11 correct (80%); latency 2.2s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n clocking in at start of shift: 10/11 correct (94%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n requesting a break during extended tasks: 8/9 correct (94%); latency 1.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n using public transportation: 12/13 correct (89%); latency 2.3s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n Property destruction: freq 2\n ABC (property destruction): A = asked to finish dressing task; B = kicked at materials; C = task paused for 3 min\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Stereotypic behavior during inter-trial intervals\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Behavior Function Hypothesis\nProperty destruction: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 56.0%.\n- Trend direction: upward (first session: 78%, last session: 81%).\n- Property destruction: mean 1.8/session (early-window mean 1.3 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "681f21eb7e5e1e1d", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {"Property destruction": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "27525"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects items by multiple features (round and red)\nCurriculum: LRFFC L2\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by multiple features (round and red)\n- Curriculum: LRFFC L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: apple, plate, glue\nDistractor stimuli: towel, ball, car, book\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "44ae267c347689ef", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by multiple features (round and red)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "55121"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3587\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading\nDate range: Sessions 1\u20138 across 16 days (2026-01-20 to 2026-02-05)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. requesting preferred items using a full sentence (Mand)\n Method: NET\n3. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n4. requesting a break during non-preferred tasks (Mand)\n Method: NET\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: tangible\n2. Self-injurious behavior (SIB) \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-20 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 10/13 correct (74%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d79\n requesting preferred items using a full sentence: 7/13 correct (54%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n indicating 'all done' when an activity is complete: 8/15 correct (56%); latency 3.2s; prompts FP\u00d75, PP\u00d75, G\u00d72, I\u00d73\n requesting a break during non-preferred tasks: 8/14 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (62%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Tantrum: freq 2, duration 8m total\n Self-injurious behavior (SIB): freq 4\n\nSession 2 \u2014 2026-01-22 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 9/13 correct (70%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n requesting preferred items using a full sentence: 7/12 correct (56%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n indicating 'all done' when an activity is complete: 6/11 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n requesting a break during non-preferred tasks: 7/11 correct (61%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (61%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n Tantrum: freq 2, duration 8m total\n Self-injurious behavior (SIB): freq 4\n\nSession 3 \u2014 2026-01-24 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 7/10 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n requesting preferred items using a full sentence: 7/13 correct (56%); latency 3.2s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n indicating 'all done' when an activity is complete: 7/13 correct (56%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n requesting a break during non-preferred tasks: 5/8 correct (56%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (60%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n Tantrum: freq 7, duration 35m total\n Self-injurious behavior (SIB): freq 9\n\nSession 4 \u2014 2026-01-26 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 9/13 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n requesting preferred items using a full sentence: 8/13 correct (59%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n indicating 'all done' when an activity is complete: 5/8 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n requesting a break during non-preferred tasks: 9/14 correct (67%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d714\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (69%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n Tantrum: freq 7, duration 21m total\n Self-injurious behavior (SIB): freq 9\n\nSession 5 \u2014 2026-01-28 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 9/13 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n requesting preferred items using a full sentence: 5/10 correct (54%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n indicating 'all done' when an activity is complete: 9/15 correct (60%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d76\n requesting a break during non-preferred tasks: 10/15 correct (67%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (70%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n Tantrum: freq 3, duration 6m total\n Self-injurious behavior (SIB): freq 3\n\nSession 6 \u2014 2026-01-30 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 7/9 correct (75%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n requesting preferred items using a full sentence: 8/15 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d75, G\u00d73, I\u00d75\n indicating 'all done' when an activity is complete: 5/8 correct (64%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting a break during non-preferred tasks: 10/15 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (67%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n Tantrum: freq 3, duration 9m total\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-02-01 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 9/13 correct (72%); latency 2.8s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d74\n requesting preferred items using a full sentence: 5/9 correct (53%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n indicating 'all done' when an activity is complete: 5/9 correct (61%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n requesting a break during non-preferred tasks: 6/9 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/14 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d76\n Tantrum: freq 2, duration 8m total\n Self-injurious behavior (SIB): freq 4\n\nSession 8 \u2014 2026-02-03 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 12/15 correct (81%); latency 2.6s; prompts FP\u00d74, PP\u00d72, G\u00d75, I\u00d74\n requesting preferred items using a full sentence: 6/10 correct (62%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n indicating 'all done' when an activity is complete: 6/10 correct (59%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n requesting a break during non-preferred tasks: 8/12 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (74%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n Tantrum: freq 2, duration 4m total\n Self-injurious behavior (SIB): freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Peer aggression (hitting or pushing peers during group activities)\n- Vocal refusal ('no', 'I don't want to')\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nTantrum: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\nSelf-injurious behavior (SIB): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 72.2%.\n- Trend direction: upward (first session: 74%, last session: 81%).\n- Tantrum: mean 3.5/session (early-window mean 2.0 -> late-window mean 2.0).\n- Self-injurious behavior (SIB): mean 4.9/session (early-window mean 4.0 -> late-window mean 3.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "3a35ed7991f57d89", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Tantrum": "tangible", "Self-injurious behavior (SIB)": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["tantrum", "sib"], "behavior_functions": ["tangible", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "20202"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: completes 8\u201312 piece puzzles\nCurriculum Reference: Visual Perceptual Skills & Matching-to-Sample L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completes 8\u201312 piece puzzles\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L2\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: jacket, blanket, book\nDistractor stimuli: car, bottle, crayon, glue\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "df2e926d82f29b0f", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L2", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "completes 8\u201312 piece puzzles", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "26013"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5628\nProfile: Adult Learner (18+ years)\nCurricula: Community + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u201312 across 24 days (2026-10-22 to 2026-11-15)\n\nACCELERATION PROGRAMS\n1. clocking in at start of shift (Vocational)\n Method: Task Analysis\n2. using public transportation (Community)\n Method: Task Analysis\n3. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-22 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 7/9 correct (75%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n using public transportation: 9/10 correct (85%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n wait calmly when a preferred activity is denied: 12/14 correct (87%); latency 2.3s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n Verbal aggression: freq 0\n\nSession 2 \u2014 2026-10-24 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 2/8 correct (28%); latency 4.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n using public transportation: 3/9 correct (36%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 4/8 correct (47%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Verbal aggression: freq 1\n\nSession 3 \u2014 2026-10-26 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 9/12 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n using public transportation: 9/10 correct (87%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 10/12 correct (82%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n Verbal aggression: freq 1\n ABC (verbal aggression): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 4 \u2014 2026-10-28 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 3/13 correct (25%); latency 5.1s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n using public transportation: 4/8 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 6/12 correct (48%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d77\n Verbal aggression: freq 0\n\nSession 5 \u2014 2026-10-30 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 8/10 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n using public transportation: 11/13 correct (87%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n wait calmly when a preferred activity is denied: 7/8 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Verbal aggression: freq 0\n\nSession 6 \u2014 2026-11-01 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 4/12 correct (35%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n using public transportation: 4/11 correct (36%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 6/15 correct (41%); latency 4.4s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d79\n Verbal aggression: freq 0\n\nSession 7 \u2014 2026-11-03 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 7/9 correct (77%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n using public transportation: 13/15 correct (84%); latency 1.9s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d79\n wait calmly when a preferred activity is denied: 11/13 correct (87%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n Verbal aggression: freq 1\n ABC (verbal aggression): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 8 \u2014 2026-11-05 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 3/8 correct (32%); latency 4.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n using public transportation: 6/14 correct (42%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 6/13 correct (43%); latency 4.3s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n Verbal aggression: freq 0\n\nSession 9 \u2014 2026-11-07 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 10/14 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d710\n using public transportation: 8/10 correct (84%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 7/8 correct (90%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n Verbal aggression: freq 0\n\nSession 10 \u2014 2026-11-09 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 3/13 correct (26%); latency 4.8s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d77\n using public transportation: 6/15 correct (42%); latency 4.3s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d712\n wait calmly when a preferred activity is denied: 7/15 correct (45%); latency 3.8s; prompts FP\u00d75, PP\u00d71, G\u00d74, I\u00d75\n Verbal aggression: freq 0\n\nSession 11 \u2014 2026-11-11 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 6/8 correct (72%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n using public transportation: 8/10 correct (82%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 11/12 correct (89%); latency 2.1s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n Verbal aggression: freq 0\n\nSession 12 \u2014 2026-11-13 \u2014 44 min \u2014 1 observer\n clocking in at start of shift: 4/11 correct (33%); latency 4.6s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n using public transportation: 6/13 correct (44%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d77\n wait calmly when a preferred activity is denied: 4/8 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n Verbal aggression: freq 0\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Behavior Function Hypothesis\nVerbal aggression: attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 52.8%.\n- Trend direction: downward (first session: 75%, last session: 33%).\n- Verbal aggression: mean 0.2/session (early-window mean 0.5 -> late-window mean 0.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "528ab18e07269e19", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {"Verbal aggression": "attention"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 12, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["verbal_aggression"], "behavior_functions": ["attention"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "43875"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: identifies all upper-case letters\nCurriculum: Reading L2\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies all upper-case letters\n- Curriculum: Reading L2\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: Q, C, V\nDistractor stimuli: T, J, R, E\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c5030be1cf3f140d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L2", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies all upper-case letters", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "67967"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tolerating hair washing by caregiver\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Near mastery\nTeaching Method: Toleration / Systematic Desensitization\n\nThe goal is for the learner to allow the activity to be performed (not to perform it independently). Provide the full program structure including shaping steps and refusal/safety procedures."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating hair washing by caregiver\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for hair washing / shampooing performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull hair-washing routine completed by caregiver without refusal or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 5 seconds of water on hair\n2. 15 seconds of water on hair\n3. Shampoo applied without lathering\n4. 30 seconds of gentle lathering\n5. 60 seconds of lathering\n6. Full rinse\n7. Full routine: wet -> shampoo -> lather -> rinse\n\nCurrent shaping step: At the near mastery stage, begin at shaping step 4 (\"30 seconds of gentle lathering\") and advance through step 5 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal duration tolerated across 3 consecutive sessions with no safety behaviors (refusal, crying, SIB, aggression, property destruction).\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 2 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ca0ed5017900e8bb", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating hair washing by caregiver", "program_type": "toleration", "learner_profile": "adolescent", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating hair washing by caregiver", "module": "basic_living", "mastery_criterion": "mc_toleration_full_duration", "shaping_n_steps": 7}, "teacher_model": null, "seed_tag": "6035"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by negative features (not round, not food)\nCurriculum: LRFFC L3\nLearner: Adolescent Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by negative features (not round, not food)\n- Curriculum: LRFFC L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: spoon, block, scissors\nDistractor stimuli: car, blanket, glue, phone, hat\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "415529e159dc7106", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by negative features (not round, not food)", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "vr3", "error_correction": "simple_correction", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "89319"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: completing assigned tasks in order\nCurriculum: AFLS Vocational Skills\nLearner: Adolescent Learner\nMastery Status: Near mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Review the task list for the shift.\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "64242a99c9db119a", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "near", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_self_initiation", "chain_type": "backward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "6490"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by feature (find something red)\nCurriculum: Listener Responding L2\nLearner: Early Learner\nMastery Status: Approaching mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by feature (find something red)\n- Curriculum: Listener Responding L2\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: sock, hat, glue\nDistractor stimuli: backpack, block, car, book\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "192bece913e456b9", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L2", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by feature (find something red)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "70400"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by class with exclusion (animal but not a pet)\nCurriculum: LRFFC L3\nLearner: Adolescent Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by class with exclusion (animal but not a pet)\n- Curriculum: LRFFC L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: cow, dog, horse\nDistractor stimuli: lion, cat, frog, bird, pig\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "73c028cd108ed571", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by class with exclusion (animal but not a pet)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "vr3", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "87673"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: initiates social interactions with peers\nCurriculum Reference: Social Behavior & Social Play L2\nLearner Profile: Early Learner\nCurrent Mastery: Generalization phase\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: initiates social interactions with peers\n- Curriculum: Social Behavior & Social Play L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "384001046494b09e", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "initiates social interactions with peers", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "75835"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5920\nProfile: Adult Learner (18+ years)\nCurricula: Community + Independent Living + Self-Care + Vocational\nDate range: Sessions 1\u20139 across 18 days (2026-02-04 to 2026-02-22)\n\nACCELERATION PROGRAMS\n1. self-advocacy statements (Independent Living)\n Method: BST\n2. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n3. washing hands independently (Self-Care)\n Method: Task Analysis\n4. clocking in at start of shift (Vocational)\n Method: Task Analysis\n5. using public transportation (Community)\n Method: Task Analysis\n6. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-04 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 5/11 correct (49%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n scheduling medical appointments: 7/15 correct (45%); latency 3.9s; prompts FP\u00d75, PP\u00d73, G\u00d73, I\u00d74\n washing hands independently: 7/15 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d713\n clocking in at start of shift: 6/14 correct (43%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n using public transportation: 3/8 correct (34%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n completing assigned tasks in order: 6/14 correct (42%); latency 4.3s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n\nSession 2 \u2014 2026-02-06 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 5/9 correct (53%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n scheduling medical appointments: 6/13 correct (49%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d711\n washing hands independently: 5/9 correct (55%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n clocking in at start of shift: 7/15 correct (50%); latency 3.9s; prompts FP\u00d75, PP\u00d74, G\u00d70, I\u00d76\n using public transportation: 3/8 correct (37%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n completing assigned tasks in order: 8/15 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n\nSession 3 \u2014 2026-02-08 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 7/12 correct (58%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n scheduling medical appointments: 7/12 correct (57%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n washing hands independently: 6/10 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n clocking in at start of shift: 6/11 correct (55%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n using public transportation: 6/12 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d76\n completing assigned tasks in order: 4/8 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n\nSession 4 \u2014 2026-02-10 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 8/12 correct (65%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n scheduling medical appointments: 9/15 correct (62%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n washing hands independently: 9/12 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n clocking in at start of shift: 6/9 correct (65%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n using public transportation: 7/14 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d710\n completing assigned tasks in order: 6/11 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n\nSession 5 \u2014 2026-02-12 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 6/8 correct (74%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n scheduling medical appointments: 9/13 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n washing hands independently: 7/9 correct (76%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n clocking in at start of shift: 6/9 correct (71%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n using public transportation: 5/8 correct (59%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n completing assigned tasks in order: 8/12 correct (67%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n\nSession 6 \u2014 2026-02-14 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 11/14 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d711\n scheduling medical appointments: 10/14 correct (71%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n washing hands independently: 11/13 correct (83%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d79\n clocking in at start of shift: 6/8 correct (76%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n using public transportation: 7/10 correct (65%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n completing assigned tasks in order: 10/15 correct (66%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n\nSession 7 \u2014 2026-02-16 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 11/13 correct (82%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n scheduling medical appointments: 8/11 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n washing hands independently: 10/11 correct (93%); latency 1.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n clocking in at start of shift: 10/12 correct (84%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n using public transportation: 9/14 correct (67%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d74\n completing assigned tasks in order: 7/9 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n\nSession 8 \u2014 2026-02-18 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 9/9 correct (95%); latency 1.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n scheduling medical appointments: 12/15 correct (83%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d710\n washing hands independently: 12/13 correct (95%); latency 1.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n clocking in at start of shift: 14/15 correct (91%); latency 2.2s; prompts FP\u00d74, PP\u00d73, G\u00d75, I\u00d73\n using public transportation: 8/11 correct (77%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n completing assigned tasks in order: 7/9 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n\nSession 9 \u2014 2026-02-20 \u2014 51 min \u2014 1 observer\n self-advocacy statements: 8/8 correct (95%); latency 1.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n scheduling medical appointments: 9/11 correct (86%); latency 2.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n washing hands independently: 10/11 correct (95%); latency 1.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n clocking in at start of shift: 10/10 correct (95%); latency 1.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n using public transportation: 7/9 correct (80%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n completing assigned tasks in order: 9/11 correct (82%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Smiling during reinforcement delivery\n- Appropriate sitting posture maintained"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 49% to 95% across 9 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 72.4%.\n- Trend direction: upward (first session: 49%, last session: 95%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "ada523a1b619f056", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "adult", "n_sessions": 9, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "82738"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4848\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Social Behavior\nDate range: Sessions 1\u20138 across 16 days (2026-09-28 to 2026-10-14)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n4. requesting a break during non-preferred tasks (Mand)\n Method: NET\n5. requesting preferred items using a full sentence (Mand)\n Method: NET\n6. reads 10\u201320 sight words (Reading)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-28 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 7/9 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/9 correct (86%); latency 2.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n takes turns during structured activities: 9/12 correct (78%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n requesting a break during non-preferred tasks: 12/15 correct (77%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n requesting preferred items using a full sentence: 13/15 correct (84%); latency 2.3s; prompts FP\u00d75, PP\u00d71, G\u00d72, I\u00d77\n reads 10\u201320 sight words: 9/12 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d76\n\nSession 2 \u2014 2026-09-30 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 9/12 correct (77%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/13 correct (88%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d79\n takes turns during structured activities: 9/11 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n requesting a break during non-preferred tasks: 12/15 correct (81%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n requesting preferred items using a full sentence: 12/14 correct (84%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n reads 10\u201320 sight words: 11/14 correct (78%); latency 2.2s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d74\n\nSession 3 \u2014 2026-10-02 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 10/14 correct (75%); latency 2.8s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/10 correct (85%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n takes turns during structured activities: 6/8 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n requesting a break during non-preferred tasks: 9/10 correct (85%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n requesting preferred items using a full sentence: 8/9 correct (85%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 11/14 correct (79%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n\nSession 4 \u2014 2026-10-04 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 10/13 correct (78%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (88%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n takes turns during structured activities: 10/12 correct (82%); latency 2.2s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n requesting a break during non-preferred tasks: 8/9 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n requesting preferred items using a full sentence: 8/9 correct (91%); latency 1.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n reads 10\u201320 sight words: 8/10 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n\nSession 5 \u2014 2026-10-06 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 8/11 correct (77%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n takes turns during structured activities: 9/11 correct (82%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n requesting a break during non-preferred tasks: 8/10 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n requesting preferred items using a full sentence: 7/8 correct (93%); latency 1.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 12/15 correct (81%); latency 2.5s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d76\n\nSession 6 \u2014 2026-10-08 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 7/8 correct (82%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/11 correct (81%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n takes turns during structured activities: 11/13 correct (81%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n requesting a break during non-preferred tasks: 7/8 correct (86%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n requesting preferred items using a full sentence: 8/9 correct (91%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n reads 10\u201320 sight words: 6/8 correct (80%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 7 \u2014 2026-10-10 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 6/8 correct (81%); latency 2.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/12 correct (83%); latency 2.2s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n takes turns during structured activities: 12/14 correct (84%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n requesting a break during non-preferred tasks: 12/14 correct (88%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n requesting preferred items using a full sentence: 8/9 correct (89%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n reads 10\u201320 sight words: 7/8 correct (84%); latency 2.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 8 \u2014 2026-10-12 \u2014 37 min \u2014 1 observer\n identifies numerals 1\u201320: 9/11 correct (81%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/13 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d79\n takes turns during structured activities: 7/8 correct (86%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n requesting a break during non-preferred tasks: 10/11 correct (92%); latency 1.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n requesting preferred items using a full sentence: 11/12 correct (93%); latency 2.0s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d77\n reads 10\u201320 sight words: 8/10 correct (83%); latency 2.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (74%\u201381%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 78.1%.\n- Trend direction: upward (first session: 74%, last session: 81%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "74a6988f11847ff8", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "83942"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously asks questions about the environment\nCurriculum: Spontaneous Vocal Behavior L2\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously asks questions about the environment\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: activity setup.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's activity setup routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "93d21e0d53ad53eb", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously asks questions about the environment", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "77480"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9879\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Motor Imitation + Writing\nDate range: Sessions 1\u20139 across 18 days (2026-05-03 to 2026-05-21)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n3. selects items by feature (Listener Responding)\n Method: DTT\n4. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n5. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-03 \u2014 47 min \u2014 1 observer\n copies simple words from model: 12/15 correct (81%); latency 2.4s; prompts FP\u00d74, PP\u00d75, G\u00d74, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/8 correct (87%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n selects items by feature: 10/12 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 8/11 correct (76%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n imitates 3-step motor sequences: 10/13 correct (79%); latency 2.7s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n Elopement: freq 2\n\nSession 2 \u2014 2026-05-05 \u2014 47 min \u2014 1 observer\n copies simple words from model: 7/10 correct (74%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/10 correct (82%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n selects items by feature: 6/9 correct (72%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 9/12 correct (74%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d78\n imitates 3-step motor sequences: 8/11 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n Elopement: freq 2\n\nSession 3 \u2014 2026-05-07 \u2014 47 min \u2014 1 observer\n copies simple words from model: 5/8 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/11 correct (78%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n selects items by feature: 7/10 correct (67%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 5/8 correct (67%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n imitates 3-step motor sequences: 8/12 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n Elopement: freq 3\n\nSession 4 \u2014 2026-05-09 \u2014 47 min \u2014 1 observer\n copies simple words from model: 7/10 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/8 correct (69%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n selects items by feature: 7/12 correct (57%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 6/10 correct (55%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n imitates 3-step motor sequences: 10/15 correct (68%); latency 2.8s; prompts FP\u00d74, PP\u00d75, G\u00d74, I\u00d72\n Elopement: freq 3\n\nSession 5 \u2014 2026-05-11 \u2014 47 min \u2014 1 observer\n copies simple words from model: 5/9 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/13 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n selects items by feature: 5/11 correct (49%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 4/9 correct (48%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n imitates 3-step motor sequences: 7/13 correct (53%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n Elopement: freq 5\n\nSession 6 \u2014 2026-05-13 \u2014 47 min \u2014 1 observer\n copies simple words from model: 6/12 correct (49%); latency 4.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (58%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d710\n selects items by feature: 6/15 correct (40%); latency 4.2s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 5/11 correct (42%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n imitates 3-step motor sequences: 6/12 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n Elopement: freq 4\n\nSession 7 \u2014 2026-05-15 \u2014 47 min \u2014 1 observer\n copies simple words from model: 3/8 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/9 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n selects items by feature: 3/10 correct (32%); latency 4.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 6/15 correct (39%); latency 4.3s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d76\n imitates 3-step motor sequences: 4/8 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n Elopement: freq 4\n\nSession 8 \u2014 2026-05-17 \u2014 47 min \u2014 1 observer\n copies simple words from model: 4/11 correct (38%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/14 correct (51%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n selects items by feature: 3/9 correct (33%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 4/12 correct (30%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n imitates 3-step motor sequences: 4/9 correct (46%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n Elopement: freq 5\n\nSession 9 \u2014 2026-05-19 \u2014 47 min \u2014 1 observer\n copies simple words from model: 3/8 correct (35%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/13 correct (46%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n selects items by feature: 3/14 correct (18%); latency 5.1s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 3/11 correct (26%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n imitates 3-step motor sequences: 3/9 correct (37%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n Elopement: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Attempts at property destruction (throwing materials, swiping items off the surface)\n- Elopement from the work area\n- Pushing materials away from the work area"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (81% -> 35%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 57.1%.\n- Trend direction: downward (first session: 81%, last session: 35%).\n- Elopement: mean 3.8/session (early-window mean 2.3 -> late-window mean 5.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "cbdf4b89769de62b", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Elopement": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "early", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "20942"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: taking medication as prescribed\nCurriculum Reference: AFLS Independent Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Generalization phase\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: taking medication as prescribed\n- Curriculum: AFLS Independent Living Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Check the medication schedule at the scheduled time. <- Learner-taught step (forward chaining starts here)\n2. Collect the correct medication bottle.\n3. Confirm the dose against the label.\n4. Measure or count the correct dose.\n5. Take the medication with water.\n6. Log the dose in the medication tracker.\n7. Return the bottle to its storage location.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fe5e7f4a54c58cfb", "gold_labels": {"method": "task_analysis", "domain": "AFLS.independent_living", "skill": "taking medication as prescribed", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "taking medication as prescribed", "module": "independent_living", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "75542"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4370\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Vocational\nDate range: Sessions 1\u201312 across 24 days (2026-04-11 to 2026-05-05)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. managing a budget (Independent Living)\n Method: Task Analysis\n3. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-11 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 10/13 correct (75%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n managing a budget: 8/11 correct (75%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n scheduling medical appointments: 12/14 correct (86%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d711\n Property destruction: freq 1\n ABC (property destruction): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 2 \u2014 2026-04-13 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 8/12 correct (66%); latency 3.2s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n managing a budget: 6/9 correct (67%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n scheduling medical appointments: 8/10 correct (77%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Property destruction: freq 2\n\nSession 3 \u2014 2026-04-15 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 8/13 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n managing a budget: 8/14 correct (59%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n scheduling medical appointments: 12/15 correct (79%); latency 2.3s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d77\n Property destruction: freq 1\n\nSession 4 \u2014 2026-04-17 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 9/15 correct (59%); latency 3.4s; prompts FP\u00d73, PP\u00d75, G\u00d74, I\u00d73\n managing a budget: 8/14 correct (58%); latency 3.7s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d77\n scheduling medical appointments: 7/11 correct (67%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n Property destruction: freq 2\n ABC (property destruction): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 5 \u2014 2026-04-19 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 6/12 correct (54%); latency 3.4s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n managing a budget: 3/8 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n scheduling medical appointments: 8/13 correct (61%); latency 3.0s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d75\n Property destruction: freq 3\n\nSession 6 \u2014 2026-04-21 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 5/10 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n managing a budget: 4/10 correct (42%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n scheduling medical appointments: 5/9 correct (59%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n Property destruction: freq 3\n\nSession 7 \u2014 2026-04-23 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 6/14 correct (45%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n managing a budget: 4/13 correct (31%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n scheduling medical appointments: 5/9 correct (60%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n Property destruction: freq 3\n ABC (property destruction): A = staff engaged in paperwork nearby; B = yelled name + tapped staff; C = staff attended and redirected\n\nSession 8 \u2014 2026-04-25 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 4/9 correct (45%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n managing a budget: 4/14 correct (26%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n scheduling medical appointments: 7/13 correct (52%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n Property destruction: freq 3\n\nSession 9 \u2014 2026-04-27 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 4/11 correct (37%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n managing a budget: 2/9 correct (22%); latency 4.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n scheduling medical appointments: 6/13 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d77\n Property destruction: freq 3\n\nSession 10 \u2014 2026-04-29 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 3/12 correct (27%); latency 5.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n managing a budget: 1/8 correct (17%); latency 5.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n scheduling medical appointments: 5/13 correct (38%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n Property destruction: freq 3\n\nSession 11 \u2014 2026-05-01 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 2/8 correct (22%); latency 5.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n managing a budget: 1/11 correct (10%); latency 5.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n scheduling medical appointments: 4/10 correct (40%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n Property destruction: freq 5\n\nSession 12 \u2014 2026-05-03 \u2014 56 min \u2014 1 observer\n completing assigned tasks in order: 2/10 correct (18%); latency 5.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n managing a budget: 1/11 correct (10%); latency 5.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n scheduling medical appointments: 4/12 correct (37%); latency 4.2s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d78\n Property destruction: freq 4\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-stimulatory behavior increase during demand periods\n- Elopement from the work area\n- Crying during demand presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 75% to 18% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nProperty destruction: attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 47.3%.\n- Trend direction: downward (first session: 75%, last session: 18%).\n- Property destruction: mean 2.8/session (early-window mean 1.5 -> late-window mean 3.8).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "d4c4852748046c1f", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Property destruction": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adult", "n_sessions": 12, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["attention"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "71491"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: identifies coins by name\nCurriculum: Math L2\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies coins by name\n- Curriculum: Math L2\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: book, phone, cup\nDistractor stimuli: brush, sock, plate, car\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0f6185c89d45e3f4", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L2", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies coins by name", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "20226"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: negotiates and compromises during group activities\nCurriculum Reference: Social Behavior & Social Play L3\nLearner Profile: School-Age Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: negotiates and compromises during group activities\n- Curriculum: Social Behavior & Social Play L3\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "dcc3298d58b6c0c6", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L3", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "negotiates and compromises during group activities", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "time_delay", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "30191"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3264\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care\nDate range: Sessions 1\u20139 across 18 days (2026-07-06 to 2026-07-24)\n\nACCELERATION PROGRAMS\n1. washing hands independently (Self-Care)\n Method: Task Analysis\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. requesting a break during extended tasks (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Vocal stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-06 \u2014 40 min \u2014 1 observer\n washing hands independently: 6/14 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n self-advocacy statements: 5/10 correct (47%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n requesting a break during extended tasks: 8/11 correct (73%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n Vocal stereotypy: freq 0; PIR 2%\n\nSession 2 \u2014 2026-07-08 \u2014 40 min \u2014 1 observer\n washing hands independently: 4/9 correct (49%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n self-advocacy statements: 6/8 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n requesting a break during extended tasks: 6/9 correct (72%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n Vocal stereotypy: freq 3; PIR 19%\n\nSession 3 \u2014 2026-07-10 \u2014 40 min \u2014 1 observer\n washing hands independently: 6/9 correct (72%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n self-advocacy statements: 10/14 correct (71%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d713\n requesting a break during extended tasks: 7/12 correct (59%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n Vocal stereotypy: freq 4; PIR 21%\n\nSession 4 \u2014 2026-07-12 \u2014 40 min \u2014 1 observer\n washing hands independently: 5/14 correct (34%); latency 4.5s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n self-advocacy statements: 9/15 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n requesting a break during extended tasks: 9/15 correct (62%); latency 3.5s; prompts FP\u00d75, PP\u00d73, G\u00d72, I\u00d75\n Vocal stereotypy: freq 1; PIR 9%\n\nSession 5 \u2014 2026-07-14 \u2014 IOA SESSION \u2014 2 observers\n washing hands independently: 58% (primary); IOA 91% agreement\n self-advocacy statements: 40% (primary); IOA 91% agreement\n requesting a break during extended tasks: 51% (primary); IOA 91% agreement\n Vocal stereotypy: freq 4; PIR 20%; IOA 91% agreement\n\nSession 6 \u2014 2026-07-16 \u2014 40 min \u2014 1 observer\n washing hands independently: 4/8 correct (47%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n self-advocacy statements: 8/12 correct (68%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d79\n requesting a break during extended tasks: 7/10 correct (74%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n Vocal stereotypy: freq 2; PIR 17%\n\nSession 7 \u2014 2026-07-18 \u2014 40 min \u2014 1 observer\n washing hands independently: 9/13 correct (70%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n self-advocacy statements: 6/10 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n requesting a break during extended tasks: 9/11 correct (78%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n Vocal stereotypy: freq 4; PIR 28%\n\nSession 8 \u2014 2026-07-20 \u2014 40 min \u2014 1 observer\n washing hands independently: 10/15 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d712\n self-advocacy statements: 7/11 correct (61%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n requesting a break during extended tasks: 5/8 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n Vocal stereotypy: freq 1; PIR 3%\n\nSession 9 \u2014 2026-07-22 \u2014 40 min \u2014 1 observer\n washing hands independently: 6/15 correct (43%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d76\n self-advocacy statements: 7/11 correct (65%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n requesting a break during extended tasks: 7/12 correct (54%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n Vocal stereotypy: freq 2; PIR 9%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Leaving seat without permission\n- Playing with non-task materials\n- Echolalic responding (repeating SD rather than answering)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 45% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nVocal stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 54.2%.\n- Trend direction: downward (first session: 46%, last session: 43%).\n- Vocal stereotypy: mean 2.3/session (early-window mean 2.3 -> late-window mean 2.3).\n- IOA data: 91% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "1d8c7a27726d1f40", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Vocal stereotypy": "automatic"}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adult", "n_sessions": 9, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["vocal_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "20321"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4052\nProfile: Adult Learner (18+ years)\nCurricula: Self-Care + Vocational\nDate range: Sessions 1\u20137 across 14 days (2026-02-06 to 2026-02-20)\n\nACCELERATION PROGRAMS\n1. clocking in at start of shift (Vocational)\n Method: Task Analysis\n2. washing hands independently (Self-Care)\n Method: Task Analysis\n3. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-06 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 3/8 correct (33%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n washing hands independently: 3/12 correct (27%); latency 4.5s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d75\n dressing (pants, shirt): 3/13 correct (25%); latency 5.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n\nSession 2 \u2014 2026-02-08 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 4/9 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n washing hands independently: 4/11 correct (40%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n dressing (pants, shirt): 4/11 correct (33%); latency 4.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n\nSession 3 \u2014 2026-02-10 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 4/9 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n washing hands independently: 7/15 correct (45%); latency 4.2s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d76\n dressing (pants, shirt): 4/9 correct (40%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n\nSession 4 \u2014 2026-02-12 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 5/10 correct (53%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n washing hands independently: 9/15 correct (57%); latency 3.2s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d78\n dressing (pants, shirt): 7/14 correct (51%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n\nSession 5 \u2014 2026-02-14 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 7/12 correct (62%); latency 3.1s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n washing hands independently: 7/11 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n dressing (pants, shirt): 6/10 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n\nSession 6 \u2014 2026-02-16 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 10/15 correct (70%); latency 2.9s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n washing hands independently: 8/10 correct (78%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n dressing (pants, shirt): 10/15 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n\nSession 7 \u2014 2026-02-18 \u2014 35 min \u2014 1 observer\n clocking in at start of shift: 9/12 correct (78%); latency 2.3s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d74\n washing hands independently: 7/8 correct (88%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n dressing (pants, shirt): 10/14 correct (72%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Eye contact maintained throughout trials\n- Quick response latency (within 2\u20133 seconds)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 33% to 78% across 7 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 55.3%.\n- Trend direction: upward (first session: 33%, last session: 78%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "b46965baacde72dd", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "adult", "n_sessions": 7, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "96053"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4608\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Self-Care\nDate range: Sessions 1\u20137 across 14 days (2026-04-07 to 2026-04-21)\n\nACCELERATION PROGRAMS\n1. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n2. requesting help when a task is difficult (Mand)\n Method: NET\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-07 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 4/8 correct (46%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n requesting help when a task is difficult: 6/12 correct (50%); latency 3.6s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n brushing teeth independently: 6/15 correct (38%); latency 4.6s; prompts FP\u00d75, PP\u00d74, G\u00d73, I\u00d73\n Non-compliance: freq 2\n\nSession 2 \u2014 2026-04-09 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 6/12 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n requesting help when a task is difficult: 7/11 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n brushing teeth independently: 5/10 correct (48%); latency 4.0s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n Non-compliance: freq 2\n\nSession 3 \u2014 2026-04-11 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 7/11 correct (62%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n requesting help when a task is difficult: 7/10 correct (66%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n brushing teeth independently: 6/11 correct (57%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d79\n Non-compliance: freq 2\n\nSession 4 \u2014 2026-04-13 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 8/11 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n requesting help when a task is difficult: 10/13 correct (76%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n brushing teeth independently: 10/14 correct (68%); latency 2.9s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n Non-compliance: freq 2\n\nSession 5 \u2014 2026-04-15 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 9/11 correct (78%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n requesting help when a task is difficult: 9/11 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n brushing teeth independently: 9/12 correct (76%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n Non-compliance: freq 2\n\nSession 6 \u2014 2026-04-17 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 7/9 correct (82%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n requesting help when a task is difficult: 8/9 correct (84%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n brushing teeth independently: 11/14 correct (76%); latency 2.5s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n Non-compliance: freq 1\n\nSession 7 \u2014 2026-04-19 \u2014 45 min \u2014 1 observer\n dressing (pants, shirt): 11/13 correct (82%); latency 2.4s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d74\n requesting help when a task is difficult: 11/13 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n brushing teeth independently: 10/13 correct (76%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d76\n Non-compliance: freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 46% at return, recovering to 82%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 67.8%.\n- Trend direction: upward (first session: 46%, last session: 82%).\n- Non-compliance: mean 2.0/session (early-window mean 2.0 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "2ed1e4d10b470ad7", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Non-compliance": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adolescent", "n_sessions": 7, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["non_compliance"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "99687"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: responds to own name\nCurriculum: Social Behavior & Social Play L1\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: responds to own name\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: peer play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's peer play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "c8e9b465cb3fbf39", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "responds to own name", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "time_delay", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "75835"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: using public transportation\nCurriculum: AFLS Community Participation Skills\nLearner: Adult Learner\nMastery Status: Emerging\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: using public transportation\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Confirm the correct stop or station. <- Learner-taught step (forward chaining starts here)\n2. Purchase or have the fare ready.\n3. Wait at the designated boarding area.\n4. Board the vehicle when it arrives.\n5. Pay the fare or present the pass.\n6. Sit or stand safely.\n7. Monitor the route for the correct stop.\n8. Exit at the correct stop.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the emerging stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "23b8b87f077c7021", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "using public transportation", "program_type": "independence", "learner_profile": "adult", "mastery_state": "emerging", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "using public transportation", "module": "community", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "91439"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1299\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-10-17 to 2026-11-08)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n3. mands for missing items needed to complete an activity (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-17 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/12 correct (88%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n follows 1-step motor instructions: 9/12 correct (77%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n mands for missing items needed to complete an activity: 7/9 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 2 \u2014 2026-10-19 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/11 correct (90%); latency 2.0s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n follows 1-step motor instructions: 7/9 correct (82%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n mands for missing items needed to complete an activity: 11/14 correct (76%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-10-21 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/9 correct (91%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n follows 1-step motor instructions: 10/13 correct (79%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n mands for missing items needed to complete an activity: 11/15 correct (76%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d75, I\u00d76\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 4 \u2014 2026-10-23 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/9 correct (92%); latency 1.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n follows 1-step motor instructions: 11/14 correct (80%); latency 2.4s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n mands for missing items needed to complete an activity: 9/12 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 5 \u2014 2026-10-25 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/12 correct (95%); latency 2.0s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d71\n follows 1-step motor instructions: 11/13 correct (82%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n mands for missing items needed to complete an activity: 11/15 correct (74%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 6 \u2014 2026-10-27 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/12 correct (92%); latency 1.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n follows 1-step motor instructions: 10/12 correct (80%); latency 2.6s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d74\n mands for missing items needed to complete an activity: 8/11 correct (73%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n Toileting accident (urine or bowel): urine: 5 in-toilet / 0 accidents; BM: 1 in-toilet / 1 accidents\n\nSession 7 \u2014 2026-10-29 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/8 correct (97%); latency 1.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n follows 1-step motor instructions: 10/12 correct (86%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n mands for missing items needed to complete an activity: 7/9 correct (74%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 8 \u2014 2026-10-31 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/9 correct (95%); latency 1.5s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n follows 1-step motor instructions: 10/12 correct (82%); latency 2.3s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n mands for missing items needed to complete an activity: 5/8 correct (66%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 9 \u2014 2026-11-02 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 12/13 correct (94%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d76\n follows 1-step motor instructions: 12/14 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n mands for missing items needed to complete an activity: 6/9 correct (66%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 10 \u2014 2026-11-04 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/8 correct (98%); latency 1.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n follows 1-step motor instructions: 13/15 correct (83%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n mands for missing items needed to complete an activity: 6/9 correct (63%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 11 \u2014 2026-11-06 \u2014 44 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 15/15 correct (98%); latency 1.7s; prompts FP\u00d72, PP\u00d75, G\u00d72, I\u00d76\n follows 1-step motor instructions: 8/10 correct (83%); latency 2.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n mands for missing items needed to complete an activity: 5/8 correct (66%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Toileting accident (urine or bowel): urine: 5 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (88%\u201398%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 93.7%.\n- Trend direction: upward (first session: 88%, last session: 98%).\n- Toileting accident (urine or bowel): mean 1.6/session (early-window mean 1.7 -> late-window mean 1.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1ad58f4e88363430", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Toileting accident (urine or bowel)": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "early", "n_sessions": 11, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "74744"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands for information using \"when\"\nCurriculum Reference: Mand L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for information using \"when\"\n- Curriculum: Mand L3\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: leisure time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "c9872efb9c14b070", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L3", "learner_profile": "adolescent", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for information using \"when\"", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "36908"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects items by class (find an animal)\nCurriculum Reference: Listener Responding L2\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by class (find an animal)\n- Curriculum: Listener Responding L2\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: dog, chicken, duck\nDistractor stimuli: bear, lion, rabbit, monkey\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "9d204ccd2a2af8db", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L2", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by class (find an animal)", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "four_step", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "87673"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies basic shapes in math context\nCurriculum Reference: Math L1\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies basic shapes in math context\n- Curriculum: Math L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: star, heart, triangle\nDistractor stimuli: diamond, rectangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "47203bda3945e9c0", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies basic shapes in math context", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "69141"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: reproduces block designs from model\nCurriculum Reference: Visual Perceptual Skills & Matching-to-Sample L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Developing\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: reproduces block designs from model\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L3\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: block, spoon, apple\nDistractor stimuli: towel, backpack, glue, jacket, car\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "31ed54f1f680c114", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L3", "learner_profile": "adolescent", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "reproduces block designs from model", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "26013"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5232\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Social Behavior\nDate range: Sessions 1\u201310 across 20 days (2026-06-25 to 2026-07-15)\n\nACCELERATION PROGRAMS\n1. counts objects 1\u201320 (Math)\n Method: DTT\n2. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n3. takes turns during structured activities (Social Behavior)\n Method: PRT\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-25 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 5/14 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d710\n indicating 'all done' when an activity is complete: 5/15 correct (36%); latency 4.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n takes turns during structured activities: 6/13 correct (47%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n\nSession 2 \u2014 2026-06-27 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 5/12 correct (44%); latency 4.2s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d72\n indicating 'all done' when an activity is complete: 3/8 correct (41%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n takes turns during structured activities: 4/8 correct (56%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-06-29 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 6/12 correct (50%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n indicating 'all done' when an activity is complete: 6/12 correct (47%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d75\n takes turns during structured activities: 5/8 correct (64%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n\nSession 4 \u2014 2026-07-01 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 5/9 correct (59%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n indicating 'all done' when an activity is complete: 6/11 correct (52%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n takes turns during structured activities: 7/10 correct (70%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n\nSession 5 \u2014 2026-07-03 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 7/10 correct (68%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n indicating 'all done' when an activity is complete: 6/10 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n takes turns during structured activities: 6/8 correct (71%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n\nSession 6 \u2014 2026-07-05 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 10/13 correct (75%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n indicating 'all done' when an activity is complete: 8/12 correct (69%); latency 3.2s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n takes turns during structured activities: 6/8 correct (71%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 7 \u2014 2026-07-07 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 9/11 correct (77%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n indicating 'all done' when an activity is complete: 11/14 correct (76%); latency 2.9s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d79\n takes turns during structured activities: 7/10 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n\nSession 8 \u2014 2026-07-09 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 10/13 correct (77%); latency 2.3s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 8/11 correct (77%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n takes turns during structured activities: 8/11 correct (71%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n\nSession 9 \u2014 2026-07-11 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 9/12 correct (77%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d77\n indicating 'all done' when an activity is complete: 11/14 correct (77%); latency 2.6s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n takes turns during structured activities: 11/15 correct (71%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d77\n\nSession 10 \u2014 2026-07-13 \u2014 40 min \u2014 1 observer\n counts objects 1\u201320: 10/13 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n indicating 'all done' when an activity is complete: 8/11 correct (77%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n takes turns during structured activities: 10/14 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 37% at return, recovering to 77%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 64.3%.\n- Trend direction: upward (first session: 37%, last session: 77%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "bd40b4ac9a754f4a", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "46674"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: sequences 3\u20134 step picture sequences\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L3\nLearner Level: Adolescent Learner\nBaseline: Mastered at current level\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: sequences 3\u20134 step picture sequences\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: shoe, cup, ball\nDistractor stimuli: car, apple, glue, pencil, backpack\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6284cf37ac9d828d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "sequences 3\u20134 step picture sequences", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "61305"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner: Adult Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 120s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "29258e58319befc3", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "6035"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: putting on a shirt (pullover)\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adolescent Learner\nBaseline: Generalization phase\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: putting on a shirt (pullover)\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Hold the shirt with the front facing away.\n2. Place both arms into the sleeves.\n3. Pull the opening over the head.\n4. Push arms through the sleeves fully.\n5. Pull the bottom of the shirt down to straighten it.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 120s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c1f6dc90eda66233", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "putting on a shirt (pullover)", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "putting on a shirt (pullover)", "module": "basic_living", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "44637"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: identifies all upper-case letters\nCurriculum: Reading L2\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies all upper-case letters\n- Curriculum: Reading L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: X, I, S\nDistractor stimuli: K, T, U, A\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0343089ca4769f77", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies all upper-case letters", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "35263"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3715\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u20138 across 16 days (2026-07-16 to 2026-08-01)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-16 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 12/14 correct (87%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d78\n wait calmly when access to preferred item is denied: 7/10 correct (74%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/12 correct (74%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n Tantrum: freq 2, duration 4m total\n\nSession 2 \u2014 2026-07-18 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 12/15 correct (78%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n wait calmly when access to preferred item is denied: 8/11 correct (72%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/10 correct (69%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n Tantrum: freq 4, duration 12m total\n\nSession 3 \u2014 2026-07-20 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 7/9 correct (78%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 8/11 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/12 correct (60%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d75\n Tantrum: freq 5, duration 25m total\n\nSession 4 \u2014 2026-07-22 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 9/13 correct (69%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n wait calmly when access to preferred item is denied: 8/13 correct (64%); latency 3.0s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/11 correct (47%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n Tantrum: freq 5, duration 15m total\n\nSession 5 \u2014 2026-07-24 \u2014 IOA SESSION \u2014 2 observers\n reads 10\u201320 sight words: 61% (primary); IOA 96% agreement\n wait calmly when access to preferred item is denied: 61% (primary); IOA 96% agreement\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 47% (primary); IOA 96% agreement\n Tantrum: freq 4, duration 8m total; IOA 96% agreement\n\nSession 6 \u2014 2026-07-26 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 7/11 correct (63%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 7/12 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/9 correct (38%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n Tantrum: freq 6, duration 30m total\n\nSession 7 \u2014 2026-07-28 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 7/11 correct (59%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 6/12 correct (52%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/12 correct (26%); latency 4.9s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n Tantrum: freq 7, duration 14m total\n\nSession 8 \u2014 2026-07-30 \u2014 59 min \u2014 1 observer\n reads 10\u201320 sight words: 6/12 correct (49%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n wait calmly when access to preferred item is denied: 4/9 correct (41%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 2/12 correct (17%); latency 4.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n Tantrum: freq 8, duration 24m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Elopement from the work area"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (87% -> 49%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 68.1%.\n- Trend direction: downward (first session: 87%, last session: 49%).\n- Tantrum: mean 5.1/session (early-window mean 3.0 -> late-window mean 7.5).\n- IOA data: 96% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "5d15065360c3bac4", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Tantrum": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "87221"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: engages in social games (peek-a-boo, tickle)\nCurriculum: Social Behavior & Social Play L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: engages in social games (peek-a-boo, tickle)\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "0f39cad4ee379441", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "engages in social games (peek-a-boo, tickle)", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "time_delay", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "83467"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: copies basic shapes from model\nCurriculum: Writing L1\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies basic shapes from model\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: diamond, square, rectangle\nDistractor stimuli: heart, triangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8f62fe68a9d67641", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies basic shapes from model", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "27367"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects items by function (What do you drink from?)\nCurriculum Reference: LRFFC L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function (What do you drink from?)\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: toothbrush, crayon, napkin\nDistractor stimuli: hat, bottle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "047d27b94f192273", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function (What do you drink from?)", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "92219"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7307\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u20137 across 14 days (2026-07-25 to 2026-08-08)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. mands for missing items needed to complete an activity (Mand)\n Method: NET\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. single-word requests for preferred items (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-25 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 6/14 correct (43%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n mands for missing items needed to complete an activity: 7/15 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n wait calmly when access to preferred item is denied: 7/13 correct (53%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n single-word requests for preferred items: 6/15 correct (42%); latency 4.0s; prompts FP\u00d72, PP\u00d75, G\u00d72, I\u00d76\n\nSession 2 \u2014 2026-07-27 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 8/15 correct (53%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d712\n mands for missing items needed to complete an activity: 5/8 correct (56%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 7/11 correct (63%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n single-word requests for preferred items: 5/10 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n\nSession 3 \u2014 2026-07-29 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 5/8 correct (58%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 8/13 correct (65%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d73\n wait calmly when access to preferred item is denied: 6/9 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n single-word requests for preferred items: 6/10 correct (60%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n\nSession 4 \u2014 2026-07-31 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 8/12 correct (64%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n mands for missing items needed to complete an activity: 11/15 correct (71%); latency 3.0s; prompts FP\u00d74, PP\u00d73, G\u00d75, I\u00d73\n wait calmly when access to preferred item is denied: 9/12 correct (77%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n single-word requests for preferred items: 5/8 correct (65%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n\nSession 5 \u2014 2026-08-02 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 6/8 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n mands for missing items needed to complete an activity: 12/15 correct (78%); latency 2.8s; prompts FP\u00d73, PP\u00d75, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 7/9 correct (83%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n single-word requests for preferred items: 6/8 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n\nSession 6 \u2014 2026-08-04 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 8/10 correct (78%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n mands for missing items needed to complete an activity: 10/12 correct (83%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d77\n wait calmly when access to preferred item is denied: 12/15 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n single-word requests for preferred items: 10/13 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n\nSession 7 \u2014 2026-08-06 \u2014 60 min \u2014 1 observer\n sounds out CVC words: 6/8 correct (78%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n mands for missing items needed to complete an activity: 12/14 correct (83%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 10/12 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n single-word requests for preferred items: 8/10 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 43% at return, recovering to 78%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 63.7%.\n- Trend direction: upward (first session: 43%, last session: 78%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "605eb128dcfdee94", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "school_age", "n_sessions": 7, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "7865"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8784\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Reading + Tact\nDate range: Sessions 1\u201310 across 20 days (2026-07-11 to 2026-07-31)\n\nACCELERATION PROGRAMS\n1. tacts colors of objects (Tact)\n Method: DTT\n2. sounds out CVC words (Reading)\n Method: DTT\n3. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-11 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 5/10 correct (50%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 7/15 correct (49%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d710\n answers WH-questions about familiar topics: 8/13 correct (63%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n\nSession 2 \u2014 2026-07-13 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 5/10 correct (52%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n sounds out CVC words: 6/10 correct (64%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 6/12 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n\nSession 3 \u2014 2026-07-15 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 7/12 correct (54%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d72\n sounds out CVC words: 8/13 correct (60%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n answers WH-questions about familiar topics: 5/9 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n\nSession 4 \u2014 2026-07-17 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 9/14 correct (61%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n sounds out CVC words: 6/10 correct (63%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n answers WH-questions about familiar topics: 5/8 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n\nSession 5 \u2014 2026-07-19 \u2014 IOA SESSION \u2014 2 observers\n tacts colors of objects: 58% (primary); IOA 90% agreement\n sounds out CVC words: 60% (primary); IOA 90% agreement\n answers WH-questions about familiar topics: 55% (primary); IOA 90% agreement\n\nSession 6 \u2014 2026-07-21 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 8/13 correct (63%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n sounds out CVC words: 8/11 correct (69%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n answers WH-questions about familiar topics: 6/10 correct (57%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n\nSession 7 \u2014 2026-07-23 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 8/13 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d75\n sounds out CVC words: 6/9 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n answers WH-questions about familiar topics: 6/10 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n\nSession 8 \u2014 2026-07-25 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 7/12 correct (62%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n sounds out CVC words: 8/13 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 7/13 correct (56%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d73\n\nSession 9 \u2014 2026-07-27 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 6/10 correct (62%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n sounds out CVC words: 8/11 correct (74%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n answers WH-questions about familiar topics: 5/11 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n\nSession 10 \u2014 2026-07-29 \u2014 33 min \u2014 1 observer\n tacts colors of objects: 8/11 correct (71%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n sounds out CVC words: 8/12 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n answers WH-questions about familiar topics: 7/15 correct (49%); latency 3.7s; prompts FP\u00d72, PP\u00d75, G\u00d70, I\u00d78\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 60% for 10 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 59.5%.\n- Trend direction: upward (first session: 50%, last session: 71%).\n- IOA data: 90% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "2350389104a74c98", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "early", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "4427"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7486\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Reading\nDate range: Sessions 1\u201312 across 24 days (2026-08-03 to 2026-08-27)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. sounds out CVC words (Reading)\n Method: DTT\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-03 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 11/15 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n sounds out CVC words: 11/13 correct (87%); latency 2.0s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/14 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n Tantrum: freq 1, duration 5m total\n\nSession 2 \u2014 2026-08-05 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (66%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n sounds out CVC words: 11/14 correct (76%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (79%); latency 2.3s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d75\n Tantrum: freq 1, duration 4m total\n\nSession 3 \u2014 2026-08-07 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/14 correct (63%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n sounds out CVC words: 8/11 correct (74%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (70%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n Tantrum: freq 2, duration 10m total\n\nSession 4 \u2014 2026-08-09 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (52%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n sounds out CVC words: 6/9 correct (66%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (71%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n Tantrum: freq 2, duration 4m total\n\nSession 5 \u2014 2026-08-11 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n sounds out CVC words: 8/14 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (60%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n Tantrum: freq 1, duration 3m total\n\nSession 6 \u2014 2026-08-13 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/12 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n sounds out CVC words: 6/10 correct (58%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n Tantrum: freq 3, duration 6m total\n\nSession 7 \u2014 2026-08-15 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/15 correct (39%); latency 4.4s; prompts FP\u00d75, PP\u00d72, G\u00d71, I\u00d77\n sounds out CVC words: 6/12 correct (51%); latency 3.7s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n Tantrum: freq 2, duration 10m total\n\nSession 8 \u2014 2026-08-17 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/12 correct (31%); latency 4.8s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n sounds out CVC words: 6/14 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (51%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n Tantrum: freq 2, duration 8m total\n\nSession 9 \u2014 2026-08-19 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/9 correct (21%); latency 5.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n sounds out CVC words: 6/15 correct (37%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/10 correct (42%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n Tantrum: freq 3, duration 9m total\n\nSession 10 \u2014 2026-08-21 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/15 correct (13%); latency 5.5s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n sounds out CVC words: 2/9 correct (27%); latency 4.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/11 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n Tantrum: freq 4, duration 12m total\n\nSession 11 \u2014 2026-08-23 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 1/9 correct (14%); latency 5.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n sounds out CVC words: 2/8 correct (23%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (30%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n Tantrum: freq 4, duration 8m total\n\nSession 12 \u2014 2026-08-25 \u2014 51 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 1/13 correct (10%); latency 5.5s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n sounds out CVC words: 3/13 correct (24%); latency 5.0s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (29%); latency 4.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n Tantrum: freq 5, duration 10m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Pushing materials away from the work area\n- Elopement from the work area\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 71% to 10% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 39.7%.\n- Trend direction: downward (first session: 71%, last session: 10%).\n- Tantrum: mean 2.5/session (early-window mean 1.5 -> late-window mean 4.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "275422cd61681fb0", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Tantrum": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 12, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "80849"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6069\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand\nDate range: Sessions 1\u20139 across 18 days (2026-03-22 to 2026-04-09)\n\nACCELERATION PROGRAMS\n1. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n\nDECELERATION TARGETS\n1. Aggression \u2014 function hypothesized: tangible\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-22 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 10/12 correct (82%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n answers WH-questions about familiar topics: 7/11 correct (61%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n follows 1-step motor instructions: 8/10 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n Aggression: freq 3\n ABC (aggression): A = peer took preferred toy; B = grabbed toy back + cried; C = staff mediated and returned item\n\nSession 2 \u2014 2026-03-24 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 8/14 correct (56%); latency 3.6s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d74\n answers WH-questions about familiar topics: 6/11 correct (51%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n follows 1-step motor instructions: 7/14 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d74\n Aggression: freq 0\n\nSession 3 \u2014 2026-03-26 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 5/9 correct (58%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n answers WH-questions about familiar topics: 6/14 correct (40%); latency 4.3s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n follows 1-step motor instructions: 7/9 correct (73%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n Aggression: freq 4\n ABC (aggression): A = peer took preferred toy; B = grabbed toy back + cried; C = staff mediated and returned item\n\nSession 4 \u2014 2026-03-28 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 8/11 correct (74%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n answers WH-questions about familiar topics: 10/14 correct (70%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n follows 1-step motor instructions: 11/15 correct (71%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d75, I\u00d75\n Aggression: freq 1\n\nSession 5 \u2014 2026-03-30 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 7/11 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 5/9 correct (56%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n follows 1-step motor instructions: 12/15 correct (83%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d712\n Aggression: freq 3\n\nSession 6 \u2014 2026-04-01 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 6/11 correct (52%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n answers WH-questions about familiar topics: 3/10 correct (34%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n follows 1-step motor instructions: 8/11 correct (73%); latency 2.7s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n Aggression: freq 1\n ABC (aggression): A = preferred iPad removed at end of break; B = screamed + reached for device; C = staff redirected to alternative\n\nSession 7 \u2014 2026-04-03 \u2014 IOA SESSION \u2014 2 observers\n requesting a break ('break please' or AAC 'I need a break'): 64% (primary); IOA 85% agreement\n answers WH-questions about familiar topics: 58% (primary); IOA 85% agreement\n follows 1-step motor instructions: 81% (primary); IOA 85% agreement\n Aggression: freq 0; IOA 85% agreement\n\nSession 8 \u2014 2026-04-05 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 9/11 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n answers WH-questions about familiar topics: 7/14 correct (47%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n follows 1-step motor instructions: 11/15 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d76\n Aggression: freq 3\n\nSession 9 \u2014 2026-04-07 \u2014 57 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 10/14 correct (68%); latency 3.0s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n answers WH-questions about familiar topics: 10/15 correct (64%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d75, I\u00d79\n follows 1-step motor instructions: 10/13 correct (79%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n Aggression: freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Delayed responding (>5 seconds consistently)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 75% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nAggression: tangible\n Evidence: ABC events in the log support the tangible function; behavior occurs when preferred items are restricted and is followed by access to the item.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 66.5%.\n- Trend direction: downward (first session: 82%, last session: 68%).\n- Aggression: mean 2.0/session (early-window mean 2.3 -> late-window mean 2.0).\n- IOA data: 85% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "1247171dd3ca44d5", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Aggression": "tangible"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "early", "n_sessions": 9, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["aggression"], "behavior_functions": ["tangible"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "11965"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1527\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Mand + Motor Imitation + Reading + Tact\nDate range: Sessions 1\u201311 across 22 days (2026-04-09 to 2026-05-01)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. tacts colors of objects (Tact)\n Method: DTT\n3. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n4. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n5. sounds out CVC words (Reading)\n Method: DTT\n6. echoes 2-word phrases (Echoic)\n Method: DTT variant\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-09 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 4/8 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n tacts colors of objects: 8/15 correct (53%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d714\n imitates 3-step motor sequences: 4/9 correct (41%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n indicating completion ('all done', 'finished'): 8/15 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d75, G\u00d74, I\u00d72\n sounds out CVC words: 4/8 correct (51%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n echoes 2-word phrases: 8/14 correct (56%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n\nSession 2 \u2014 2026-04-11 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 8/13 correct (59%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n tacts colors of objects: 4/9 correct (45%); latency 4.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n imitates 3-step motor sequences: 4/10 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 5/10 correct (52%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n sounds out CVC words: 5/9 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n echoes 2-word phrases: 6/11 correct (55%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n\nSession 3 \u2014 2026-04-13 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 6/10 correct (60%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n tacts colors of objects: 6/15 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d78\n imitates 3-step motor sequences: 6/11 correct (56%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n indicating completion ('all done', 'finished'): 4/10 correct (43%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n sounds out CVC words: 7/13 correct (51%); latency 3.8s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n echoes 2-word phrases: 5/10 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n\nSession 4 \u2014 2026-04-15 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 4/8 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n tacts colors of objects: 6/14 correct (41%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n imitates 3-step motor sequences: 7/14 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n indicating completion ('all done', 'finished'): 7/13 correct (54%); latency 3.3s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n sounds out CVC words: 4/8 correct (52%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n echoes 2-word phrases: 9/14 correct (65%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d78\n\nSession 5 \u2014 2026-04-17 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 8/14 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n tacts colors of objects: 6/12 correct (46%); latency 4.0s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n imitates 3-step motor sequences: 8/15 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d75, I\u00d78\n indicating completion ('all done', 'finished'): 5/10 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 7/12 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d74\n echoes 2-word phrases: 6/13 correct (50%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n\nSession 6 \u2014 2026-04-19 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 5/11 correct (44%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n tacts colors of objects: 5/11 correct (49%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n imitates 3-step motor sequences: 6/12 correct (46%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n indicating completion ('all done', 'finished'): 5/11 correct (43%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n sounds out CVC words: 6/13 correct (42%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n echoes 2-word phrases: 6/11 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n\nSession 7 \u2014 2026-04-21 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 8/14 correct (54%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n tacts colors of objects: 8/15 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n imitates 3-step motor sequences: 7/14 correct (48%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d77\n indicating completion ('all done', 'finished'): 8/14 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n sounds out CVC words: 5/11 correct (44%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n echoes 2-word phrases: 9/14 correct (64%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n\nSession 8 \u2014 2026-04-23 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 8/14 correct (57%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n tacts colors of objects: 4/10 correct (41%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n imitates 3-step motor sequences: 7/15 correct (50%); latency 3.9s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n indicating completion ('all done', 'finished'): 4/9 correct (50%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n sounds out CVC words: 6/12 correct (50%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n echoes 2-word phrases: 7/13 correct (56%); latency 3.4s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n\nSession 9 \u2014 2026-04-25 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 4/8 correct (44%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n tacts colors of objects: 5/12 correct (43%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n imitates 3-step motor sequences: 6/15 correct (42%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n indicating completion ('all done', 'finished'): 4/9 correct (50%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n sounds out CVC words: 4/8 correct (53%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n echoes 2-word phrases: 5/8 correct (62%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n\nSession 10 \u2014 2026-04-27 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 7/14 correct (48%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d77\n tacts colors of objects: 4/8 correct (47%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n imitates 3-step motor sequences: 8/15 correct (50%); latency 3.6s; prompts FP\u00d74, PP\u00d74, G\u00d75, I\u00d72\n indicating completion ('all done', 'finished'): 8/15 correct (52%); latency 3.7s; prompts FP\u00d75, PP\u00d71, G\u00d70, I\u00d79\n sounds out CVC words: 3/8 correct (39%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n echoes 2-word phrases: 9/15 correct (63%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d77\n\nSession 11 \u2014 2026-04-29 \u2014 32 min \u2014 1 observer\n reads 10\u201320 sight words: 5/9 correct (52%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n tacts colors of objects: 3/8 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n imitates 3-step motor sequences: 5/10 correct (51%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n indicating completion ('all done', 'finished'): 7/15 correct (48%); latency 3.7s; prompts FP\u00d75, PP\u00d70, G\u00d70, I\u00d710\n sounds out CVC words: 7/15 correct (45%); latency 3.7s; prompts FP\u00d75, PP\u00d74, G\u00d75, I\u00d71\n echoes 2-word phrases: 7/11 correct (63%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Leaving seat without permission\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 50% for 11 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 52.4%.\n- Trend direction: upward (first session: 47%, last session: 52%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "26fd25f89043f5c9", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "early", "n_sessions": 11, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "55785"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: initiates social interactions with peers\nCurriculum Reference: Social Behavior & Social Play L2\nLearner Profile: Early Learner\nCurrent Mastery: Developing\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: initiates social interactions with peers\n- Curriculum: Social Behavior & Social Play L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: peer play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's peer play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "100758a5bfb54a73", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "initiates social interactions with peers", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "70185"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3610\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care\nDate range: Sessions 1\u20139 across 18 days (2026-07-21 to 2026-08-08)\n\nACCELERATION PROGRAMS\n1. requesting assistance using AAC device (Mand)\n Method: NET\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n4. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-21 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 11/14 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n self-advocacy statements: 9/12 correct (75%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n brushing teeth independently: 6/8 correct (74%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n scheduling medical appointments: 10/11 correct (88%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 2 \u2014 2026-07-23 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 12/15 correct (78%); latency 2.4s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d75\n self-advocacy statements: 9/14 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n brushing teeth independently: 6/9 correct (67%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n scheduling medical appointments: 7/10 correct (72%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-07-25 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 10/15 correct (67%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d713\n self-advocacy statements: 6/10 correct (58%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n brushing teeth independently: 6/11 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n scheduling medical appointments: 5/8 correct (66%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 4 \u2014 2026-07-27 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 8/12 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n self-advocacy statements: 6/10 correct (60%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n brushing teeth independently: 9/15 correct (58%); latency 3.3s; prompts FP\u00d72, PP\u00d75, G\u00d71, I\u00d77\n scheduling medical appointments: 7/10 correct (65%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n Toileting accident (urine or bowel): urine: 0 in-toilet / 5 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 5 \u2014 2026-07-29 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 7/11 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n self-advocacy statements: 5/9 correct (54%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n brushing teeth independently: 7/14 correct (51%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n scheduling medical appointments: 5/8 correct (58%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 0 in-toilet / 4 accidents; BM: 0 in-toilet / 2 accidents\n\nSession 6 \u2014 2026-07-31 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 6/10 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n self-advocacy statements: 6/13 correct (42%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n brushing teeth independently: 4/8 correct (45%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n scheduling medical appointments: 4/9 correct (44%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n Toileting accident (urine or bowel): urine: 1 in-toilet / 6 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 7 \u2014 2026-08-02 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 5/10 correct (47%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n self-advocacy statements: 5/12 correct (42%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d78\n brushing teeth independently: 4/13 correct (34%); latency 4.2s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n scheduling medical appointments: 3/8 correct (41%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n Toileting accident (urine or bowel): urine: 2 in-toilet / 6 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 8 \u2014 2026-08-04 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 6/11 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n self-advocacy statements: 3/9 correct (39%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n brushing teeth independently: 4/14 correct (28%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d710\n scheduling medical appointments: 5/15 correct (35%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d75, I\u00d74\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 2 accidents\n\nSession 9 \u2014 2026-08-06 \u2014 44 min \u2014 1 observer\n requesting assistance using AAC device: 4/8 correct (44%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n self-advocacy statements: 2/8 correct (27%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n brushing teeth independently: 2/10 correct (23%); latency 4.9s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n scheduling medical appointments: 3/14 correct (21%); latency 5.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n Toileting accident (urine or bowel): urine: 0 in-toilet / 7 accidents; BM: 0 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Pushing materials away from the work area\n- Crying during demand presentation\n- Turning away from task or staff"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (81% -> 44%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 62.0%.\n- Trend direction: downward (first session: 81%, last session: 44%).\n- Toileting accident (urine or bowel): mean 5.1/session (early-window mean 3.3 -> late-window mean 6.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "7ab4a36f97d4b956", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Toileting accident (urine or bowel)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adult", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "6686"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner Level: Adolescent Learner\nBaseline: Developing\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the developing stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "d28df9505938b55d", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_generalization", "chain_type": "total_task", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "44637"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3091\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math\nDate range: Sessions 1\u20135 across 10 days (2026-02-12 to 2026-02-22)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. counts objects 1\u201320 (Math)\n Method: DTT\n4. requesting preferred items using a full sentence (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-12 \u2014 58 min \u2014 1 observer\n identifies numerals 1\u201320: 4/12 correct (31%); latency 4.7s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/11 correct (22%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n counts objects 1\u201320: 3/11 correct (29%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n requesting preferred items using a full sentence: 4/12 correct (35%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n\nSession 2 \u2014 2026-02-14 \u2014 58 min \u2014 1 observer\n identifies numerals 1\u201320: 4/9 correct (40%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (37%); latency 4.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n counts objects 1\u201320: 5/14 correct (34%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d78\n requesting preferred items using a full sentence: 6/15 correct (41%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d77\n\nSession 3 \u2014 2026-02-16 \u2014 58 min \u2014 1 observer\n identifies numerals 1\u201320: 7/12 correct (55%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (41%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n counts objects 1\u201320: 6/15 correct (42%); latency 3.8s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n requesting preferred items using a full sentence: 8/15 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n\nSession 4 \u2014 2026-02-18 \u2014 58 min \u2014 1 observer\n identifies numerals 1\u201320: 7/12 correct (60%); latency 3.5s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/15 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n counts objects 1\u201320: 6/13 correct (49%); latency 4.0s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n requesting preferred items using a full sentence: 7/10 correct (66%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n\nSession 5 \u2014 2026-02-20 \u2014 58 min \u2014 1 observer\n identifies numerals 1\u201320: 7/10 correct (74%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n counts objects 1\u201320: 5/8 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n requesting preferred items using a full sentence: 7/9 correct (73%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Smiling during reinforcement delivery"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 31% to 74% across 5 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 5 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 5 sessions: 52.0%.\n- Trend direction: upward (first session: 31%, last session: 74%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "38523ee2605e059d", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "school_age", "n_sessions": 5, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "42504"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects correct item from an array of 2\nCurriculum: Listener Responding L1\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from an array of 2\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: backpack, block, scissors\nDistractor stimuli: bottle, ball\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6f400751336bfa39", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from an array of 2", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "70400"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3089\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Vocational\nDate range: Sessions 1\u201311 across 22 days (2026-10-13 to 2026-11-04)\n\nACCELERATION PROGRAMS\n1. managing a budget (Independent Living)\n Method: Task Analysis\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. clocking in at start of shift (Vocational)\n Method: Task Analysis\n4. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n5. doing laundry (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-13 \u2014 32 min \u2014 1 observer\n managing a budget: 7/8 correct (91%); latency 1.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n self-advocacy statements: 10/11 correct (88%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n clocking in at start of shift: 11/14 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n scheduling medical appointments: 11/13 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n doing laundry: 12/13 correct (95%); latency 1.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d78\n\nSession 2 \u2014 2026-10-15 \u2014 32 min \u2014 1 observer\n managing a budget: 8/8 correct (94%); latency 1.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n self-advocacy statements: 12/13 correct (92%); latency 1.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d710\n clocking in at start of shift: 10/13 correct (79%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d76\n scheduling medical appointments: 13/15 correct (85%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n doing laundry: 10/11 correct (94%); latency 1.9s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n\nSession 3 \u2014 2026-10-17 \u2014 32 min \u2014 1 observer\n managing a budget: 9/9 correct (95%); latency 1.4s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n self-advocacy statements: 12/14 correct (89%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d78\n clocking in at start of shift: 6/8 correct (79%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n scheduling medical appointments: 10/12 correct (82%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n doing laundry: 14/15 correct (92%); latency 1.9s; prompts FP\u00d75, PP\u00d74, G\u00d71, I\u00d75\n\nSession 4 \u2014 2026-10-19 \u2014 32 min \u2014 1 observer\n managing a budget: 14/15 correct (93%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d712\n self-advocacy statements: 11/12 correct (90%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n clocking in at start of shift: 7/9 correct (80%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n scheduling medical appointments: 9/11 correct (85%); latency 2.0s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n doing laundry: 10/11 correct (95%); latency 1.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n\nSession 5 \u2014 2026-10-21 \u2014 32 min \u2014 1 observer\n managing a budget: 10/11 correct (87%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n self-advocacy statements: 13/14 correct (90%); latency 2.2s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d711\n clocking in at start of shift: 8/10 correct (76%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n scheduling medical appointments: 12/15 correct (78%); latency 2.4s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n doing laundry: 9/10 correct (93%); latency 1.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n\nSession 6 \u2014 2026-10-23 \u2014 32 min \u2014 1 observer\n managing a budget: 9/10 correct (91%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n self-advocacy statements: 7/8 correct (89%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n clocking in at start of shift: 9/13 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n scheduling medical appointments: 7/8 correct (82%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n doing laundry: 10/11 correct (94%); latency 1.8s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n\nSession 7 \u2014 2026-10-25 \u2014 32 min \u2014 1 observer\n managing a budget: 8/9 correct (88%); latency 1.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n self-advocacy statements: 8/9 correct (92%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n clocking in at start of shift: 8/11 correct (72%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n scheduling medical appointments: 10/12 correct (82%); latency 2.1s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n doing laundry: 12/13 correct (91%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d710\n\nSession 8 \u2014 2026-10-27 \u2014 32 min \u2014 1 observer\n managing a budget: 13/15 correct (87%); latency 2.2s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n self-advocacy statements: 10/11 correct (92%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n clocking in at start of shift: 11/15 correct (73%); latency 2.5s; prompts FP\u00d75, PP\u00d71, G\u00d74, I\u00d75\n scheduling medical appointments: 6/8 correct (76%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n doing laundry: 11/12 correct (94%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d710\n\nSession 9 \u2014 2026-10-29 \u2014 32 min \u2014 1 observer\n managing a budget: 11/12 correct (92%); latency 1.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d78\n self-advocacy statements: 7/8 correct (91%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n clocking in at start of shift: 10/13 correct (75%); latency 2.3s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n scheduling medical appointments: 10/13 correct (78%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n doing laundry: 8/9 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n\nSession 10 \u2014 2026-10-31 \u2014 32 min \u2014 1 observer\n managing a budget: 12/14 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d711\n self-advocacy statements: 14/15 correct (97%); latency 1.7s; prompts FP\u00d74, PP\u00d71, G\u00d75, I\u00d75\n clocking in at start of shift: 10/13 correct (75%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n scheduling medical appointments: 8/10 correct (75%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n doing laundry: 9/10 correct (92%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n\nSession 11 \u2014 2026-11-02 \u2014 32 min \u2014 1 observer\n managing a budget: 12/13 correct (89%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d76\n self-advocacy statements: 9/9 correct (98%); latency 1.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n clocking in at start of shift: 8/11 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n scheduling medical appointments: 12/15 correct (82%); latency 2.2s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n doing laundry: 8/9 correct (92%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (91%\u201389%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 90.3%.\n- Trend direction: downward (first session: 91%, last session: 89%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "acb13453f24cdb34", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adult", "n_sessions": 11, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "39407"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2332\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u201312 across 24 days (2026-01-07 to 2026-01-31)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. sounds out CVC words (Reading)\n Method: DTT\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. single-word requests for preferred items (Mand)\n Method: NET\n5. mands for missing items needed to complete an activity (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n2. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-07 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/10 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n sounds out CVC words: 9/13 correct (68%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (75%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n single-word requests for preferred items: 7/9 correct (76%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n mands for missing items needed to complete an activity: 12/14 correct (86%); latency 2.3s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n Property destruction: freq 2\n\nSession 2 \u2014 2026-01-09 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/10 correct (76%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n sounds out CVC words: 6/9 correct (72%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/9 correct (73%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n single-word requests for preferred items: 10/15 correct (66%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n mands for missing items needed to complete an activity: 9/12 correct (78%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d79\n Toileting accident (urine or bowel): urine: 1 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 3\n\nSession 3 \u2014 2026-01-11 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/11 correct (73%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n sounds out CVC words: 6/10 correct (59%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n single-word requests for preferred items: 9/13 correct (70%); latency 2.9s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n mands for missing items needed to complete an activity: 10/14 correct (72%); latency 3.0s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n Toileting accident (urine or bowel): urine: 5 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 2\n\nSession 4 \u2014 2026-01-13 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/12 correct (66%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n sounds out CVC words: 7/11 correct (60%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n single-word requests for preferred items: 8/13 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n mands for missing items needed to complete an activity: 8/11 correct (74%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n Toileting accident (urine or bowel): urine: 3 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 3\n\nSession 5 \u2014 2026-01-15 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/15 correct (58%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d78\n sounds out CVC words: 4/8 correct (56%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (57%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n single-word requests for preferred items: 9/15 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d75, I\u00d79\n mands for missing items needed to complete an activity: 6/9 correct (65%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n Property destruction: freq 3\n\nSession 6 \u2014 2026-01-17 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/11 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n sounds out CVC words: 4/8 correct (47%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/15 correct (54%); latency 3.5s; prompts FP\u00d75, PP\u00d73, G\u00d74, I\u00d73\n single-word requests for preferred items: 6/12 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 9/14 correct (61%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 3\n\nSession 7 \u2014 2026-01-19 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/11 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n sounds out CVC words: 5/11 correct (46%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/15 correct (54%); latency 3.8s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n single-word requests for preferred items: 6/11 correct (50%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n mands for missing items needed to complete an activity: 7/12 correct (60%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 1 accidents\n Property destruction: freq 3\n\nSession 8 \u2014 2026-01-21 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/11 correct (38%); latency 4.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n sounds out CVC words: 5/13 correct (37%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n single-word requests for preferred items: 5/12 correct (45%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 8/15 correct (52%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d78\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 1 in-toilet / 0 accidents\n Property destruction: freq 4\n\nSession 9 \u2014 2026-01-23 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 5/15 correct (32%); latency 4.7s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n sounds out CVC words: 6/15 correct (38%); latency 4.3s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n single-word requests for preferred items: 5/11 correct (42%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n mands for missing items needed to complete an activity: 5/10 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n Toileting accident (urine or bowel): urine: 0 in-toilet / 4 accidents; BM: 0 in-toilet / 1 accidents\n Property destruction: freq 5\n\nSession 10 \u2014 2026-01-25 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 3/12 correct (27%); latency 4.9s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n sounds out CVC words: 3/10 correct (30%); latency 4.9s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/14 correct (46%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d75\n single-word requests for preferred items: 5/13 correct (38%); latency 4.2s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n mands for missing items needed to complete an activity: 6/14 correct (45%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n Toileting accident (urine or bowel): urine: 2 in-toilet / 4 accidents; BM: 0 in-toilet / 1 accidents\n Property destruction: freq 5\n\nSession 11 \u2014 2026-01-27 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 2/11 correct (18%); latency 4.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n sounds out CVC words: 3/9 correct (31%); latency 4.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (34%); latency 4.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n single-word requests for preferred items: 3/9 correct (31%); latency 4.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n mands for missing items needed to complete an activity: 4/10 correct (39%); latency 4.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 2 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 5\n\nSession 12 \u2014 2026-01-29 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 2/13 correct (12%); latency 5.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n sounds out CVC words: 3/14 correct (24%); latency 5.0s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/13 correct (37%); latency 4.3s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n single-word requests for preferred items: 2/11 correct (22%); latency 4.9s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 3/8 correct (40%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 0 in-toilet / 5 accidents; BM: 1 in-toilet / 1 accidents\n Property destruction: freq 5\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Turning away from task or staff\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 78% to 12% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nProperty destruction: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 47.4%.\n- Trend direction: downward (first session: 78%, last session: 12%).\n- Toileting accident (urine or bowel): mean 3.9/session (early-window mean 2.8 -> late-window mean 5.0).\n- Property destruction: mean 3.6/session (early-window mean 2.5 -> late-window mean 5.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "f380ff3d198961e7", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Toileting accident (urine or bowel)": "escape", "Property destruction": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "school_age", "n_sessions": 12, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["toileting_accident", "property_destruction"], "behavior_functions": ["escape", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "6062"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies place value (ones, tens)\nCurriculum Reference: Math L3\nLearner Profile: School-Age Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies place value (ones, tens)\n- Curriculum: Math L3\n- Learner: School-Age Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: towel, scissors, hat\nDistractor stimuli: pencil, jacket, backpack, apple, cup\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "55532e0b6b52e83a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L3", "learner_profile": "school_age", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies place value (ones, tens)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "token_economy", "error_correction": "errorless", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "28081"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts abstract concepts (same/different, first/last)\nCurriculum: Tact L3\nLearner Level: Adolescent Learner\nBaseline: Mastered at current level\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts abstract concepts (same/different, first/last)\n- Curriculum: Tact L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: plate, bottle, hat\nDistractor stimuli: blanket, sock, backpack, scissors, napkin\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "d056a9483678d00f", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts abstract concepts (same/different, first/last)", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "four_step", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "70400"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: provides definitions of words\nCurriculum: Intraverbal L3\nLearner: School-Age Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: provides definitions of words\n- Curriculum: Intraverbal L3\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: post-activity debrief.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's post-activity debrief routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "550f8896125a0c2a", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "provides definitions of words", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "54871"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5254\nProfile: Adult Learner (18+ years)\nCurricula: Community + Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-01-10 to 2026-01-26)\n\nACCELERATION PROGRAMS\n1. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n2. using public transportation (Community)\n Method: Task Analysis\n3. requesting assistance using AAC device (Mand)\n Method: NET\n4. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n5. washing hands independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Fecal smearing (scatolia) \u2014 function hypothesized: attention\n2. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-10 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 9/13 correct (69%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d710\n using public transportation: 8/10 correct (84%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n requesting assistance using AAC device: 7/10 correct (67%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n completing assigned tasks in order: 7/10 correct (72%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n washing hands independently: 7/9 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 2\n\nSession 2 \u2014 2026-01-12 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 10/14 correct (69%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n using public transportation: 6/8 correct (78%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n requesting assistance using AAC device: 5/8 correct (63%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n completing assigned tasks in order: 8/12 correct (68%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n washing hands independently: 6/9 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 2\n\nSession 3 \u2014 2026-01-14 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 8/12 correct (63%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n using public transportation: 11/15 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n requesting assistance using AAC device: 7/12 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d75\n completing assigned tasks in order: 9/14 correct (63%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n washing hands independently: 8/13 correct (65%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n Fecal smearing (scatolia): attempts 5 (4 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 2\n\nSession 4 \u2014 2026-01-16 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 4/8 correct (53%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n using public transportation: 8/12 correct (70%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n requesting assistance using AAC device: 6/13 correct (49%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d77\n completing assigned tasks in order: 5/10 correct (55%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n washing hands independently: 8/14 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d710\n Fecal smearing (scatolia): attempts 4 (4 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 5\n\nSession 5 \u2014 2026-01-18 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 7/15 correct (48%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n using public transportation: 6/10 correct (62%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n requesting assistance using AAC device: 7/14 correct (48%); latency 4.1s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n completing assigned tasks in order: 6/13 correct (47%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n washing hands independently: 4/9 correct (44%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n Fecal smearing (scatolia): attempts 6 (4 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 4\n\nSession 6 \u2014 2026-01-20 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/10 correct (46%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n using public transportation: 8/14 correct (58%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d76\n requesting assistance using AAC device: 5/13 correct (41%); latency 4.3s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n completing assigned tasks in order: 6/13 correct (45%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n washing hands independently: 6/13 correct (44%); latency 4.2s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n Fecal smearing (scatolia): attempts 6 (5 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 5\n\nSession 7 \u2014 2026-01-22 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/12 correct (39%); latency 4.0s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n using public transportation: 8/14 correct (54%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d76\n requesting assistance using AAC device: 3/8 correct (38%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n completing assigned tasks in order: 3/10 correct (32%); latency 4.9s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n washing hands independently: 4/15 correct (29%); latency 4.7s; prompts FP\u00d73, PP\u00d75, G\u00d74, I\u00d73\n Fecal smearing (scatolia): attempts 7 (5 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 5\n\nSession 8 \u2014 2026-01-24 \u2014 43 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 3/8 correct (39%); latency 4.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n using public transportation: 5/9 correct (53%); latency 3.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n requesting assistance using AAC device: 3/10 correct (29%); latency 4.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n completing assigned tasks in order: 4/13 correct (29%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n washing hands independently: 3/14 correct (25%); latency 4.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n Fecal smearing (scatolia): attempts 8 (5 intercepted \u2014 staff redirected before smearing; 3 completed \u2014 feces transferred to skin, clothing, or surface)\n Self-injurious behavior (SIB): freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Pushing materials away from the work area\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Increased response latency (3\u20135\u00d7 baseline)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (69% -> 39%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nFecal smearing (scatolia): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 53.1%.\n- Trend direction: downward (first session: 69%, last session: 39%).\n- Fecal smearing (scatolia): mean 5.2/session (early-window mean 3.0 -> late-window mean 7.5).\n- Self-injurious behavior (SIB): mean 3.9/session (early-window mean 2.0 -> late-window mean 5.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "ce183713e1ab3651", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Fecal smearing (scatolia)": "attention", "Self-injurious behavior (SIB)": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adult", "n_sessions": 8, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["fecal_smearing", "sib"], "behavior_functions": ["attention", "escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "55369"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: ordering in a restaurant\nCurriculum Reference: AFLS Community Participation Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Near mastery\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: ordering in a restaurant\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Approach the counter or wait to be seated.\n2. Review the menu.\n3. Decide on a choice.\n4. Place the order verbally (or with an AAC device).\n5. Respond to clarifying questions.\n6. Pay using money or a card.\n7. Wait for the order and collect it. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "bbcedfdf0100038b", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "ordering in a restaurant", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "near", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "ordering in a restaurant", "module": "community", "mastery_criterion": "mc_fluent_chain", "chain_type": "backward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "97806"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3841\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Reading + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-10-04 to 2026-10-20)\n\nACCELERATION PROGRAMS\n1. requesting preferred items using AAC device (Mand)\n Method: NET\n2. identifying community safety signs (Community)\n Method: DTT\n3. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. sounds out CVC words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-04 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 8/11 correct (71%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n identifying community safety signs: 11/15 correct (76%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n meal cleanup after eating: 10/13 correct (78%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n brushing teeth independently: 7/9 correct (80%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n sounds out CVC words: 12/15 correct (81%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n Property destruction: freq 2\n\nSession 2 \u2014 2026-10-06 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 6/8 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n identifying community safety signs: 7/9 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n meal cleanup after eating: 8/12 correct (65%); latency 3.2s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n brushing teeth independently: 11/15 correct (72%); latency 2.5s; prompts FP\u00d75, PP\u00d73, G\u00d71, I\u00d76\n sounds out CVC words: 8/12 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n Property destruction: freq 3\n\nSession 3 \u2014 2026-10-08 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 8/14 correct (59%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n identifying community safety signs: 9/14 correct (62%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d79\n meal cleanup after eating: 7/12 correct (58%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n brushing teeth independently: 6/9 correct (66%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 5/8 correct (59%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Property destruction: freq 4\n\nSession 4 \u2014 2026-10-10 \u2014 IOA SESSION \u2014 2 observers\n requesting preferred items using AAC device: 54% (primary); IOA 82% agreement\n identifying community safety signs: 62% (primary); IOA 82% agreement\n meal cleanup after eating: 52% (primary); IOA 82% agreement\n brushing teeth independently: 54% (primary); IOA 82% agreement\n sounds out CVC words: 55% (primary); IOA 82% agreement\n Property destruction: freq 4; IOA 82% agreement\n\nSession 5 \u2014 2026-10-12 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 7/13 correct (53%); latency 3.8s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n identifying community safety signs: 7/13 correct (54%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d76\n meal cleanup after eating: 4/8 correct (46%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n brushing teeth independently: 7/13 correct (55%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n sounds out CVC words: 4/9 correct (43%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n Property destruction: freq 5\n\nSession 6 \u2014 2026-10-14 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 7/14 correct (49%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d77\n identifying community safety signs: 6/15 correct (43%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d79\n meal cleanup after eating: 5/11 correct (45%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n brushing teeth independently: 6/12 correct (47%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n sounds out CVC words: 4/10 correct (40%); latency 4.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n Property destruction: freq 4\n\nSession 7 \u2014 2026-10-16 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 3/8 correct (37%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n identifying community safety signs: 4/12 correct (35%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d710\n meal cleanup after eating: 4/11 correct (38%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n brushing teeth independently: 5/15 correct (33%); latency 4.5s; prompts FP\u00d75, PP\u00d70, G\u00d74, I\u00d76\n sounds out CVC words: 3/9 correct (29%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n Property destruction: freq 6\n\nSession 8 \u2014 2026-10-18 \u2014 34 min \u2014 1 observer\n requesting preferred items using AAC device: 4/14 correct (30%); latency 4.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n identifying community safety signs: 3/9 correct (29%); latency 4.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n meal cleanup after eating: 4/12 correct (35%); latency 4.7s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n brushing teeth independently: 2/8 correct (27%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n sounds out CVC words: 2/9 correct (22%); latency 5.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n Property destruction: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-stimulatory behavior increase during demand periods\n- Turning away from task or staff\n- Elopement from the work area\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 71% to 30% across 8 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nProperty destruction: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 52.9%.\n- Trend direction: downward (first session: 71%, last session: 30%).\n- Property destruction: mean 4.2/session (early-window mean 2.5 -> late-window mean 6.0).\n- IOA data: 82% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "8ac09633bbe9844a", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Property destruction": "attention"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "54692"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: maintains friendships over time\nCurriculum Reference: Social Behavior & Social Play L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: maintains friendships over time\n- Curriculum: Social Behavior & Social Play L3\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: peer play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's peer play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "7a5ab2737dbc8bb6", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L3", "learner_profile": "adolescent", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "maintains friendships over time", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "76914"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9381\nProfile: Adult Learner (18+ years)\nCurricula: Community + Independent Living + Mand + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-06-05 to 2026-06-21)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. requesting a break during extended tasks (Mand)\n Method: NET\n4. using public transportation (Community)\n Method: Task Analysis\n5. indicating 'all done' at task completion (Mand)\n Method: NET\n6. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-05 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 2/9 correct (23%); latency 5.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n self-advocacy statements: 3/10 correct (32%); latency 4.6s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n requesting a break during extended tasks: 3/9 correct (30%); latency 4.9s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n using public transportation: 3/14 correct (22%); latency 5.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d79\n indicating 'all done' at task completion: 3/11 correct (28%); latency 4.5s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n wait calmly when a preferred activity is denied: 3/8 correct (36%); latency 4.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n\nSession 2 \u2014 2026-06-07 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 5/14 correct (34%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n self-advocacy statements: 5/13 correct (38%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n requesting a break during extended tasks: 5/15 correct (33%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d710\n using public transportation: 4/12 correct (31%); latency 4.8s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n indicating 'all done' at task completion: 5/13 correct (40%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n wait calmly when a preferred activity is denied: 4/8 correct (50%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n\nSession 3 \u2014 2026-06-09 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 4/11 correct (41%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n self-advocacy statements: 5/10 correct (51%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n requesting a break during extended tasks: 4/10 correct (42%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n using public transportation: 6/15 correct (40%); latency 4.3s; prompts FP\u00d73, PP\u00d75, G\u00d73, I\u00d74\n indicating 'all done' at task completion: 6/12 correct (51%); latency 3.8s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n wait calmly when a preferred activity is denied: 5/9 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n\nSession 4 \u2014 2026-06-11 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 7/14 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d710\n self-advocacy statements: 6/9 correct (65%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n requesting a break during extended tasks: 6/12 correct (53%); latency 3.9s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n using public transportation: 7/14 correct (53%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n indicating 'all done' at task completion: 6/10 correct (63%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 11/15 correct (73%); latency 2.4s; prompts FP\u00d72, PP\u00d75, G\u00d74, I\u00d74\n\nSession 5 \u2014 2026-06-13 \u2014 IOA SESSION \u2014 2 observers\n completing assigned tasks in order: 62% (primary); IOA 84% agreement\n self-advocacy statements: 73% (primary); IOA 84% agreement\n requesting a break during extended tasks: 59% (primary); IOA 84% agreement\n using public transportation: 67% (primary); IOA 84% agreement\n indicating 'all done' at task completion: 74% (primary); IOA 84% agreement\n wait calmly when a preferred activity is denied: 81% (primary); IOA 84% agreement\n\nSession 6 \u2014 2026-06-15 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 6/8 correct (73%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n self-advocacy statements: 9/11 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n requesting a break during extended tasks: 6/9 correct (69%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n using public transportation: 8/10 correct (77%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n indicating 'all done' at task completion: 13/15 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n wait calmly when a preferred activity is denied: 10/11 correct (92%); latency 1.6s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n\nSession 7 \u2014 2026-06-17 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 8/9 correct (86%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n self-advocacy statements: 11/12 correct (93%); latency 1.8s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n requesting a break during extended tasks: 11/14 correct (79%); latency 2.4s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d76\n using public transportation: 9/10 correct (90%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n indicating 'all done' at task completion: 7/8 correct (94%); latency 1.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when a preferred activity is denied: 10/10 correct (97%); latency 1.6s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n\nSession 8 \u2014 2026-06-19 \u2014 54 min \u2014 1 observer\n completing assigned tasks in order: 9/9 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n self-advocacy statements: 9/9 correct (97%); latency 1.5s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n requesting a break during extended tasks: 11/12 correct (91%); latency 1.8s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n using public transportation: 13/13 correct (97%); latency 1.7s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n indicating 'all done' at task completion: 10/10 correct (97%); latency 1.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 8/8 correct (97%); latency 1.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Appropriate sitting posture maintained\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 23% to 95% across 8 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nhigh\n\nPattern is well-supported: 8 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 58.2%.\n- Trend direction: upward (first session: 23%, last session: 95%).\n- IOA data: 84% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "828fa6936749f212", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "adult", "n_sessions": 8, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "42406"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: brushing teeth\nCurriculum: AFLS Basic Living Skills\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: brushing teeth\n- Curriculum: AFLS Basic Living Skills\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Collect toothbrush and toothpaste.\n2. Wet the toothbrush under water.\n3. Apply a pea-sized amount of toothpaste.\n4. Brush upper-front teeth in small circles for 30 seconds.\n5. Brush lower-front teeth in small circles for 30 seconds.\n6. Brush upper-back teeth (left and right).\n7. Brush lower-back teeth (left and right).\n8. Spit into the sink and rinse mouth.\n9. Rinse the toothbrush and return it to the holder.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 180s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "213d35bdc90930a3", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "brushing teeth", "program_type": "independence", "learner_profile": "school_age", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "brushing teeth", "module": "basic_living", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "75542"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8111\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u201312 across 24 days (2026-05-19 to 2026-06-12)\n\nACCELERATION PROGRAMS\n1. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. self-advocacy statements (Independent Living)\n Method: BST\n4. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-19 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 7/12 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 11/15 correct (72%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d75, I\u00d76\n self-advocacy statements: 6/15 correct (40%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n simple meal preparation: 8/13 correct (59%); latency 3.5s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d74\n\nSession 2 \u2014 2026-05-21 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 3/8 correct (35%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 5/9 correct (50%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n self-advocacy statements: 5/11 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n simple meal preparation: 10/14 correct (70%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n\nSession 3 \u2014 2026-05-23 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 7/10 correct (66%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n wait calmly when a preferred activity is denied: 8/15 correct (54%); latency 3.3s; prompts FP\u00d75, PP\u00d75, G\u00d71, I\u00d74\n self-advocacy statements: 7/15 correct (50%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n simple meal preparation: 8/11 correct (72%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n\nSession 4 \u2014 2026-05-25 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 3/8 correct (31%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n wait calmly when a preferred activity is denied: 6/12 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d77\n self-advocacy statements: 4/8 correct (56%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n simple meal preparation: 7/15 correct (46%); latency 3.7s; prompts FP\u00d75, PP\u00d74, G\u00d75, I\u00d71\n\nSession 5 \u2014 2026-05-27 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 10/15 correct (65%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d77\n wait calmly when a preferred activity is denied: 5/10 correct (48%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n self-advocacy statements: 7/11 correct (61%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n simple meal preparation: 4/11 correct (40%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n\nSession 6 \u2014 2026-05-29 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 4/15 correct (29%); latency 4.9s; prompts FP\u00d71, PP\u00d75, G\u00d72, I\u00d77\n wait calmly when a preferred activity is denied: 8/10 correct (81%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n self-advocacy statements: 5/13 correct (40%); latency 4.3s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n simple meal preparation: 11/15 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n\nSession 7 \u2014 2026-05-31 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 4/8 correct (54%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 6/13 correct (46%); latency 4.2s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d74\n self-advocacy statements: 4/10 correct (41%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n simple meal preparation: 4/10 correct (41%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n\nSession 8 \u2014 2026-06-02 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 10/15 correct (66%); latency 3.1s; prompts FP\u00d75, PP\u00d75, G\u00d70, I\u00d75\n wait calmly when a preferred activity is denied: 7/14 correct (48%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n self-advocacy statements: 7/15 correct (49%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d79\n simple meal preparation: 6/15 correct (42%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d78\n\nSession 9 \u2014 2026-06-04 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 7/10 correct (66%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n wait calmly when a preferred activity is denied: 6/12 correct (48%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d76\n self-advocacy statements: 9/13 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n simple meal preparation: 4/11 correct (34%); latency 4.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n\nSession 10 \u2014 2026-06-06 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 7/14 correct (47%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d77\n wait calmly when a preferred activity is denied: 7/15 correct (47%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d79\n self-advocacy statements: 10/14 correct (71%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n simple meal preparation: 8/12 correct (70%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n\nSession 11 \u2014 2026-06-08 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 3/9 correct (37%); latency 4.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n wait calmly when a preferred activity is denied: 6/10 correct (65%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n self-advocacy statements: 4/10 correct (39%); latency 4.3s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d74\n simple meal preparation: 8/14 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n\nSession 12 \u2014 2026-06-10 \u2014 36 min \u2014 1 observer\n completing assigned tasks in order: 3/9 correct (32%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n wait calmly when a preferred activity is denied: 11/15 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d714\n self-advocacy statements: 5/10 correct (48%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n simple meal preparation: 6/12 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 44% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 48.7%.\n- Trend direction: downward (first session: 55%, last session: 32%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e19fa23e4d579ded", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adult", "n_sessions": 12, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "76678"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: scheduling a medical appointment\nCurriculum Reference: AFLS Independent Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: scheduling a medical appointment\n- Curriculum: AFLS Independent Living Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Identify the reason for the appointment. <- Learner-taught step (forward chaining starts here)\n2. Locate the provider's contact information.\n3. Call or use the online system.\n4. Provide name, date of birth, and insurance information.\n5. Request an available appointment time.\n6. Confirm the appointment and save it to the calendar.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "55bb0787e41697b0", "gold_labels": {"method": "task_analysis", "domain": "AFLS.independent_living", "skill": "scheduling a medical appointment", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "scheduling a medical appointment", "module": "independent_living", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "98892"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4502\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Mand + Reading + Tact + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-09-21 to 2026-10-09)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n5. tacts common objects (Tact)\n Method: DTT\n6. echoes 2-word phrases (Echoic)\n Method: DTT variant\n\nDECELERATION TARGETS\n1. Pica \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-21 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 5/8 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 4/8 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n wait calmly when access to preferred item is denied: 6/11 correct (56%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n tacts common objects: 7/13 correct (53%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n echoes 2-word phrases: 7/11 correct (64%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 2 \u2014 2026-09-23 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 8/13 correct (63%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (66%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 8/15 correct (54%); latency 3.3s; prompts FP\u00d75, PP\u00d73, G\u00d74, I\u00d73\n wait calmly when access to preferred item is denied: 6/13 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n tacts common objects: 8/15 correct (56%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n echoes 2-word phrases: 9/14 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n Pica: attempts 2 (0 unsuccessful \u2014 staff retrieved item before ingestion; 2 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-09-25 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 6/10 correct (62%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/15 correct (70%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d711\n requesting a break ('break please' or AAC 'I need a break'): 7/13 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 4/8 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n tacts common objects: 4/8 correct (55%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n echoes 2-word phrases: 8/12 correct (68%); latency 3.1s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n Pica: attempts 5 (4 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-09-27 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 8/11 correct (70%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/15 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n requesting a break ('break please' or AAC 'I need a break'): 8/12 correct (65%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 7/12 correct (58%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n tacts common objects: 5/8 correct (57%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n echoes 2-word phrases: 10/14 correct (73%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n Pica: attempts 5 (4 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-09-29 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 8/11 correct (73%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/15 correct (76%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 5/8 correct (65%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 4/8 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n tacts common objects: 5/9 correct (53%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n echoes 2-word phrases: 10/15 correct (67%); latency 3.1s; prompts FP\u00d72, PP\u00d75, G\u00d73, I\u00d75\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-10-01 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 9/12 correct (72%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/11 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 10/14 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 8/14 correct (54%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d76\n tacts common objects: 8/14 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n echoes 2-word phrases: 11/15 correct (74%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d75, I\u00d77\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-10-03 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 10/14 correct (75%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/10 correct (78%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 9/13 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 7/14 correct (53%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d710\n tacts common objects: 7/13 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n echoes 2-word phrases: 12/15 correct (77%); latency 2.4s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-10-05 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 7/9 correct (78%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/14 correct (84%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 7/10 correct (70%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n wait calmly when access to preferred item is denied: 6/12 correct (53%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n tacts common objects: 4/8 correct (53%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n echoes 2-word phrases: 10/14 correct (74%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d711\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 9 \u2014 2026-10-07 \u2014 51 min \u2014 1 observer\n reads 10\u201320 sight words: 8/10 correct (82%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/12 correct (81%); latency 2.5s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 12/15 correct (78%); latency 2.4s; prompts FP\u00d74, PP\u00d75, G\u00d74, I\u00d72\n wait calmly when access to preferred item is denied: 7/12 correct (57%); latency 3.5s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n tacts common objects: 5/9 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n echoes 2-word phrases: 7/9 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Attempts at property destruction (throwing materials, swiping items off the surface)\n- Pushing materials away from the work area\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nPica: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 70.9%.\n- Trend direction: upward (first session: 64%, last session: 82%).\n- Pica: mean 2.4/session (early-window mean 3.0 -> late-window mean 1.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "b248cd78c0ea1c32", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Pica": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "early", "n_sessions": 9, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["pica"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "10409"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5619\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Self-Care + Tolerating Denied Access + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-03-08 to 2026-04-01)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n3. brushing teeth independently (Self-Care)\n Method: Task Analysis\n4. making the bed (Home Skills)\n Method: Task Analysis\n5. copies simple words from model (Writing)\n Method: DTT\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-08 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 13/15 correct (84%); latency 2.1s; prompts FP\u00d75, PP\u00d70, G\u00d71, I\u00d79\n meal cleanup after eating: 7/8 correct (90%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n brushing teeth independently: 9/12 correct (77%); latency 2.3s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n making the bed: 10/11 correct (91%); latency 1.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n copies simple words from model: 12/15 correct (78%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d78\n Property destruction: freq 2\n\nSession 2 \u2014 2026-03-10 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 13/14 correct (90%); latency 2.0s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n meal cleanup after eating: 7/8 correct (85%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n brushing teeth independently: 12/15 correct (80%); latency 2.4s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d78\n making the bed: 9/10 correct (93%); latency 1.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n copies simple words from model: 12/15 correct (82%); latency 2.0s; prompts FP\u00d75, PP\u00d70, G\u00d74, I\u00d76\n Property destruction: freq 2\n\nSession 3 \u2014 2026-03-12 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/13 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n meal cleanup after eating: 10/12 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n brushing teeth independently: 7/9 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n making the bed: 8/9 correct (90%); latency 1.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n copies simple words from model: 10/13 correct (76%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d710\n Property destruction: freq 2\n\nSession 4 \u2014 2026-03-14 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/10 correct (89%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n meal cleanup after eating: 11/12 correct (90%); latency 1.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n brushing teeth independently: 11/15 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d78\n making the bed: 12/13 correct (90%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n copies simple words from model: 9/12 correct (79%); latency 2.4s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n Property destruction: freq 2\n\nSession 5 \u2014 2026-03-16 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 12/14 correct (88%); latency 2.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n meal cleanup after eating: 8/9 correct (87%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n brushing teeth independently: 6/8 correct (73%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n making the bed: 8/9 correct (90%); latency 1.7s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n copies simple words from model: 9/12 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n Property destruction: freq 3\n\nSession 6 \u2014 2026-03-18 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/11 correct (89%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n meal cleanup after eating: 10/11 correct (89%); latency 1.8s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n brushing teeth independently: 6/8 correct (70%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n making the bed: 8/9 correct (90%); latency 1.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n copies simple words from model: 8/12 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n Property destruction: freq 3\n\nSession 7 \u2014 2026-03-20 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 12/14 correct (88%); latency 2.2s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n meal cleanup after eating: 12/14 correct (87%); latency 1.9s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n brushing teeth independently: 8/11 correct (72%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n making the bed: 10/11 correct (89%); latency 1.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n copies simple words from model: 7/10 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n Property destruction: freq 2\n\nSession 8 \u2014 2026-03-22 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 12/14 correct (85%); latency 2.4s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n meal cleanup after eating: 10/11 correct (89%); latency 2.0s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n brushing teeth independently: 8/12 correct (66%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n making the bed: 9/11 correct (86%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n copies simple words from model: 6/9 correct (66%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n Property destruction: freq 2\n\nSession 9 \u2014 2026-03-24 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/11 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n meal cleanup after eating: 8/9 correct (88%); latency 1.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n brushing teeth independently: 9/14 correct (67%); latency 2.7s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n making the bed: 13/15 correct (85%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d79\n copies simple words from model: 10/14 correct (71%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d710\n Property destruction: freq 3\n\nSession 10 \u2014 2026-03-26 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/10 correct (88%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n meal cleanup after eating: 10/11 correct (92%); latency 1.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n brushing teeth independently: 9/14 correct (65%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d76\n making the bed: 8/10 correct (82%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n copies simple words from model: 8/12 correct (63%); latency 3.1s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d72\n Property destruction: freq 3\n\nSession 11 \u2014 2026-03-28 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/11 correct (90%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n meal cleanup after eating: 10/11 correct (91%); latency 1.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n brushing teeth independently: 9/14 correct (67%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n making the bed: 9/11 correct (77%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n copies simple words from model: 6/9 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n Property destruction: freq 3\n\nSession 12 \u2014 2026-03-30 \u2014 60 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/9 correct (87%); latency 1.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n meal cleanup after eating: 8/9 correct (87%); latency 1.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n brushing teeth independently: 6/10 correct (60%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n making the bed: 8/11 correct (76%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n copies simple words from model: 6/10 correct (60%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n Property destruction: freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (84%\u201387%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nProperty destruction: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 87.2%.\n- Trend direction: upward (first session: 84%, last session: 87%).\n- Property destruction: mean 2.5/session (early-window mean 2.0 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "44f157d318bb92cb", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Property destruction": "attention"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "49562"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects by single function from array of 3\nCurriculum Reference: LRFFC L1\nLearner Profile: Early Learner\nCurrent Mastery: Developing\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects by single function from array of 3\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: cup, block, shoe\nDistractor stimuli: key, napkin\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3bd890922b748d88", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects by single function from array of 3", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "26013"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9424\nProfile: Adult Learner (18+ years)\nCurricula: Community + Home Skills + Independent Living + Mand + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-02-04 to 2026-02-20)\n\nACCELERATION PROGRAMS\n1. using public transportation (Community)\n Method: Task Analysis\n2. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n3. doing laundry (Home Skills)\n Method: Task Analysis\n4. requesting a break during extended tasks (Mand)\n Method: NET\n5. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-04 \u2014 46 min \u2014 1 observer\n using public transportation: 5/8 correct (68%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n scheduling medical appointments: 7/12 correct (61%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n doing laundry: 8/12 correct (64%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d78\n requesting a break during extended tasks: 12/15 correct (80%); latency 2.6s; prompts FP\u00d75, PP\u00d73, G\u00d70, I\u00d77\n dressing (pants, shirt): 8/11 correct (69%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n\nSession 2 \u2014 2026-02-06 \u2014 46 min \u2014 1 observer\n using public transportation: 9/13 correct (66%); latency 3.2s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d74\n scheduling medical appointments: 8/11 correct (70%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n doing laundry: 9/14 correct (64%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n requesting a break during extended tasks: 8/11 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n dressing (pants, shirt): 8/11 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n\nSession 3 \u2014 2026-02-08 \u2014 46 min \u2014 1 observer\n using public transportation: 8/12 correct (68%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n scheduling medical appointments: 9/14 correct (66%); latency 2.9s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d75\n doing laundry: 8/12 correct (70%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n requesting a break during extended tasks: 8/10 correct (79%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n dressing (pants, shirt): 6/9 correct (71%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n\nSession 4 \u2014 2026-02-10 \u2014 46 min \u2014 1 observer\n using public transportation: 5/9 correct (53%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n scheduling medical appointments: 4/12 correct (32%); latency 4.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n doing laundry: 4/9 correct (48%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n requesting a break during extended tasks: 6/13 correct (50%); latency 3.5s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d74\n dressing (pants, shirt): 6/15 correct (43%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n\nSession 5 \u2014 2026-02-12 \u2014 46 min \u2014 1 observer\n using public transportation: 7/11 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n scheduling medical appointments: 4/10 correct (41%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d77\n doing laundry: 7/13 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n requesting a break during extended tasks: 7/11 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n dressing (pants, shirt): 7/15 correct (47%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n\nSession 6 \u2014 2026-02-14 \u2014 46 min \u2014 1 observer\n using public transportation: 9/13 correct (69%); latency 2.6s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d77\n scheduling medical appointments: 3/8 correct (44%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n doing laundry: 5/9 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n requesting a break during extended tasks: 9/13 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d711\n dressing (pants, shirt): 6/11 correct (51%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n\nSession 7 \u2014 2026-02-16 \u2014 46 min \u2014 1 observer\n using public transportation: 10/15 correct (69%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d79\n scheduling medical appointments: 8/15 correct (51%); latency 3.9s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d77\n doing laundry: 5/9 correct (60%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n requesting a break during extended tasks: 8/11 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n dressing (pants, shirt): 6/10 correct (63%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n\nSession 8 \u2014 2026-02-18 \u2014 46 min \u2014 1 observer\n using public transportation: 8/12 correct (69%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n scheduling medical appointments: 5/10 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n doing laundry: 6/9 correct (64%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n requesting a break during extended tasks: 11/15 correct (75%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d75, I\u00d75\n dressing (pants, shirt): 6/9 correct (67%); latency 3.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 65.3%.\n- Trend direction: upward (first session: 68%, last session: 69%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "d0b834159ead6a38", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adult", "n_sessions": 8, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "65927"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: writes own name from model\nCurriculum: Writing L1\nLearner Level: Early Learner\nBaseline: Generalization phase\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: writes own name from model\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: blanket, key, glue\nDistractor stimuli: sock, ball\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "33649723e10238f2", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "writes own name from model", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "vr3", "error_correction": "simple_correction", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "61305"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8199\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Intraverbal + Listener Responding + Tolerating Denied Access\nDate range: Sessions 1\u20137 across 14 days (2026-09-17 to 2026-10-01)\n\nACCELERATION PROGRAMS\n1. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. echoes 2-word phrases (Echoic)\n Method: DTT variant\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-17 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 3/9 correct (36%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n answers WH-questions about familiar topics: 2/8 correct (30%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 2/8 correct (23%); latency 4.7s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n echoes 2-word phrases: 3/13 correct (22%); latency 5.3s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n\nSession 2 \u2014 2026-09-19 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 5/11 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n answers WH-questions about familiar topics: 6/14 correct (42%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n wait calmly when access to preferred item is denied: 5/14 correct (38%); latency 4.2s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n echoes 2-word phrases: 3/11 correct (31%); latency 4.5s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-09-21 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 9/15 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d78\n answers WH-questions about familiar topics: 6/12 correct (49%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n wait calmly when access to preferred item is denied: 6/15 correct (42%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n echoes 2-word phrases: 3/9 correct (39%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n\nSession 4 \u2014 2026-09-23 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 7/9 correct (73%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n answers WH-questions about familiar topics: 5/9 correct (60%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 7/12 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n echoes 2-word phrases: 7/14 correct (46%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n\nSession 5 \u2014 2026-09-25 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 11/13 correct (82%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 6/9 correct (72%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 9/14 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d78\n echoes 2-word phrases: 4/8 correct (55%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n\nSession 6 \u2014 2026-09-27 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 11/12 correct (93%); latency 1.7s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n answers WH-questions about familiar topics: 12/14 correct (83%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n wait calmly when access to preferred item is denied: 6/8 correct (79%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n echoes 2-word phrases: 9/15 correct (63%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n\nSession 7 \u2014 2026-09-29 \u2014 43 min \u2014 1 observer\n follows 1-step motor instructions: 9/9 correct (97%); latency 1.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n answers WH-questions about familiar topics: 7/8 correct (93%); latency 1.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n wait calmly when access to preferred item is denied: 7/8 correct (84%); latency 2.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n echoes 2-word phrases: 7/10 correct (72%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Quick response latency (within 2\u20133 seconds)\n- Requesting more trials verbally\n- Smiling during reinforcement delivery"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 36% to 97% across 7 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 69.9%.\n- Trend direction: upward (first session: 36%, last session: 97%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "db498b635738696c", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "early", "n_sessions": 7, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "23006"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: copies basic shapes from model\nCurriculum Reference: Writing L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies basic shapes from model\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: triangle, heart, square\nDistractor stimuli: star, rectangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fc4b2855638cbd8f", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies basic shapes from model", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "93978"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: identifies basic shapes in math context\nCurriculum Reference: Math L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies basic shapes in math context\n- Curriculum: Math L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: heart, diamond, square\nDistractor stimuli: circle, rectangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "2f1b8725440a7f2a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies basic shapes in math context", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "vr3", "error_correction": "backstep", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "60118"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: reads simple 2\u20133 word phrases\nCurriculum: Reading L2\nLearner Level: Early Learner\nBaseline: Developing\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: reads simple 2\u20133 word phrases\n- Curriculum: Reading L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What word?\"\nVariations: \"Read this word\"; \"What does it say?\"\nPresentation: Present the sight-word card and deliver the SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: bottle, book, apple\nDistractor stimuli: brush, phone, toothbrush, key\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0cb2c347cb31bddf", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "reads simple 2\u20133 word phrases", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "97982"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: follows instructions involving temporal concepts (before, after)\nCurriculum: Listener Responding L3\nLearner: Adolescent Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: follows instructions involving temporal concepts (before, after)\n- Curriculum: Listener Responding L3\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: phone, backpack, toothbrush\nDistractor stimuli: bottle, book, towel, key, napkin\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "4fabd8df5b3a4f19", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L3", "learner_profile": "adolescent", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "follows instructions involving temporal concepts (before, after)", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "85854"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5719\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Self-Care + Writing\nDate range: Sessions 1\u20137 across 14 days (2026-01-20 to 2026-02-03)\n\nACCELERATION PROGRAMS\n1. washing hands independently (Self-Care)\n Method: Task Analysis\n2. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n3. copies simple words from model (Writing)\n Method: DTT\n4. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n5. identifying community safety signs (Community)\n Method: DTT\n6. requesting a break during non-preferred activities (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-20 \u2014 37 min \u2014 1 observer\n washing hands independently: 6/9 correct (68%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n meal cleanup after eating: 9/15 correct (57%); latency 3.7s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n copies simple words from model: 6/11 correct (51%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n dressing (pants, shirt): 6/10 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n identifying community safety signs: 7/10 correct (66%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n requesting a break during non-preferred activities: 9/15 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d77\n Tantrum: freq 1, duration 3m total\n\nSession 2 \u2014 2026-01-22 \u2014 37 min \u2014 1 observer\n washing hands independently: 5/8 correct (65%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n meal cleanup after eating: 7/13 correct (54%); latency 3.3s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d77\n copies simple words from model: 4/8 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n dressing (pants, shirt): 5/8 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n identifying community safety signs: 8/11 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n requesting a break during non-preferred activities: 8/15 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d75, I\u00d77\n Tantrum: freq 2, duration 10m total\n\nSession 3 \u2014 2026-01-24 \u2014 37 min \u2014 1 observer\n washing hands independently: 8/11 correct (75%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n meal cleanup after eating: 8/14 correct (57%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d711\n copies simple words from model: 7/13 correct (54%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d78\n dressing (pants, shirt): 7/11 correct (60%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n identifying community safety signs: 8/11 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d75\n requesting a break during non-preferred activities: 7/12 correct (60%); latency 3.0s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n Tantrum: freq 5, duration 25m total\n\nSession 4 \u2014 2026-01-26 \u2014 37 min \u2014 1 observer\n washing hands independently: 7/9 correct (73%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n meal cleanup after eating: 5/9 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n copies simple words from model: 8/14 correct (55%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n dressing (pants, shirt): 7/12 correct (58%); latency 3.5s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d70\n identifying community safety signs: 8/11 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n requesting a break during non-preferred activities: 6/11 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n Tantrum: freq 7, duration 28m total\n\nSession 5 \u2014 2026-01-28 \u2014 37 min \u2014 1 observer\n washing hands independently: 7/10 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n meal cleanup after eating: 7/12 correct (58%); latency 3.5s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d72\n copies simple words from model: 5/8 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n dressing (pants, shirt): 5/8 correct (63%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n identifying community safety signs: 6/8 correct (81%); latency 2.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n requesting a break during non-preferred activities: 7/12 correct (62%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d75\n Tantrum: freq 3, duration 15m total\n\nSession 6 \u2014 2026-01-30 \u2014 37 min \u2014 1 observer\n washing hands independently: 6/8 correct (75%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n meal cleanup after eating: 7/13 correct (56%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n copies simple words from model: 5/8 correct (57%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n dressing (pants, shirt): 6/9 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n identifying community safety signs: 13/15 correct (85%); latency 2.4s; prompts FP\u00d74, PP\u00d72, G\u00d75, I\u00d74\n requesting a break during non-preferred activities: 7/11 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n Tantrum: freq 1, duration 2m total\n\nSession 7 \u2014 2026-02-01 \u2014 37 min \u2014 1 observer\n washing hands independently: 7/9 correct (83%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n meal cleanup after eating: 7/12 correct (57%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n copies simple words from model: 9/13 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n dressing (pants, shirt): 8/12 correct (63%); latency 3.4s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d76\n identifying community safety signs: 11/14 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n requesting a break during non-preferred activities: 6/10 correct (61%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n Tantrum: freq 3, duration 12m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Vocal refusal ('no', 'I don't want to')\n- Self-stimulatory behavior increase during demand periods"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 73.2%.\n- Trend direction: upward (first session: 68%, last session: 83%).\n- Tantrum: mean 3.1/session (early-window mean 1.5 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a9ab83f3f24f7c8e", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Tantrum": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adolescent", "n_sessions": 7, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "76065"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands for a break (e.g., \"break please\", AAC \"I need a break\")\nCurriculum Reference: Mand L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for a break (e.g., \"break please\", AAC \"I need a break\")\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nPresent a brief non-preferred task or demand at a level the learner can tolerate but may wish to pause. Wait for a request-for-break topography; honor the request immediately on first acquisition (CRF), thin later.\nPrimary natural context: academic period.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's academic period routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: immediate brief escape from the demand (30\u201360 seconds) to a neutral or preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "ac03b61c7496c605", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for a break (e.g., \"break please\", AAC \"I need a break\")", "mo_category": "mand_break", "mo_arrangement": "break_opportunity", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "75835"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: spontaneously uses humor appropriately\nCurriculum: Spontaneous Vocal Behavior L3\nLearner Level: School-Age Learner\nBaseline: Near mastery\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously uses humor appropriately\n- Curriculum: Spontaneous Vocal Behavior L3\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "0ac76b29391f279e", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L3", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously uses humor appropriately", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "36908"}, "curation": {"target_split": "test"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands using adjectives (big, little, more)\nCurriculum Reference: Mand L2\nLearner Profile: Early Learner\nCurrent Mastery: Generalization phase\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands using adjectives (big, little, more)\n- Curriculum: Mand L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: snack time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's snack time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "4cd2613946218aec", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands using adjectives (big, little, more)", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "time_delay", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "3768"}, "curation": {"target_split": "test"}}} diff --git a/data/splits/train.jsonl b/data/splits/train.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..1c2a70815bb09b27fd96c93ae46c03dfa162993f --- /dev/null +++ b/data/splits/train.jsonl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3553e10b4b04249f047fb5c035677c1cef5e32e89e3aad8686aa09b7ac2b5ca8 +size 16283798 diff --git a/data/splits/valid.jsonl b/data/splits/valid.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..b04a4ba1b37c6ca2c11bd87a0dc41798d6a1419c --- /dev/null +++ b/data/splits/valid.jsonl @@ -0,0 +1,149 @@ +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: shows items to others spontaneously\nCurriculum: Social Behavior & Social Play L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: shows items to others spontaneously\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "f993fb2de6f7b4ea", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "shows items to others spontaneously", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "3768"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: copies basic shapes from model\nCurriculum: Writing L1\nLearner Level: Early Learner\nBaseline: Generalization phase\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies basic shapes from model\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: square, star, heart\nDistractor stimuli: diamond, circle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3a150a8ce4d327a4", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies basic shapes from model", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "token_economy", "error_correction": "four_step", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "46080"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: putting on pants\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adolescent Learner\nBaseline: Approaching mastery\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: putting on pants\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Sit or stand securely.\n2. Hold pants with the front facing away.\n3. Place one leg into the corresponding pant leg.\n4. Place the other leg into the other pant leg.\n5. Pull pants up to the waist.\n6. Fasten the button / zipper / velcro. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "779fceb52fb328c9", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "putting on pants", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "approaching", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "putting on pants", "module": "basic_living", "mastery_criterion": "mc_all_steps_independent", "chain_type": "backward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "42817"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects correct item from array of 4\u20136\nCurriculum: Listener Responding L2\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from array of 4\u20136\n- Curriculum: Listener Responding L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: blanket, bottle, napkin\nDistractor stimuli: block, sock, jacket, plate\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e57ec2e9f483e923", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from array of 4\u20136", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "58800"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts common objects (ball, cup, shoe)\nCurriculum: Tact L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common objects (ball, cup, shoe)\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: scissors, plate, shoe\nDistractor stimuli: jacket, cup\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "da2c5e36be0f9539", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common objects (ball, cup, shoe)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "67967"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands using complete sentences with correct grammar\nCurriculum Reference: Mand L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Developing\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands using complete sentences with correct grammar\n- Curriculum: Mand L3\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: free play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's free play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "768edce6c7154f0f", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L3", "learner_profile": "adolescent", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands using complete sentences with correct grammar", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "16366"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8403\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Reading + Tact + Tolerating Denied Access + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-09-18 to 2026-10-12)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n4. tacts colors of objects (Tact)\n Method: DTT\n5. copies simple words from model (Writing)\n Method: DTT\n6. tacts common objects (Tact)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-18 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 9/15 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d75, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 5/8 correct (65%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n indicating completion ('all done', 'finished'): 10/13 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n tacts colors of objects: 7/11 correct (60%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n copies simple words from model: 9/15 correct (58%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d76\n tacts common objects: 9/12 correct (73%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n\nSession 2 \u2014 2026-09-20 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 5/9 correct (60%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 6/10 correct (63%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n indicating completion ('all done', 'finished'): 10/13 correct (76%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n tacts colors of objects: 9/15 correct (60%); latency 3.4s; prompts FP\u00d74, PP\u00d73, G\u00d75, I\u00d73\n copies simple words from model: 4/8 correct (56%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n tacts common objects: 9/14 correct (68%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n\nSession 3 \u2014 2026-09-22 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 8/14 correct (61%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 8/13 correct (65%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d79\n indicating completion ('all done', 'finished'): 7/10 correct (69%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n tacts colors of objects: 6/10 correct (58%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n copies simple words from model: 7/13 correct (56%); latency 3.7s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d74\n tacts common objects: 8/11 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n\nSession 4 \u2014 2026-09-24 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 8/15 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d711\n wait calmly when access to preferred item is denied: 9/15 correct (62%); latency 3.3s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d75\n indicating completion ('all done', 'finished'): 11/15 correct (73%); latency 2.4s; prompts FP\u00d72, PP\u00d75, G\u00d75, I\u00d73\n tacts colors of objects: 5/9 correct (54%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n copies simple words from model: 6/11 correct (56%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n tacts common objects: 11/15 correct (73%); latency 2.5s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d76\n\nSession 5 \u2014 2026-09-26 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 3/9 correct (30%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 3/8 correct (34%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n indicating completion ('all done', 'finished'): 6/12 correct (47%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n tacts colors of objects: 3/10 correct (31%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n copies simple words from model: 3/9 correct (33%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n tacts common objects: 5/10 correct (46%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n\nSession 6 \u2014 2026-09-28 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 5/15 correct (34%); latency 4.3s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d77\n wait calmly when access to preferred item is denied: 5/13 correct (40%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n indicating completion ('all done', 'finished'): 6/12 correct (51%); latency 4.0s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d73\n tacts colors of objects: 5/15 correct (35%); latency 4.7s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n copies simple words from model: 6/15 correct (38%); latency 4.5s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d76\n tacts common objects: 5/10 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n\nSession 7 \u2014 2026-09-30 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 3/8 correct (37%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 6/12 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d77\n indicating completion ('all done', 'finished'): 4/8 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n tacts colors of objects: 3/8 correct (37%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n copies simple words from model: 4/10 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n tacts common objects: 7/13 correct (54%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n\nSession 8 \u2014 2026-10-02 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 3/8 correct (41%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 7/15 correct (47%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n indicating completion ('all done', 'finished'): 7/13 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n tacts colors of objects: 6/14 correct (44%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n copies simple words from model: 7/15 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d75, I\u00d75\n tacts common objects: 8/12 correct (63%); latency 3.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n\nSession 9 \u2014 2026-10-04 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 5/10 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 6/12 correct (50%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d75\n indicating completion ('all done', 'finished'): 9/14 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d78\n tacts colors of objects: 4/10 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n copies simple words from model: 4/8 correct (56%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n tacts common objects: 11/15 correct (72%); latency 2.8s; prompts FP\u00d75, PP\u00d72, G\u00d73, I\u00d75\n\nSession 10 \u2014 2026-10-06 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 9/15 correct (58%); latency 3.6s; prompts FP\u00d75, PP\u00d71, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 6/13 correct (50%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d74\n indicating completion ('all done', 'finished'): 9/13 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d79\n tacts colors of objects: 5/10 correct (50%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n copies simple words from model: 8/14 correct (59%); latency 3.4s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d76\n tacts common objects: 9/13 correct (72%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d77\n\nSession 11 \u2014 2026-10-08 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 5/8 correct (58%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 7/13 correct (56%); latency 3.7s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n indicating completion ('all done', 'finished'): 7/9 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n tacts colors of objects: 5/9 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n copies simple words from model: 5/8 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n tacts common objects: 7/10 correct (72%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n\nSession 12 \u2014 2026-10-10 \u2014 35 min \u2014 1 observer\n sounds out CVC words: 6/10 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 5/8 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n indicating completion ('all done', 'finished'): 11/15 correct (74%); latency 2.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n tacts colors of objects: 8/14 correct (57%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n copies simple words from model: 7/12 correct (60%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n tacts common objects: 6/8 correct (72%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 50.3%.\n- Trend direction: upward (first session: 57%, last session: 58%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "630f4040bd7bbee9", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "early", "n_sessions": 12, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "62739"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6581\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20136 across 12 days (2026-02-20 to 2026-03-04)\n\nACCELERATION PROGRAMS\n1. clocking in at start of shift (Vocational)\n Method: Task Analysis\n2. brushing teeth independently (Self-Care)\n Method: Task Analysis\n3. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n4. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n5. requesting a break during extended tasks (Mand)\n Method: NET\n6. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-20 \u2014 48 min \u2014 1 observer\n clocking in at start of shift: 5/10 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n brushing teeth independently: 5/10 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n dressing (pants, shirt): 6/12 correct (49%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d76\n scheduling medical appointments: 7/14 correct (47%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n requesting a break during extended tasks: 4/9 correct (43%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n wait calmly when a preferred activity is denied: 8/15 correct (54%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d79\n Elopement: freq 3\n\nSession 2 \u2014 2026-02-22 \u2014 48 min \u2014 1 observer\n clocking in at start of shift: 5/8 correct (58%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n brushing teeth independently: 5/10 correct (50%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n dressing (pants, shirt): 6/10 correct (59%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n scheduling medical appointments: 6/11 correct (52%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n requesting a break during extended tasks: 4/8 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n wait calmly when a preferred activity is denied: 6/9 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n Elopement: freq 2\n\nSession 3 \u2014 2026-02-24 \u2014 48 min \u2014 1 observer\n clocking in at start of shift: 7/10 correct (68%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n brushing teeth independently: 6/10 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n dressing (pants, shirt): 7/10 correct (68%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n scheduling medical appointments: 8/15 correct (56%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n requesting a break during extended tasks: 9/15 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d75\n wait calmly when a preferred activity is denied: 6/9 correct (71%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n Elopement: freq 2\n\nSession 4 \u2014 2026-02-26 \u2014 48 min \u2014 1 observer\n clocking in at start of shift: 8/10 correct (77%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n brushing teeth independently: 8/12 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d77\n dressing (pants, shirt): 8/10 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n scheduling medical appointments: 9/14 correct (65%); latency 2.8s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d78\n requesting a break during extended tasks: 8/11 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n wait calmly when a preferred activity is denied: 9/11 correct (78%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n Elopement: freq 2\n\nSession 5 \u2014 2026-02-28 \u2014 48 min \u2014 1 observer\n clocking in at start of shift: 11/13 correct (82%); latency 2.4s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d78\n brushing teeth independently: 10/14 correct (69%); latency 3.1s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n dressing (pants, shirt): 10/13 correct (75%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d79\n scheduling medical appointments: 9/12 correct (72%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n requesting a break during extended tasks: 8/11 correct (74%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 8/9 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n Elopement: freq 3\n\nSession 6 \u2014 2026-03-02 \u2014 48 min \u2014 1 observer\n clocking in at start of shift: 11/13 correct (82%); latency 2.4s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n brushing teeth independently: 9/13 correct (72%); latency 2.5s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n dressing (pants, shirt): 8/11 correct (75%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n scheduling medical appointments: 6/8 correct (77%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n requesting a break during extended tasks: 8/11 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n wait calmly when a preferred activity is denied: 13/15 correct (83%); latency 2.4s; prompts FP\u00d73, PP\u00d75, G\u00d71, I\u00d76\n Elopement: freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Delayed responding (>5 seconds consistently)\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 49% at return, recovering to 82%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nElopement: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 6 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 69.5%.\n- Trend direction: upward (first session: 49%, last session: 82%).\n- Elopement: mean 2.5/session (early-window mean 2.5 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "12c0b9f047fa91a7", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Elopement": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adult", "n_sessions": 6, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "31866"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: answers why- and how-questions\nCurriculum Reference: Intraverbal L3\nLearner Profile: School-Age Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers why- and how-questions\n- Curriculum: Intraverbal L3\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "5e32f0bcc5a24008", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers why- and how-questions", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "time_delay", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "44373"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: matches identical objects\nCurriculum Reference: Visual Perceptual Skills & Matching-to-Sample L1\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches identical objects\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: sock, spoon, glue\nDistractor stimuli: scissors, blanket\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0aba22c68def3d9b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches identical objects", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "28081"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tolerates proximity of peers during parallel play\nCurriculum Reference: Social Behavior & Social Play L1\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerates proximity of peers during parallel play\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "e05844d91c863b3a", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "tolerates proximity of peers during parallel play", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "67055"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3672\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Mand + Motor Imitation + Tact\nDate range: Sessions 1\u20136 across 12 days (2026-10-25 to 2026-11-06)\n\nACCELERATION PROGRAMS\n1. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n4. tacts colors of objects (Tact)\n Method: DTT\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-25 \u2014 59 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 4/8 correct (50%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 9/15 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n imitates 3-step motor sequences: 6/12 correct (53%); latency 3.8s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n tacts colors of objects: 6/11 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 2\n\nSession 2 \u2014 2026-10-27 \u2014 59 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 6/12 correct (48%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n answers WH-questions about familiar topics: 5/8 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n imitates 3-step motor sequences: 6/10 correct (62%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n tacts colors of objects: 4/8 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n Self-injurious behavior (SIB): freq 4\n\nSession 3 \u2014 2026-10-29 \u2014 59 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 7/14 correct (53%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d714\n answers WH-questions about familiar topics: 5/8 correct (59%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n imitates 3-step motor sequences: 6/9 correct (64%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n tacts colors of objects: 7/12 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n Self-injurious behavior (SIB): freq 11\n\nSession 4 \u2014 2026-10-31 \u2014 59 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 4/9 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n answers WH-questions about familiar topics: 5/9 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n imitates 3-step motor sequences: 4/8 correct (55%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n tacts colors of objects: 6/11 correct (56%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n Self-injurious behavior (SIB): freq 11\n\nSession 5 \u2014 2026-11-02 \u2014 IOA SESSION \u2014 2 observers\n indicating completion ('all done', 'finished'): 51% (primary); IOA 85% agreement\n answers WH-questions about familiar topics: 66% (primary); IOA 85% agreement\n imitates 3-step motor sequences: 59% (primary); IOA 85% agreement\n tacts colors of objects: 53% (primary); IOA 85% agreement\n Self-injurious behavior (SIB): freq 2; IOA 85% agreement\n\nSession 6 \u2014 2026-11-04 \u2014 59 min \u2014 1 observer\n indicating completion ('all done', 'finished'): 5/8 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n answers WH-questions about familiar topics: 8/13 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n imitates 3-step motor sequences: 7/11 correct (64%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n tacts colors of objects: 9/15 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n Self-injurious behavior (SIB): freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Elopement from the work area"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 6 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 52.0%.\n- Trend direction: upward (first session: 50%, last session: 59%).\n- Self-injurious behavior (SIB): mean 5.3/session (early-window mean 3.0 -> late-window mean 2.0).\n- IOA data: 85% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "cf5ed11fed5f44f7", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "early", "n_sessions": 6, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "19932"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by class from array of 5\u20138\nCurriculum: LRFFC L2\nLearner: School-Age Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by class from array of 5\u20138\n- Curriculum: LRFFC L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: block, pencil, ball\nDistractor stimuli: toothbrush, blanket, car, towel\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "b97d2da22ddad7e6", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by class from array of 5\u20138", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "11906"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9140\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Writing\nDate range: Sessions 1\u201311 across 22 days (2026-01-21 to 2026-02-12)\n\nACCELERATION PROGRAMS\n1. selects items by feature (Listener Responding)\n Method: DTT\n2. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n3. copies simple words from model (Writing)\n Method: DTT\n4. sounds out CVC words (Reading)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-21 \u2014 49 min \u2014 1 observer\n selects items by feature: 7/8 correct (82%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n indicating completion ('all done', 'finished'): 9/10 correct (90%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n copies simple words from model: 11/13 correct (84%); latency 2.2s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n sounds out CVC words: 6/8 correct (81%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n\nSession 2 \u2014 2026-01-23 \u2014 49 min \u2014 1 observer\n selects items by feature: 4/13 correct (30%); latency 4.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n indicating completion ('all done', 'finished'): 2/11 correct (21%); latency 4.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n copies simple words from model: 3/11 correct (30%); latency 4.9s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 5/13 correct (38%); latency 4.2s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n\nSession 3 \u2014 2026-01-25 \u2014 49 min \u2014 1 observer\n selects items by feature: 7/9 correct (81%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n indicating completion ('all done', 'finished'): 9/10 correct (90%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n copies simple words from model: 10/13 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d78\n sounds out CVC words: 12/15 correct (79%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n\nSession 4 \u2014 2026-01-27 \u2014 49 min \u2014 1 observer\n selects items by feature: 4/13 correct (31%); latency 4.7s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n indicating completion ('all done', 'finished'): 4/14 correct (30%); latency 4.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n copies simple words from model: 4/10 correct (37%); latency 4.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n sounds out CVC words: 4/11 correct (34%); latency 4.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n\nSession 5 \u2014 2026-01-29 \u2014 49 min \u2014 1 observer\n selects items by feature: 8/10 correct (80%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n indicating completion ('all done', 'finished'): 10/12 correct (87%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n copies simple words from model: 11/13 correct (81%); latency 2.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d79\n sounds out CVC words: 8/10 correct (80%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n\nSession 6 \u2014 2026-01-31 \u2014 49 min \u2014 1 observer\n selects items by feature: 4/10 correct (35%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n indicating completion ('all done', 'finished'): 2/10 correct (24%); latency 4.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n copies simple words from model: 4/13 correct (33%); latency 4.4s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n sounds out CVC words: 3/11 correct (30%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n\nSession 7 \u2014 2026-02-02 \u2014 49 min \u2014 1 observer\n selects items by feature: 6/8 correct (77%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n indicating completion ('all done', 'finished'): 9/10 correct (89%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n copies simple words from model: 11/14 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n sounds out CVC words: 6/8 correct (79%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n\nSession 8 \u2014 2026-02-04 \u2014 49 min \u2014 1 observer\n selects items by feature: 5/14 correct (33%); latency 4.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n indicating completion ('all done', 'finished'): 3/9 correct (28%); latency 4.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n copies simple words from model: 3/9 correct (32%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n sounds out CVC words: 5/12 correct (38%); latency 4.4s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n\nSession 9 \u2014 2026-02-06 \u2014 49 min \u2014 1 observer\n selects items by feature: 11/13 correct (81%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d710\n indicating completion ('all done', 'finished'): 10/11 correct (91%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n copies simple words from model: 7/9 correct (81%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n sounds out CVC words: 7/8 correct (85%); latency 1.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 10 \u2014 2026-02-08 \u2014 49 min \u2014 1 observer\n selects items by feature: 5/14 correct (33%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n indicating completion ('all done', 'finished'): 3/13 correct (22%); latency 5.0s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n copies simple words from model: 4/10 correct (39%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n sounds out CVC words: 4/14 correct (32%); latency 4.4s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n\nSession 11 \u2014 2026-02-10 \u2014 49 min \u2014 1 observer\n selects items by feature: 12/15 correct (83%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n indicating completion ('all done', 'finished'): 13/15 correct (85%); latency 2.4s; prompts FP\u00d75, PP\u00d74, G\u00d75, I\u00d71\n copies simple words from model: 8/11 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 8/9 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 58.5%.\n- Trend direction: upward (first session: 82%, last session: 83%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "88414bb5aaaa6542", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "early", "n_sessions": 11, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "56533"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: maintains friendships over time\nCurriculum: Social Behavior & Social Play L3\nLearner: School-Age Learner\nMastery Status: Near mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: maintains friendships over time\n- Curriculum: Social Behavior & Social Play L3\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: circle time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's circle time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "4475ce08c4608b83", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L3", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "maintains friendships over time", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "40151"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6284\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand\nDate range: Sessions 1\u20139 across 18 days (2026-09-14 to 2026-10-02)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n3. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-14 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/11 correct (82%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n follows 1-step motor instructions: 9/11 correct (83%); latency 2.0s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n indicating completion ('all done', 'finished'): 13/15 correct (84%); latency 2.0s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n Tantrum: freq 3, duration 15m total\n\nSession 2 \u2014 2026-09-16 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/13 correct (29%); latency 4.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n follows 1-step motor instructions: 3/12 correct (29%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n indicating completion ('all done', 'finished'): 3/11 correct (30%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n Tantrum: freq 4, duration 8m total\n\nSession 3 \u2014 2026-09-18 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/8 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n follows 1-step motor instructions: 10/12 correct (83%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n indicating completion ('all done', 'finished'): 10/11 correct (90%); latency 1.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n Tantrum: freq 3, duration 9m total\n\nSession 4 \u2014 2026-09-20 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/14 correct (30%); latency 4.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n follows 1-step motor instructions: 3/9 correct (36%); latency 4.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n indicating completion ('all done', 'finished'): 2/9 correct (21%); latency 4.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n Tantrum: freq 3, duration 6m total\n\nSession 5 \u2014 2026-09-22 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/8 correct (85%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n follows 1-step motor instructions: 10/12 correct (83%); latency 2.0s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n indicating completion ('all done', 'finished'): 12/14 correct (86%); latency 1.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n Tantrum: freq 3, duration 9m total\n\nSession 6 \u2014 2026-09-24 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/14 correct (34%); latency 4.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d79\n follows 1-step motor instructions: 2/8 correct (31%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n indicating completion ('all done', 'finished'): 3/13 correct (27%); latency 4.7s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n Tantrum: freq 4, duration 20m total\n\nSession 7 \u2014 2026-09-26 \u2014 IOA SESSION \u2014 2 observers\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 81% (primary); IOA 84% agreement\n follows 1-step motor instructions: 82% (primary); IOA 84% agreement\n indicating completion ('all done', 'finished'): 85% (primary); IOA 84% agreement\n Tantrum: freq 3, duration 15m total; IOA 84% agreement\n\nSession 8 \u2014 2026-09-28 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/12 correct (37%); latency 4.6s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d74\n follows 1-step motor instructions: 4/14 correct (29%); latency 4.9s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n indicating completion ('all done', 'finished'): 3/13 correct (22%); latency 5.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n Tantrum: freq 3, duration 15m total\n\nSession 9 \u2014 2026-09-30 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/13 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n follows 1-step motor instructions: 8/10 correct (79%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n indicating completion ('all done', 'finished'): 7/8 correct (86%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n Tantrum: freq 2, duration 8m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 60.2%.\n- Trend direction: downward (first session: 82%, last session: 80%).\n- Tantrum: mean 3.1/session (early-window mean 3.3 -> late-window mean 2.7).\n- IOA data: 84% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "c7e377304547b39b", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {"Tantrum": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "early", "n_sessions": 9, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "40245"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects items by class (Find the animal)\nCurriculum: LRFFC L1\nLearner Level: Early Learner\nBaseline: Developing\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by class (Find the animal)\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: elephant, duck, frog\nDistractor stimuli: rabbit, cat\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5fd55f18b66986e6", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by class (Find the animal)", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "54357"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by function when item not visible\nCurriculum: LRFFC L2\nLearner: School-Age Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function when item not visible\n- Curriculum: LRFFC L2\n- Learner: School-Age Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: shoe, glue, napkin\nDistractor stimuli: scissors, phone, hat, cup\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "536ca68ab76c5567", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L2", "learner_profile": "school_age", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function when item not visible", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "errorless", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "92219"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9881\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Reading + Self-Care + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-06-20 to 2026-07-14)\n\nACCELERATION PROGRAMS\n1. sounds out CVC words (Reading)\n Method: DTT\n2. copies simple words from model (Writing)\n Method: DTT\n3. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n4. washing hands independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Fecal smearing (scatolia) \u2014 function hypothesized: attention\n2. Tantrum \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-20 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 12/15 correct (77%); latency 2.5s; prompts FP\u00d75, PP\u00d73, G\u00d70, I\u00d77\n copies simple words from model: 9/12 correct (72%); latency 2.9s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n meal cleanup after eating: 7/9 correct (77%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n washing hands independently: 10/13 correct (81%); latency 2.5s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 2, duration 8m total\n\nSession 2 \u2014 2026-06-22 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 12/15 correct (78%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d710\n copies simple words from model: 8/12 correct (69%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n meal cleanup after eating: 7/9 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n washing hands independently: 9/11 correct (83%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n Fecal smearing (scatolia): attempts 2 (2 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 1, duration 2m total\n\nSession 3 \u2014 2026-06-24 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 7/9 correct (75%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n copies simple words from model: 7/10 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n meal cleanup after eating: 9/12 correct (74%); latency 2.5s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d72\n washing hands independently: 8/11 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 2, duration 6m total\n\nSession 4 \u2014 2026-06-26 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 7/10 correct (70%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n copies simple words from model: 6/10 correct (57%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n meal cleanup after eating: 7/9 correct (72%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n washing hands independently: 8/12 correct (65%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d77\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 3, duration 15m total\n\nSession 5 \u2014 2026-06-28 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 8/12 correct (66%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n copies simple words from model: 4/8 correct (47%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n meal cleanup after eating: 7/12 correct (60%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d79\n washing hands independently: 8/13 correct (61%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n Fecal smearing (scatolia): attempts 3 (3 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 3, duration 9m total\n\nSession 6 \u2014 2026-06-30 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 8/13 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d75\n copies simple words from model: 6/13 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n meal cleanup after eating: 9/15 correct (61%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d77\n washing hands independently: 4/8 correct (56%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Fecal smearing (scatolia): attempts 5 (2 intercepted \u2014 staff redirected before smearing; 3 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 5, duration 25m total\n\nSession 7 \u2014 2026-07-02 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 6/10 correct (55%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n copies simple words from model: 4/9 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n meal cleanup after eating: 6/11 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n washing hands independently: 6/11 correct (52%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n Fecal smearing (scatolia): attempts 6 (4 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 6, duration 12m total\n\nSession 8 \u2014 2026-07-04 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 7/13 correct (52%); latency 3.8s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n copies simple words from model: 3/8 correct (33%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n meal cleanup after eating: 7/14 correct (51%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n washing hands independently: 5/10 correct (48%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n Fecal smearing (scatolia): attempts 7 (7 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 5, duration 20m total\n\nSession 9 \u2014 2026-07-06 \u2014 IOA SESSION \u2014 2 observers\n sounds out CVC words: 38% (primary); IOA 87% agreement\n copies simple words from model: 23% (primary); IOA 87% agreement\n meal cleanup after eating: 48% (primary); IOA 87% agreement\n washing hands independently: 45% (primary); IOA 87% agreement\n Fecal smearing (scatolia): attempts 6 (4 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface); IOA 87% agreement\n Tantrum: freq 6, duration 30m total; IOA 87% agreement\n\nSession 10 \u2014 2026-07-08 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 5/12 correct (41%); latency 4.4s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n copies simple words from model: 2/13 correct (17%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n meal cleanup after eating: 4/9 correct (48%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n washing hands independently: 3/8 correct (33%); latency 4.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n Fecal smearing (scatolia): attempts 8 (7 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 7, duration 35m total\n\nSession 11 \u2014 2026-07-10 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 3/9 correct (35%); latency 4.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n copies simple words from model: 1/11 correct (11%); latency 5.7s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n meal cleanup after eating: 4/9 correct (40%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n washing hands independently: 3/8 correct (32%); latency 4.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n Fecal smearing (scatolia): attempts 8 (5 intercepted \u2014 staff redirected before smearing; 3 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 8, duration 24m total\n\nSession 12 \u2014 2026-07-12 \u2014 53 min \u2014 1 observer\n sounds out CVC words: 3/11 correct (23%); latency 5.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n copies simple words from model: 2/13 correct (14%); latency 5.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n meal cleanup after eating: 5/12 correct (40%); latency 4.5s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d75\n washing hands independently: 3/13 correct (26%); latency 4.9s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n Fecal smearing (scatolia): attempts 9 (9 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n Tantrum: freq 8, duration 24m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Crying during demand presentation\n- Peer aggression (hitting or pushing peers during group activities)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (77% -> 23%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nFecal smearing (scatolia): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\nTantrum: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nhigh\n\nPattern is well-supported: 12 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 55.8%.\n- Trend direction: downward (first session: 77%, last session: 23%).\n- Fecal smearing (scatolia): mean 5.0/session (early-window mean 2.0 -> late-window mean 7.8).\n- Tantrum: mean 4.7/session (early-window mean 2.0 -> late-window mean 7.2).\n- IOA data: 87% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "d5d2af054bcfff58", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Fecal smearing (scatolia)": "attention", "Tantrum": "attention"}, "escalation_level": 4, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adolescent", "n_sessions": 12, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["fecal_smearing", "tantrum"], "behavior_functions": ["attention", "attention"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "3991"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: reproduces block designs from model\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L3\nLearner: Adolescent Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: reproduces block designs from model\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L3\n- Learner: Adolescent Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: plate, apple, glue\nDistractor stimuli: towel, bottle, jacket, sock, book\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7ba0688302c1f43f", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L3", "learner_profile": "adolescent", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "reproduces block designs from model", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "transfer_trial", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "38800"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: identifies 5\u201310 upper-case letters\nCurriculum: Reading L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies 5\u201310 upper-case letters\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: I, H, D\nDistractor stimuli: G, A\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a7083e5e4134648a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies 5\u201310 upper-case letters", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "15942"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: reads 10\u201320 sight words\nCurriculum Reference: Reading L2\nLearner Profile: Early Learner\nCurrent Mastery: Developing\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: reads 10\u201320 sight words\n- Curriculum: Reading L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What word?\"\nVariations: \"Read this word\"; \"What does it say?\"\nPresentation: Present the sight-word card and deliver the SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: jacket, bottle, glue\nDistractor stimuli: block, phone, car, ball\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "03a2478f1757992a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "reads 10\u201320 sight words", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "17197"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8461\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding + Mand + Reading\nDate range: Sessions 1\u201312 across 24 days (2026-06-24 to 2026-07-18)\n\nACCELERATION PROGRAMS\n1. selects items by feature (Listener Responding)\n Method: DTT\n2. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n5. sounds out CVC words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Mouthing \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-24 \u2014 41 min \u2014 1 observer\n selects items by feature: 6/14 correct (41%); latency 4.0s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d73\n follows 1-step motor instructions: 5/15 correct (36%); latency 4.6s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/15 correct (45%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d78\n answers WH-questions about familiar topics: 4/11 correct (37%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n sounds out CVC words: 3/12 correct (26%); latency 4.6s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n Mouthing: freq 3; PIR 16%\n\nSession 2 \u2014 2026-06-26 \u2014 41 min \u2014 1 observer\n selects items by feature: 6/14 correct (45%); latency 4.3s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d73\n follows 1-step motor instructions: 5/12 correct (43%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (57%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n answers WH-questions about familiar topics: 6/14 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d79\n sounds out CVC words: 5/14 correct (38%); latency 4.3s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n Mouthing: freq 3; PIR 21%\n\nSession 3 \u2014 2026-06-28 \u2014 41 min \u2014 1 observer\n selects items by feature: 7/11 correct (60%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n follows 1-step motor instructions: 4/9 correct (46%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/14 correct (60%); latency 3.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n answers WH-questions about familiar topics: 6/14 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d712\n sounds out CVC words: 6/14 correct (43%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n Mouthing: freq 3; PIR 19%\n\nSession 4 \u2014 2026-06-30 \u2014 41 min \u2014 1 observer\n selects items by feature: 8/14 correct (59%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d77\n follows 1-step motor instructions: 8/14 correct (55%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (63%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n answers WH-questions about familiar topics: 4/8 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n sounds out CVC words: 4/10 correct (38%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n Mouthing: freq 3; PIR 12%\n\nSession 5 \u2014 2026-07-02 \u2014 41 min \u2014 1 observer\n selects items by feature: 6/8 correct (73%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n follows 1-step motor instructions: 7/12 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (72%); latency 2.9s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n answers WH-questions about familiar topics: 9/15 correct (60%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d75, I\u00d72\n sounds out CVC words: 4/9 correct (46%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n Mouthing: freq 2; PIR 9%\n\nSession 6 \u2014 2026-07-04 \u2014 41 min \u2014 1 observer\n selects items by feature: 7/10 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n follows 1-step motor instructions: 8/11 correct (68%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/11 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n answers WH-questions about familiar topics: 9/14 correct (64%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n sounds out CVC words: 4/8 correct (47%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Mouthing: freq 3; PIR 18%\n\nSession 7 \u2014 2026-07-06 \u2014 41 min \u2014 1 observer\n selects items by feature: 10/12 correct (83%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d75\n follows 1-step motor instructions: 6/9 correct (72%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/9 correct (83%); latency 2.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n answers WH-questions about familiar topics: 9/13 correct (65%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n sounds out CVC words: 7/13 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n Mouthing: freq 3; PIR 21%\n\nSession 8 \u2014 2026-07-08 \u2014 41 min \u2014 1 observer\n selects items by feature: 9/10 correct (93%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n follows 1-step motor instructions: 10/13 correct (77%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (92%); latency 2.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n answers WH-questions about familiar topics: 10/15 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d78\n sounds out CVC words: 7/13 correct (57%); latency 3.7s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d77\n Mouthing: freq 4; PIR 21%\n\nSession 9 \u2014 2026-07-10 \u2014 41 min \u2014 1 observer\n selects items by feature: 11/12 correct (95%); latency 1.8s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n follows 1-step motor instructions: 8/10 correct (80%); latency 2.2s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/11 correct (95%); latency 1.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n answers WH-questions about familiar topics: 10/14 correct (75%); latency 2.9s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n sounds out CVC words: 5/8 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n Mouthing: freq 4; PIR 21%\n\nSession 10 \u2014 2026-07-12 \u2014 41 min \u2014 1 observer\n selects items by feature: 14/15 correct (95%); latency 1.6s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n follows 1-step motor instructions: 12/13 correct (90%); latency 2.2s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 13/14 correct (95%); latency 2.0s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n answers WH-questions about familiar topics: 11/13 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n sounds out CVC words: 7/11 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Mouthing: freq 3; PIR 13%\n\nSession 11 \u2014 2026-07-14 \u2014 41 min \u2014 1 observer\n selects items by feature: 9/9 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n follows 1-step motor instructions: 9/9 correct (95%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 14/15 correct (95%); latency 1.9s; prompts FP\u00d75, PP\u00d75, G\u00d70, I\u00d75\n answers WH-questions about familiar topics: 13/14 correct (95%); latency 1.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n sounds out CVC words: 11/15 correct (71%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d711\n Mouthing: freq 4; PIR 18%\n\nSession 12 \u2014 2026-07-16 \u2014 41 min \u2014 1 observer\n selects items by feature: 14/15 correct (95%); latency 1.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n follows 1-step motor instructions: 10/10 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/8 correct (95%); latency 1.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n answers WH-questions about familiar topics: 10/11 correct (95%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n sounds out CVC words: 11/15 correct (76%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n Mouthing: freq 3; PIR 23%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Quick response latency (within 2\u20133 seconds)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 41% to 95% across 12 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Behavior Function Hypothesis\nMouthing: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 75.6%.\n- Trend direction: upward (first session: 41%, last session: 95%).\n- Mouthing: mean 3.2/session (early-window mean 3.0 -> late-window mean 3.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "c46cbc73518f1c7a", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {"Mouthing": "automatic"}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "early", "n_sessions": 12, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["mouthing"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "12611"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts common objects (ball, cup, shoe)\nCurriculum: Tact L1\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common objects (ball, cup, shoe)\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: book, spoon, towel\nDistractor stimuli: hat, apple\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f5ccec5c5b46e706", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common objects (ball, cup, shoe)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "5667"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: putting on pants\nCurriculum: AFLS Basic Living Skills\nLearner: Adolescent Learner\nMastery Status: Developing\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: putting on pants\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Sit or stand securely.\n2. Hold pants with the front facing away.\n3. Place one leg into the corresponding pant leg.\n4. Place the other leg into the other pant leg.\n5. Pull pants up to the waist.\n6. Fasten the button / zipper / velcro. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "4cf62bfc8853a5b1", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "putting on pants", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "developing", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "putting on pants", "module": "basic_living", "mastery_criterion": "mc_self_initiation", "chain_type": "backward", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "57808"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9529\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20138 across 16 days (2026-05-16 to 2026-06-01)\n\nACCELERATION PROGRAMS\n1. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n2. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n3. self-advocacy statements (Independent Living)\n Method: BST\n4. requesting assistance using AAC device (Mand)\n Method: NET\n5. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Pica \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-16 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 11/14 correct (76%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d75\n wait calmly when a preferred activity is denied: 6/9 correct (63%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n self-advocacy statements: 6/11 correct (59%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n requesting assistance using AAC device: 7/13 correct (51%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n completing assigned tasks in order: 4/8 correct (54%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 2 \u2014 2026-05-18 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 11/15 correct (74%); latency 2.5s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d75\n wait calmly when a preferred activity is denied: 5/9 correct (59%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n self-advocacy statements: 6/10 correct (56%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n requesting assistance using AAC device: 8/14 correct (56%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n completing assigned tasks in order: 6/10 correct (56%); latency 3.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-05-20 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 9/12 correct (73%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n wait calmly when a preferred activity is denied: 5/8 correct (57%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n self-advocacy statements: 9/15 correct (62%); latency 3.5s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d74\n requesting assistance using AAC device: 5/10 correct (53%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n completing assigned tasks in order: 6/10 correct (62%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Pica: attempts 2 (0 unsuccessful \u2014 staff retrieved item before ingestion; 2 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-05-22 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 4/10 correct (43%); latency 4.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n wait calmly when a preferred activity is denied: 4/15 correct (28%); latency 4.7s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n self-advocacy statements: 3/9 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n requesting assistance using AAC device: 3/11 correct (28%); latency 4.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n completing assigned tasks in order: 5/14 correct (37%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d712\n Pica: attempts 5 (5 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-05-24 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 5/10 correct (45%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n wait calmly when a preferred activity is denied: 4/15 correct (29%); latency 4.8s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d79\n self-advocacy statements: 5/13 correct (40%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d710\n requesting assistance using AAC device: 3/8 correct (35%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n completing assigned tasks in order: 7/14 correct (47%); latency 3.7s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n Pica: attempts 4 (3 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-05-26 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 7/14 correct (48%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d713\n wait calmly when a preferred activity is denied: 3/8 correct (33%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n self-advocacy statements: 4/8 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n requesting assistance using AAC device: 4/12 correct (35%); latency 4.7s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n completing assigned tasks in order: 8/15 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d712\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-05-28 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 4/8 correct (53%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n wait calmly when a preferred activity is denied: 5/12 correct (38%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d73\n self-advocacy statements: 7/13 correct (51%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d77\n requesting assistance using AAC device: 5/12 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n completing assigned tasks in order: 5/8 correct (57%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-05-30 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 8/15 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n wait calmly when a preferred activity is denied: 4/9 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n self-advocacy statements: 5/9 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n requesting assistance using AAC device: 5/11 correct (41%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n completing assigned tasks in order: 6/10 correct (57%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n- Leaving seat without permission\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nPica: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n### Crisis plan\n- If the setting event involves a medical concern (seizure-like activity, injury, illness): stop the program immediately, follow the facility's medical-incident protocol, and refer to the appropriate medical professional. Medical events are NOT behavioral targets and must not be treated as ABA programming issues.\n- If behavior escalates during the setting-event window: use verbal de-escalation and environmental safety first; do not add demand. Only staff currently certified in the facility's crisis-prevention framework may use physical intervention, and only if imminent safety risk warrants it AND the learner's BIP specifically authorizes it (contraindications and BIP restrictions may rule it out).\n- Coordinate with caregivers, the medical team, and the supervising BCBA; document the setting event and the associated behavior for pattern analysis. Review the BIP in light of the setting-event data.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 58.3%.\n- Trend direction: downward (first session: 76%, last session: 54%).\n- Pica: mean 2.1/session (early-window mean 1.0 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "23e3474b05ccc9f2", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Pica": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adult", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["pica"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "67790"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner Level: Adult Learner\nBaseline: Developing\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the developing stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ea8edaf7ac6903ad", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adult", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "44637"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: identifies 5\u201310 upper-case letters\nCurriculum: Reading L1\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies 5\u201310 upper-case letters\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: B, F, I\nDistractor stimuli: J, D\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f636b7e53e4ed638", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies 5\u201310 upper-case letters", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "vr3", "error_correction": "simple_correction", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "69141"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3969\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Motor Imitation + Tact + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-03-08 to 2026-03-30)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n5. tacts colors of objects (Tact)\n Method: DTT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-08 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/15 correct (40%); latency 4.3s; prompts FP\u00d70, PP\u00d74, G\u00d75, I\u00d76\n imitates 3-step motor sequences: 4/9 correct (50%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 7/15 correct (45%); latency 3.7s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d711\n requesting a break ('break please' or AAC 'I need a break'): 8/15 correct (53%); latency 3.6s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d78\n tacts colors of objects: 6/12 correct (49%); latency 3.7s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n Elopement: freq 2\n\nSession 2 \u2014 2026-03-10 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (52%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d76\n imitates 3-step motor sequences: 4/10 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n wait calmly when access to preferred item is denied: 9/14 correct (63%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 5/12 correct (42%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n tacts colors of objects: 10/12 correct (84%); latency 1.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n Elopement: freq 3\n\nSession 3 \u2014 2026-03-12 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (46%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n imitates 3-step motor sequences: 8/13 correct (62%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 5/9 correct (56%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 4/11 correct (33%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n tacts colors of objects: 7/9 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n Elopement: freq 4\n\nSession 4 \u2014 2026-03-14 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (33%); latency 4.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n imitates 3-step motor sequences: 4/8 correct (51%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 6/10 correct (63%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 3/8 correct (40%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n tacts colors of objects: 10/14 correct (74%); latency 2.7s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n Elopement: freq 7\n\nSession 5 \u2014 2026-03-16 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (51%); latency 3.6s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n imitates 3-step motor sequences: 4/15 correct (24%); latency 4.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n wait calmly when access to preferred item is denied: 5/14 correct (37%); latency 4.1s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 7/13 correct (57%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n tacts colors of objects: 5/10 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n Elopement: freq 3\n\nSession 6 \u2014 2026-03-18 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n imitates 3-step motor sequences: 6/15 correct (39%); latency 4.1s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 8/15 correct (52%); latency 3.6s; prompts FP\u00d71, PP\u00d75, G\u00d70, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 5/10 correct (55%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n tacts colors of objects: 9/11 correct (82%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n Elopement: freq 2\n\nSession 7 \u2014 2026-03-20 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/15 correct (56%); latency 3.3s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d76\n imitates 3-step motor sequences: 5/8 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 5/9 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n requesting a break ('break please' or AAC 'I need a break'): 8/14 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n tacts colors of objects: 7/8 correct (82%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Elopement: freq 3\n\nSession 8 \u2014 2026-03-22 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n imitates 3-step motor sequences: 5/12 correct (43%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n wait calmly when access to preferred item is denied: 3/9 correct (35%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 4/13 correct (31%); latency 4.5s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n tacts colors of objects: 7/11 correct (63%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n Elopement: freq 7\n\nSession 9 \u2014 2026-03-24 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n imitates 3-step motor sequences: 5/8 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 4/14 correct (28%); latency 4.7s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 5/10 correct (47%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n tacts colors of objects: 12/14 correct (84%); latency 2.2s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d77\n Elopement: freq 7\n\nSession 10 \u2014 2026-03-26 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n imitates 3-step motor sequences: 5/11 correct (44%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 5/11 correct (43%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 4/9 correct (41%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n tacts colors of objects: 9/12 correct (74%); latency 2.4s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n Elopement: freq 7\n\nSession 11 \u2014 2026-03-28 \u2014 37 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (64%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n imitates 3-step motor sequences: 6/10 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 8/13 correct (60%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 4/15 correct (26%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n tacts colors of objects: 6/12 correct (50%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n Elopement: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Leaving seat without permission\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 52% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nElopement: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 51.9%.\n- Trend direction: upward (first session: 40%, last session: 64%).\n- Elopement: mean 4.6/session (early-window mean 3.0 -> late-window mean 6.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "cd6be09a745baf15", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Elopement": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "early", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "8363"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3823\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand\nDate range: Sessions 1\u201310 across 20 days (2026-10-26 to 2026-11-15)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-26 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/10 correct (82%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 11/12 correct (89%); latency 2.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/9 correct (91%); latency 1.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n\nSession 2 \u2014 2026-10-28 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/13 correct (24%); latency 5.0s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/9 correct (41%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/9 correct (22%); latency 5.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n\nSession 3 \u2014 2026-10-30 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/11 correct (83%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 13/15 correct (89%); latency 1.9s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 14/15 correct (92%); latency 2.0s; prompts FP\u00d71, PP\u00d75, G\u00d70, I\u00d79\n\nSession 4 \u2014 2026-11-01 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/14 correct (31%); latency 4.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d711\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/14 correct (41%); latency 4.0s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/11 correct (28%); latency 5.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n\nSession 5 \u2014 2026-11-03 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/10 correct (84%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/10 correct (92%); latency 1.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/12 correct (89%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n\nSession 6 \u2014 2026-11-05 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/15 correct (23%); latency 5.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/8 correct (36%); latency 4.3s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/13 correct (27%); latency 4.6s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n\nSession 7 \u2014 2026-11-07 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/11 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 10/12 correct (87%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/13 correct (92%); latency 1.6s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d73\n\nSession 8 \u2014 2026-11-09 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (26%); latency 4.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/15 correct (38%); latency 4.3s; prompts FP\u00d73, PP\u00d75, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/13 correct (21%); latency 5.1s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n\nSession 9 \u2014 2026-11-11 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/8 correct (83%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 12/13 correct (92%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 13/14 correct (91%); latency 1.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d712\n\nSession 10 \u2014 2026-11-13 \u2014 47 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (30%); latency 4.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/10 correct (38%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/11 correct (21%); latency 5.3s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 54.6%.\n- Trend direction: downward (first session: 82%, last session: 30%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "48e8f95de9ec59fb", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "2267"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects correct item from array of 4\u20136\nCurriculum Reference: Listener Responding L2\nLearner Profile: School-Age Learner\nCurrent Mastery: Generalization phase\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from array of 4\u20136\n- Curriculum: Listener Responding L2\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: sock, apple, cup\nDistractor stimuli: towel, phone, glue, bottle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c2d0ee2a88b2a486", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L2", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from array of 4\u20136", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "token_economy", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "5667"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: traces letters of own name\nCurriculum: Writing L1\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: traces letters of own name\n- Curriculum: Writing L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: H, U, C\nDistractor stimuli: V, T\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "9982362ac71a598d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "traces letters of own name", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "68189"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7482\nProfile: Adult Learner (18+ years)\nCurricula: Community + Mand + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-03-14 to 2026-04-05)\n\nACCELERATION PROGRAMS\n1. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n2. using public transportation (Community)\n Method: Task Analysis\n3. requesting assistance using AAC device (Mand)\n Method: NET\n4. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Aggression \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-14 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 8/14 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d711\n using public transportation: 11/14 correct (76%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n requesting assistance using AAC device: 6/9 correct (67%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n dressing (pants, shirt): 6/12 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d72\n Aggression: freq 7\n\nSession 2 \u2014 2026-03-16 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 7/9 correct (78%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n using public transportation: 7/15 correct (46%); latency 4.1s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d78\n requesting assistance using AAC device: 8/13 correct (64%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n dressing (pants, shirt): 4/13 correct (29%); latency 5.0s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n Aggression: freq 5\n\nSession 3 \u2014 2026-03-18 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 6/8 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n using public transportation: 10/14 correct (74%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n requesting assistance using AAC device: 6/10 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n dressing (pants, shirt): 4/9 correct (40%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n Aggression: freq 7\n\nSession 4 \u2014 2026-03-20 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 11/15 correct (72%); latency 3.0s; prompts FP\u00d74, PP\u00d71, G\u00d75, I\u00d75\n using public transportation: 5/8 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n requesting assistance using AAC device: 9/15 correct (57%); latency 3.4s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d78\n dressing (pants, shirt): 4/12 correct (30%); latency 4.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n Aggression: freq 6\n\nSession 5 \u2014 2026-03-22 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 7/15 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d711\n using public transportation: 4/8 correct (52%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n requesting assistance using AAC device: 5/10 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n dressing (pants, shirt): 5/9 correct (59%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n Aggression: freq 7\n\nSession 6 \u2014 2026-03-24 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/8 correct (66%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n using public transportation: 6/13 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d79\n requesting assistance using AAC device: 8/12 correct (70%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n dressing (pants, shirt): 4/14 correct (31%); latency 4.5s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n Aggression: freq 2\n\nSession 7 \u2014 2026-03-26 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 10/15 correct (68%); latency 3.0s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n using public transportation: 9/12 correct (73%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d79\n requesting assistance using AAC device: 9/13 correct (67%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d79\n dressing (pants, shirt): 6/12 correct (48%); latency 4.0s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n Aggression: freq 8\n\nSession 8 \u2014 2026-03-28 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 6/10 correct (63%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n using public transportation: 10/13 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d78\n requesting assistance using AAC device: 7/13 correct (56%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d78\n dressing (pants, shirt): 5/11 correct (45%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n Aggression: freq 4\n\nSession 9 \u2014 2026-03-30 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/9 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n using public transportation: 7/9 correct (76%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n requesting assistance using AAC device: 10/14 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n dressing (pants, shirt): 5/11 correct (41%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n Aggression: freq 7\n\nSession 10 \u2014 2026-04-01 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 13/15 correct (84%); latency 2.3s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n using public transportation: 5/11 correct (49%); latency 3.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n requesting assistance using AAC device: 5/9 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n dressing (pants, shirt): 6/10 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n Aggression: freq 6\n\nSession 11 \u2014 2026-04-03 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 7/8 correct (82%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n using public transportation: 8/12 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d75\n requesting assistance using AAC device: 5/11 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n dressing (pants, shirt): 4/14 correct (31%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d711\n Aggression: freq 5\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Stereotypic behavior during inter-trial intervals\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 69% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nAggression: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 68.1%.\n- Trend direction: upward (first session: 56%, last session: 82%).\n- Aggression: mean 5.8/session (early-window mean 6.3 -> late-window mean 6.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1586abc2fed90213", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Aggression": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "adult", "n_sessions": 11, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["aggression"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "52825"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: names items in categories when given the category\nCurriculum: Intraverbal L2\nLearner Level: School-Age Learner\nBaseline: Emerging\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: names items in categories when given the category\n- Curriculum: Intraverbal L2\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "da8e05750afeae84", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "names items in categories when given the category", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "16366"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2536\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u201312 across 24 days (2026-02-15 to 2026-03-11)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. sounds out CVC words (Reading)\n Method: DTT\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n5. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n6. mands for missing items needed to complete an activity (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-15 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/11 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n sounds out CVC words: 11/15 correct (75%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d79\n wait calmly when access to preferred item is denied: 8/10 correct (76%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n indicating 'all done' when an activity is complete: 9/12 correct (76%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/9 correct (74%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n mands for missing items needed to complete an activity: 13/15 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d75, I\u00d710\n Elopement: freq 2\n\nSession 2 \u2014 2026-02-17 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n sounds out CVC words: 10/13 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 6/8 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n indicating 'all done' when an activity is complete: 7/10 correct (72%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/11 correct (72%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n mands for missing items needed to complete an activity: 12/15 correct (79%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n Elopement: freq 2\n\nSession 3 \u2014 2026-02-19 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/15 correct (74%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n sounds out CVC words: 9/13 correct (70%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 6/9 correct (64%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 7/10 correct (69%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/15 correct (60%); latency 3.4s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 9/12 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d79\n Elopement: freq 3\n\nSession 4 \u2014 2026-02-21 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (63%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n sounds out CVC words: 9/14 correct (62%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 7/12 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n indicating 'all done' when an activity is complete: 7/11 correct (64%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/9 correct (57%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 9/14 correct (67%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d710\n Elopement: freq 3\n\nSession 5 \u2014 2026-02-23 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n sounds out CVC words: 8/15 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d712\n wait calmly when access to preferred item is denied: 5/10 correct (51%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n indicating 'all done' when an activity is complete: 6/9 correct (62%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 5/9 correct (54%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n mands for missing items needed to complete an activity: 9/15 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n Elopement: freq 4\n\nSession 6 \u2014 2026-02-25 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n sounds out CVC words: 4/9 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 7/15 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n indicating 'all done' when an activity is complete: 6/11 correct (50%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/13 correct (50%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n mands for missing items needed to complete an activity: 9/15 correct (58%); latency 3.3s; prompts FP\u00d75, PP\u00d75, G\u00d74, I\u00d71\n Elopement: freq 4\n\nSession 7 \u2014 2026-02-27 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/13 correct (46%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n sounds out CVC words: 6/13 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 4/13 correct (32%); latency 4.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d78\n indicating 'all done' when an activity is complete: 4/10 correct (42%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/14 correct (45%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d77\n mands for missing items needed to complete an activity: 8/15 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d75, G\u00d70, I\u00d77\n Elopement: freq 5\n\nSession 8 \u2014 2026-03-01 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/14 correct (41%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d710\n sounds out CVC words: 3/9 correct (38%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n wait calmly when access to preferred item is denied: 3/14 correct (21%); latency 4.8s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n indicating 'all done' when an activity is complete: 4/11 correct (39%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/10 correct (34%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n mands for missing items needed to complete an activity: 3/8 correct (43%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n Elopement: freq 4\n\nSession 9 \u2014 2026-03-03 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/15 correct (42%); latency 4.0s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n sounds out CVC words: 2/8 correct (26%); latency 5.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 2/11 correct (18%); latency 5.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n indicating 'all done' when an activity is complete: 4/10 correct (40%); latency 4.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/11 correct (32%); latency 4.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n mands for missing items needed to complete an activity: 4/10 correct (40%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n Elopement: freq 5\n\nSession 10 \u2014 2026-03-05 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/12 correct (33%); latency 4.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n sounds out CVC words: 2/11 correct (19%); latency 5.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 1/13 correct (10%); latency 5.3s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n indicating 'all done' when an activity is complete: 4/11 correct (33%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 3/10 correct (31%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n mands for missing items needed to complete an activity: 4/13 correct (32%); latency 4.6s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n Elopement: freq 5\n\nSession 11 \u2014 2026-03-07 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/12 correct (30%); latency 4.5s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n sounds out CVC words: 2/11 correct (19%); latency 4.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 2/15 correct (10%); latency 5.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n indicating 'all done' when an activity is complete: 2/8 correct (27%); latency 4.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 2/12 correct (17%); latency 5.0s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n mands for missing items needed to complete an activity: 3/9 correct (30%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n Elopement: freq 5\n\nSession 12 \u2014 2026-03-09 \u2014 47 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/11 correct (18%); latency 5.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n sounds out CVC words: 2/13 correct (12%); latency 5.7s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 1/8 correct (10%); latency 5.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n indicating 'all done' when an activity is complete: 2/11 correct (21%); latency 5.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 2/11 correct (20%); latency 5.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 3/14 correct (19%); latency 5.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d710\n Elopement: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Elopement from the work area\n- Peer aggression (hitting or pushing peers during group activities)\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 75% to 18% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 50.7%.\n- Trend direction: downward (first session: 75%, last session: 18%).\n- Elopement: mean 4.0/session (early-window mean 2.5 -> late-window mean 5.2).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "6424c21e648282b1", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Elopement": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "school_age", "n_sessions": 12, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "23269"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: spontaneously reports past events\nCurriculum: Spontaneous Vocal Behavior L2\nLearner: School-Age Learner\nMastery Status: Emerging\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously reports past events\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: room transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's room transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "d4156698d1388125", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously reports past events", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "16366"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4235\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand\nDate range: Sessions 1\u201310 across 20 days (2026-03-11 to 2026-03-31)\n\nACCELERATION PROGRAMS\n1. requesting assistance using AAC device (Mand)\n Method: NET\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. doing laundry (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-11 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 12/14 correct (86%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n self-advocacy statements: 8/11 correct (77%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n doing laundry: 7/8 correct (91%); latency 1.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n\nSession 2 \u2014 2026-03-13 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 5/15 correct (35%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d710\n self-advocacy statements: 2/8 correct (26%); latency 4.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n doing laundry: 5/11 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n\nSession 3 \u2014 2026-03-15 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 10/12 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n self-advocacy statements: 11/15 correct (74%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d76\n doing laundry: 13/15 correct (86%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d79\n\nSession 4 \u2014 2026-03-17 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 5/13 correct (36%); latency 4.1s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n self-advocacy statements: 3/9 correct (29%); latency 4.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n doing laundry: 5/12 correct (45%); latency 4.1s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n\nSession 5 \u2014 2026-03-19 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 11/12 correct (88%); latency 1.7s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d78\n self-advocacy statements: 11/14 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n doing laundry: 13/14 correct (90%); latency 1.8s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n\nSession 6 \u2014 2026-03-21 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 4/13 correct (34%); latency 4.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n self-advocacy statements: 2/8 correct (25%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n doing laundry: 5/12 correct (40%); latency 4.4s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n\nSession 7 \u2014 2026-03-23 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 7/8 correct (84%); latency 2.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n self-advocacy statements: 8/10 correct (76%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n doing laundry: 11/13 correct (86%); latency 2.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n\nSession 8 \u2014 2026-03-25 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 4/10 correct (36%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n self-advocacy statements: 3/9 correct (32%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n doing laundry: 4/11 correct (39%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n\nSession 9 \u2014 2026-03-27 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 10/11 correct (87%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n self-advocacy statements: 12/15 correct (79%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d712\n doing laundry: 14/15 correct (90%); latency 2.1s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d78\n\nSession 10 \u2014 2026-03-29 \u2014 53 min \u2014 1 observer\n requesting assistance using AAC device: 3/8 correct (36%); latency 4.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n self-advocacy statements: 3/9 correct (33%); latency 4.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n doing laundry: 7/15 correct (45%); latency 3.8s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 60.2%.\n- Trend direction: downward (first session: 86%, last session: 36%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "3c5e7680c8cc8f3d", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "adult", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "87443"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: answers why- and how-questions\nCurriculum Reference: Intraverbal L3\nLearner Profile: School-Age Learner\nCurrent Mastery: Generalization phase\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers why- and how-questions\n- Curriculum: Intraverbal L3\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "ddc3f331563724a9", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers why- and how-questions", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "25180"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: spontaneously names items in the environment\nCurriculum Reference: Spontaneous Vocal Behavior L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously names items in the environment\n- Curriculum: Spontaneous Vocal Behavior L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "bdf6eee4c3cbaaf7", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously names items in the environment", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "time_delay", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "33330"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: answers WH-questions about familiar topics\nCurriculum: Intraverbal L2\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers WH-questions about familiar topics\n- Curriculum: Intraverbal L2\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "536404e01abab2b3", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers WH-questions about familiar topics", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "time_delay", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "54871"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner: Adult Learner\nMastery Status: Emerging\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the emerging stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ef77ab06dc3d2645", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adult", "mastery_state": "emerging", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "44637"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9630\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Social Behavior\nDate range: Sessions 1\u20139 across 18 days (2026-10-22 to 2026-11-09)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. reads 10\u201320 sight words (Reading)\n Method: DTT\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. takes turns during structured activities (Social Behavior)\n Method: PRT\n\nDECELERATION TARGETS\n1. Pica \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-22 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 5/9 correct (53%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 5/12 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (44%); latency 4.2s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d74\n takes turns during structured activities: 5/11 correct (49%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n ABC (pica): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 2 \u2014 2026-10-24 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 9/15 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d75, I\u00d75\n reads 10\u201320 sight words: 7/13 correct (54%); latency 3.4s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (54%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d76\n takes turns during structured activities: 6/11 correct (56%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-10-26 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 7/10 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n reads 10\u201320 sight words: 6/10 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/12 correct (66%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d74\n takes turns during structured activities: 10/15 correct (65%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d78\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-10-28 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 10/13 correct (80%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n reads 10\u201320 sight words: 6/8 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (75%); latency 2.9s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n takes turns during structured activities: 9/12 correct (71%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n Pica: attempts 1 (0 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-10-30 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 12/14 correct (84%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d711\n reads 10\u201320 sight words: 9/12 correct (77%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/10 correct (75%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n takes turns during structured activities: 9/12 correct (79%); latency 2.7s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d75\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n ABC (pica): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 6 \u2014 2026-11-01 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 8/10 correct (84%); latency 2.0s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 10/12 correct (80%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (75%); latency 2.5s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d73\n takes turns during structured activities: 9/12 correct (79%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n ABC (pica): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 7 \u2014 2026-11-03 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 8/9 correct (84%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 7/9 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/13 correct (75%); latency 2.6s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d74\n takes turns during structured activities: 9/11 correct (79%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-11-05 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 13/15 correct (84%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n reads 10\u201320 sight words: 10/13 correct (80%); latency 2.5s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/13 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d76\n takes turns during structured activities: 10/13 correct (79%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n Pica: attempts 0\n\nSession 9 \u2014 2026-11-07 \u2014 42 min \u2014 1 observer\n identifies numerals 1\u201320: 7/8 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 8/10 correct (80%); latency 2.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/11 correct (75%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n takes turns during structured activities: 8/10 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 53% at return, recovering to 84%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nPica: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 76.4%.\n- Trend direction: upward (first session: 53%, last session: 84%).\n- Pica: mean 1.1/session (early-window mean 1.3 -> late-window mean 0.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "de00c6cfcea68056", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Pica": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "school_age", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["pica"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "881"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands using adjectives (big, little, more)\nCurriculum: Mand L2\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands using adjectives (big, little, more)\n- Curriculum: Mand L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: leisure time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "849d75f431a8a82f", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands using adjectives (big, little, more)", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "75835"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6233\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Listener Responding + Mand + Motor Imitation + Tact + Tolerating Denied Access\nDate range: Sessions 1\u20137 across 14 days (2026-08-28 to 2026-09-11)\n\nACCELERATION PROGRAMS\n1. echoes 2-word phrases (Echoic)\n Method: DTT variant\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n4. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n5. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n6. tacts colors of objects (Tact)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-28 \u2014 54 min \u2014 1 observer\n echoes 2-word phrases: 2/10 correct (24%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 2/9 correct (25%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 4/11 correct (34%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n imitates 3-step motor sequences: 4/14 correct (28%); latency 4.8s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d710\n follows 1-step motor instructions: 3/15 correct (20%); latency 5.1s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n tacts colors of objects: 3/11 correct (31%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n\nSession 2 \u2014 2026-08-30 \u2014 54 min \u2014 1 observer\n echoes 2-word phrases: 3/9 correct (35%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 3/8 correct (38%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 4/9 correct (43%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d76\n imitates 3-step motor sequences: 3/8 correct (40%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n follows 1-step motor instructions: 5/14 correct (33%); latency 4.6s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d75\n tacts colors of objects: 5/15 correct (36%); latency 4.5s; prompts FP\u00d75, PP\u00d74, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-09-01 \u2014 IOA SESSION \u2014 2 observers\n echoes 2-word phrases: 42% (primary); IOA 89% agreement\n wait calmly when access to preferred item is denied: 45% (primary); IOA 89% agreement\n requesting a break ('break please' or AAC 'I need a break'): 53% (primary); IOA 89% agreement\n imitates 3-step motor sequences: 46% (primary); IOA 89% agreement\n follows 1-step motor instructions: 41% (primary); IOA 89% agreement\n tacts colors of objects: 45% (primary); IOA 89% agreement\n\nSession 4 \u2014 2026-09-03 \u2014 54 min \u2014 1 observer\n echoes 2-word phrases: 6/12 correct (46%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n wait calmly when access to preferred item is denied: 7/13 correct (57%); latency 3.4s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 7/11 correct (62%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n imitates 3-step motor sequences: 6/10 correct (61%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n follows 1-step motor instructions: 5/10 correct (54%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n tacts colors of objects: 4/8 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 5 \u2014 2026-09-05 \u2014 54 min \u2014 1 observer\n echoes 2-word phrases: 8/15 correct (56%); latency 3.7s; prompts FP\u00d75, PP\u00d74, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 8/12 correct (70%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 6/8 correct (71%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n imitates 3-step motor sequences: 7/11 correct (66%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n follows 1-step motor instructions: 6/9 correct (67%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n tacts colors of objects: 7/11 correct (60%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n\nSession 6 \u2014 2026-09-07 \u2014 54 min \u2014 1 observer\n echoes 2-word phrases: 6/10 correct (64%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 10/13 correct (77%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d75\n requesting a break ('break please' or AAC 'I need a break'): 9/11 correct (82%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n imitates 3-step motor sequences: 9/11 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n follows 1-step motor instructions: 10/13 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n tacts colors of objects: 10/15 correct (68%); latency 3.2s; prompts FP\u00d75, PP\u00d71, G\u00d71, I\u00d78\n\nSession 7 \u2014 2026-09-09 \u2014 54 min \u2014 1 observer\n echoes 2-word phrases: 8/10 correct (76%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 12/14 correct (89%); latency 1.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 11/12 correct (92%); latency 1.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d79\n imitates 3-step motor sequences: 8/9 correct (88%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n follows 1-step motor instructions: 10/11 correct (90%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n tacts colors of objects: 6/8 correct (78%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Appropriate sitting posture maintained"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 24% to 76% across 7 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 49.1%.\n- Trend direction: upward (first session: 24%, last session: 76%).\n- IOA data: 89% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "a84d758c45f04388", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "early", "n_sessions": 7, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "60899"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6858\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand + Self-Care\nDate range: Sessions 1\u20139 across 18 days (2026-10-08 to 2026-10-26)\n\nACCELERATION PROGRAMS\n1. ordering in a restaurant (Community)\n Method: Task Analysis\n2. indicating 'all done' with a step or activity (Mand)\n Method: NET\n3. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. requesting a break during non-preferred activities (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Pica \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-08 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 11/13 correct (81%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d78\n indicating 'all done' with a step or activity: 9/10 correct (90%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n meal cleanup after eating: 10/12 correct (81%); latency 2.1s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d77\n brushing teeth independently: 11/13 correct (84%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n requesting a break during non-preferred activities: 7/9 correct (77%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n Pica: attempts 1 (0 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 2 \u2014 2026-10-10 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 10/12 correct (82%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d710\n indicating 'all done' with a step or activity: 10/11 correct (93%); latency 2.0s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n meal cleanup after eating: 12/15 correct (79%); latency 2.5s; prompts FP\u00d74, PP\u00d75, G\u00d75, I\u00d71\n brushing teeth independently: 7/8 correct (87%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n requesting a break during non-preferred activities: 7/9 correct (76%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-10-12 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 11/13 correct (81%); latency 2.2s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d77\n indicating 'all done' with a step or activity: 10/12 correct (87%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n meal cleanup after eating: 10/12 correct (85%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d74\n brushing teeth independently: 9/11 correct (84%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n requesting a break during non-preferred activities: 10/13 correct (75%); latency 2.6s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d76\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-10-14 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 12/15 correct (80%); latency 2.2s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d74\n indicating 'all done' with a step or activity: 10/11 correct (90%); latency 2.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n meal cleanup after eating: 9/10 correct (88%); latency 2.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n brushing teeth independently: 8/9 correct (88%); latency 2.3s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n requesting a break during non-preferred activities: 7/9 correct (80%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-10-16 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 9/12 correct (72%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n indicating 'all done' with a step or activity: 12/13 correct (89%); latency 1.8s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n meal cleanup after eating: 8/9 correct (86%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n brushing teeth independently: 11/12 correct (91%); latency 2.0s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n requesting a break during non-preferred activities: 11/14 correct (77%); latency 2.5s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-10-18 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 7/10 correct (73%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 9/11 correct (85%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n meal cleanup after eating: 13/15 correct (89%); latency 1.9s; prompts FP\u00d75, PP\u00d74, G\u00d73, I\u00d73\n brushing teeth independently: 13/14 correct (90%); latency 1.7s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n requesting a break during non-preferred activities: 7/8 correct (85%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Pica: attempts 1 (0 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-10-20 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 10/14 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n indicating 'all done' with a step or activity: 8/9 correct (86%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n meal cleanup after eating: 9/10 correct (91%); latency 2.0s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n brushing teeth independently: 12/13 correct (92%); latency 1.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n requesting a break during non-preferred activities: 12/15 correct (83%); latency 2.3s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-10-22 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 11/15 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n indicating 'all done' with a step or activity: 7/9 correct (77%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n meal cleanup after eating: 13/15 correct (89%); latency 2.1s; prompts FP\u00d75, PP\u00d73, G\u00d72, I\u00d75\n brushing teeth independently: 12/13 correct (93%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n requesting a break during non-preferred activities: 12/14 correct (84%); latency 2.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 9 \u2014 2026-10-24 \u2014 54 min \u2014 1 observer\n ordering in a restaurant: 8/12 correct (71%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d74\n indicating 'all done' with a step or activity: 11/14 correct (77%); latency 2.4s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n meal cleanup after eating: 9/10 correct (94%); latency 1.8s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n brushing teeth independently: 13/14 correct (95%); latency 1.7s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n requesting a break during non-preferred activities: 12/14 correct (84%); latency 2.2s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (81%\u201371%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nPica: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 76.3%.\n- Trend direction: downward (first session: 81%, last session: 71%).\n- Pica: mean 1.4/session (early-window mean 1.3 -> late-window mean 1.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "d87fb2360c196582", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Pica": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adolescent", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["pica"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "17449"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4995\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading\nDate range: Sessions 1\u201310 across 20 days (2026-06-19 to 2026-07-09)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. counts objects 1\u201320 (Math)\n Method: DTT\n3. single-word requests for preferred items (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-19 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 4/12 correct (36%); latency 4.3s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n counts objects 1\u201320: 7/13 correct (52%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n single-word requests for preferred items: 5/11 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n\nSession 2 \u2014 2026-06-21 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 5/11 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n counts objects 1\u201320: 6/11 correct (59%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n single-word requests for preferred items: 7/14 correct (47%); latency 3.9s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d77\n\nSession 3 \u2014 2026-06-23 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 4/8 correct (53%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n counts objects 1\u201320: 7/11 correct (68%); latency 2.7s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n single-word requests for preferred items: 6/10 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n\nSession 4 \u2014 2026-06-25 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 7/13 correct (55%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n counts objects 1\u201320: 6/8 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n single-word requests for preferred items: 7/11 correct (63%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n\nSession 5 \u2014 2026-06-27 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 8/12 correct (69%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d75\n counts objects 1\u201320: 10/14 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d78\n single-word requests for preferred items: 7/10 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n\nSession 6 \u2014 2026-06-29 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 10/13 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n counts objects 1\u201320: 9/11 correct (82%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n single-word requests for preferred items: 6/9 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n\nSession 7 \u2014 2026-07-01 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 7/9 correct (76%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n counts objects 1\u201320: 12/13 correct (95%); latency 1.6s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n single-word requests for preferred items: 9/13 correct (72%); latency 2.7s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n\nSession 8 \u2014 2026-07-03 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 11/13 correct (86%); latency 2.0s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n counts objects 1\u201320: 10/10 correct (95%); latency 1.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n single-word requests for preferred items: 11/14 correct (81%); latency 2.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d710\n\nSession 9 \u2014 2026-07-05 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 9/9 correct (95%); latency 2.0s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n counts objects 1\u201320: 13/14 correct (95%); latency 1.8s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d79\n single-word requests for preferred items: 6/8 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n\nSession 10 \u2014 2026-07-07 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 13/14 correct (95%); latency 1.6s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d77\n counts objects 1\u201320: 8/8 correct (95%); latency 1.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n single-word requests for preferred items: 10/12 correct (84%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Quick response latency (within 2\u20133 seconds)\n- Appropriate sitting posture maintained\n- Eye contact maintained throughout trials"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 36% to 95% across 10 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 68.6%.\n- Trend direction: upward (first session: 36%, last session: 95%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "070356e88f879074", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "31112"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects correct item from an array of 2\nCurriculum: Listener Responding L1\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from an array of 2\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: plate, bottle, shoe\nDistractor stimuli: toothbrush, backpack\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a3f28dac970a8a7a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from an array of 2", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "58800"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5099\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-06-02 to 2026-06-20)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n4. mands for missing items needed to complete an activity (Mand)\n Method: NET\n5. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-02 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (64%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n single-word requests for preferred items: 6/10 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n indicating 'all done' when an activity is complete: 5/8 correct (58%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n mands for missing items needed to complete an activity: 8/14 correct (57%); latency 3.5s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n wait calmly when access to preferred item is denied: 8/13 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d76\n\nSession 2 \u2014 2026-06-04 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/14 correct (59%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d78\n single-word requests for preferred items: 9/14 correct (61%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n indicating 'all done' when an activity is complete: 5/8 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 9/15 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d75, I\u00d76\n wait calmly when access to preferred item is denied: 5/11 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n\nSession 3 \u2014 2026-06-06 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/14 correct (65%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d77\n single-word requests for preferred items: 7/14 correct (50%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n indicating 'all done' when an activity is complete: 6/10 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 5/12 correct (45%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 7/11 correct (60%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n\nSession 4 \u2014 2026-06-08 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (69%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n single-word requests for preferred items: 5/9 correct (57%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 7/12 correct (58%); latency 3.2s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n mands for missing items needed to complete an activity: 8/14 correct (56%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n wait calmly when access to preferred item is denied: 6/9 correct (62%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n\nSession 5 \u2014 2026-06-10 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/15 correct (69%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d710\n single-word requests for preferred items: 5/9 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n indicating 'all done' when an activity is complete: 7/15 correct (48%); latency 3.8s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d75\n mands for missing items needed to complete an activity: 6/13 correct (46%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 9/14 correct (61%); latency 3.2s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d77\n\nSession 6 \u2014 2026-06-12 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/15 correct (64%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d75, I\u00d78\n single-word requests for preferred items: 6/10 correct (62%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 7/14 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d710\n mands for missing items needed to complete an activity: 7/15 correct (45%); latency 4.2s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d74\n wait calmly when access to preferred item is denied: 8/13 correct (59%); latency 3.3s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d75\n\nSession 7 \u2014 2026-06-14 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n single-word requests for preferred items: 4/8 correct (52%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n indicating 'all done' when an activity is complete: 8/15 correct (53%); latency 3.8s; prompts FP\u00d75, PP\u00d71, G\u00d72, I\u00d77\n mands for missing items needed to complete an activity: 6/13 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d73\n wait calmly when access to preferred item is denied: 11/15 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n\nSession 8 \u2014 2026-06-16 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n single-word requests for preferred items: 6/10 correct (61%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n indicating 'all done' when an activity is complete: 6/11 correct (52%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d78\n mands for missing items needed to complete an activity: 6/15 correct (40%); latency 4.0s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 10/13 correct (76%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d75\n\nSession 9 \u2014 2026-06-18 \u2014 52 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (71%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n single-word requests for preferred items: 6/9 correct (68%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d73\n indicating 'all done' when an activity is complete: 6/14 correct (43%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n mands for missing items needed to complete an activity: 4/9 correct (46%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n wait calmly when access to preferred item is denied: 10/15 correct (66%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d711\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Leaving seat without permission\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 67% for 9 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 64.4%.\n- Trend direction: upward (first session: 64%, last session: 71%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "348b08b948c49ef8", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "school_age", "n_sessions": 9, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "90839"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: spontaneously uses social phrases appropriately\nCurriculum: Spontaneous Vocal Behavior L2\nLearner Level: Early Learner\nBaseline: Emerging\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously uses social phrases appropriately\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: activity setup.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's activity setup routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "b1e97188fa788184", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously uses social phrases appropriately", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "16366"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4715\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Home Skills + Mand\nDate range: Sessions 1\u20138 across 16 days (2026-01-25 to 2026-02-10)\n\nACCELERATION PROGRAMS\n1. making the bed (Home Skills)\n Method: Task Analysis\n2. requesting preferred items using AAC device (Mand)\n Method: NET\n3. ordering in a restaurant (Community)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-25 \u2014 48 min \u2014 1 observer\n making the bed: 6/8 correct (75%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n requesting preferred items using AAC device: 11/15 correct (75%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n ordering in a restaurant: 7/9 correct (75%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n Motor stereotypy: freq 3; PIR 15%\n\nSession 2 \u2014 2026-01-27 \u2014 48 min \u2014 1 observer\n making the bed: 6/8 correct (79%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n requesting preferred items using AAC device: 6/9 correct (70%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n ordering in a restaurant: 7/10 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n Motor stereotypy: freq 2; PIR 14%\n\nSession 3 \u2014 2026-01-29 \u2014 48 min \u2014 1 observer\n making the bed: 5/12 correct (39%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n requesting preferred items using AAC device: 4/12 correct (30%); latency 4.7s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d75\n ordering in a restaurant: 4/9 correct (46%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n Motor stereotypy: freq 3; PIR 15%\n\nSession 4 \u2014 2026-01-31 \u2014 48 min \u2014 1 observer\n making the bed: 6/15 correct (41%); latency 3.9s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d73\n requesting preferred items using AAC device: 3/10 correct (33%); latency 4.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d76\n ordering in a restaurant: 4/9 correct (42%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n Motor stereotypy: freq 2; PIR 14%\n\nSession 5 \u2014 2026-02-02 \u2014 48 min \u2014 1 observer\n making the bed: 5/13 correct (38%); latency 4.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n requesting preferred items using AAC device: 4/10 correct (37%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n ordering in a restaurant: 4/8 correct (45%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n Motor stereotypy: freq 2; PIR 14%\n\nSession 6 \u2014 2026-02-04 \u2014 48 min \u2014 1 observer\n making the bed: 4/10 correct (42%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n requesting preferred items using AAC device: 4/9 correct (40%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d76\n ordering in a restaurant: 4/9 correct (49%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n Motor stereotypy: freq 2; PIR 16%\n\nSession 7 \u2014 2026-02-06 \u2014 48 min \u2014 1 observer\n making the bed: 4/8 correct (51%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n requesting preferred items using AAC device: 4/10 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n ordering in a restaurant: 5/9 correct (51%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n Motor stereotypy: freq 3; PIR 16%\n\nSession 8 \u2014 2026-02-08 \u2014 48 min \u2014 1 observer\n making the bed: 7/12 correct (55%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n requesting preferred items using AAC device: 6/12 correct (48%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n ordering in a restaurant: 6/10 correct (59%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n Motor stereotypy: freq 2; PIR 16%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 52.7%.\n- Trend direction: downward (first session: 75%, last session: 55%).\n- Motor stereotypy: mean 2.4/session (early-window mean 2.5 -> late-window mean 2.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "3add9ea4abad3aa5", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Motor stereotypy": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["motor_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "50359"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7461\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Tact\nDate range: Sessions 1\u201311 across 22 days (2026-04-25 to 2026-05-17)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. sounds out CVC words (Reading)\n Method: DTT\n3. tacts common objects (Tact)\n Method: DTT\n4. selects items by feature (Listener Responding)\n Method: DTT\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-25 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/12 correct (45%); latency 4.0s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d76\n sounds out CVC words: 3/9 correct (37%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n tacts common objects: 4/10 correct (36%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n selects items by feature: 7/13 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/11 correct (38%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = during group activity without requesting bathroom; B = voided (urine) in seat; C = staff removed learner from group, changed clothing, re-entered after clean-up\n\nSession 2 \u2014 2026-04-27 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/13 correct (47%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n sounds out CVC words: 6/14 correct (40%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n tacts common objects: 4/12 correct (36%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n selects items by feature: 5/11 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/14 correct (41%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d710\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 3 \u2014 2026-04-29 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (55%); latency 3.4s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d74\n sounds out CVC words: 4/8 correct (48%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n tacts common objects: 7/13 correct (51%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n selects items by feature: 8/14 correct (55%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 4 \u2014 2026-05-01 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/14 correct (51%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d79\n sounds out CVC words: 5/10 correct (47%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n tacts common objects: 9/15 correct (58%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d77\n selects items by feature: 6/9 correct (66%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (60%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n Toileting accident (urine or bowel): urine: 1 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = end of 90-minute block without scheduled toileting; B = soiled (bowel) at the work table; C = staff initiated clean-up and hygiene routine; program paused\n\nSession 5 \u2014 2026-05-03 \u2014 IOA SESSION \u2014 2 observers\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 54% (primary); IOA 92% agreement\n sounds out CVC words: 55% (primary); IOA 92% agreement\n tacts common objects: 66% (primary); IOA 92% agreement\n selects items by feature: 68% (primary); IOA 92% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 66% (primary); IOA 92% agreement\n Toileting accident (urine or bowel): urine: 5 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents; IOA 92% agreement\n\nSession 6 \u2014 2026-05-05 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/13 correct (60%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d76\n sounds out CVC words: 6/10 correct (64%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n tacts common objects: 10/15 correct (67%); latency 3.1s; prompts FP\u00d75, PP\u00d74, G\u00d75, I\u00d71\n selects items by feature: 7/10 correct (66%); latency 3.1s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/14 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 7 \u2014 2026-05-07 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/11 correct (64%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n sounds out CVC words: 7/10 correct (72%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n tacts common objects: 8/10 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n selects items by feature: 12/15 correct (78%); latency 2.8s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (81%); latency 2.1s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n Toileting accident (urine or bowel): urine: 3 in-toilet / 3 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 8 \u2014 2026-05-09 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/13 correct (68%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n sounds out CVC words: 10/13 correct (80%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d79\n tacts common objects: 7/9 correct (82%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n selects items by feature: 9/11 correct (83%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/10 correct (81%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 9 \u2014 2026-05-11 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (81%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n sounds out CVC words: 10/13 correct (79%); latency 2.6s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n tacts common objects: 13/14 correct (92%); latency 1.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d711\n selects items by feature: 8/10 correct (81%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/12 correct (94%); latency 1.9s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d74\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 1 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = during demand sequence at the work table; B = voided (urine) without signaling need; C = staff paused programming, changed clothing, resumed after 10 min\n\nSession 10 \u2014 2026-05-13 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 12/15 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d711\n sounds out CVC words: 9/10 correct (86%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n tacts common objects: 13/14 correct (95%); latency 1.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d710\n selects items by feature: 9/11 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (93%); latency 2.1s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 1 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 11 \u2014 2026-05-15 \u2014 57 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (80%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n sounds out CVC words: 8/8 correct (95%); latency 1.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n tacts common objects: 11/12 correct (95%); latency 1.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d75\n selects items by feature: 10/11 correct (90%); latency 2.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 14/15 correct (95%); latency 2.0s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d711\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 1 in-toilet / 1 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Requesting more trials verbally\n- Quick response latency (within 2\u20133 seconds)\n- Appropriate sitting posture maintained\n- Eye contact maintained throughout trials"}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 45% to 80% across 11 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 62.2%.\n- Trend direction: upward (first session: 45%, last session: 80%).\n- Toileting accident (urine or bowel): mean 2.7/session (early-window mean 2.3 -> late-window mean 2.7).\n- IOA data: 92% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "83ff9f9b3a9ac06a", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {"Toileting accident (urine or bowel)": "automatic"}, "escalation_level": 1, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "early", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "52257"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: meal cleanup\nCurriculum: AFLS Home Skills\nLearner: Adult Learner\nMastery Status: Mastered at current level\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: meal cleanup\n- Curriculum: AFLS Home Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Stack used dishes on one side of the table. <- Learner-taught step (forward chaining starts here)\n2. Carry dishes to the sink or dishwasher.\n3. Rinse food residue into the trash or disposal.\n4. Load dishes into the dishwasher (or wash by hand).\n5. Wipe the table with a damp cloth.\n6. Put away any leftover food into containers.\n7. Place containers in the refrigerator.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "7db921d677e6f699", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "meal cleanup", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "meal cleanup", "module": "home", "mastery_criterion": "mc_all_steps_independent", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "92010"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8783\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-10-11 to 2026-10-31)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n3. selects items by feature (Listener Responding)\n Method: DTT\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n2. Property destruction \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-11 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/11 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d75\n follows 1-step motor instructions: 6/12 correct (50%); latency 3.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n selects items by feature: 7/15 correct (48%); latency 3.8s; prompts FP\u00d75, PP\u00d74, G\u00d71, I\u00d75\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n Property destruction: freq 2\n\nSession 2 \u2014 2026-10-13 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/12 correct (70%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d76\n follows 1-step motor instructions: 8/14 correct (58%); latency 3.3s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d77\n selects items by feature: 7/13 correct (57%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 2\n\nSession 3 \u2014 2026-10-15 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/15 correct (72%); latency 3.0s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n follows 1-step motor instructions: 8/14 correct (56%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n selects items by feature: 8/13 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d75\n Toileting accident (urine or bowel): urine: 5 in-toilet / 0 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 3\n\nSession 4 \u2014 2026-10-17 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/14 correct (66%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d74\n follows 1-step motor instructions: 6/10 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n selects items by feature: 6/9 correct (64%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 8\n\nSession 5 \u2014 2026-10-19 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/9 correct (72%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n follows 1-step motor instructions: 6/11 correct (59%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n selects items by feature: 6/9 correct (63%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n Toileting accident (urine or bowel): urine: 0 in-toilet / 5 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 8\n\nSession 6 \u2014 2026-10-21 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/10 correct (70%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n follows 1-step motor instructions: 8/13 correct (61%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d77\n selects items by feature: 9/14 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n Toileting accident (urine or bowel): urine: 4 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 3\n\nSession 7 \u2014 2026-10-23 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/13 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n follows 1-step motor instructions: 9/13 correct (66%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n selects items by feature: 9/12 correct (72%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n Toileting accident (urine or bowel): urine: 2 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 2\n\nSession 8 \u2014 2026-10-25 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 11/15 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d712\n follows 1-step motor instructions: 9/14 correct (66%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n selects items by feature: 11/15 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d712\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 2\n\nSession 9 \u2014 2026-10-27 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 7/9 correct (76%); latency 2.3s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n follows 1-step motor instructions: 9/13 correct (69%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d75\n selects items by feature: 10/13 correct (77%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n Toileting accident (urine or bowel): urine: 5 in-toilet / 2 accidents; BM: 0 in-toilet / 0 accidents\n Property destruction: freq 2\n\nSession 10 \u2014 2026-10-29 \u2014 52 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/9 correct (71%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n follows 1-step motor instructions: 10/15 correct (68%); latency 2.9s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n selects items by feature: 7/9 correct (75%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n Property destruction: freq 3\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Peer aggression (hitting or pushing peers during group activities)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nProperty destruction: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 70.1%.\n- Trend direction: upward (first session: 65%, last session: 71%).\n- Toileting accident (urine or bowel): mean 2.1/session (early-window mean 1.3 -> late-window mean 1.7).\n- Property destruction: mean 3.5/session (early-window mean 2.3 -> late-window mean 2.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "76afd6bd032dccd2", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Toileting accident (urine or bowel)": "escape", "Property destruction": "attention"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "early", "n_sessions": 10, "n_programs": 3, "n_behaviors": 2, "behavior_ids": ["toileting_accident", "property_destruction"], "behavior_functions": ["escape", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "84135"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3400\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading\nDate range: Sessions 1\u201311 across 22 days (2026-09-27 to 2026-10-19)\n\nACCELERATION PROGRAMS\n1. mands for missing items needed to complete an activity (Mand)\n Method: NET\n2. reads 10\u201320 sight words (Reading)\n Method: DTT\n3. sounds out CVC words (Reading)\n Method: DTT\n4. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n5. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n2. Toileting accident (urine or bowel) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-27 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 4/8 correct (48%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n reads 10\u201320 sight words: 11/14 correct (82%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n sounds out CVC words: 10/13 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n follows 1-step motor instructions: 9/14 correct (64%); latency 3.1s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/11 correct (47%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n Non-compliance: freq 5\n Toileting accident (urine or bowel): urine: 0 in-toilet / 4 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 2 \u2014 2026-09-29 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/14 correct (38%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n reads 10\u201320 sight words: 7/9 correct (81%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n sounds out CVC words: 6/10 correct (58%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n follows 1-step motor instructions: 5/8 correct (61%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/14 correct (46%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n Non-compliance: freq 7\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 3 \u2014 2026-10-01 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/9 correct (66%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 7/11 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n sounds out CVC words: 10/12 correct (81%); latency 2.1s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n follows 1-step motor instructions: 6/14 correct (45%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/9 correct (46%); latency 4.1s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n Non-compliance: freq 5\n Toileting accident (urine or bowel): urine: 0 in-toilet / 5 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 4 \u2014 2026-10-03 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 8/11 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 6/11 correct (51%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n sounds out CVC words: 9/12 correct (74%); latency 2.7s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n follows 1-step motor instructions: 5/8 correct (65%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/10 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n Non-compliance: freq 3\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 5 \u2014 2026-10-05 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 9/14 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 11/15 correct (70%); latency 2.7s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n sounds out CVC words: 7/10 correct (68%); latency 2.9s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n follows 1-step motor instructions: 4/9 correct (40%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/14 correct (41%); latency 4.1s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n Non-compliance: freq 2\n Toileting accident (urine or bowel): urine: 3 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 6 \u2014 2026-10-07 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/11 correct (53%); latency 3.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n reads 10\u201320 sight words: 6/10 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n sounds out CVC words: 8/12 correct (63%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n follows 1-step motor instructions: 8/14 correct (60%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (70%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n Non-compliance: freq 1\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 7 \u2014 2026-10-09 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/13 correct (49%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 5/9 correct (61%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d74\n sounds out CVC words: 10/14 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n follows 1-step motor instructions: 3/8 correct (35%); latency 4.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (57%); latency 3.6s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n Non-compliance: freq 3\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 8 \u2014 2026-10-11 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 10/15 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d79\n reads 10\u201320 sight words: 11/12 correct (88%); latency 2.1s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n sounds out CVC words: 7/13 correct (54%); latency 3.3s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n follows 1-step motor instructions: 7/15 correct (46%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d711\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/10 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d72\n Non-compliance: freq 5\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 9 \u2014 2026-10-13 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/8 correct (62%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 7/14 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n sounds out CVC words: 10/13 correct (80%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d78\n follows 1-step motor instructions: 9/13 correct (71%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d710\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Non-compliance: freq 6\n Toileting accident (urine or bowel): urine: 3 in-toilet / 3 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 10 \u2014 2026-10-15 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/10 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n reads 10\u201320 sight words: 7/10 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n sounds out CVC words: 8/14 correct (56%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n follows 1-step motor instructions: 7/14 correct (51%); latency 3.5s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n Non-compliance: freq 5\n Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 11 \u2014 2026-10-17 \u2014 54 min \u2014 1 observer\n mands for missing items needed to complete an activity: 9/14 correct (67%); latency 3.2s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n reads 10\u201320 sight words: 8/10 correct (82%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n sounds out CVC words: 7/10 correct (71%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n follows 1-step motor instructions: 3/8 correct (37%); latency 4.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (54%); latency 3.8s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n Non-compliance: freq 3\n Toileting accident (urine or bowel): urine: 2 in-toilet / 6 accidents; BM: 0 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 58% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nToileting accident (urine or bowel): automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 58.6%.\n- Trend direction: upward (first session: 48%, last session: 67%).\n- Non-compliance: mean 4.1/session (early-window mean 5.7 -> late-window mean 4.7).\n- Toileting accident (urine or bowel): mean 3.5/session (early-window mean 4.3 -> late-window mean 4.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "cdcc61a8fc2ac2a2", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Non-compliance": "escape", "Toileting accident (urine or bowel)": "automatic"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "early", "n_sessions": 11, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["non_compliance", "toileting_accident"], "behavior_functions": ["escape", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "32021"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2656\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Listener Responding\nDate range: Sessions 1\u20136 across 12 days (2026-06-08 to 2026-06-20)\n\nACCELERATION PROGRAMS\n1. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n2. selects items by feature (Listener Responding)\n Method: DTT\n3. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-08 \u2014 41 min \u2014 1 observer\n follows 1-step motor instructions: 8/13 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n selects items by feature: 6/10 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n answers WH-questions about familiar topics: 7/11 correct (68%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 2\n ABC (self-injurious behavior (sib)): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 2 \u2014 2026-06-10 \u2014 41 min \u2014 1 observer\n follows 1-step motor instructions: 7/12 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n selects items by feature: 10/15 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d75, G\u00d70, I\u00d710\n answers WH-questions about familiar topics: 8/12 correct (63%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d79\n Self-injurious behavior (SIB): freq 2\n ABC (self-injurious behavior (sib)): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 3 \u2014 2026-06-12 \u2014 41 min \u2014 1 observer\n follows 1-step motor instructions: 6/11 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n selects items by feature: 8/14 correct (61%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d77\n answers WH-questions about familiar topics: 7/11 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n Self-injurious behavior (SIB): freq 5\n\nSession 4 \u2014 2026-06-14 \u2014 41 min \u2014 1 observer\n follows 1-step motor instructions: 9/14 correct (61%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d711\n selects items by feature: 5/8 correct (66%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n answers WH-questions about familiar topics: 10/14 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d75\n Self-injurious behavior (SIB): freq 5\n\nSession 5 \u2014 2026-06-16 \u2014 41 min \u2014 1 observer\n follows 1-step motor instructions: 9/15 correct (60%); latency 3.6s; prompts FP\u00d71, PP\u00d75, G\u00d72, I\u00d77\n selects items by feature: 6/9 correct (64%); latency 2.8s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n answers WH-questions about familiar topics: 10/14 correct (70%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n Self-injurious behavior (SIB): freq 2\n ABC (self-injurious behavior (sib)): A = adult attention diverted to another learner; B = vocalized loudly + approached staff; C = staff turned to address behavior\n\nSession 6 \u2014 2026-06-18 \u2014 41 min \u2014 1 observer\n follows 1-step motor instructions: 5/8 correct (57%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n selects items by feature: 7/12 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d78\n answers WH-questions about familiar topics: 8/12 correct (68%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n Self-injurious behavior (SIB): freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Self-stimulatory behavior increase during demand periods\n- Increased response latency (3\u20135\u00d7 baseline)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n### Crisis plan\n- If problem-behavior intensity or topography exceeds the learner's baseline safety threshold (SIB causing tissue damage, aggression causing injury, imminent elopement into hazard): pause the extinction procedure and implement the facility's tiered crisis-prevention protocol (e.g., Safety-Care, CPI, or equivalent).\n- De-escalation and environmental safety first; physical intervention is last resort and only by staff currently certified in the facility's framework. Least-restrictive option appropriate to the risk level applies at every step. Physical intervention is used only when specifically authorized for this learner in the BIP \u2014 contraindications (medical, developmental, trauma history) may restrict the procedures available, in which case the plan is de-escalation + environmental safety + supervisor contact only.\n- Contact the supervising BCBA immediately; do not resume the extinction procedure without BCBA review. Document duration, intensity, antecedent, staff response, and post-incident state.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 6 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 58.2%.\n- Trend direction: downward (first session: 58%, last session: 57%).\n- Self-injurious behavior (SIB): mean 2.8/session (early-window mean 2.0 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "92b7f0b9ef2bd813", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Self-injurious behavior (SIB)": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "early", "n_sessions": 6, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["attention"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "86579"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: toileting routine\nCurriculum: AFLS Basic Living Skills\nLearner: Adolescent Learner\nMastery Status: Near mastery\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: toileting routine\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Recognize the need and walk to the bathroom. <- Learner-taught step (forward chaining starts here)\n2. Close the door for privacy.\n3. Lower clothing appropriately.\n4. Sit on the toilet.\n5. Complete the elimination.\n6. Use toilet paper correctly.\n7. Stand and pull clothing up.\n8. Flush the toilet.\n9. Wash hands.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "75642e594f453040", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "toileting routine", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "near", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "toileting routine", "module": "basic_living", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "75542"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: matches shapes (identical blocks)\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L1\nLearner: Early Learner\nMastery Status: Near mastery\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches shapes (identical blocks)\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Graduated Guidance\nSequence: hand over hand full -> hand over hand light -> shadow -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: oval, diamond, heart\nDistractor stimuli: circle, triangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6f4ace8f529317d4", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches shapes (identical blocks)", "prompt_hierarchy": "graduated_guidance", "reinforcement_schedule": "fr2", "error_correction": "four_step", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "55121"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4481\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Social Behavior\nDate range: Sessions 1\u20136 across 12 days (2026-05-02 to 2026-05-14)\n\nACCELERATION PROGRAMS\n1. takes turns during structured activities (Social Behavior)\n Method: PRT\n2. mands for missing items needed to complete an activity (Mand)\n Method: NET\n3. requesting a break during non-preferred tasks (Mand)\n Method: NET\n4. counts objects 1\u201320 (Math)\n Method: DTT\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-02 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 4/15 correct (27%); latency 4.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d712\n mands for missing items needed to complete an activity: 3/13 correct (27%); latency 4.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n requesting a break during non-preferred tasks: 4/11 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n counts objects 1\u201320: 4/14 correct (29%); latency 4.9s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/14 correct (21%); latency 5.1s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d74\n\nSession 2 \u2014 2026-05-04 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 4/11 correct (40%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 5/13 correct (38%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d78\n requesting a break during non-preferred tasks: 5/10 correct (49%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n counts objects 1\u201320: 4/11 correct (39%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/13 correct (33%); latency 4.8s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n\nSession 3 \u2014 2026-05-06 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 6/13 correct (50%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d711\n mands for missing items needed to complete an activity: 5/11 correct (44%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n requesting a break during non-preferred tasks: 9/15 correct (57%); latency 3.3s; prompts FP\u00d74, PP\u00d70, G\u00d75, I\u00d76\n counts objects 1\u201320: 6/12 correct (51%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (41%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n\nSession 4 \u2014 2026-05-08 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 5/8 correct (62%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n mands for missing items needed to complete an activity: 5/9 correct (56%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n requesting a break during non-preferred tasks: 6/9 correct (65%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n counts objects 1\u201320: 8/13 correct (60%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/14 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d78\n\nSession 5 \u2014 2026-05-10 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 8/11 correct (69%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n mands for missing items needed to complete an activity: 9/13 correct (66%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n requesting a break during non-preferred tasks: 12/15 correct (77%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d710\n counts objects 1\u201320: 10/13 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (57%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n\nSession 6 \u2014 2026-05-12 \u2014 31 min \u2014 1 observer\n takes turns during structured activities: 12/14 correct (83%); latency 2.3s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d75\n mands for missing items needed to complete an activity: 7/10 correct (72%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n requesting a break during non-preferred tasks: 13/15 correct (89%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n counts objects 1\u201320: 10/12 correct (87%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/8 correct (66%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Smiling during reinforcement delivery\n- Eye contact maintained throughout trials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 27% to 83% across 6 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nlow\n\nInsufficient data: only 6 sessions, high variance, and/or no IOA evidence. Recommend additional data collection before programming changes.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 55.2%.\n- Trend direction: upward (first session: 27%, last session: 83%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "0495d86a79a594cd", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "low", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "school_age", "n_sessions": 6, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "58385"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5606\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand\nDate range: Sessions 1\u20139 across 18 days (2026-08-09 to 2026-08-27)\n\nACCELERATION PROGRAMS\n1. mands for missing items needed to complete an activity (Mand)\n Method: NET\n2. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n3. single-word requests for preferred items (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-09 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 9/14 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n indicating 'all done' when an activity is complete: 5/8 correct (66%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n single-word requests for preferred items: 8/12 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n\nSession 2 \u2014 2026-08-11 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/10 correct (69%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n indicating 'all done' when an activity is complete: 8/12 correct (68%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n single-word requests for preferred items: 5/8 correct (65%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n\nSession 3 \u2014 2026-08-13 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/10 correct (67%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 9/13 correct (71%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n single-word requests for preferred items: 10/15 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d75, G\u00d73, I\u00d76\n\nSession 4 \u2014 2026-08-15 \u2014 IOA SESSION \u2014 2 observers\n mands for missing items needed to complete an activity: 40% (primary); IOA 86% agreement\n indicating 'all done' when an activity is complete: 39% (primary); IOA 86% agreement\n single-word requests for preferred items: 44% (primary); IOA 86% agreement\n\nSession 5 \u2014 2026-08-17 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/12 correct (44%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n indicating 'all done' when an activity is complete: 6/15 correct (42%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d75, I\u00d77\n single-word requests for preferred items: 6/12 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n\nSession 6 \u2014 2026-08-19 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/14 correct (48%); latency 3.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n indicating 'all done' when an activity is complete: 8/15 correct (54%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d75, I\u00d78\n single-word requests for preferred items: 8/15 correct (55%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n\nSession 7 \u2014 2026-08-21 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 8/13 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d77\n indicating 'all done' when an activity is complete: 6/11 correct (56%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n single-word requests for preferred items: 6/10 correct (65%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n\nSession 8 \u2014 2026-08-23 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 10/15 correct (65%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d75, I\u00d76\n indicating 'all done' when an activity is complete: 5/8 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n single-word requests for preferred items: 6/9 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n\nSession 9 \u2014 2026-08-25 \u2014 53 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/11 correct (65%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n indicating 'all done' when an activity is complete: 7/10 correct (66%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n single-word requests for preferred items: 8/12 correct (66%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 57.4%.\n- Trend direction: upward (first session: 61%, last session: 65%).\n- IOA data: 86% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "e2c2c2c7a7251049", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "school_age", "n_sessions": 9, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "64009"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: identifies basic shapes in math context\nCurriculum: Math L1\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies basic shapes in math context\n- Curriculum: Math L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What number is this?\"\nVariations: \"Which number?\"; \"Tell me the number\"\nPresentation: Present the numeral card.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: circle, diamond, heart\nDistractor stimuli: star, triangle\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "2a5ea4c09a8d75e0", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies basic shapes in math context", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "15942"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: doing laundry\nCurriculum Reference: AFLS Home Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Generalization phase\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: doing laundry\n- Curriculum: AFLS Home Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Sort laundry by color (lights / darks / whites).\n2. Load the washing machine with one color group.\n3. Add detergent according to instructions.\n4. Select the correct wash cycle and start the machine.\n5. When complete, move clothes to the dryer.\n6. Start the dryer on the correct cycle.\n7. When dry, fold clothes by category.\n8. Put folded clothes away in drawers or closet.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "b166688ea3565579", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "doing laundry", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "doing laundry", "module": "home", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "68917"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts emotions in self and others\nCurriculum Reference: Tact L2\nLearner Profile: Early Learner\nCurrent Mastery: Generalization phase\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts emotions in self and others\n- Curriculum: Tact L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How does he feel?\"\nVariations: \"What emotion?\"; \"How is she feeling?\"\nPresentation: Present a photograph or emoji card depicting the target emotion.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: crayon, pencil, ball\nDistractor stimuli: hat, shoe, bottle, napkin\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "30600014115d587a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts emotions in self and others", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "backstep", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "47307"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: spontaneously reports past events\nCurriculum: Spontaneous Vocal Behavior L2\nLearner Level: School-Age Learner\nBaseline: Generalization phase\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously reports past events\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: room transition.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's room transition routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "429cde9baede7e5b", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously reports past events", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "time_delay", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "75835"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: answers WH-questions about familiar topics\nCurriculum: Intraverbal L2\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: answers WH-questions about familiar topics\n- Curriculum: Intraverbal L2\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: leisure chat.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure chat routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "e854ede8260f414d", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L2", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "answers WH-questions about familiar topics", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "67055"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2202\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-08-24 to 2026-09-13)\n\nACCELERATION PROGRAMS\n1. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. sounds out CVC words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-24 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 8/15 correct (52%); latency 3.8s; prompts FP\u00d75, PP\u00d75, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 5/10 correct (46%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (45%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n sounds out CVC words: 5/10 correct (48%); latency 3.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n Elopement: freq 0\n\nSession 2 \u2014 2026-08-26 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 8/14 correct (61%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 6/11 correct (52%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/11 correct (50%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n sounds out CVC words: 8/13 correct (59%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n Elopement: freq 0\n\nSession 3 \u2014 2026-08-28 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 7/10 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 8/14 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (59%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 5/8 correct (67%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n Elopement: freq 0\n\nSession 4 \u2014 2026-08-30 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 6/8 correct (81%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 8/13 correct (61%); latency 3.5s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/12 correct (65%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n sounds out CVC words: 10/13 correct (78%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n Elopement: freq 0\n\nSession 5 \u2014 2026-09-01 \u2014 IOA SESSION \u2014 2 observers\n answers WH-questions about familiar topics: 83% (primary); IOA 97% agreement\n wait calmly when access to preferred item is denied: 69% (primary); IOA 97% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 69% (primary); IOA 97% agreement\n sounds out CVC words: 84% (primary); IOA 97% agreement\n Elopement: freq 1; IOA 97% agreement\n\nSession 6 \u2014 2026-09-03 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 7/8 correct (83%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 11/15 correct (72%); latency 2.5s; prompts FP\u00d75, PP\u00d75, G\u00d75, I\u00d70\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/14 correct (77%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n sounds out CVC words: 9/11 correct (84%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n Elopement: freq 0\n\nSession 7 \u2014 2026-09-05 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 12/14 correct (83%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 8/11 correct (72%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/15 correct (77%); latency 2.5s; prompts FP\u00d75, PP\u00d72, G\u00d72, I\u00d76\n sounds out CVC words: 10/12 correct (84%); latency 2.3s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d74\n Elopement: freq 0\n\nSession 8 \u2014 2026-09-07 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 12/14 correct (83%); latency 2.3s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 9/13 correct (72%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/15 correct (77%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d711\n sounds out CVC words: 13/15 correct (84%); latency 2.2s; prompts FP\u00d75, PP\u00d71, G\u00d71, I\u00d78\n Elopement: freq 1\n ABC (elopement): A = asked to finish dressing task; B = kicked at materials; C = task paused for 3 min\n\nSession 9 \u2014 2026-09-09 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 7/9 correct (83%); latency 2.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 6/8 correct (72%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n sounds out CVC words: 9/11 correct (84%); latency 2.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n Elopement: freq 1\n\nSession 10 \u2014 2026-09-11 \u2014 30 min \u2014 1 observer\n answers WH-questions about familiar topics: 8/10 correct (83%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 8/11 correct (72%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (77%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d74\n sounds out CVC words: 8/10 correct (84%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n Elopement: freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials\n- Echolalic responding (repeating SD rather than answering)\n- Delayed responding (>5 seconds consistently)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 52% at return, recovering to 83%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 76.3%.\n- Trend direction: upward (first session: 52%, last session: 83%).\n- Elopement: mean 0.4/session (early-window mean 0.0 -> late-window mean 1.0).\n- IOA data: 97% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "a2b9efa8a8c2f43b", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Elopement": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "early", "n_sessions": 10, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["elopement"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "13643"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: simple meal preparation (sandwich)\nCurriculum: AFLS Home Skills\nLearner Level: Adult Learner\nBaseline: Approaching mastery\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: simple meal preparation (sandwich)\n- Curriculum: AFLS Home Skills\n- Learner: Adult Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Collect all ingredients (bread, filling, spread).\n2. Place two slices of bread on the plate.\n3. Apply spread to one slice.\n4. Place filling on the spread slice.\n5. Place the second slice on top.\n6. Cut the sandwich in half if desired.\n7. Place on a plate to serve. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "dec9bb9893f9ee6a", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "simple meal preparation (sandwich)", "program_type": "independence", "learner_profile": "adult", "mastery_state": "approaching", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "simple meal preparation (sandwich)", "module": "home", "mastery_criterion": "mc_all_steps_independent", "chain_type": "backward", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "50630"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9539\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand + Self-Care + Vocational\nDate range: Sessions 1\u201310 across 20 days (2026-03-15 to 2026-04-04)\n\nACCELERATION PROGRAMS\n1. simple meal preparation (Home Skills)\n Method: Task Analysis\n2. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n3. completing assigned tasks in order (Vocational)\n Method: Task Analysis\n4. requesting a break during extended tasks (Mand)\n Method: NET\n5. washing hands independently (Self-Care)\n Method: Task Analysis\n6. self-advocacy statements (Independent Living)\n Method: BST\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n2. Aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-15 \u2014 50 min \u2014 1 observer\n simple meal preparation: 12/15 correct (83%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n scheduling medical appointments: 10/14 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n completing assigned tasks in order: 7/10 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n requesting a break during extended tasks: 7/8 correct (83%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n washing hands independently: 8/10 correct (80%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n self-advocacy statements: 11/15 correct (74%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d711\n Non-compliance: freq 1\n Aggression: freq 3\n\nSession 2 \u2014 2026-03-17 \u2014 50 min \u2014 1 observer\n simple meal preparation: 8/11 correct (72%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n scheduling medical appointments: 6/9 correct (68%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n completing assigned tasks in order: 10/15 correct (68%); latency 3.0s; prompts FP\u00d72, PP\u00d75, G\u00d70, I\u00d78\n requesting a break during extended tasks: 7/9 correct (75%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n washing hands independently: 11/15 correct (72%); latency 3.0s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n self-advocacy statements: 6/10 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n Non-compliance: freq 3\n Aggression: freq 2\n\nSession 3 \u2014 2026-03-19 \u2014 50 min \u2014 1 observer\n simple meal preparation: 6/9 correct (70%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n scheduling medical appointments: 4/8 correct (53%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n completing assigned tasks in order: 6/11 correct (57%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n requesting a break during extended tasks: 8/10 correct (77%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d74\n washing hands independently: 5/8 correct (64%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n self-advocacy statements: 8/13 correct (58%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d711\n Non-compliance: freq 3\n Aggression: freq 3\n ABC (aggression): A = demand to complete math problem; B = threw materials and vocalized 'no'; C = staff removed demand, redirected\n\nSession 4 \u2014 2026-03-21 \u2014 50 min \u2014 1 observer\n simple meal preparation: 5/8 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n scheduling medical appointments: 7/14 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d79\n completing assigned tasks in order: 5/11 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n requesting a break during extended tasks: 7/11 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d78\n washing hands independently: 7/12 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n self-advocacy statements: 6/12 correct (54%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d79\n Non-compliance: freq 3\n Aggression: freq 2\n\nSession 5 \u2014 2026-03-23 \u2014 50 min \u2014 1 observer\n simple meal preparation: 5/8 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n scheduling medical appointments: 6/12 correct (46%); latency 4.1s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n completing assigned tasks in order: 3/9 correct (37%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n requesting a break during extended tasks: 6/9 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n washing hands independently: 9/15 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n self-advocacy statements: 6/14 correct (40%); latency 4.4s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n Non-compliance: freq 3\n Aggression: freq 3\n\nSession 6 \u2014 2026-03-25 \u2014 50 min \u2014 1 observer\n simple meal preparation: 9/14 correct (63%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d711\n scheduling medical appointments: 3/9 correct (33%); latency 4.8s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n completing assigned tasks in order: 6/15 correct (38%); latency 4.1s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d710\n requesting a break during extended tasks: 8/14 correct (60%); latency 3.6s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d79\n washing hands independently: 7/13 correct (50%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n self-advocacy statements: 5/14 correct (38%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d78\n Non-compliance: freq 3\n Aggression: freq 3\n\nSession 7 \u2014 2026-03-27 \u2014 50 min \u2014 1 observer\n simple meal preparation: 5/9 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n scheduling medical appointments: 3/11 correct (29%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n completing assigned tasks in order: 2/10 correct (20%); latency 5.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n requesting a break during extended tasks: 7/13 correct (55%); latency 3.3s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n washing hands independently: 4/8 correct (46%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n self-advocacy statements: 4/15 correct (25%); latency 5.1s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d77\n Non-compliance: freq 4\n Aggression: freq 4\n\nSession 8 \u2014 2026-03-29 \u2014 50 min \u2014 1 observer\n simple meal preparation: 7/13 correct (56%); latency 3.5s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n scheduling medical appointments: 2/15 correct (15%); latency 5.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d710\n completing assigned tasks in order: 2/11 correct (20%); latency 4.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n requesting a break during extended tasks: 7/13 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n washing hands independently: 4/11 correct (41%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n self-advocacy statements: 2/11 correct (19%); latency 5.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n Non-compliance: freq 3\n Aggression: freq 5\n\nSession 9 \u2014 2026-03-31 \u2014 50 min \u2014 1 observer\n simple meal preparation: 7/14 correct (49%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n scheduling medical appointments: 1/10 correct (10%); latency 5.5s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n completing assigned tasks in order: 2/12 correct (14%); latency 5.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n requesting a break during extended tasks: 5/12 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d77\n washing hands independently: 3/8 correct (35%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n self-advocacy statements: 1/14 correct (11%); latency 5.4s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d79\n Non-compliance: freq 5\n Aggression: freq 4\n\nSession 10 \u2014 2026-04-02 \u2014 50 min \u2014 1 observer\n simple meal preparation: 5/13 correct (41%); latency 4.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d710\n scheduling medical appointments: 2/15 correct (10%); latency 5.3s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d76\n completing assigned tasks in order: 1/9 correct (10%); latency 5.3s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n requesting a break during extended tasks: 3/9 correct (37%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n washing hands independently: 4/14 correct (30%); latency 4.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n self-advocacy statements: 1/13 correct (10%); latency 5.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n Non-compliance: freq 5\n Aggression: freq 5\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 83% to 41% across 10 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\nAggression: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 62.1%.\n- Trend direction: downward (first session: 83%, last session: 41%).\n- Non-compliance: mean 3.3/session (early-window mean 2.3 -> late-window mean 4.3).\n- Aggression: mean 3.4/session (early-window mean 2.7 -> late-window mean 4.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "23589742e8d67259", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Non-compliance": "escape", "Aggression": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adult", "n_sessions": 10, "n_programs": 6, "n_behaviors": 2, "behavior_ids": ["non_compliance", "aggression"], "behavior_functions": ["escape", "escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "397"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6128\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Motor Imitation + Reading\nDate range: Sessions 1\u201311 across 22 days (2026-07-24 to 2026-08-15)\n\nACCELERATION PROGRAMS\n1. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n2. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n3. sounds out CVC words (Reading)\n Method: DTT\n4. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n2. Fecal smearing (scatolia) \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-24 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 5/14 correct (33%); latency 4.5s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d79\n follows 1-step motor instructions: 7/13 correct (51%); latency 3.8s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n sounds out CVC words: 8/13 correct (64%); latency 3.0s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/9 correct (31%); latency 4.8s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n Non-compliance: freq 4\n Fecal smearing (scatolia): attempts 0\n\nSession 2 \u2014 2026-07-26 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 6/13 correct (49%); latency 3.6s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n follows 1-step motor instructions: 5/9 correct (59%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n sounds out CVC words: 6/10 correct (60%); latency 3.1s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/11 correct (44%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n Non-compliance: freq 5\n Fecal smearing (scatolia): attempts 1 (0 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 3 \u2014 2026-07-28 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 8/13 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d77\n follows 1-step motor instructions: 4/10 correct (36%); latency 4.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n sounds out CVC words: 7/15 correct (44%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d712\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/13 correct (46%); latency 3.9s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n Non-compliance: freq 3\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 4 \u2014 2026-07-30 \u2014 IOA SESSION \u2014 2 observers\n imitates 3-step motor sequences: 37% (primary); IOA 97% agreement\n follows 1-step motor instructions: 50% (primary); IOA 97% agreement\n sounds out CVC words: 49% (primary); IOA 97% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 26% (primary); IOA 97% agreement\n Non-compliance: freq 7; IOA 97% agreement\n Fecal smearing (scatolia): attempts 0; IOA 97% agreement\n\nSession 5 \u2014 2026-08-01 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 8/14 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d77\n follows 1-step motor instructions: 4/12 correct (32%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n sounds out CVC words: 3/9 correct (37%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/9 correct (30%); latency 4.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n Non-compliance: freq 5\n Fecal smearing (scatolia): attempts 4 (2 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 6 \u2014 2026-08-03 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 4/11 correct (35%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n follows 1-step motor instructions: 5/10 correct (51%); latency 3.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n sounds out CVC words: 3/8 correct (33%); latency 4.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (43%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d76\n Non-compliance: freq 2\n Fecal smearing (scatolia): attempts 0\n\nSession 7 \u2014 2026-08-05 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 6/11 correct (54%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n follows 1-step motor instructions: 6/13 correct (49%); latency 4.1s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n sounds out CVC words: 4/14 correct (30%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d713\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (40%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n Non-compliance: freq 7\n Fecal smearing (scatolia): attempts 0\n\nSession 8 \u2014 2026-08-07 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 4/13 correct (31%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n follows 1-step motor instructions: 5/11 correct (41%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n sounds out CVC words: 6/11 correct (53%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (39%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n Non-compliance: freq 6\n Fecal smearing (scatolia): attempts 0\n\nSession 9 \u2014 2026-08-09 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 5/9 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n follows 1-step motor instructions: 3/8 correct (32%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n sounds out CVC words: 4/11 correct (38%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/9 correct (48%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n Non-compliance: freq 6\n Fecal smearing (scatolia): attempts 0\n\nSession 10 \u2014 2026-08-11 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 8/12 correct (64%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n follows 1-step motor instructions: 9/13 correct (68%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d78\n sounds out CVC words: 3/11 correct (27%); latency 4.7s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/8 correct (39%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n Non-compliance: freq 6\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 11 \u2014 2026-08-13 \u2014 57 min \u2014 1 observer\n imitates 3-step motor sequences: 6/12 correct (52%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d73\n follows 1-step motor instructions: 5/8 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n sounds out CVC words: 4/9 correct (47%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (56%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n Non-compliance: freq 6\n Fecal smearing (scatolia): attempts 4 (2 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Looking away during SD presentation\n- Leaving seat without permission\n- Delayed responding (>5 seconds consistently)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 43% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\nFecal smearing (scatolia): attention\n Evidence: ABC events in the log support the attention function; behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 48.4%.\n- Trend direction: upward (first session: 33%, last session: 52%).\n- Non-compliance: mean 5.2/session (early-window mean 4.0 -> late-window mean 6.0).\n- Fecal smearing (scatolia): mean 1.1/session (early-window mean 0.7 -> late-window mean 2.0).\n- IOA data: 97% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "4c8437840295598d", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Non-compliance": "escape", "Fecal smearing (scatolia)": "attention"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "early", "n_sessions": 11, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["non_compliance", "fecal_smearing"], "behavior_functions": ["escape", "attention"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "6059"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: mands using polite social conventions\nCurriculum Reference: Mand L3\nLearner Profile: Adolescent Learner\nCurrent Mastery: Emerging\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands using polite social conventions\n- Curriculum: Mand L3\n- Learner: Adolescent Learner\n- Current mastery: Emerging\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: leisure time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the emerging stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "c02d31e4c16dc51f", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L3", "learner_profile": "adolescent", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands using polite social conventions", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "83467"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: responds to peer initiations appropriately\nCurriculum: Social Behavior & Social Play L2\nLearner Level: Early Learner\nBaseline: Developing\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: responds to peer initiations appropriately\n- Curriculum: Social Behavior & Social Play L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the developing stage, begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "ca10ca90e13a9cd3", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "responds to peer initiations appropriately", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "76914"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-3069\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Motor Imitation + Reading\nDate range: Sessions 1\u201311 across 22 days (2026-06-09 to 2026-07-01)\n\nACCELERATION PROGRAMS\n1. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n2. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n3. mands for missing items needed to complete an activity (Mand)\n Method: NET\n4. selects items by feature (Listener Responding)\n Method: DTT\n5. sounds out CVC words (Reading)\n Method: DTT\n\nDECELERATION TARGETS\n1. Motor stereotypy \u2014 function hypothesized: automatic\n2. Aggression \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-09 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 11/15 correct (74%); latency 2.6s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/14 correct (72%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 6/8 correct (70%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n selects items by feature: 6/9 correct (65%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n sounds out CVC words: 11/14 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d710\n Motor stereotypy: freq 2; PIR 20%\n Aggression: freq 1\n\nSession 2 \u2014 2026-06-11 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 9/13 correct (70%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/11 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 6/8 correct (71%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n selects items by feature: 5/8 correct (68%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n sounds out CVC words: 12/15 correct (78%); latency 2.3s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d79\n Motor stereotypy: freq 2; PIR 19%\n Aggression: freq 1\n\nSession 3 \u2014 2026-06-13 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 9/13 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 5/8 correct (69%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n selects items by feature: 9/14 correct (66%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d76\n sounds out CVC words: 7/9 correct (79%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n Motor stereotypy: freq 2; PIR 18%\n Aggression: freq 1\n\nSession 4 \u2014 2026-06-15 \u2014 IOA SESSION \u2014 2 observers\n imitates 3-step motor sequences: 68% (primary); IOA 97% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 75% (primary); IOA 97% agreement\n mands for missing items needed to complete an activity: 65% (primary); IOA 97% agreement\n selects items by feature: 63% (primary); IOA 97% agreement\n sounds out CVC words: 72% (primary); IOA 97% agreement\n Motor stereotypy: freq 3; PIR 15%; IOA 97% agreement\n Aggression: freq 1; IOA 97% agreement\n\nSession 5 \u2014 2026-06-17 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 5/10 correct (47%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/11 correct (49%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n mands for missing items needed to complete an activity: 5/10 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d73\n selects items by feature: 5/12 correct (39%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n sounds out CVC words: 7/15 correct (48%); latency 4.1s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n Motor stereotypy: freq 8; PIR 36%\n Aggression: freq 6\n\nSession 6 \u2014 2026-06-19 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 6/12 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (56%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 5/9 correct (56%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n selects items by feature: 5/11 correct (46%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n sounds out CVC words: 7/13 correct (57%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d78\n Motor stereotypy: freq 10; PIR 50%\n Aggression: freq 5\n\nSession 7 \u2014 2026-06-21 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 8/15 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (65%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n mands for missing items needed to complete an activity: 8/13 correct (60%); latency 3.2s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d73\n selects items by feature: 7/12 correct (58%); latency 3.7s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d75\n sounds out CVC words: 10/15 correct (65%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d710\n Motor stereotypy: freq 3; PIR 9%\n Aggression: freq 2\n\nSession 8 \u2014 2026-06-23 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 4/8 correct (55%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/11 correct (68%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n mands for missing items needed to complete an activity: 7/11 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n selects items by feature: 8/13 correct (61%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d77\n sounds out CVC words: 10/13 correct (74%); latency 2.6s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n Motor stereotypy: freq 4; PIR 24%\n Aggression: freq 2\n\nSession 9 \u2014 2026-06-25 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 6/9 correct (64%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 10/14 correct (72%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d73\n mands for missing items needed to complete an activity: 9/13 correct (67%); latency 3.1s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n selects items by feature: 7/11 correct (63%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n sounds out CVC words: 9/12 correct (74%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n Motor stereotypy: freq 3; PIR 22%\n Aggression: freq 2\n\nSession 10 \u2014 2026-06-27 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 9/13 correct (71%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n mands for missing items needed to complete an activity: 10/15 correct (69%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d711\n selects items by feature: 6/9 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n sounds out CVC words: 6/8 correct (74%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Motor stereotypy: freq 2; PIR 16%\n Aggression: freq 1\n\nSession 11 \u2014 2026-06-29 \u2014 48 min \u2014 1 observer\n imitates 3-step motor sequences: 6/8 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 10/14 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d710\n selects items by feature: 8/12 correct (63%); latency 3.0s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n sounds out CVC words: 7/10 correct (74%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n Motor stereotypy: freq 2; PIR 9%\n Aggression: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\nAggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n### Crisis plan\n- If the setting event involves a medical concern (seizure-like activity, injury, illness): stop the program immediately, follow the facility's medical-incident protocol, and refer to the appropriate medical professional. Medical events are NOT behavioral targets and must not be treated as ABA programming issues.\n- If behavior escalates during the setting-event window: use verbal de-escalation and environmental safety first; do not add demand. Only staff currently certified in the facility's crisis-prevention framework may use physical intervention, and only if imminent safety risk warrants it AND the learner's BIP specifically authorizes it (contraindications and BIP restrictions may rule it out).\n- Coordinate with caregivers, the medical team, and the supervising BCBA; document the setting event and the associated behavior for pattern analysis. Review the BIP in light of the setting-event data.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 62.3%.\n- Trend direction: downward (first session: 74%, last session: 71%).\n- Motor stereotypy: mean 3.7/session (early-window mean 2.0 -> late-window mean 2.3).\n- Aggression: mean 2.2/session (early-window mean 1.0 -> late-window mean 1.7).\n- IOA data: 97% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "b086659bdcc392f4", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Motor stereotypy": "automatic", "Aggression": "escape"}, "escalation_level": 3, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "early", "n_sessions": 11, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["motor_stereotypy", "aggression"], "behavior_functions": ["automatic", "escape"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "6568"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: points to named body parts\nCurriculum: Listener Responding L1\nLearner: Early Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: points to named body parts\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: glue, phone, brush\nDistractor stimuli: ball, key\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8dfc52390c072d2e", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "points to named body parts", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "vr3", "error_correction": "transfer_trial", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "85854"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8854\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Listener Responding + Mand + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-06-04 to 2026-06-26)\n\nACCELERATION PROGRAMS\n1. mands for missing items needed to complete an activity (Mand)\n Method: NET\n2. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n3. echoes 2-word phrases (Echoic)\n Method: DTT variant\n4. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n5. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: tangible\n\nSESSION DATA\n\nSession 1 \u2014 2026-06-04 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 11/15 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d75, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/14 correct (73%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n echoes 2-word phrases: 9/11 correct (85%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n follows 1-step motor instructions: 13/15 correct (85%); latency 2.1s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d711\n wait calmly when access to preferred item is denied: 5/8 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n Tantrum: freq 0\n\nSession 2 \u2014 2026-06-06 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 9/11 correct (78%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 11/15 correct (72%); latency 2.7s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d77\n echoes 2-word phrases: 9/10 correct (86%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n follows 1-step motor instructions: 8/10 correct (78%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 6/8 correct (70%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n Tantrum: freq 0\n\nSession 3 \u2014 2026-06-08 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/9 correct (78%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n echoes 2-word phrases: 12/15 correct (83%); latency 2.1s; prompts FP\u00d75, PP\u00d75, G\u00d70, I\u00d75\n follows 1-step motor instructions: 9/11 correct (81%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d73\n wait calmly when access to preferred item is denied: 7/10 correct (74%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n Tantrum: freq 0\n\nSession 4 \u2014 2026-06-10 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 3/8 correct (42%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n echoes 2-word phrases: 4/10 correct (43%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n follows 1-step motor instructions: 5/11 correct (48%); latency 3.9s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 3/9 correct (30%); latency 4.5s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n Tantrum: freq 1, duration 4m total\n\nSession 5 \u2014 2026-06-12 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 3/8 correct (40%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/11 correct (48%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n echoes 2-word phrases: 6/12 correct (49%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n follows 1-step motor instructions: 5/10 correct (46%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 4/15 correct (29%); latency 4.6s; prompts FP\u00d72, PP\u00d75, G\u00d72, I\u00d76\n Tantrum: freq 1, duration 4m total\n\nSession 6 \u2014 2026-06-14 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/15 correct (40%); latency 4.5s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/13 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n echoes 2-word phrases: 5/11 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n follows 1-step motor instructions: 5/11 correct (46%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 3/8 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n Tantrum: freq 1, duration 5m total\n\nSession 7 \u2014 2026-06-16 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/14 correct (38%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/11 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n echoes 2-word phrases: 4/9 correct (46%); latency 4.2s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n follows 1-step motor instructions: 7/13 correct (53%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n wait calmly when access to preferred item is denied: 6/13 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d75\n Tantrum: freq 0\n\nSession 8 \u2014 2026-06-18 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/14 correct (46%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (60%); latency 3.3s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n echoes 2-word phrases: 7/13 correct (56%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d711\n follows 1-step motor instructions: 5/8 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n wait calmly when access to preferred item is denied: 5/10 correct (50%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n Tantrum: freq 0\n\nSession 9 \u2014 2026-06-20 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/11 correct (52%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/14 correct (64%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d77\n echoes 2-word phrases: 7/12 correct (62%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d78\n follows 1-step motor instructions: 8/12 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n wait calmly when access to preferred item is denied: 5/10 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n Tantrum: freq 0\n\nSession 10 \u2014 2026-06-22 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/11 correct (60%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/10 correct (71%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n echoes 2-word phrases: 7/11 correct (66%); latency 3.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n follows 1-step motor instructions: 6/9 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 5/8 correct (57%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n Tantrum: freq 0\n\nSession 11 \u2014 2026-06-24 \u2014 44 min \u2014 1 observer\n mands for missing items needed to complete an activity: 8/13 correct (63%); latency 3.4s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/12 correct (74%); latency 2.8s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n echoes 2-word phrases: 11/15 correct (76%); latency 2.7s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d710\n follows 1-step motor instructions: 8/11 correct (73%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 6/10 correct (64%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n Tantrum: freq 0\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nTantrum: tangible\n Evidence: behavior occurs when preferred items are restricted and is followed by access to the item.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 55.6%.\n- Trend direction: downward (first session: 74%, last session: 63%).\n- Tantrum: mean 0.3/session (early-window mean 0.0 -> late-window mean 0.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "9813d9a17c166011", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Tantrum": "tangible"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "early", "n_sessions": 11, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["tangible"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "10517"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4545\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Motor Imitation\nDate range: Sessions 1\u20138 across 16 days (2026-10-15 to 2026-10-31)\n\nACCELERATION PROGRAMS\n1. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n2. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n3. indicating completion ('all done', 'finished') (Mand)\n Method: NET\n4. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-15 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/12 correct (69%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n imitates 3-step motor sequences: 11/15 correct (76%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d711\n indicating completion ('all done', 'finished'): 10/13 correct (74%); latency 3.0s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 9/11 correct (82%); latency 2.2s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/13 correct (86%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n Verbal aggression: freq 3\n\nSession 2 \u2014 2026-10-17 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/14 correct (66%); latency 3.1s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n imitates 3-step motor sequences: 7/10 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 9/14 correct (66%); latency 3.0s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 11/14 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/11 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n Verbal aggression: freq 2\n\nSession 3 \u2014 2026-10-19 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/10 correct (63%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d74\n imitates 3-step motor sequences: 10/15 correct (67%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n indicating completion ('all done', 'finished'): 8/13 correct (60%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 5/8 correct (69%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (66%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n Verbal aggression: freq 3\n\nSession 4 \u2014 2026-10-21 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/11 correct (48%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n imitates 3-step motor sequences: 5/8 correct (59%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n indicating completion ('all done', 'finished'): 4/8 correct (46%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 9/15 correct (62%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (61%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n Verbal aggression: freq 5\n\nSession 5 \u2014 2026-10-23 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/11 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n imitates 3-step motor sequences: 4/8 correct (52%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n indicating completion ('all done', 'finished'): 6/14 correct (41%); latency 4.4s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 5/9 correct (61%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/13 correct (54%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d79\n Verbal aggression: freq 4\n\nSession 6 \u2014 2026-10-25 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/14 correct (38%); latency 4.3s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n imitates 3-step motor sequences: 6/15 correct (39%); latency 4.4s; prompts FP\u00d72, PP\u00d75, G\u00d72, I\u00d76\n indicating completion ('all done', 'finished'): 4/9 correct (40%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 6/12 correct (52%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (50%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d76\n Verbal aggression: freq 5\n\nSession 7 \u2014 2026-10-27 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/12 correct (32%); latency 4.8s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n imitates 3-step motor sequences: 4/11 correct (36%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n indicating completion ('all done', 'finished'): 3/11 correct (25%); latency 4.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 5/11 correct (48%); latency 3.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/14 correct (37%); latency 4.1s; prompts FP\u00d74, PP\u00d73, G\u00d70, I\u00d77\n Verbal aggression: freq 6\n\nSession 8 \u2014 2026-10-29 \u2014 49 min \u2014 1 observer\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (26%); latency 5.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n imitates 3-step motor sequences: 3/10 correct (26%); latency 5.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n indicating completion ('all done', 'finished'): 2/8 correct (23%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 5/13 correct (35%); latency 4.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (41%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d76\n Verbal aggression: freq 6\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Vocal refusal ('no', 'I don't want to')\n- Self-stimulatory behavior increase during demand periods\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 69% to 26% across 8 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nVerbal aggression: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 48.8%.\n- Trend direction: downward (first session: 69%, last session: 26%).\n- Verbal aggression: mean 4.2/session (early-window mean 2.5 -> late-window mean 6.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a667eb86c0c304a9", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Verbal aggression": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 8, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["verbal_aggression"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "51985"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2092\nProfile: Adult Learner (18+ years)\nCurricula: Community + Mand + Self-Care\nDate range: Sessions 1\u20137 across 14 days (2026-09-21 to 2026-10-05)\n\nACCELERATION PROGRAMS\n1. indicating 'all done' at task completion (Mand)\n Method: NET\n2. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n3. washing hands independently (Self-Care)\n Method: Task Analysis\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n5. requesting a break during extended tasks (Mand)\n Method: NET\n6. using public transportation (Community)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-21 \u2014 30 min \u2014 1 observer\n indicating 'all done' at task completion: 8/12 correct (63%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n dressing (pants, shirt): 7/12 correct (59%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n washing hands independently: 6/9 correct (67%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d75\n brushing teeth independently: 8/12 correct (63%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n requesting a break during extended tasks: 5/8 correct (65%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n using public transportation: 11/15 correct (70%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d75, I\u00d76\n Tantrum: freq 1, duration 5m total\n\nSession 2 \u2014 2026-09-23 \u2014 30 min \u2014 1 observer\n indicating 'all done' at task completion: 6/9 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n dressing (pants, shirt): 6/10 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n washing hands independently: 10/15 correct (68%); latency 3.1s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d79\n brushing teeth independently: 5/8 correct (65%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n requesting a break during extended tasks: 7/10 correct (70%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d75\n using public transportation: 5/8 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Tantrum: freq 1, duration 5m total\n\nSession 3 \u2014 2026-09-25 \u2014 IOA SESSION \u2014 2 observers\n indicating 'all done' at task completion: 61% (primary); IOA 96% agreement\n dressing (pants, shirt): 70% (primary); IOA 96% agreement\n washing hands independently: 71% (primary); IOA 96% agreement\n brushing teeth independently: 74% (primary); IOA 96% agreement\n requesting a break during extended tasks: 74% (primary); IOA 96% agreement\n using public transportation: 74% (primary); IOA 96% agreement\n Tantrum: freq 4, duration 8m total; IOA 96% agreement\n\nSession 4 \u2014 2026-09-27 \u2014 30 min \u2014 1 observer\n indicating 'all done' at task completion: 8/12 correct (66%); latency 3.1s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d74\n dressing (pants, shirt): 5/8 correct (63%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n washing hands independently: 6/8 correct (75%); latency 2.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n brushing teeth independently: 6/9 correct (71%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n requesting a break during extended tasks: 7/9 correct (74%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n using public transportation: 9/12 correct (74%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d76\n Tantrum: freq 5, duration 15m total\n\nSession 5 \u2014 2026-09-29 \u2014 30 min \u2014 1 observer\n indicating 'all done' at task completion: 9/13 correct (67%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n dressing (pants, shirt): 10/14 correct (72%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d710\n washing hands independently: 6/8 correct (77%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n brushing teeth independently: 12/15 correct (77%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d76\n requesting a break during extended tasks: 6/8 correct (81%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n using public transportation: 8/12 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d79\n Tantrum: freq 0\n\nSession 6 \u2014 2026-10-01 \u2014 30 min \u2014 1 observer\n indicating 'all done' at task completion: 10/14 correct (71%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n dressing (pants, shirt): 10/15 correct (67%); latency 3.2s; prompts FP\u00d71, PP\u00d75, G\u00d73, I\u00d76\n washing hands independently: 9/11 correct (79%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n brushing teeth independently: 8/10 correct (76%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n requesting a break during extended tasks: 8/9 correct (86%); latency 2.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n using public transportation: 7/10 correct (73%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n Tantrum: freq 1, duration 4m total\n\nSession 7 \u2014 2026-10-03 \u2014 30 min \u2014 1 observer\n indicating 'all done' at task completion: 11/15 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n dressing (pants, shirt): 8/10 correct (77%); latency 2.8s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n washing hands independently: 10/13 correct (78%); latency 2.6s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d78\n brushing teeth independently: 8/10 correct (79%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d77\n requesting a break during extended tasks: 8/10 correct (81%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n using public transportation: 6/8 correct (75%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n Tantrum: freq 1, duration 5m total\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n- Elopement from the work area\n- Vocal refusal ('no', 'I don't want to')\n- Peer aggression (hitting or pushing peers during group activities)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nTantrum: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 65.5%.\n- Trend direction: upward (first session: 63%, last session: 70%).\n- Tantrum: mean 1.9/session (early-window mean 1.0 -> late-window mean 1.0).\n- IOA data: 96% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "c31ac167bc96029e", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Tantrum": "attention"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adult", "n_sessions": 7, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "47050"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands for a break during a non-preferred task (\"break please, I need a break\")\nCurriculum: Mand L2\nLearner: School-Age Learner\nMastery Status: Approaching mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for a break during a non-preferred task (\"break please, I need a break\")\n- Curriculum: Mand L2\n- Learner: School-Age Learner\n- Current mastery: Approaching mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nPresent a brief non-preferred task or demand at a level the learner can tolerate but may wish to pause. Wait for a request-for-break topography; honor the request immediately on first acquisition (CRF), thin later.\nPrimary natural context: academic period.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's academic period routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the approaching mastery stage, use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially.\n\n## Natural Reinforcer\nNatural reinforcer: immediate brief escape from the demand (30\u201360 seconds) to a neutral or preferred activity.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "9c661bb2acc71bdd", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "school_age", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for a break during a non-preferred task (\"break please, I need a break\")", "mo_category": "mand_break", "mo_arrangement": "break_opportunity", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "33330"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6263\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading\nDate range: Sessions 1\u201311 across 22 days (2026-09-17 to 2026-10-09)\n\nACCELERATION PROGRAMS\n1. mands for missing items needed to complete an activity (Mand)\n Method: NET\n2. counts objects 1\u201320 (Math)\n Method: DTT\n3. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n4. sounds out CVC words (Reading)\n Method: DTT\n5. identifies numerals 1\u201320 (Math)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-17 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/12 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d72\n counts objects 1\u201320: 4/8 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/8 correct (55%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n sounds out CVC words: 6/11 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n identifies numerals 1\u201320: 8/11 correct (71%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n\nSession 2 \u2014 2026-09-19 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 7/11 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n counts objects 1\u201320: 6/12 correct (52%); latency 3.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (55%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d72\n sounds out CVC words: 5/8 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n identifies numerals 1\u201320: 7/11 correct (59%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d76\n\nSession 3 \u2014 2026-09-21 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/8 correct (58%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n counts objects 1\u201320: 5/10 correct (48%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (57%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n sounds out CVC words: 8/15 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d710\n identifies numerals 1\u201320: 9/13 correct (68%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n\nSession 4 \u2014 2026-09-23 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/12 correct (54%); latency 3.4s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n counts objects 1\u201320: 3/8 correct (39%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (52%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n sounds out CVC words: 6/10 correct (65%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d74\n identifies numerals 1\u201320: 5/9 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n\nSession 5 \u2014 2026-09-25 \u2014 IOA SESSION \u2014 2 observers\n mands for missing items needed to complete an activity: 51% (primary); IOA 93% agreement\n counts objects 1\u201320: 49% (primary); IOA 93% agreement\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 45% (primary); IOA 93% agreement\n sounds out CVC words: 63% (primary); IOA 93% agreement\n identifies numerals 1\u201320: 56% (primary); IOA 93% agreement\n\nSession 6 \u2014 2026-09-27 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 4/9 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n counts objects 1\u201320: 6/14 correct (43%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/15 correct (49%); latency 3.7s; prompts FP\u00d72, PP\u00d70, G\u00d75, I\u00d78\n sounds out CVC words: 5/10 correct (51%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d76\n identifies numerals 1\u201320: 7/11 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n\nSession 7 \u2014 2026-09-29 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/9 correct (51%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n counts objects 1\u201320: 5/10 correct (46%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/13 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d78\n sounds out CVC words: 5/8 correct (61%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n identifies numerals 1\u201320: 6/12 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n\nSession 8 \u2014 2026-10-01 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 6/10 correct (63%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n counts objects 1\u201320: 5/14 correct (37%); latency 4.3s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/11 correct (48%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n sounds out CVC words: 8/14 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d712\n identifies numerals 1\u201320: 8/13 correct (59%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d78\n\nSession 9 \u2014 2026-10-03 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 5/9 correct (57%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n counts objects 1\u201320: 3/8 correct (40%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/14 correct (46%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n sounds out CVC words: 8/14 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n identifies numerals 1\u201320: 7/13 correct (53%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n\nSession 10 \u2014 2026-10-05 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 8/15 correct (52%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n counts objects 1\u201320: 4/13 correct (34%); latency 4.5s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/10 correct (43%); latency 4.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n sounds out CVC words: 8/15 correct (55%); latency 3.4s; prompts FP\u00d72, PP\u00d71, G\u00d75, I\u00d77\n identifies numerals 1\u201320: 7/14 correct (48%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n\nSession 11 \u2014 2026-10-07 \u2014 52 min \u2014 1 observer\n mands for missing items needed to complete an activity: 8/13 correct (58%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n counts objects 1\u201320: 4/12 correct (36%); latency 4.3s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/10 correct (40%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n sounds out CVC words: 5/9 correct (55%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n identifies numerals 1\u201320: 5/10 correct (49%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Stereotypic behavior during inter-trial intervals\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 52% for 11 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 11 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 54.3%.\n- Trend direction: upward (first session: 46%, last session: 58%).\n- IOA data: 93% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "4e05fcee44f06284", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "school_age", "n_sessions": 11, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "98683"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: completing assigned tasks in order\nCurriculum: AFLS Vocational Skills\nLearner Level: Adult Learner\nBaseline: Generalization phase\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Review the task list for the shift. <- Learner-taught step (forward chaining starts here)\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 60s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fb4cfe601aed627d", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "75542"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: counts objects 1\u20135 with 1:1 correspondence\nCurriculum: Math L1\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: counts objects 1\u20135 with 1:1 correspondence\n- Curriculum: Math L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How many?\"\nVariations: \"Count these\"; \"How many are there?\"\nPresentation: Present the set of objects to count.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: 2, 5, 4\nDistractor stimuli: 1, 3\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "1cc54eb8bff7942f", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "counts objects 1\u20135 with 1:1 correspondence", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "66255"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4021\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-07-10 to 2026-07-28)\n\nACCELERATION PROGRAMS\n1. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n2. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n3. indicating 'all done' at task completion (Mand)\n Method: NET\n4. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Non-compliance \u2014 function hypothesized: escape\n2. Pica \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-10 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 7/9 correct (78%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n dressing (pants, shirt): 11/15 correct (75%); latency 2.6s; prompts FP\u00d70, PP\u00d75, G\u00d71, I\u00d79\n indicating 'all done' at task completion: 10/14 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n scheduling medical appointments: 9/11 correct (80%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n Non-compliance: freq 1\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 2 \u2014 2026-07-12 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 8/12 correct (70%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n dressing (pants, shirt): 6/10 correct (63%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n indicating 'all done' at task completion: 8/11 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n scheduling medical appointments: 9/12 correct (78%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d78\n Non-compliance: freq 3\n Pica: attempts 3 (2 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-07-14 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/8 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n dressing (pants, shirt): 9/15 correct (57%); latency 3.2s; prompts FP\u00d71, PP\u00d75, G\u00d70, I\u00d79\n indicating 'all done' at task completion: 8/13 correct (64%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d710\n scheduling medical appointments: 11/15 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d77\n Non-compliance: freq 2\n Pica: attempts 3 (3 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-07-16 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/9 correct (51%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n dressing (pants, shirt): 5/11 correct (48%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n indicating 'all done' at task completion: 7/13 correct (56%); latency 3.3s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n scheduling medical appointments: 8/12 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d77\n Non-compliance: freq 4\n Pica: attempts 4 (4 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-07-18 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/12 correct (45%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d710\n dressing (pants, shirt): 3/8 correct (41%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n indicating 'all done' at task completion: 5/11 correct (42%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n scheduling medical appointments: 5/10 correct (52%); latency 3.5s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n Non-compliance: freq 5\n Pica: attempts 5 (4 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-07-20 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/11 correct (45%); latency 4.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d77\n dressing (pants, shirt): 5/13 correct (39%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n indicating 'all done' at task completion: 4/10 correct (37%); latency 4.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n scheduling medical appointments: 6/12 correct (48%); latency 3.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d79\n Non-compliance: freq 5\n Pica: attempts 6 (6 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-07-22 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 4/13 correct (33%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n dressing (pants, shirt): 4/12 correct (32%); latency 4.6s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n indicating 'all done' at task completion: 4/13 correct (33%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d78\n scheduling medical appointments: 4/10 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n Non-compliance: freq 5\n Pica: attempts 6 (5 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-07-24 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 5/14 correct (34%); latency 4.2s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n dressing (pants, shirt): 2/10 correct (19%); latency 5.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n indicating 'all done' at task completion: 3/14 correct (21%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d79\n scheduling medical appointments: 3/9 correct (36%); latency 4.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n Non-compliance: freq 5\n Pica: attempts 8 (5 unsuccessful \u2014 staff retrieved item before ingestion; 3 successful \u2014 item ingested)\n\nSession 9 \u2014 2026-07-26 \u2014 53 min \u2014 1 observer\n wait calmly when a preferred activity is denied: 2/8 correct (23%); latency 4.9s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n dressing (pants, shirt): 2/10 correct (16%); latency 5.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d74\n indicating 'all done' at task completion: 3/14 correct (19%); latency 5.0s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d76\n scheduling medical appointments: 3/15 correct (21%); latency 5.0s; prompts FP\u00d72, PP\u00d75, G\u00d74, I\u00d74\n Non-compliance: freq 6\n Pica: attempts 7 (6 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Self-stimulatory behavior increase during demand periods\n- Increased response latency (3\u20135\u00d7 baseline)\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (78% -> 23%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nPica: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 48.8%.\n- Trend direction: downward (first session: 78%, last session: 23%).\n- Non-compliance: mean 4.0/session (early-window mean 2.0 -> late-window mean 5.3).\n- Pica: mean 4.9/session (early-window mean 2.7 -> late-window mean 7.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "5842093b0c8c7273", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Non-compliance": "escape", "Pica": "attention"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adult", "n_sessions": 9, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["non_compliance", "pica"], "behavior_functions": ["escape", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "28057"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8120\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Reading + Self-Care\nDate range: Sessions 1\u20138 across 16 days (2026-09-19 to 2026-10-05)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. tolerating haircut (Self-Care)\n Method: Task Analysis\n3. identifying community safety signs (Community)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-19 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 3/8 correct (33%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n tolerating haircut: 4/15 correct (28%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n identifying community safety signs: 3/12 correct (23%); latency 4.7s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n\nSession 2 \u2014 2026-09-21 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 4/11 correct (40%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n tolerating haircut: 4/11 correct (36%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n identifying community safety signs: 4/12 correct (34%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d77\n\nSession 3 \u2014 2026-09-23 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 4/8 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n tolerating haircut: 6/13 correct (47%); latency 3.8s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n identifying community safety signs: 4/8 correct (45%); latency 4.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n\nSession 4 \u2014 2026-09-25 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 5/9 correct (55%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n tolerating haircut: 6/10 correct (55%); latency 3.2s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n identifying community safety signs: 6/10 correct (55%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n\nSession 5 \u2014 2026-09-27 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 6/10 correct (64%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d76\n tolerating haircut: 9/14 correct (65%); latency 3.2s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n identifying community safety signs: 9/15 correct (61%); latency 3.5s; prompts FP\u00d75, PP\u00d75, G\u00d71, I\u00d74\n\nSession 6 \u2014 2026-09-29 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 6/8 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n tolerating haircut: 6/8 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n identifying community safety signs: 7/9 correct (73%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n\nSession 7 \u2014 2026-10-01 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 8/10 correct (78%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n tolerating haircut: 10/12 correct (87%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d710\n identifying community safety signs: 8/9 correct (85%); latency 2.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n\nSession 8 \u2014 2026-10-03 \u2014 46 min \u2014 1 observer\n reads 10\u201320 sight words: 9/10 correct (90%); latency 1.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n tolerating haircut: 14/15 correct (95%); latency 1.8s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d76\n identifying community safety signs: 10/11 correct (92%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Reaching for materials proactively\n- Quick response latency (within 2\u20133 seconds)\n- Appropriate sitting posture maintained\n- Requesting more trials verbally"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 33% to 90% across 8 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 60.1%.\n- Trend direction: upward (first session: 33%, last session: 90%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a345e9e2f41bf249", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "adolescent", "n_sessions": 8, "n_programs": 3, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "28505"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2637\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Reading + Tolerating Denied Access\nDate range: Sessions 1\u20137 across 14 days (2026-03-22 to 2026-04-05)\n\nACCELERATION PROGRAMS\n1. requesting a break during non-preferred activities (Mand)\n Method: NET\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. requesting help when a task is difficult (Mand)\n Method: NET\n4. sounds out CVC words (Reading)\n Method: DTT\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-22 \u2014 34 min \u2014 1 observer\n requesting a break during non-preferred activities: 3/11 correct (26%); latency 4.7s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 3/15 correct (22%); latency 5.0s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d79\n requesting help when a task is difficult: 4/13 correct (31%); latency 4.8s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d78\n sounds out CVC words: 4/14 correct (26%); latency 5.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n\nSession 2 \u2014 2026-03-24 \u2014 34 min \u2014 1 observer\n requesting a break during non-preferred activities: 3/10 correct (33%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n wait calmly when access to preferred item is denied: 4/12 correct (34%); latency 4.5s; prompts FP\u00d74, PP\u00d72, G\u00d74, I\u00d72\n requesting help when a task is difficult: 6/14 correct (44%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d79\n sounds out CVC words: 3/10 correct (34%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d75\n\nSession 3 \u2014 2026-03-26 \u2014 34 min \u2014 1 observer\n requesting a break during non-preferred activities: 5/11 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 5/12 correct (42%); latency 4.2s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d75\n requesting help when a task is difficult: 6/12 correct (51%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n sounds out CVC words: 5/10 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n\nSession 4 \u2014 2026-03-28 \u2014 IOA SESSION \u2014 2 observers\n requesting a break during non-preferred activities: 54% (primary); IOA 94% agreement\n wait calmly when access to preferred item is denied: 51% (primary); IOA 94% agreement\n requesting help when a task is difficult: 61% (primary); IOA 94% agreement\n sounds out CVC words: 50% (primary); IOA 94% agreement\n\nSession 5 \u2014 2026-03-30 \u2014 34 min \u2014 1 observer\n requesting a break during non-preferred activities: 9/14 correct (68%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n wait calmly when access to preferred item is denied: 7/12 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d76\n requesting help when a task is difficult: 8/11 correct (76%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n sounds out CVC words: 8/13 correct (64%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d79\n\nSession 6 \u2014 2026-04-01 \u2014 34 min \u2014 1 observer\n requesting a break during non-preferred activities: 11/15 correct (76%); latency 2.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d712\n wait calmly when access to preferred item is denied: 8/11 correct (70%); latency 2.6s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n requesting help when a task is difficult: 9/10 correct (87%); latency 2.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n sounds out CVC words: 6/9 correct (71%); latency 2.6s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n\nSession 7 \u2014 2026-04-03 \u2014 34 min \u2014 1 observer\n requesting a break during non-preferred activities: 7/8 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n wait calmly when access to preferred item is denied: 10/12 correct (81%); latency 2.4s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n requesting help when a task is difficult: 10/11 correct (95%); latency 1.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n sounds out CVC words: 8/10 correct (81%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Eye contact maintained throughout trials\n- Appropriate sitting posture maintained\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 26% to 84% across 7 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data with IOA evidence.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 55.0%.\n- Trend direction: upward (first session: 26%, last session: 84%).\n- IOA data: 94% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "78d21e46e7f6ccb9", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "adolescent", "n_sessions": 7, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "39857"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: completing assigned tasks in order\nCurriculum Reference: AFLS Vocational Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Near mastery\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: completing assigned tasks in order\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Review the task list for the shift. <- Learner-taught step (forward chaining starts here)\n2. Gather needed materials for the first task.\n3. Complete the first task to the specified standard.\n4. Mark the first task complete on the list.\n5. Return materials to storage.\n6. Move to the next task on the list and repeat.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nDifferential per-step + terminal \u2014 Small per-step reinforcement for independent responses; larger reinforcer at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 90s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3fa8ce87d7187f19", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "completing assigned tasks in order", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "near", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "completing assigned tasks in order", "module": "vocational", "mastery_criterion": "mc_fluent_chain", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "differential_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "50550"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: showering\nCurriculum: AFLS Basic Living Skills\nLearner: School-Age Learner\nMastery Status: Developing\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: showering\n- Curriculum: AFLS Basic Living Skills\n- Learner: School-Age Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Backward chaining\nComplete all steps for the learner except the last; teach the last step first, then work backward. The final reinforcer (task completion) is always present for the learner.\n\nSteps:\n1. Enter the shower and close the curtain.\n2. Turn on the water and adjust temperature.\n3. Wet body completely.\n4. Apply soap to a washcloth.\n5. Wash body systematically (face -> torso -> arms -> legs -> feet).\n6. Rinse all soap off.\n7. Apply shampoo and rub into hair.\n8. Rinse hair completely.\n9. Turn off the water.\n10. Exit the shower and dry off with a towel. <- Learner-taught step (backward chaining starts here)\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6db98ef5362d1f51", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "showering", "program_type": "independence", "learner_profile": "school_age", "mastery_state": "developing", "chain_type": "backward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "showering", "module": "basic_living", "mastery_criterion": "mc_self_initiation", "chain_type": "backward", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "98892"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5912\nProfile: Adult Learner (18+ years)\nCurricula: Independent Living + Mand + Self-Care\nDate range: Sessions 1\u201312 across 24 days (2026-04-09 to 2026-05-03)\n\nACCELERATION PROGRAMS\n1. managing a budget (Independent Living)\n Method: Task Analysis\n2. requesting a break during extended tasks (Mand)\n Method: NET\n3. self-advocacy statements (Independent Living)\n Method: BST\n4. brushing teeth independently (Self-Care)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Property destruction \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-09 \u2014 32 min \u2014 1 observer\n managing a budget: 7/11 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n requesting a break during extended tasks: 7/12 correct (54%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d73\n self-advocacy statements: 5/8 correct (61%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n brushing teeth independently: 6/12 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d710\n Property destruction: freq 0\n\nSession 2 \u2014 2026-04-11 \u2014 32 min \u2014 1 observer\n managing a budget: 9/14 correct (65%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d76\n requesting a break during extended tasks: 4/9 correct (48%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n self-advocacy statements: 7/13 correct (53%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n brushing teeth independently: 8/15 correct (51%); latency 3.8s; prompts FP\u00d75, PP\u00d71, G\u00d70, I\u00d79\n Property destruction: freq 1\n\nSession 3 \u2014 2026-04-13 \u2014 32 min \u2014 1 observer\n managing a budget: 5/8 correct (67%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting a break during extended tasks: 4/10 correct (40%); latency 4.4s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n self-advocacy statements: 7/12 correct (62%); latency 2.9s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n brushing teeth independently: 6/13 correct (48%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d79\n Property destruction: freq 0\n\nSession 4 \u2014 2026-04-15 \u2014 32 min \u2014 1 observer\n managing a budget: 4/8 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n requesting a break during extended tasks: 7/14 correct (48%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d77\n self-advocacy statements: 7/14 correct (51%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n brushing teeth independently: 6/14 correct (45%); latency 4.0s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n Property destruction: freq 0\n\nSession 5 \u2014 2026-04-17 \u2014 32 min \u2014 1 observer\n managing a budget: 8/12 correct (69%); latency 3.2s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n requesting a break during extended tasks: 4/10 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n self-advocacy statements: 5/9 correct (53%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n brushing teeth independently: 6/14 correct (42%); latency 4.1s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n Property destruction: freq 0\n\nSession 6 \u2014 2026-04-19 \u2014 32 min \u2014 1 observer\n managing a budget: 7/12 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d75\n requesting a break during extended tasks: 5/13 correct (42%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d712\n self-advocacy statements: 8/13 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n brushing teeth independently: 5/13 correct (40%); latency 4.2s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n Property destruction: freq 0\n\nSession 7 \u2014 2026-04-21 \u2014 32 min \u2014 1 observer\n managing a budget: 7/10 correct (66%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n requesting a break during extended tasks: 6/11 correct (51%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n self-advocacy statements: 4/8 correct (51%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n brushing teeth independently: 4/11 correct (40%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n Property destruction: freq 0\n\nSession 8 \u2014 2026-04-23 \u2014 32 min \u2014 1 observer\n managing a budget: 5/8 correct (67%); latency 3.0s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n requesting a break during extended tasks: 6/15 correct (42%); latency 4.2s; prompts FP\u00d75, PP\u00d70, G\u00d75, I\u00d75\n self-advocacy statements: 8/14 correct (57%); latency 3.2s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n brushing teeth independently: 3/9 correct (37%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d75\n Property destruction: freq 1\n\nSession 9 \u2014 2026-04-25 \u2014 32 min \u2014 1 observer\n managing a budget: 9/15 correct (63%); latency 3.1s; prompts FP\u00d75, PP\u00d74, G\u00d74, I\u00d72\n requesting a break during extended tasks: 5/15 correct (35%); latency 4.5s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d79\n self-advocacy statements: 6/11 correct (55%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n brushing teeth independently: 6/15 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d75, I\u00d79\n Property destruction: freq 0\n\nSession 10 \u2014 2026-04-27 \u2014 32 min \u2014 1 observer\n managing a budget: 6/12 correct (54%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n requesting a break during extended tasks: 5/10 correct (46%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n self-advocacy statements: 8/13 correct (63%); latency 3.3s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n brushing teeth independently: 5/12 correct (43%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n Property destruction: freq 0\n\nSession 11 \u2014 2026-04-29 \u2014 32 min \u2014 1 observer\n managing a budget: 8/14 correct (60%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n requesting a break during extended tasks: 4/10 correct (43%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n self-advocacy statements: 8/12 correct (64%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d76\n brushing teeth independently: 4/9 correct (42%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n Property destruction: freq 0\n\nSession 12 \u2014 2026-05-01 \u2014 32 min \u2014 1 observer\n managing a budget: 10/15 correct (68%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d713\n requesting a break during extended tasks: 4/8 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n self-advocacy statements: 7/13 correct (56%); latency 3.2s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n brushing teeth independently: 6/13 correct (46%); latency 4.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d711\n Property destruction: freq 1\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 66% for 12 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Behavior Function Hypothesis\nProperty destruction: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 62.8%.\n- Trend direction: upward (first session: 63%, last session: 68%).\n- Property destruction: mean 0.2/session (early-window mean 0.2 -> late-window mean 0.2).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "fc4776969c277e44", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {"Property destruction": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "adult", "n_sessions": 12, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["property_destruction"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "41470"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a systematic desensitization / toleration program for the following.\n\nSkill: tolerating medical exam\nCurriculum: AFLS Basic Living Skills\nLearner Level: Adult Learner\nBaseline: Emerging\n\nThe learner currently refuses or exhibits safety behaviors during the activity; the program should build tolerance gradually."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating medical exam\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for routine medical exam by clinician through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull routine exam (stethoscope, blood pressure cuff, otoscope, temperature) completed without refusal or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. Enter and sit in the exam room\n2. Tolerate stethoscope contact for 5 seconds\n3. Tolerate blood pressure cuff on arm\n4. Tolerate otoscope in ear\n5. Tolerate temperature reading\n6. Tolerate full routine exam\n\nCurrent shaping step: At the emerging stage, begin at shaping step 1 (\"Enter and sit in the exam room\") and advance through step 2 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated with 2 different caregivers (primary caregiver + one other) across 2 consecutive weeks.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "1973fc35ace783d7", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating medical exam", "program_type": "toleration", "learner_profile": "adult", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating medical exam", "module": "basic_living", "mastery_criterion": "mc_toleration_caregivers", "shaping_n_steps": 6}, "teacher_model": null, "seed_tag": "91439"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8917\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Echoic + Listener Responding + Mand + Tolerating Denied Access\nDate range: Sessions 1\u20139 across 18 days (2026-08-17 to 2026-09-04)\n\nACCELERATION PROGRAMS\n1. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n4. echoes 2-word phrases (Echoic)\n Method: DTT variant\n\nDECELERATION TARGETS\n1. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-17 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 10/12 correct (81%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 8/10 correct (83%); latency 2.0s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n follows 1-step motor instructions: 11/13 correct (84%); latency 2.4s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n echoes 2-word phrases: 11/15 correct (75%); latency 2.8s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d710\n Motor stereotypy: freq 2; PIR 15%\n\nSession 2 \u2014 2026-08-19 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 11/14 correct (81%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d76\n wait calmly when access to preferred item is denied: 6/8 correct (80%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n follows 1-step motor instructions: 8/11 correct (74%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n echoes 2-word phrases: 10/15 correct (64%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d710\n Motor stereotypy: freq 3; PIR 15%\n\nSession 3 \u2014 2026-08-21 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 9/14 correct (65%); latency 3.1s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 10/13 correct (78%); latency 2.5s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d74\n follows 1-step motor instructions: 8/11 correct (71%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n echoes 2-word phrases: 6/11 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d77\n Motor stereotypy: freq 4; PIR 23%\n\nSession 4 \u2014 2026-08-23 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 6/11 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 7/10 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n follows 1-step motor instructions: 6/9 correct (67%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d72\n echoes 2-word phrases: 6/11 correct (59%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d76\n Motor stereotypy: freq 4; PIR 17%\n\nSession 5 \u2014 2026-08-25 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 6/11 correct (55%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 6/10 correct (61%); latency 3.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n follows 1-step motor instructions: 9/15 correct (61%); latency 3.5s; prompts FP\u00d75, PP\u00d70, G\u00d71, I\u00d79\n echoes 2-word phrases: 7/14 correct (48%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d77\n Motor stereotypy: freq 4; PIR 17%\n\nSession 6 \u2014 2026-08-27 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 4/9 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 7/11 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d76\n follows 1-step motor instructions: 8/14 correct (55%); latency 3.3s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d75\n echoes 2-word phrases: 4/9 correct (45%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n Motor stereotypy: freq 4; PIR 23%\n\nSession 7 \u2014 2026-08-29 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 5/11 correct (44%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 8/14 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n follows 1-step motor instructions: 6/12 correct (54%); latency 3.3s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n echoes 2-word phrases: 4/11 correct (35%); latency 4.6s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d77\n Motor stereotypy: freq 6; PIR 18%\n\nSession 8 \u2014 2026-08-31 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 4/10 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n wait calmly when access to preferred item is denied: 6/12 correct (51%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d73\n follows 1-step motor instructions: 5/10 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d74\n echoes 2-word phrases: 3/9 correct (30%); latency 4.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n Motor stereotypy: freq 6; PIR 37%\n\nSession 9 \u2014 2026-09-02 \u2014 54 min \u2014 1 observer\n requesting a break ('break please' or AAC 'I need a break'): 3/11 correct (28%); latency 4.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n wait calmly when access to preferred item is denied: 4/8 correct (46%); latency 3.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n follows 1-step motor instructions: 3/8 correct (40%); latency 4.4s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n echoes 2-word phrases: 3/9 correct (31%); latency 4.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n Motor stereotypy: freq 6; PIR 24%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Elopement from the work area\n- Crying during demand presentation\n- Aggression attempt directed at staff (hit, kick, or scratch toward adult)\n- Self-stimulatory behavior increase during demand periods\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 81% to 28% across 9 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nMotor stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 54.7%.\n- Trend direction: downward (first session: 81%, last session: 28%).\n- Motor stereotypy: mean 4.3/session (early-window mean 3.0 -> late-window mean 6.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "566f84e5360e7067", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Motor stereotypy": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 9, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["motor_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "79971"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts common animals\nCurriculum Reference: Tact L1\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common animals\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What animal is this?\"\nVariations: \"What is this?\"; \"Name this animal\"\nPresentation: Present an animal picture card and ask while pointing.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: lion, fish, duck\nDistractor stimuli: chicken, monkey\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "07369fedb08ac60b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common animals", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "97982"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1903\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Social Behavior\nDate range: Sessions 1\u20137 across 14 days (2026-02-18 to 2026-03-04)\n\nACCELERATION PROGRAMS\n1. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n2. mands for missing items needed to complete an activity (Mand)\n Method: NET\n3. reads 10\u201320 sight words (Reading)\n Method: DTT\n4. takes turns during structured activities (Social Behavior)\n Method: PRT\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-18 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 2/10 correct (22%); latency 5.3s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n mands for missing items needed to complete an activity: 3/14 correct (22%); latency 5.1s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n reads 10\u201320 sight words: 5/14 correct (38%); latency 4.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d712\n takes turns during structured activities: 3/12 correct (22%); latency 5.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d79\n\nSession 2 \u2014 2026-02-20 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 4/11 correct (34%); latency 4.6s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 3/10 correct (33%); latency 4.8s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 4/9 correct (43%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n takes turns during structured activities: 4/13 correct (28%); latency 5.0s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d713\n\nSession 3 \u2014 2026-02-22 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 6/13 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 5/12 correct (42%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n reads 10\u201320 sight words: 7/13 correct (52%); latency 3.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n takes turns during structured activities: 4/11 correct (38%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d78\n\nSession 4 \u2014 2026-02-24 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 8/15 correct (54%); latency 3.6s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n mands for missing items needed to complete an activity: 5/9 correct (52%); latency 3.6s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n reads 10\u201320 sight words: 7/12 correct (59%); latency 3.1s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d74\n takes turns during structured activities: 5/10 correct (50%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n\nSession 5 \u2014 2026-02-26 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 5/8 correct (61%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 5/9 correct (59%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 7/10 correct (73%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n takes turns during structured activities: 5/9 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n\nSession 6 \u2014 2026-02-28 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 6/9 correct (71%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n mands for missing items needed to complete an activity: 8/11 correct (70%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 12/15 correct (81%); latency 2.4s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d710\n takes turns during structured activities: 7/11 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d75\n\nSession 7 \u2014 2026-03-02 \u2014 35 min \u2014 1 observer\n indicating 'all done' when an activity is complete: 12/15 correct (77%); latency 2.6s; prompts FP\u00d71, PP\u00d75, G\u00d72, I\u00d77\n mands for missing items needed to complete an activity: 7/9 correct (79%); latency 2.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n reads 10\u201320 sight words: 11/12 correct (90%); latency 1.8s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d71\n takes turns during structured activities: 12/15 correct (82%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d75, I\u00d76\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Quick response latency (within 2\u20133 seconds)\n- Eye contact maintained throughout trials\n- Smiling during reinforcement delivery\n- Reaching for materials proactively"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Rapid acquisition**: accelerated mastery from 22% to 77% across 7 sessions.\n- Acquisition has exceeded expected timeline.\n\n## Pattern Classification\nrapid_acquisition\n\nAccuracy ascended steeply beyond expected timeline for this learner's profile.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Advance to next skill in the sequence immediately.\n- Begin generalization probes across people, settings, and materials.\n\n### Consequence strategies\n- Move to maintenance schedule (probe monthly) and redirect session time to acquisition targets.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 52.0%.\n- Trend direction: upward (first session: 22%, last session: 77%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "a6f9bd2a6c0d066f", "gold_labels": {"pattern_class": "rapid_acquisition", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "rapid_acquisition", "learner_profile": "school_age", "n_sessions": 7, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "8968"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5244\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Tact + Tolerating Denied Access\nDate range: Sessions 1\u201312 across 24 days (2026-04-09 to 2026-05-03)\n\nACCELERATION PROGRAMS\n1. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n2. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n3. tacts colors of objects (Tact)\n Method: DTT\n\nDECELERATION TARGETS\n1. Fecal smearing (scatolia) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-09 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/8 correct (72%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 11/13 correct (87%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d78\n tacts colors of objects: 13/15 correct (84%); latency 2.0s; prompts FP\u00d73, PP\u00d75, G\u00d70, I\u00d77\n Fecal smearing (scatolia): attempts 2 (2 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 2 \u2014 2026-04-11 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/8 correct (78%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 7/9 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d76\n tacts colors of objects: 9/11 correct (82%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n Fecal smearing (scatolia): attempts 2 (2 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 3 \u2014 2026-04-13 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 10/14 correct (72%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d77\n answers WH-questions about familiar topics: 12/15 correct (83%); latency 2.5s; prompts FP\u00d73, PP\u00d75, G\u00d73, I\u00d74\n tacts colors of objects: 8/10 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 4 \u2014 2026-04-15 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 9/12 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d76\n answers WH-questions about familiar topics: 10/12 correct (82%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d78\n tacts colors of objects: 12/15 correct (82%); latency 2.1s; prompts FP\u00d75, PP\u00d70, G\u00d74, I\u00d76\n Fecal smearing (scatolia): attempts 2 (0 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 5 \u2014 2026-04-17 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/9 correct (41%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d72\n answers WH-questions about familiar topics: 6/13 correct (48%); latency 3.7s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d76\n tacts colors of objects: 8/15 correct (56%); latency 3.5s; prompts FP\u00d73, PP\u00d75, G\u00d74, I\u00d73\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 6 \u2014 2026-04-19 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 3/8 correct (37%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n answers WH-questions about familiar topics: 4/8 correct (53%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n tacts colors of objects: 5/10 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 7 \u2014 2026-04-21 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/15 correct (37%); latency 4.2s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d710\n answers WH-questions about familiar topics: 9/14 correct (62%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d710\n tacts colors of objects: 8/15 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d75, I\u00d77\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 8 \u2014 2026-04-23 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/10 correct (38%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n answers WH-questions about familiar topics: 10/15 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d79\n tacts colors of objects: 6/11 correct (55%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n Fecal smearing (scatolia): attempts 1 (0 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 9 \u2014 2026-04-25 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 4/9 correct (46%); latency 3.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n answers WH-questions about familiar topics: 6/8 correct (77%); latency 2.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n tacts colors of objects: 7/11 correct (61%); latency 3.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n Fecal smearing (scatolia): attempts 2 (0 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 10 \u2014 2026-04-27 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 6/13 correct (50%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n answers WH-questions about familiar topics: 11/13 correct (83%); latency 2.5s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n tacts colors of objects: 7/10 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d75\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 11 \u2014 2026-04-29 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/14 correct (59%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d714\n answers WH-questions about familiar topics: 9/11 correct (83%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n tacts colors of objects: 8/11 correct (70%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n Fecal smearing (scatolia): attempts 2 (2 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 12 \u2014 2026-05-01 \u2014 46 min \u2014 1 observer\n wait calmly when access to preferred item is denied: 8/12 correct (63%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d74\n answers WH-questions about familiar topics: 12/14 correct (83%); latency 2.5s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d75\n tacts colors of objects: 11/14 correct (77%); latency 2.3s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Playing with non-task materials\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Behavior Function Hypothesis\nFecal smearing (scatolia): automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 55.5%.\n- Trend direction: downward (first session: 72%, last session: 63%).\n- Fecal smearing (scatolia): mean 1.5/session (early-window mean 1.8 -> late-window mean 1.5).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "21b9ae2297258521", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {"Fecal smearing (scatolia)": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "early", "n_sessions": 12, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["fecal_smearing"], "behavior_functions": ["automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "64183"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4415\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading\nDate range: Sessions 1\u20138 across 16 days (2026-03-14 to 2026-03-30)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n3. requesting preferred items using a full sentence (Mand)\n Method: NET\n4. sounds out CVC words (Reading)\n Method: DTT\n5. mands for missing items needed to complete an activity (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-14 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 7/8 correct (90%); latency 1.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 11/14 correct (75%); latency 2.9s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d73\n requesting preferred items using a full sentence: 10/11 correct (94%); latency 1.9s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n sounds out CVC words: 9/10 correct (90%); latency 1.8s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 10/11 correct (88%); latency 2.0s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n\nSession 2 \u2014 2026-03-16 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 9/10 correct (92%); latency 1.9s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 9/12 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d76\n requesting preferred items using a full sentence: 8/9 correct (92%); latency 1.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n sounds out CVC words: 8/9 correct (88%); latency 2.2s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n mands for missing items needed to complete an activity: 10/11 correct (88%); latency 1.8s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d79\n\nSession 3 \u2014 2026-03-18 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 8/9 correct (90%); latency 1.8s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/10 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n requesting preferred items using a full sentence: 15/15 correct (98%); latency 1.5s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d76\n sounds out CVC words: 7/8 correct (86%); latency 1.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n mands for missing items needed to complete an activity: 13/15 correct (89%); latency 2.3s; prompts FP\u00d74, PP\u00d75, G\u00d75, I\u00d71\n\nSession 4 \u2014 2026-03-20 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 8/9 correct (88%); latency 1.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/10 correct (76%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n requesting preferred items using a full sentence: 9/9 correct (98%); latency 1.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n sounds out CVC words: 8/10 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n mands for missing items needed to complete an activity: 7/8 correct (88%); latency 1.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n\nSession 5 \u2014 2026-03-22 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 10/12 correct (87%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/9 correct (74%); latency 2.9s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d75\n requesting preferred items using a full sentence: 8/8 correct (98%); latency 1.8s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n sounds out CVC words: 10/12 correct (83%); latency 2.5s; prompts FP\u00d74, PP\u00d71, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 11/13 correct (83%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n\nSession 6 \u2014 2026-03-24 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 11/13 correct (85%); latency 2.0s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/10 correct (74%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n requesting preferred items using a full sentence: 10/10 correct (98%); latency 1.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n sounds out CVC words: 8/10 correct (80%); latency 2.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n mands for missing items needed to complete an activity: 10/12 correct (82%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n\nSession 7 \u2014 2026-03-26 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 8/9 correct (85%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 10/15 correct (66%); latency 2.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d79\n requesting preferred items using a full sentence: 14/14 correct (98%); latency 1.7s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d79\n sounds out CVC words: 6/8 correct (80%); latency 2.2s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n mands for missing items needed to complete an activity: 10/12 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d79\n\nSession 8 \u2014 2026-03-28 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 7/8 correct (88%); latency 1.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 8/11 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d78\n requesting preferred items using a full sentence: 13/13 correct (98%); latency 1.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d711\n sounds out CVC words: 10/12 correct (80%); latency 2.7s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d74\n mands for missing items needed to complete an activity: 9/10 correct (86%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Leaving seat without permission\n- Stereotypic behavior during inter-trial intervals\n- Looking away during SD presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (90%\u201388%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 88.1%.\n- Trend direction: downward (first session: 90%, last session: 88%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "b7835a0ad551ec9b", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "71251"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: making the bed\nCurriculum: AFLS Home Skills\nLearner Level: Adolescent Learner\nBaseline: Mastered at current level\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: making the bed\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Strip any disheveled linens and lay them flat. <- Learner-taught step (forward chaining starts here)\n2. Straighten the bottom sheet across the mattress.\n3. Place the top sheet evenly over the mattress.\n4. Tuck the top sheet in at the foot of the bed.\n5. Place the comforter over the top sheet.\n6. Place pillows at the head of the bed.\n7. Smooth the comforter surface.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8e307f2d19f6191c", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "making the bed", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "making the bed", "module": "home", "mastery_criterion": "mc_all_steps_independent", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "6490"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8866\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Reading + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-02-09 to 2026-03-05)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n3. sounds out CVC words (Reading)\n Method: DTT\n4. single-word requests for preferred items (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: attention\n2. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-09 \u2014 59 min \u2014 1 observer\n copies simple words from model: 7/10 correct (72%); latency 2.9s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d74\n indicating 'all done' when an activity is complete: 10/12 correct (79%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d75\n sounds out CVC words: 10/13 correct (75%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n single-word requests for preferred items: 12/15 correct (81%); latency 2.3s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d78\n Self-injurious behavior (SIB): freq 2\n Non-compliance: freq 3\n\nSession 2 \u2014 2026-02-11 \u2014 59 min \u2014 1 observer\n copies simple words from model: 9/14 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n indicating 'all done' when an activity is complete: 7/10 correct (72%); latency 2.5s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 8/12 correct (66%); latency 2.9s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n single-word requests for preferred items: 8/11 correct (70%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d75\n Self-injurious behavior (SIB): freq 1\n Non-compliance: freq 4\n\nSession 3 \u2014 2026-02-13 \u2014 59 min \u2014 1 observer\n copies simple words from model: 7/11 correct (65%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d79\n indicating 'all done' when an activity is complete: 6/8 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n sounds out CVC words: 5/9 correct (59%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n single-word requests for preferred items: 11/15 correct (70%); latency 2.9s; prompts FP\u00d75, PP\u00d75, G\u00d71, I\u00d74\n Self-injurious behavior (SIB): freq 3\n Non-compliance: freq 5\n\nSession 4 \u2014 2026-02-15 \u2014 59 min \u2014 1 observer\n copies simple words from model: 7/13 correct (56%); latency 3.4s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n indicating 'all done' when an activity is complete: 8/12 correct (63%); latency 2.9s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d74\n sounds out CVC words: 5/9 correct (57%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n single-word requests for preferred items: 7/12 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 2\n Non-compliance: freq 5\n\nSession 5 \u2014 2026-02-17 \u2014 59 min \u2014 1 observer\n copies simple words from model: 5/9 correct (53%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n indicating 'all done' when an activity is complete: 7/12 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n sounds out CVC words: 4/10 correct (45%); latency 4.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n single-word requests for preferred items: 9/15 correct (57%); latency 3.7s; prompts FP\u00d75, PP\u00d75, G\u00d74, I\u00d71\n Self-injurious behavior (SIB): freq 3\n Non-compliance: freq 6\n\nSession 6 \u2014 2026-02-19 \u2014 59 min \u2014 1 observer\n copies simple words from model: 5/10 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n indicating 'all done' when an activity is complete: 4/8 correct (48%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n sounds out CVC words: 4/10 correct (42%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n single-word requests for preferred items: 4/8 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n Self-injurious behavior (SIB): freq 5\n Non-compliance: freq 6\n\nSession 7 \u2014 2026-02-21 \u2014 59 min \u2014 1 observer\n copies simple words from model: 6/14 correct (45%); latency 3.8s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n indicating 'all done' when an activity is complete: 5/13 correct (39%); latency 4.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d77\n sounds out CVC words: 5/15 correct (33%); latency 4.7s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d710\n single-word requests for preferred items: 6/13 correct (46%); latency 3.9s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d74\n Self-injurious behavior (SIB): freq 4\n Non-compliance: freq 5\n\nSession 8 \u2014 2026-02-23 \u2014 59 min \u2014 1 observer\n copies simple words from model: 3/9 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d74\n indicating 'all done' when an activity is complete: 4/12 correct (36%); latency 4.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d79\n sounds out CVC words: 3/11 correct (28%); latency 4.8s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n single-word requests for preferred items: 4/10 correct (43%); latency 4.2s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n Self-injurious behavior (SIB): freq 5\n Non-compliance: freq 7\n\nSession 9 \u2014 2026-02-25 \u2014 59 min \u2014 1 observer\n copies simple words from model: 4/11 correct (33%); latency 4.8s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n indicating 'all done' when an activity is complete: 4/14 correct (31%); latency 4.8s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d77\n sounds out CVC words: 3/14 correct (23%); latency 5.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d76\n single-word requests for preferred items: 4/11 correct (38%); latency 4.5s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n Self-injurious behavior (SIB): freq 6\n Non-compliance: freq 7\n\nSession 10 \u2014 2026-02-27 \u2014 59 min \u2014 1 observer\n copies simple words from model: 2/8 correct (22%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 2/11 correct (20%); latency 5.3s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n sounds out CVC words: 2/14 correct (14%); latency 5.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d713\n single-word requests for preferred items: 6/14 correct (41%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d711\n Self-injurious behavior (SIB): freq 7\n Non-compliance: freq 7\n\nSession 11 \u2014 2026-03-01 \u2014 59 min \u2014 1 observer\n copies simple words from model: 2/9 correct (24%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n indicating 'all done' when an activity is complete: 2/13 correct (16%); latency 5.5s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n sounds out CVC words: 1/12 correct (10%); latency 5.8s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d75\n single-word requests for preferred items: 3/9 correct (32%); latency 4.7s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 7\n Non-compliance: freq 9\n\nSession 12 \u2014 2026-03-03 \u2014 59 min \u2014 1 observer\n copies simple words from model: 2/13 correct (14%); latency 5.1s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d74\n indicating 'all done' when an activity is complete: 1/9 correct (10%); latency 5.8s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n sounds out CVC words: 1/12 correct (10%); latency 5.8s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d74\n single-word requests for preferred items: 3/11 correct (23%); latency 4.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 8\n Non-compliance: freq 10\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Pushing materials away from the work area\n- Increased response latency (3\u20135\u00d7 baseline)\n- Self-stimulatory behavior increase during demand periods\n- Vocal refusal ('no', 'I don't want to')"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (72% -> 14%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 44.7%.\n- Trend direction: downward (first session: 72%, last session: 14%).\n- Self-injurious behavior (SIB): mean 4.4/session (early-window mean 2.0 -> late-window mean 7.0).\n- Non-compliance: mean 6.2/session (early-window mean 4.2 -> late-window mean 8.2).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "9e7c17d4b50815fb", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Self-injurious behavior (SIB)": "attention", "Non-compliance": "escape"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "school_age", "n_sessions": 12, "n_programs": 4, "n_behaviors": 2, "behavior_ids": ["sib", "non_compliance"], "behavior_functions": ["attention", "escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "10903"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: shows items to others spontaneously\nCurriculum Reference: Social Behavior & Social Play L1\nLearner Profile: Early Learner\nCurrent Mastery: Near mastery\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: shows items to others spontaneously\n- Curriculum: Social Behavior & Social Play L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: shared snack.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's shared snack routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: peer engages back; activity continues with both.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "1d6145bfe6bf7681", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "shows items to others spontaneously", "mo_category": "social_initiation", "mo_arrangement": "peer_presence", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "40151"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tolerating haircut\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Emerging\nTeaching Method: Toleration / Systematic Desensitization\n\nThe goal is for the learner to allow the activity to be performed (not to perform it independently). Provide the full program structure including shaping steps and refusal/safety procedures."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating haircut\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for haircut by caregiver or stylist through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nfull haircut (10+ minutes) completed without refusal, leaving the chair, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. Sit in the chair with cape on for 30 seconds\n2. Tolerate 1 minute of quiet sitting with cape\n3. Tolerate 2 minutes of brief clipping\n4. Tolerate 5 minutes of clipping\n5. Tolerate 10 minutes of clipping\n6. Tolerate full haircut routine\n\nCurrent shaping step: At the emerging stage, begin at shaping step 1 (\"Sit in the chair with cape on for 30 seconds\") and advance through step 2 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 3 different settings and with 2 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "fafefbd5fea085f6", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating haircut", "program_type": "toleration", "learner_profile": "adult", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating haircut", "module": "basic_living", "mastery_criterion": "mc_toleration_settings", "shaping_n_steps": 6}, "teacher_model": null, "seed_tag": "21081"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: mands for others to stop an action\nCurriculum: Mand L2\nLearner Level: School-Age Learner\nBaseline: Near mastery\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for others to stop an action\n- Curriculum: Mand L2\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: free play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's free play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: performance of the requested action.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "615bd3e8494db3d8", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L2", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for others to stop an action", "mo_category": "mand_action", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "87353"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: single-word mands for preferred items\nCurriculum: Mand L1\nLearner Level: Early Learner\nBaseline: Generalization phase\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: single-word mands for preferred items\n- Curriculum: Mand L1\n- Learner: Early Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: leisure time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "4998a34220fe2a6f", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L1", "learner_profile": "early", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "single-word mands for preferred items", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "time_delay", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "30191"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tracks print left to right\nCurriculum: Reading L1\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tracks print left to right\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter or word on a flashcard.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: phone, shoe, block\nDistractor stimuli: jacket, car\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c79cba32b4d039a4", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tracks print left to right", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "four_step", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "56565"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tells time to the hour and half-hour\nCurriculum: Math L3\nLearner Level: School-Age Learner\nBaseline: Developing\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tells time to the hour and half-hour\n- Curriculum: Math L3\n- Learner: School-Age Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How many?\"\nVariations: \"Count these\"; \"What number is this?\"\nPresentation: Present the numeral card or a set of objects to count.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: scissors, plate, ball\nDistractor stimuli: jacket, brush, toothbrush, glue, cup\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "70eb9095f277047e", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.math", "level": "L3", "learner_profile": "school_age", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tells time to the hour and half-hour", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "transfer_trial", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "97982"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4235\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Tolerating Denied Access + Writing\nDate range: Sessions 1\u201312 across 24 days (2026-09-01 to 2026-09-25)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. reads 10\u201320 sight words (Reading)\n Method: DTT\n3. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n4. copies simple words from model (Writing)\n Method: DTT\n5. follows 1-step motor instructions (Listener Responding)\n Method: DTT\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: automatic\n2. Vocal stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-09-01 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 6/8 correct (72%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 7/9 correct (78%); latency 2.5s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d75\n copies simple words from model: 12/14 correct (84%); latency 2.5s; prompts FP\u00d73, PP\u00d74, G\u00d74, I\u00d73\n follows 1-step motor instructions: 10/13 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 2; PIR 17%\n\nSession 2 \u2014 2026-09-03 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (69%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 5/8 correct (64%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 10/14 correct (71%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d711\n copies simple words from model: 10/13 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n follows 1-step motor instructions: 10/14 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n Self-injurious behavior (SIB): freq 2\n Vocal stereotypy: freq 2; PIR 15%\n\nSession 3 \u2014 2026-09-05 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/13 correct (65%); latency 3.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n reads 10\u201320 sight words: 7/12 correct (59%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 9/15 correct (61%); latency 3.4s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d75\n copies simple words from model: 10/14 correct (69%); latency 3.1s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d78\n follows 1-step motor instructions: 7/10 correct (66%); latency 3.0s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d77\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 4; PIR 29%\n\nSession 4 \u2014 2026-09-07 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/13 correct (55%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d711\n reads 10\u201320 sight words: 6/12 correct (52%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 5/8 correct (59%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n copies simple words from model: 7/12 correct (60%); latency 3.3s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n follows 1-step motor instructions: 7/13 correct (53%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d77\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 4; PIR 26%\n\nSession 5 \u2014 2026-09-09 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/9 correct (49%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n reads 10\u201320 sight words: 4/11 correct (40%); latency 4.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 7/13 correct (51%); latency 3.6s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n copies simple words from model: 9/15 correct (60%); latency 3.2s; prompts FP\u00d73, PP\u00d75, G\u00d71, I\u00d76\n follows 1-step motor instructions: 5/10 correct (51%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 3; PIR 16%\n\nSession 6 \u2014 2026-09-11 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 3/8 correct (32%); latency 4.8s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n wait calmly when access to preferred item is denied: 5/12 correct (44%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n copies simple words from model: 8/15 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d75, I\u00d77\n follows 1-step motor instructions: 4/9 correct (43%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 4; PIR 16%\n\nSession 7 \u2014 2026-09-13 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/11 correct (44%); latency 4.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d77\n reads 10\u201320 sight words: 3/14 correct (23%); latency 5.2s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n wait calmly when access to preferred item is denied: 4/11 correct (35%); latency 4.2s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n copies simple words from model: 4/9 correct (45%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d73\n follows 1-step motor instructions: 3/10 correct (34%); latency 4.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 4; PIR 20%\n\nSession 8 \u2014 2026-09-15 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/9 correct (38%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n reads 10\u201320 sight words: 3/13 correct (20%); latency 4.9s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n wait calmly when access to preferred item is denied: 3/9 correct (30%); latency 4.8s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d74\n copies simple words from model: 5/12 correct (45%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n follows 1-step motor instructions: 4/15 correct (24%); latency 4.7s; prompts FP\u00d75, PP\u00d75, G\u00d74, I\u00d71\n Self-injurious behavior (SIB): freq 3\n Vocal stereotypy: freq 6; PIR 23%\n\nSession 9 \u2014 2026-09-17 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/15 correct (23%); latency 5.2s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d710\n reads 10\u201320 sight words: 1/10 correct (11%); latency 5.4s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n wait calmly when access to preferred item is denied: 3/13 correct (23%); latency 4.9s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n copies simple words from model: 5/13 correct (35%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n follows 1-step motor instructions: 1/9 correct (15%); latency 5.1s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n Self-injurious behavior (SIB): freq 5\n Vocal stereotypy: freq 5; PIR 35%\n\nSession 10 \u2014 2026-09-19 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 2/10 correct (19%); latency 5.1s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n reads 10\u201320 sight words: 1/9 correct (10%); latency 5.6s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n wait calmly when access to preferred item is denied: 2/8 correct (22%); latency 4.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n copies simple words from model: 2/8 correct (28%); latency 4.4s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n follows 1-step motor instructions: 2/14 correct (12%); latency 5.7s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d76\n Self-injurious behavior (SIB): freq 5\n Vocal stereotypy: freq 6; PIR 28%\n\nSession 11 \u2014 2026-09-21 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 1/9 correct (11%); latency 5.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d77\n reads 10\u201320 sight words: 1/8 correct (10%); latency 5.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 1/10 correct (14%); latency 5.3s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n copies simple words from model: 2/14 correct (17%); latency 5.1s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d78\n follows 1-step motor instructions: 1/8 correct (10%); latency 5.4s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d73\n Self-injurious behavior (SIB): freq 4\n Vocal stereotypy: freq 6; PIR 32%\n\nSession 12 \u2014 2026-09-23 \u2014 59 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 1/9 correct (10%); latency 5.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n reads 10\u201320 sight words: 1/10 correct (10%); latency 5.5s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 1/14 correct (10%); latency 5.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d78\n copies simple words from model: 1/8 correct (18%); latency 5.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n follows 1-step motor instructions: 1/12 correct (10%); latency 5.3s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d73\n Self-injurious behavior (SIB): freq 5\n Vocal stereotypy: freq 7; PIR 38%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Pushing materials away from the work area\n- Self-stimulatory behavior increase during demand periods\n- Crying during demand presentation\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 75% to 10% across 12 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): automatic\n Evidence: behavior persists across contexts without clear social mediation.\nVocal stereotypy: automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 42.6%.\n- Trend direction: downward (first session: 75%, last session: 10%).\n- Self-injurious behavior (SIB): mean 3.5/session (early-window mean 2.8 -> late-window mean 4.8).\n- Vocal stereotypy: mean 4.4/session (early-window mean 3.0 -> late-window mean 6.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "d0cb68435060c481", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Self-injurious behavior (SIB)": "automatic", "Vocal stereotypy": "automatic"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "early", "n_sessions": 12, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["sib", "vocal_stereotypy"], "behavior_functions": ["automatic", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "78703"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects items by comparison (which is heavier)\nCurriculum: LRFFC L3\nLearner Level: School-Age Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by comparison (which is heavier)\n- Curriculum: LRFFC L3\n- Learner: School-Age Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: crayon, blanket, scissors\nDistractor stimuli: backpack, apple, napkin, brush, phone\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a76cec26df813a0c", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L3", "learner_profile": "school_age", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by comparison (which is heavier)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "vr3", "error_correction": "four_step", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "92219"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: taking medication as prescribed\nCurriculum: AFLS Independent Living Skills\nLearner Level: Adult Learner\nBaseline: Developing\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: taking medication as prescribed\n- Curriculum: AFLS Independent Living Skills\n- Learner: Adult Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Check the medication schedule at the scheduled time.\n2. Collect the correct medication bottle.\n3. Confirm the dose against the label.\n4. Measure or count the correct dose.\n5. Take the medication with water.\n6. Log the dose in the medication tracker.\n7. Return the bottle to its storage location.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the developing stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\n100% of steps completed independently (no prompts) across 3 consecutive trials of the full chain.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "02f428ff724f0602", "gold_labels": {"method": "task_analysis", "domain": "AFLS.independent_living", "skill": "taking medication as prescribed", "program_type": "independence", "learner_profile": "adult", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "taking medication as prescribed", "module": "independent_living", "mastery_criterion": "mc_all_steps_independent", "chain_type": "total_task", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "51834"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts familiar people by name\nCurriculum Reference: Tact L1\nLearner Profile: Early Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts familiar people by name\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: book, napkin, key\nDistractor stimuli: brush, backpack\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e4811832d91467b7", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts familiar people by name", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "35263"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: provides definitions of words\nCurriculum: Intraverbal L3\nLearner: Adolescent Learner\nMastery Status: Mastered at current level\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: provides definitions of words\n- Curriculum: Intraverbal L3\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nEngage the learner in a naturalistic conversation on a topic of demonstrated interest. Pause at natural turn boundaries to invite the target response.\nPrimary natural context: post-activity debrief.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's post-activity debrief routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: conversation continues with adult's naturally matched reply.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 3 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nFluent exchange (response within 3 s of MO / opportunity) with appropriate topography, 80% of opportunities across 3 sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "059cad01b1b22788", "gold_labels": {"method": "net", "domain": "VB-MAPP.intraverbal", "level": "L3", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "provides definitions of words", "mo_category": "intraverbal_multi_turn", "mo_arrangement": "reciprocal_conversation", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_fluent_exchange"}, "teacher_model": null, "seed_tag": "44373"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6454\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Mand + Vocational\nDate range: Sessions 1\u201312 across 24 days (2026-02-03 to 2026-02-27)\n\nACCELERATION PROGRAMS\n1. requesting assistance using AAC device (Mand)\n Method: NET\n2. simple meal preparation (Home Skills)\n Method: Task Analysis\n3. clocking in at start of shift (Vocational)\n Method: Task Analysis\n4. doing laundry (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-03 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 10/14 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n simple meal preparation: 9/13 correct (69%); latency 2.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d78\n clocking in at start of shift: 8/9 correct (86%); latency 1.9s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d75\n doing laundry: 11/15 correct (72%); latency 2.5s; prompts FP\u00d75, PP\u00d72, G\u00d73, I\u00d75\n\nSession 2 \u2014 2026-02-05 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 11/15 correct (71%); latency 2.8s; prompts FP\u00d75, PP\u00d75, G\u00d74, I\u00d71\n simple meal preparation: 10/14 correct (68%); latency 3.1s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n clocking in at start of shift: 11/13 correct (81%); latency 2.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d79\n doing laundry: 7/10 correct (71%); latency 2.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n\nSession 3 \u2014 2026-02-07 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 6/8 correct (74%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n simple meal preparation: 8/11 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n clocking in at start of shift: 13/15 correct (88%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d712\n doing laundry: 10/15 correct (69%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d79\n\nSession 4 \u2014 2026-02-09 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 11/15 correct (73%); latency 2.6s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d79\n simple meal preparation: 10/14 correct (72%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n clocking in at start of shift: 12/14 correct (85%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d78\n doing laundry: 10/13 correct (75%); latency 2.8s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d73\n\nSession 5 \u2014 2026-02-11 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 5/13 correct (36%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n simple meal preparation: 3/9 correct (37%); latency 4.5s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d73\n clocking in at start of shift: 7/12 correct (57%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n doing laundry: 5/12 correct (38%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n\nSession 6 \u2014 2026-02-13 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 3/11 correct (32%); latency 4.8s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n simple meal preparation: 3/9 correct (38%); latency 4.1s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n clocking in at start of shift: 6/11 correct (52%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n doing laundry: 4/12 correct (32%); latency 4.6s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d78\n\nSession 7 \u2014 2026-02-15 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 6/15 correct (37%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d710\n simple meal preparation: 5/13 correct (38%); latency 4.1s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d78\n clocking in at start of shift: 5/9 correct (53%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n doing laundry: 3/9 correct (34%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n\nSession 8 \u2014 2026-02-17 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 5/13 correct (38%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n simple meal preparation: 5/14 correct (35%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d713\n clocking in at start of shift: 7/14 correct (52%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n doing laundry: 4/9 correct (44%); latency 4.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n\nSession 9 \u2014 2026-02-19 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 4/9 correct (45%); latency 3.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n simple meal preparation: 6/15 correct (38%); latency 4.3s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n clocking in at start of shift: 5/9 correct (58%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n doing laundry: 6/12 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n\nSession 10 \u2014 2026-02-21 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 5/10 correct (52%); latency 3.6s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n simple meal preparation: 6/13 correct (45%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d76\n clocking in at start of shift: 9/15 correct (62%); latency 3.5s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d75\n doing laundry: 7/11 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d74\n\nSession 11 \u2014 2026-02-23 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 5/9 correct (57%); latency 3.7s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d74\n simple meal preparation: 5/10 correct (53%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d74\n clocking in at start of shift: 6/9 correct (68%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d70\n doing laundry: 5/8 correct (67%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n\nSession 12 \u2014 2026-02-25 \u2014 31 min \u2014 1 observer\n requesting assistance using AAC device: 10/15 correct (68%); latency 2.9s; prompts FP\u00d74, PP\u00d75, G\u00d70, I\u00d76\n simple meal preparation: 6/11 correct (55%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d78\n clocking in at start of shift: 12/15 correct (78%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d711\n doing laundry: 6/9 correct (69%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Stereotypic behavior during inter-trial intervals\n- Delayed responding (>5 seconds consistently)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 54.7%.\n- Trend direction: downward (first session: 74%, last session: 68%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "16794a03bbcb4e7a", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "adult", "n_sessions": 12, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "38361"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4411\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Mand + Reading + Writing\nDate range: Sessions 1\u201310 across 20 days (2026-10-28 to 2026-11-17)\n\nACCELERATION PROGRAMS\n1. copies simple words from model (Writing)\n Method: DTT\n2. sounds out CVC words (Reading)\n Method: DTT\n3. requesting a break ('break please' or AAC 'I need a break') (Mand)\n Method: NET\n4. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n5. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Mouthing \u2014 function hypothesized: automatic\n2. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-28 \u2014 50 min \u2014 1 observer\n copies simple words from model: 4/14 correct (31%); latency 4.5s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d77\n sounds out CVC words: 7/14 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (62%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/12 correct (62%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/10 correct (29%); latency 4.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n Mouthing: freq 2; PIR 14%\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = end of 90-minute block without scheduled toileting; B = soiled (bowel) at the work table; C = staff initiated clean-up and hygiene routine; program paused\n\nSession 2 \u2014 2026-10-30 \u2014 50 min \u2014 1 observer\n copies simple words from model: 5/8 correct (66%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n sounds out CVC words: 6/13 correct (42%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n requesting a break ('break please' or AAC 'I need a break'): 9/12 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/10 correct (27%); latency 4.7s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/12 correct (57%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d712\n Mouthing: freq 2; PIR 14%\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 3 \u2014 2026-11-01 \u2014 50 min \u2014 1 observer\n copies simple words from model: 5/9 correct (54%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n sounds out CVC words: 4/11 correct (39%); latency 4.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (65%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/10 correct (34%); latency 4.5s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/9 correct (42%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n Mouthing: freq 1; PIR 11%\n Toileting accident (urine or bowel): urine: 5 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 4 \u2014 2026-11-03 \u2014 50 min \u2014 1 observer\n copies simple words from model: 5/12 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d74\n sounds out CVC words: 4/14 correct (31%); latency 4.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d711\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (63%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/13 correct (49%); latency 3.8s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/11 correct (54%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d76\n Mouthing: freq 5; PIR 20%\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = during group activity without requesting bathroom; B = voided (urine) in seat; C = staff removed learner from group, changed clothing, re-entered after clean-up\n\nSession 5 \u2014 2026-11-05 \u2014 50 min \u2014 1 observer\n copies simple words from model: 5/9 correct (54%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 7/12 correct (55%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d79\n requesting a break ('break please' or AAC 'I need a break'): 6/9 correct (67%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (60%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (49%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n Mouthing: freq 5; PIR 20%\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 6 \u2014 2026-11-07 \u2014 50 min \u2014 1 observer\n copies simple words from model: 4/10 correct (41%); latency 4.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d74\n sounds out CVC words: 5/14 correct (38%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n requesting a break ('break please' or AAC 'I need a break'): 8/14 correct (56%); latency 3.6s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/10 correct (32%); latency 4.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (49%); latency 3.9s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d76\n Mouthing: freq 1; PIR 5%\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 7 \u2014 2026-11-09 \u2014 50 min \u2014 1 observer\n copies simple words from model: 5/10 correct (50%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n sounds out CVC words: 5/11 correct (45%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n requesting a break ('break please' or AAC 'I need a break'): 6/8 correct (75%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/8 correct (52%); latency 3.4s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/12 correct (42%); latency 3.9s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d75\n Mouthing: freq 2; PIR 8%\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 1 in-toilet / 1 accidents\n\nSession 8 \u2014 2026-11-11 \u2014 50 min \u2014 1 observer\n copies simple words from model: 7/12 correct (57%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n sounds out CVC words: 6/11 correct (55%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d77\n requesting a break ('break please' or AAC 'I need a break'): 6/8 correct (75%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/11 correct (40%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/12 correct (26%); latency 5.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n Mouthing: freq 4; PIR 20%\n Toileting accident (urine or bowel): urine: 1 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n ABC (mouthing): A = waiting in line transition; B = mouthed own hand; C = staff redirected to task materials\n\nSession 9 \u2014 2026-11-13 \u2014 50 min \u2014 1 observer\n copies simple words from model: 3/8 correct (31%); latency 4.9s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n sounds out CVC words: 3/9 correct (34%); latency 4.8s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n requesting a break ('break please' or AAC 'I need a break'): 7/13 correct (50%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 5/15 correct (36%); latency 4.6s; prompts FP\u00d73, PP\u00d73, G\u00d75, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/13 correct (35%); latency 4.2s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n Mouthing: freq 4; PIR 22%\n Toileting accident (urine or bowel): urine: 2 in-toilet / 3 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 10 \u2014 2026-11-15 \u2014 50 min \u2014 1 observer\n copies simple words from model: 7/15 correct (50%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d78\n sounds out CVC words: 8/13 correct (58%); latency 3.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d710\n requesting a break ('break please' or AAC 'I need a break'): 6/11 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/9 correct (37%); latency 4.3s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/13 correct (34%); latency 4.2s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d71\n Mouthing: freq 5; PIR 16%\n Toileting accident (urine or bowel): urine: 4 in-toilet / 1 accidents; BM: 0 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Variable performance**: accuracy fluctuating around 41% with no clear trend.\n- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.\n\n## Pattern Classification\nvariable_performance\n\nHigh session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.\n\n## Behavior Function Hypothesis\nMouthing: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\nToileting accident (urine or bowel): escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Check for environmental variables that differ across sessions (time of day, therapist, setting).\n- Standardize session structure and reinforcement delivery across therapists.\n\n### Consequence strategies\n- Implement inter-observer agreement probes to rule out measurement inconsistency.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 47.6%.\n- Trend direction: upward (first session: 31%, last session: 50%).\n- Mouthing: mean 3.1/session (early-window mean 1.7 -> late-window mean 4.3).\n- Toileting accident (urine or bowel): mean 3.3/session (early-window mean 3.3 -> late-window mean 3.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "fad6fa9d74b04496", "gold_labels": {"pattern_class": "variable_performance", "behavior_functions": {"Mouthing": "automatic", "Toileting accident (urine or bowel)": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "variable_performance", "learner_profile": "early", "n_sessions": 10, "n_programs": 5, "n_behaviors": 2, "behavior_ids": ["mouthing", "toileting_accident"], "behavior_functions": ["automatic", "escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "77020"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: writes own name independently\nCurriculum: Writing L2\nLearner: Early Learner\nMastery Status: Developing\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: writes own name independently\n- Curriculum: Writing L2\n- Learner: Early Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: book, pencil, glue\nDistractor stimuli: hat, phone, sock, backpack\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "2d062b7631ea4845", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L2", "learner_profile": "early", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "writes own name independently", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "38800"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: doing laundry\nCurriculum: AFLS Home Skills\nLearner: Adolescent Learner\nMastery Status: Developing\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: doing laundry\n- Curriculum: AFLS Home Skills\n- Learner: Adolescent Learner\n- Current mastery: Developing\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Sort laundry by color (lights / darks / whites).\n2. Load the washing machine with one color group.\n3. Add detergent according to instructions.\n4. Select the correct wash cycle and start the machine.\n5. When complete, move clothes to the dryer.\n6. Start the dryer on the correct cycle.\n7. When dry, fold clothes by category.\n8. Put folded clothes away in drawers or closet.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the developing stage, use full-physical or partial-physical prompts at each step; fade systematically.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "172352ddd62122ac", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "doing laundry", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "developing", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "doing laundry", "module": "home", "mastery_criterion": "mc_generalization", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "50630"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts shapes (circle, square, triangle)\nCurriculum Reference: Tact L2\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts shapes (circle, square, triangle)\n- Curriculum: Tact L2\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What shape?\"\nVariations: \"What shape is this?\"; \"Tell me the shape\"\nPresentation: Present a clearly-shaped object or shape card and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: square, rectangle, diamond\nDistractor stimuli: triangle, oval, circle, heart\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "ee08327a827f46a9", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts shapes (circle, square, triangle)", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "68189"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: brushing teeth\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: brushing teeth\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adult Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Collect toothbrush and toothpaste.\n2. Wet the toothbrush under water.\n3. Apply a pea-sized amount of toothpaste.\n4. Brush upper-front teeth in small circles for 30 seconds.\n5. Brush lower-front teeth in small circles for 30 seconds.\n6. Brush upper-back teeth (left and right).\n7. Brush lower-back teeth (left and right).\n8. Spit into the sink and rinse mouth.\n9. Rinse the toothbrush and return it to the holder.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nDifferential per-step + terminal \u2014 Small per-step reinforcement for independent responses; larger reinforcer at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e41778fb1ddb4a77", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "brushing teeth", "program_type": "independence", "learner_profile": "adult", "mastery_state": "approaching", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "brushing teeth", "module": "basic_living", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "most_to_least", "reinforcement_schedule": "differential_per_step", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "51834"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9750\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-08-10 to 2026-08-30)\n\nACCELERATION PROGRAMS\n1. identifies numerals 1\u201320 (Math)\n Method: DTT\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. reads 10\u201320 sight words (Reading)\n Method: DTT\n4. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n5. requesting preferred items using AAC device (e.g., \"I want cookie please\") (Mand)\n Method: NET\n6. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-10 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 5/8 correct (58%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d74\n single-word requests for preferred items: 8/15 correct (51%); latency 3.5s; prompts FP\u00d75, PP\u00d75, G\u00d75, I\u00d70\n reads 10\u201320 sight words: 7/11 correct (60%); latency 3.5s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/10 correct (53%); latency 3.8s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/11 correct (64%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 9/15 correct (59%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d75, I\u00d710\n\nSession 2 \u2014 2026-08-12 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 6/13 correct (49%); latency 3.9s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n single-word requests for preferred items: 5/10 correct (50%); latency 4.0s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 6/11 correct (59%); latency 3.5s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/15 correct (57%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d712\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/13 correct (56%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n wait calmly when access to preferred item is denied: 5/8 correct (62%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n\nSession 3 \u2014 2026-08-14 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 9/15 correct (58%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d78\n single-word requests for preferred items: 9/15 correct (57%); latency 3.7s; prompts FP\u00d73, PP\u00d75, G\u00d75, I\u00d72\n reads 10\u201320 sight words: 6/10 correct (57%); latency 3.3s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/13 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/13 correct (48%); latency 3.8s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 7/12 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d75\n\nSession 4 \u2014 2026-08-16 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 5/9 correct (50%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d78\n single-word requests for preferred items: 8/14 correct (56%); latency 3.8s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n reads 10\u201320 sight words: 6/11 correct (51%); latency 3.9s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (57%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/8 correct (50%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 5/8 correct (58%); latency 3.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n\nSession 5 \u2014 2026-08-18 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 6/10 correct (58%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n single-word requests for preferred items: 5/9 correct (58%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n reads 10\u201320 sight words: 8/13 correct (60%); latency 3.3s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 8/14 correct (54%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 7/14 correct (50%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d712\n wait calmly when access to preferred item is denied: 10/15 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d74, G\u00d75, I\u00d76\n\nSession 6 \u2014 2026-08-20 \u2014 IOA SESSION \u2014 2 observers\n identifies numerals 1\u201320: 63% (primary); IOA 97% agreement\n single-word requests for preferred items: 57% (primary); IOA 97% agreement\n reads 10\u201320 sight words: 63% (primary); IOA 97% agreement\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 54% (primary); IOA 97% agreement\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 54% (primary); IOA 97% agreement\n wait calmly when access to preferred item is denied: 65% (primary); IOA 97% agreement\n\nSession 7 \u2014 2026-08-22 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 6/10 correct (56%); latency 3.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n single-word requests for preferred items: 8/15 correct (55%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d715\n reads 10\u201320 sight words: 9/15 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d711\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/15 correct (60%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d79\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/9 correct (44%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n wait calmly when access to preferred item is denied: 6/9 correct (66%); latency 3.2s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n\nSession 8 \u2014 2026-08-24 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 10/14 correct (71%); latency 2.7s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d78\n single-word requests for preferred items: 5/9 correct (54%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d73\n reads 10\u201320 sight words: 7/12 correct (61%); latency 3.5s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d72\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (55%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/14 correct (46%); latency 3.9s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 7/13 correct (57%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d712\n\nSession 9 \u2014 2026-08-26 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 9/15 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d75, G\u00d74, I\u00d75\n single-word requests for preferred items: 8/13 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n reads 10\u201320 sight words: 6/11 correct (58%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/13 correct (68%); latency 2.7s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d75\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 4/10 correct (44%); latency 3.9s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n wait calmly when access to preferred item is denied: 10/15 correct (68%); latency 2.7s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d78\n\nSession 10 \u2014 2026-08-28 \u2014 32 min \u2014 1 observer\n identifies numerals 1\u201320: 7/12 correct (61%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n single-word requests for preferred items: 7/13 correct (56%); latency 3.7s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n reads 10\u201320 sight words: 6/12 correct (51%); latency 3.5s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/10 correct (66%); latency 2.9s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n requesting preferred items using AAC device (e.g., \"I want cookie please\"): 6/15 correct (40%); latency 4.3s; prompts FP\u00d73, PP\u00d74, G\u00d75, I\u00d73\n wait calmly when access to preferred item is denied: 7/9 correct (73%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Plateau**: accuracy has remained near 60% for 10 sessions without meaningful improvement.\n- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.\n\n## Pattern Classification\nplateau\n\nResponding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Modify teaching procedure \u2014 switch prompt type (e.g., from most-to-least to time delay).\n- Conduct preference assessment and update reinforcer menu.\n- Break the current skill into smaller component steps if possible.\n- Increase practice opportunities by embedding targets in natural-environment training.\n\n### Consequence strategies\n- Check for inadvertent reinforcement of near-misses; tighten differential reinforcement.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nhigh\n\nPattern is well-supported: 10 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 58.7%.\n- Trend direction: upward (first session: 58%, last session: 61%).\n- IOA data: 97% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "e82727bf5ad9ab01", "gold_labels": {"pattern_class": "plateau", "behavior_functions": {}, "escalation_level": 2, "confidence": "high", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "plateau", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "67658"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: using public transportation\nCurriculum: AFLS Community Participation Skills\nLearner: Adult Learner\nMastery Status: Emerging\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: using public transportation\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Emerging\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Confirm the correct stop or station. <- Learner-taught step (forward chaining starts here)\n2. Purchase or have the fare ready.\n3. Wait at the designated boarding area.\n4. Board the vehicle when it arrives.\n5. Pay the fare or present the pass.\n6. Sit or stand safely.\n7. Monitor the route for the correct stop.\n8. Exit at the correct stop.\n\n## Prompt Strategy Per Step\nStrategy: Graduated Guidance\nHand-over-hand physical guidance with fading pressure; move to shadow, then independent.\nCurrent prompt level: At the emerging stage, provide hand-over-hand guidance at each step; reduce physical pressure across sessions.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f5253653d1de243f", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "using public transportation", "program_type": "independence", "learner_profile": "adult", "mastery_state": "emerging", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "using public transportation", "module": "community", "mastery_criterion": "mc_generalization", "chain_type": "forward", "prompt_strategy": "graduated_guidance", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "50630"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tolerating nail clipping\nCurriculum Reference: AFLS Basic Living Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: Toleration / Systematic Desensitization\n\nThe goal is for the learner to allow the activity to be performed (not to perform it independently). Provide the full program structure including shaping steps and refusal/safety procedures."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tolerating nail clipping\n- Curriculum: AFLS Basic Living Skills\n- Learner: Adolescent Learner\n- Current mastery: Mastered at current level\n- Approach: Toleration / systematic desensitization (caregiver-delivered activity)\n\nThis program builds tolerance for nail clipping performed by caregiver through systematic shaping. The learner is not being taught to perform the skill independently \u2014 the goal is to allow the caregiver or staff to perform it across an incrementally increasing duration, without refusal, crying, self-injurious behavior, or aggression.\n\n## Toleration End Goal\nall 10 fingernails (and optionally toenails) clipped by caregiver without refusal, pulling hand away, or safety behaviors\n\n## Shaping Progression\nTeach by moving through these shaping steps in order. Advance to the next step only when the current step is tolerated for 3 consecutive sessions without safety behaviors.\n\n1. 1 fingernail clipped\n2. 2 fingernails clipped in one sitting\n3. 5 fingernails clipped (one hand)\n4. All 10 fingernails clipped\n5. All 10 fingernails + 10 toenails clipped\n\nCurrent shaping step: At the mastered stage, begin at shaping step 5 (\"All 10 fingernails + 10 toenails clipped\") and advance through step 5 across sessions as tolerance stabilizes.\n\n## Antecedent Strategies\n- Pre-warn the learner before the activity begins (verbal or visual schedule).\n- Offer a preferred item or sensory tool to hold during the activity.\n- Where possible, offer a choice of position, location, or starting point.\n- Keep the environment calm and predictable; avoid other demands immediately before or after.\n\n## Reinforcement\nDeliver high-magnitude, high-preference reinforcement contingent on successful toleration of the target duration. Pair with effusive praise. Reinforcement is continuous (CRF) during acquisition of each shaping step.\n\n## Refusal and Safety Procedures\n- If the learner refuses or turns away: pause briefly (5\u201310 seconds), re-present the activity at the current or previous shaping step.\n- If problem behavior (crying, property destruction, SIB, aggression, peer aggression) emerges: stop the activity immediately, implement the facility crisis plan, and document the incident.\n- Do NOT remove the activity contingent on safety behaviors in a way that establishes an escape-maintained pattern \u2014 consult supervising BCBA for the specific escape-extinction protocol appropriate to this learner.\n\n## Mastery Criteria\nEnd-goal activity tolerated across 2 different settings (e.g., home and clinic) across 2 consecutive weeks without safety behaviors.\n\n## Data Collection\nPer-trial: duration tolerated before refusal, presence / absence of safety behaviors (crying, property destruction, SIB, aggression), staff support level, and shaping step attempted. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full routine across 2 different settings and with 3 different adults (including the primary caregiver if relevant). Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "c4325b6c5991fb6f", "gold_labels": {"method": "task_analysis", "domain": "AFLS.basic_living", "skill": "tolerating nail clipping", "program_type": "toleration", "learner_profile": "adolescent", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_toleration_v1", "taxonomy_cells": {"skill_target": "tolerating nail clipping", "module": "basic_living", "mastery_criterion": "mc_toleration_settings", "shaping_n_steps": 5}, "teacher_model": null, "seed_tag": "92010"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8503\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Listener Responding + Mand + Reading + Tact\nDate range: Sessions 1\u201312 across 24 days (2026-10-03 to 2026-10-27)\n\nACCELERATION PROGRAMS\n1. selects items by feature (Listener Responding)\n Method: DTT\n2. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n3. tacts colors of objects (Tact)\n Method: DTT\n4. mands for missing items needed to complete an activity (Mand)\n Method: NET\n5. sounds out CVC words (Reading)\n Method: DTT\n6. tacts common objects (Tact)\n Method: DTT\n\nDECELERATION TARGETS\n1. Elopement \u2014 function hypothesized: escape\n2. Pica \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-10-03 \u2014 43 min \u2014 1 observer\n selects items by feature: 11/15 correct (75%); latency 2.5s; prompts FP\u00d75, PP\u00d72, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/13 correct (73%); latency 3.0s; prompts FP\u00d71, PP\u00d74, G\u00d71, I\u00d77\n tacts colors of objects: 11/15 correct (71%); latency 3.1s; prompts FP\u00d72, PP\u00d75, G\u00d74, I\u00d74\n mands for missing items needed to complete an activity: 10/13 correct (79%); latency 2.6s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d74\n sounds out CVC words: 7/9 correct (73%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n tacts common objects: 9/11 correct (79%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n Elopement: freq 1\n Pica: attempts 3 (2 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 2 \u2014 2026-10-05 \u2014 43 min \u2014 1 observer\n selects items by feature: 10/15 correct (69%); latency 2.6s; prompts FP\u00d75, PP\u00d72, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/8 correct (72%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n tacts colors of objects: 7/11 correct (68%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n mands for missing items needed to complete an activity: 7/9 correct (75%); latency 2.5s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d74\n sounds out CVC words: 11/15 correct (72%); latency 2.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d710\n tacts common objects: 5/8 correct (63%); latency 3.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n Elopement: freq 1\n Pica: attempts 3 (3 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 3 \u2014 2026-10-07 \u2014 43 min \u2014 1 observer\n selects items by feature: 7/12 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/12 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n tacts colors of objects: 10/15 correct (67%); latency 2.9s; prompts FP\u00d70, PP\u00d75, G\u00d72, I\u00d78\n mands for missing items needed to complete an activity: 7/11 correct (62%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n sounds out CVC words: 5/9 correct (57%); latency 3.4s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d73\n tacts common objects: 5/8 correct (59%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n Elopement: freq 3\n Pica: attempts 3 (2 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 4 \u2014 2026-10-09 \u2014 43 min \u2014 1 observer\n selects items by feature: 5/11 correct (49%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 8/14 correct (56%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d714\n tacts colors of objects: 6/9 correct (62%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n mands for missing items needed to complete an activity: 5/9 correct (59%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d76\n sounds out CVC words: 9/15 correct (58%); latency 3.4s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d79\n tacts common objects: 5/8 correct (57%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n Elopement: freq 2\n Pica: attempts 5 (4 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 5 \u2014 2026-10-11 \u2014 43 min \u2014 1 observer\n selects items by feature: 5/10 correct (48%); latency 3.7s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/13 correct (54%); latency 3.3s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n tacts colors of objects: 5/9 correct (59%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n mands for missing items needed to complete an activity: 9/15 correct (61%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d78\n sounds out CVC words: 7/15 correct (48%); latency 3.8s; prompts FP\u00d74, PP\u00d70, G\u00d71, I\u00d710\n tacts common objects: 6/14 correct (45%); latency 3.7s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n Elopement: freq 4\n Pica: attempts 4 (3 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 6 \u2014 2026-10-13 \u2014 43 min \u2014 1 observer\n selects items by feature: 4/10 correct (37%); latency 4.6s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/13 correct (49%); latency 3.8s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d75\n tacts colors of objects: 5/9 correct (55%); latency 3.5s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n mands for missing items needed to complete an activity: 6/13 correct (48%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n sounds out CVC words: 6/14 correct (44%); latency 3.8s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d74\n tacts common objects: 6/14 correct (40%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d711\n Elopement: freq 5\n Pica: attempts 6 (3 unsuccessful \u2014 staff retrieved item before ingestion; 3 successful \u2014 item ingested)\n\nSession 7 \u2014 2026-10-15 \u2014 43 min \u2014 1 observer\n selects items by feature: 3/9 correct (32%); latency 4.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 6/13 correct (45%); latency 3.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d76\n tacts colors of objects: 7/14 correct (47%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d79\n mands for missing items needed to complete an activity: 7/13 correct (51%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d710\n sounds out CVC words: 6/13 correct (43%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d75\n tacts common objects: 3/10 correct (35%); latency 4.5s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d75\n Elopement: freq 6\n Pica: attempts 5 (4 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nSession 8 \u2014 2026-10-17 \u2014 43 min \u2014 1 observer\n selects items by feature: 4/15 correct (26%); latency 4.9s; prompts FP\u00d74, PP\u00d75, G\u00d74, I\u00d72\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/11 correct (38%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d79\n tacts colors of objects: 6/12 correct (48%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d76\n mands for missing items needed to complete an activity: 3/8 correct (39%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d76\n sounds out CVC words: 3/9 correct (39%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n tacts common objects: 2/8 correct (26%); latency 5.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n Elopement: freq 5\n Pica: attempts 6 (6 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n\nSession 9 \u2014 2026-10-19 \u2014 43 min \u2014 1 observer\n selects items by feature: 2/12 correct (13%); latency 5.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/13 correct (29%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d79\n tacts colors of objects: 5/14 correct (38%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n mands for missing items needed to complete an activity: 3/8 correct (42%); latency 4.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n sounds out CVC words: 3/10 correct (33%); latency 4.3s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d75\n tacts common objects: 2/12 correct (17%); latency 5.5s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d78\n Elopement: freq 7\n Pica: attempts 7 (4 unsuccessful \u2014 staff retrieved item before ingestion; 3 successful \u2014 item ingested)\n\nSession 10 \u2014 2026-10-21 \u2014 43 min \u2014 1 observer\n selects items by feature: 1/10 correct (11%); latency 5.5s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (28%); latency 4.7s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n tacts colors of objects: 5/13 correct (37%); latency 4.6s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d79\n mands for missing items needed to complete an activity: 3/8 correct (35%); latency 4.4s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n sounds out CVC words: 4/13 correct (29%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d712\n tacts common objects: 1/10 correct (10%); latency 5.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Elopement: freq 6\n Pica: attempts 8 (4 unsuccessful \u2014 staff retrieved item before ingestion; 4 successful \u2014 item ingested)\n\nSession 11 \u2014 2026-10-23 \u2014 43 min \u2014 1 observer\n selects items by feature: 1/13 correct (10%); latency 5.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d711\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (20%); latency 4.9s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n tacts colors of objects: 5/15 correct (37%); latency 4.4s; prompts FP\u00d73, PP\u00d74, G\u00d73, I\u00d75\n mands for missing items needed to complete an activity: 3/10 correct (33%); latency 4.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n sounds out CVC words: 3/14 correct (21%); latency 5.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d711\n tacts common objects: 1/14 correct (10%); latency 5.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n Elopement: freq 8\n Pica: attempts 8 (5 unsuccessful \u2014 staff retrieved item before ingestion; 3 successful \u2014 item ingested)\n\nSession 12 \u2014 2026-10-25 \u2014 43 min \u2014 1 observer\n selects items by feature: 1/9 correct (10%); latency 5.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 1/8 correct (15%); latency 5.4s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n tacts colors of objects: 3/12 correct (25%); latency 4.9s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d710\n mands for missing items needed to complete an activity: 3/11 correct (25%); latency 5.0s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d76\n sounds out CVC words: 2/12 correct (14%); latency 5.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d79\n tacts common objects: 1/9 correct (10%); latency 5.8s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n Elopement: freq 8\n Pica: attempts 8 (7 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Vocal refusal ('no', 'I don't want to')\n- Crying during demand presentation\n- Increased response latency (3\u20135\u00d7 baseline)\n- Self-stimulatory behavior increase during demand periods"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (75% -> 10%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nElopement: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nPica: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 12 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 12 sessions: 36.6%.\n- Trend direction: downward (first session: 75%, last session: 10%).\n- Elopement: mean 4.7/session (early-window mean 1.8 -> late-window mean 7.2).\n- Pica: mean 5.5/session (early-window mean 3.5 -> late-window mean 7.8).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "0faedff98f890d5b", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Elopement": "escape", "Pica": "attention"}, "escalation_level": 4, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "early", "n_sessions": 12, "n_programs": 6, "n_behaviors": 2, "behavior_ids": ["elopement", "pica"], "behavior_functions": ["escape", "attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "15018"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: matches letters to identical letters\nCurriculum Reference: Reading L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: matches letters to identical letters\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter flashcard.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: J, X, Z\nDistractor stimuli: T, M\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8563ce0a1a3b3b9d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "matches letters to identical letters", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_80_2sessions"}, "teacher_model": null, "seed_tag": "85854"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: selects items given 2+ features/functions simultaneously\nCurriculum Reference: LRFFC L3\nLearner Profile: School-Age Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items given 2+ features/functions simultaneously\n- Curriculum: LRFFC L3\n- Learner: School-Age Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: car, ball, toothbrush\nDistractor stimuli: pencil, book, cup, block, shoe\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nPrevent errors by using high prompts from the start; fade gradually. Paired with most-to-least prompt hierarchy.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "b2bc46884b6c81e9", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L3", "learner_profile": "school_age", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items given 2+ features/functions simultaneously", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "errorless", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "89319"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: tacts emotions in self and others\nCurriculum: Tact L2\nLearner Level: School-Age Learner\nBaseline: Generalization phase\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts emotions in self and others\n- Curriculum: Tact L2\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"How does he feel?\"\nVariations: \"What emotion?\"; \"How is she feeling?\"\nPresentation: Present a photograph or emoji card depicting the target emotion.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: scissors, shoe, brush\nDistractor stimuli: ball, bottle, toothbrush, book\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nVR-3 (variable ratio 3) \u2014 Average of every 3rd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy with 2 different therapists and 2 different settings.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "da18cb2a3954cf68", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts emotions in self and others", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "vr3", "error_correction": "four_step", "mastery_criterion": "mc_80_generalization"}, "teacher_model": null, "seed_tag": "47307"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a Task Analysis program for this multi-step skill.\n\nTarget: using public transportation\nCurriculum: AFLS Community Participation Skills\nLearner: Adult Learner\nMastery Status: Mastered at current level\n\nPlease include the task analysis (numbered steps), chain type, prompt strategy per step, error correction, reinforcement, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: using public transportation\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Mastered at current level\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Confirm the correct stop or station. <- Learner-taught step (forward chaining starts here)\n2. Purchase or have the fare ready.\n3. Wait at the designated boarding area.\n4. Board the vehicle when it arrives.\n5. Pay the fare or present the pass.\n6. Sit or stand safely.\n7. Monitor the route for the correct stop.\n8. Exit at the correct stop.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the mastered stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "86daf0b03559cdb9", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "using public transportation", "program_type": "independence", "learner_profile": "adult", "mastery_state": "mastered", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "using public transportation", "module": "community", "mastery_criterion": "mc_generalization", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "39765"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6396\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand\nDate range: Sessions 1\u20139 across 18 days (2026-05-16 to 2026-06-03)\n\nACCELERATION PROGRAMS\n1. doing laundry (Home Skills)\n Method: Task Analysis\n2. self-advocacy statements (Independent Living)\n Method: BST\n3. requesting assistance using AAC device (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Motor stereotypy \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-16 \u2014 48 min \u2014 1 observer\n doing laundry: 8/11 correct (77%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n self-advocacy statements: 6/8 correct (74%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d77\n requesting assistance using AAC device: 12/14 correct (83%); latency 2.1s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n Motor stereotypy: freq 3; PIR 17%\n\nSession 2 \u2014 2026-05-18 \u2014 48 min \u2014 1 observer\n doing laundry: 6/8 correct (71%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n self-advocacy statements: 5/9 correct (61%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n requesting assistance using AAC device: 8/10 correct (77%); latency 2.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n Motor stereotypy: freq 3; PIR 15%\n\nSession 3 \u2014 2026-05-20 \u2014 48 min \u2014 1 observer\n doing laundry: 9/13 correct (68%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d75\n self-advocacy statements: 8/14 correct (56%); latency 3.4s; prompts FP\u00d74, PP\u00d74, G\u00d70, I\u00d76\n requesting assistance using AAC device: 8/12 correct (70%); latency 3.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n Motor stereotypy: freq 3; PIR 19%\n\nSession 4 \u2014 2026-05-22 \u2014 48 min \u2014 1 observer\n doing laundry: 4/8 correct (54%); latency 3.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n self-advocacy statements: 7/13 correct (55%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d77\n requesting assistance using AAC device: 10/15 correct (69%); latency 2.8s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d79\n Motor stereotypy: freq 3; PIR 15%\n\nSession 5 \u2014 2026-05-24 \u2014 48 min \u2014 1 observer\n doing laundry: 8/15 correct (51%); latency 3.9s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n self-advocacy statements: 4/8 correct (49%); latency 3.7s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n requesting assistance using AAC device: 6/9 correct (70%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n Motor stereotypy: freq 4; PIR 30%\n ABC (motor stereotypy): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 6 \u2014 2026-05-26 \u2014 48 min \u2014 1 observer\n doing laundry: 7/14 correct (48%); latency 4.0s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d74\n self-advocacy statements: 4/9 correct (40%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d76\n requesting assistance using AAC device: 6/10 correct (65%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n Motor stereotypy: freq 3; PIR 22%\n ABC (motor stereotypy): A = quiet unstructured moment between trials; B = engaged in hand-flapping; C = no social consequence; self-terminated after 30s\n\nSession 7 \u2014 2026-05-28 \u2014 48 min \u2014 1 observer\n doing laundry: 5/13 correct (40%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d710\n self-advocacy statements: 3/10 correct (27%); latency 4.7s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d75\n requesting assistance using AAC device: 5/8 correct (60%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n Motor stereotypy: freq 5; PIR 21%\n\nSession 8 \u2014 2026-05-30 \u2014 48 min \u2014 1 observer\n doing laundry: 5/15 correct (36%); latency 4.4s; prompts FP\u00d70, PP\u00d74, G\u00d75, I\u00d76\n self-advocacy statements: 1/8 correct (18%); latency 5.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n requesting assistance using AAC device: 8/15 correct (55%); latency 3.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d712\n Motor stereotypy: freq 5; PIR 34%\n\nSession 9 \u2014 2026-06-01 \u2014 48 min \u2014 1 observer\n doing laundry: 4/13 correct (31%); latency 4.8s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d77\n self-advocacy statements: 1/12 correct (11%); latency 5.6s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d710\n requesting assistance using AAC device: 4/8 correct (48%); latency 4.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n Motor stereotypy: freq 6; PIR 38%\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Attempts at property destruction (throwing materials, swiping items off the surface)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Regression identified**: accuracy has declined from 77% to 31% across 9 sessions.\n- Accompanying behavior data suggests the decline may be function-related.\n\n## Pattern Classification\nregression\n\nPreviously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.\n\n## Behavior Function Hypothesis\nMotor stereotypy: automatic\n Evidence: ABC events in the log support the automatic function; behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct reinforcer assessment to identify current high-preference items.\n- Review recent environmental or schedule changes that may be affecting performance.\n- Return temporarily to a previously-mastered prompt level and re-fade systematically.\n\n### Replacement behavior (FCT)\n- If problem behavior accompanies regression, strengthen replacement-behavior training for the hypothesized function.\n\n### Consequence strategies\n- Increase reinforcement magnitude and schedule temporarily (return to CRF if needed).\n- Implement more frequent maintenance probes (at least weekly) after re-mastery.\n\n### Crisis plan\n- If regression co-occurs with severe behavior (SIB, aggression, property destruction) reaching the learner's safety threshold: pause all skill-acquisition programming, implement environmental safety measures (clear the area of hazards, move peers to safe distance), and contact the supervising BCBA.\n- Staff should first attempt verbal de-escalation and least-restrictive strategies; physical intervention is used only as a last resort and only by staff currently certified in the facility's crisis-prevention framework (e.g., Safety-Care, CPI, or equivalent).\n- Physical intervention is applied only when specifically authorized in this learner's BIP. Many learners have contraindications (medical conditions, developmental or age-related factors, trauma history, or explicit BIP restrictions) that rule out restrictive procedures \u2014 in those cases the crisis plan defaults to verbal de-escalation, environmental safety measures, and immediate supervisor / emergency contact without physical intervention.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 52.8%.\n- Trend direction: downward (first session: 77%, last session: 31%).\n- Motor stereotypy: mean 3.9/session (early-window mean 3.0 -> late-window mean 5.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1baf6baefd16d549", "gold_labels": {"pattern_class": "regression", "behavior_functions": {"Motor stereotypy": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "regression", "learner_profile": "adult", "n_sessions": 9, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["motor_stereotypy"], "behavior_functions": ["automatic"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "53027"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6176\nProfile: Adult Learner (18+ years)\nCurricula: Community + Home Skills + Independent Living + Mand + Self-Care\nDate range: Sessions 1\u20139 across 18 days (2026-08-01 to 2026-08-19)\n\nACCELERATION PROGRAMS\n1. scheduling medical appointments (Independent Living)\n Method: Task Analysis\n2. washing hands independently (Self-Care)\n Method: Task Analysis\n3. indicating 'all done' at task completion (Mand)\n Method: NET\n4. using public transportation (Community)\n Method: Task Analysis\n5. simple meal preparation (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-08-01 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 9/12 correct (78%); latency 2.4s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n washing hands independently: 8/10 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n indicating 'all done' at task completion: 7/9 correct (74%); latency 2.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d72\n using public transportation: 12/13 correct (92%); latency 1.7s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d711\n simple meal preparation: 12/14 correct (87%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d76\n\nSession 2 \u2014 2026-08-03 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 10/14 correct (73%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d79\n washing hands independently: 9/12 correct (76%); latency 2.8s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d79\n indicating 'all done' at task completion: 9/11 correct (77%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n using public transportation: 11/13 correct (84%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d77\n simple meal preparation: 12/15 correct (83%); latency 2.1s; prompts FP\u00d75, PP\u00d73, G\u00d73, I\u00d74\n\nSession 3 \u2014 2026-08-05 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 6/9 correct (70%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n washing hands independently: 10/13 correct (78%); latency 2.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d711\n indicating 'all done' at task completion: 8/10 correct (77%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n using public transportation: 11/13 correct (87%); latency 2.3s; prompts FP\u00d70, PP\u00d73, G\u00d74, I\u00d76\n simple meal preparation: 12/14 correct (84%); latency 2.2s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n\nSession 4 \u2014 2026-08-07 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 9/13 correct (70%); latency 3.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n washing hands independently: 6/8 correct (70%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n indicating 'all done' at task completion: 10/12 correct (81%); latency 2.3s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n using public transportation: 11/13 correct (82%); latency 2.5s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d74\n simple meal preparation: 11/13 correct (86%); latency 2.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d77\n\nSession 5 \u2014 2026-08-09 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 10/13 correct (75%); latency 2.8s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d711\n washing hands independently: 8/10 correct (76%); latency 2.4s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d78\n indicating 'all done' at task completion: 8/10 correct (78%); latency 2.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d76\n using public transportation: 6/8 correct (79%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d75\n simple meal preparation: 9/10 correct (89%); latency 1.8s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n\nSession 6 \u2014 2026-08-11 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 11/15 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d74, G\u00d75, I\u00d75\n washing hands independently: 7/10 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n indicating 'all done' at task completion: 7/9 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n using public transportation: 13/15 correct (84%); latency 2.4s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d77\n simple meal preparation: 13/14 correct (95%); latency 1.9s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d76\n\nSession 7 \u2014 2026-08-13 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 9/12 correct (73%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d77\n washing hands independently: 10/15 correct (67%); latency 2.9s; prompts FP\u00d75, PP\u00d71, G\u00d71, I\u00d78\n indicating 'all done' at task completion: 7/9 correct (77%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n using public transportation: 10/13 correct (77%); latency 2.3s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d77\n simple meal preparation: 7/8 correct (90%); latency 1.8s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n\nSession 8 \u2014 2026-08-15 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 8/12 correct (66%); latency 3.2s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d710\n washing hands independently: 6/9 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d72\n indicating 'all done' at task completion: 10/13 correct (75%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d75\n using public transportation: 11/15 correct (76%); latency 2.9s; prompts FP\u00d75, PP\u00d72, G\u00d74, I\u00d74\n simple meal preparation: 12/13 correct (96%); latency 2.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d77\n\nSession 9 \u2014 2026-08-17 \u2014 40 min \u2014 1 observer\n scheduling medical appointments: 6/9 correct (69%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n washing hands independently: 7/11 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n indicating 'all done' at task completion: 9/12 correct (73%); latency 2.7s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d73\n using public transportation: 9/13 correct (73%); latency 2.7s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d79\n simple meal preparation: 10/10 correct (98%); latency 1.8s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d72\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Echolalic responding (repeating SD rather than answering)\n- Stereotypic behavior during inter-trial intervals\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (78%\u201369%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 9 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 71.6%.\n- Trend direction: downward (first session: 78%, last session: 69%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "f5c20e9782f59dfa", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "adult", "n_sessions": 9, "n_programs": 5, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "68062"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: ordering in a restaurant\nCurriculum Reference: AFLS Community Participation Skills\nLearner Profile: Adolescent Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: ordering in a restaurant\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adolescent Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Approach the counter or wait to be seated. <- Learner-taught step (forward chaining starts here)\n2. Review the menu.\n3. Decide on a choice.\n4. Place the order verbally (or with an AAC device).\n5. Respond to clarifying questions.\n6. Pay using money or a card.\n7. Wait for the order and collect it.\n\n## Prompt Strategy Per Step\nStrategy: Most-to-Least\nStart with full-physical prompt at each step; fade to partial physical, gestural, verbal, independent.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "2f85d9ec5b610ca7", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "ordering in a restaurant", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "approaching", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "ordering in a restaurant", "module": "community", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "most_to_least", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "98892"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2204\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Home Skills + Mand + Reading + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201311 across 22 days (2026-01-06 to 2026-01-28)\n\nACCELERATION PROGRAMS\n1. meal cleanup after eating (Home Skills)\n Method: Task Analysis\n2. indicating 'all done' with a step or activity (Mand)\n Method: NET\n3. sounds out CVC words (Reading)\n Method: DTT\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n5. tolerating haircut (Self-Care)\n Method: Task Analysis\n6. making the bed (Home Skills)\n Method: Task Analysis\n\nSESSION DATA\n\nSession 1 \u2014 2026-01-06 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 4/8 correct (44%); latency 3.8s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d74\n indicating 'all done' with a step or activity: 5/15 correct (31%); latency 4.4s; prompts FP\u00d72, PP\u00d75, G\u00d71, I\u00d77\n sounds out CVC words: 3/8 correct (42%); latency 4.2s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n wait calmly when access to preferred item is denied: 5/10 correct (48%); latency 4.1s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n tolerating haircut: 5/12 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d710\n making the bed: 4/9 correct (45%); latency 4.0s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n\nSession 2 \u2014 2026-01-08 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 3/8 correct (42%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n indicating 'all done' with a step or activity: 6/13 correct (44%); latency 4.1s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n sounds out CVC words: 5/10 correct (50%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 4/8 correct (45%); latency 4.0s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d76\n tolerating haircut: 7/14 correct (49%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d710\n making the bed: 8/15 correct (55%); latency 3.5s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d711\n\nSession 3 \u2014 2026-01-10 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 8/15 correct (51%); latency 3.8s; prompts FP\u00d75, PP\u00d75, G\u00d73, I\u00d72\n indicating 'all done' with a step or activity: 5/10 correct (50%); latency 3.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d72\n sounds out CVC words: 9/15 correct (58%); latency 3.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d713\n wait calmly when access to preferred item is denied: 7/14 correct (51%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n tolerating haircut: 8/15 correct (51%); latency 4.0s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d713\n making the bed: 8/13 correct (63%); latency 3.3s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n\nSession 4 \u2014 2026-01-12 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 6/11 correct (55%); latency 3.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n indicating 'all done' with a step or activity: 8/14 correct (55%); latency 3.5s; prompts FP\u00d70, PP\u00d74, G\u00d70, I\u00d710\n sounds out CVC words: 6/10 correct (64%); latency 3.1s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 5/9 correct (55%); latency 3.4s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n tolerating haircut: 9/14 correct (62%); latency 3.0s; prompts FP\u00d71, PP\u00d71, G\u00d73, I\u00d79\n making the bed: 8/12 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d74, I\u00d75\n\nSession 5 \u2014 2026-01-14 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 6/9 correct (67%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n indicating 'all done' with a step or activity: 5/8 correct (62%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n sounds out CVC words: 7/9 correct (73%); latency 3.0s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d73\n wait calmly when access to preferred item is denied: 10/15 correct (68%); latency 3.0s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n tolerating haircut: 9/13 correct (68%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d73\n making the bed: 8/10 correct (77%); latency 2.3s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d78\n\nSession 6 \u2014 2026-01-16 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 7/11 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n indicating 'all done' with a step or activity: 9/14 correct (66%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d712\n sounds out CVC words: 9/13 correct (69%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d710\n wait calmly when access to preferred item is denied: 10/14 correct (72%); latency 3.0s; prompts FP\u00d73, PP\u00d71, G\u00d74, I\u00d76\n tolerating haircut: 6/8 correct (76%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n making the bed: 9/11 correct (80%); latency 2.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d710\n\nSession 7 \u2014 2026-01-18 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 10/13 correct (75%); latency 2.6s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 7/9 correct (76%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n sounds out CVC words: 11/14 correct (82%); latency 2.2s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 8/11 correct (74%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d75\n tolerating haircut: 12/15 correct (78%); latency 2.3s; prompts FP\u00d75, PP\u00d71, G\u00d75, I\u00d74\n making the bed: 13/14 correct (92%); latency 1.8s; prompts FP\u00d73, PP\u00d74, G\u00d71, I\u00d76\n\nSession 8 \u2014 2026-01-20 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 13/15 correct (86%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d70, I\u00d713\n indicating 'all done' with a step or activity: 10/12 correct (82%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d711\n sounds out CVC words: 10/12 correct (81%); latency 2.3s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 8/10 correct (80%); latency 2.6s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n tolerating haircut: 8/9 correct (83%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n making the bed: 10/11 correct (92%); latency 2.1s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d74\n\nSession 9 \u2014 2026-01-22 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 14/15 correct (92%); latency 2.0s; prompts FP\u00d70, PP\u00d75, G\u00d73, I\u00d77\n indicating 'all done' with a step or activity: 7/8 correct (93%); latency 2.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n sounds out CVC words: 12/14 correct (86%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 6/8 correct (78%); latency 2.7s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n tolerating haircut: 11/12 correct (91%); latency 2.0s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d76\n making the bed: 13/14 correct (95%); latency 1.7s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n\nSession 10 \u2014 2026-01-24 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 9/9 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d74\n indicating 'all done' with a step or activity: 11/12 correct (95%); latency 1.8s; prompts FP\u00d74, PP\u00d73, G\u00d72, I\u00d73\n sounds out CVC words: 14/15 correct (95%); latency 1.9s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d710\n wait calmly when access to preferred item is denied: 8/9 correct (87%); latency 2.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n tolerating haircut: 9/10 correct (90%); latency 2.0s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d73\n making the bed: 12/13 correct (95%); latency 1.9s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n\nSession 11 \u2014 2026-01-26 \u2014 53 min \u2014 1 observer\n meal cleanup after eating: 8/8 correct (95%); latency 2.0s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d77\n indicating 'all done' with a step or activity: 11/12 correct (95%); latency 1.8s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n sounds out CVC words: 13/14 correct (95%); latency 1.6s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d78\n wait calmly when access to preferred item is denied: 12/13 correct (91%); latency 2.1s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n tolerating haircut: 8/8 correct (95%); latency 1.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n making the bed: 8/8 correct (95%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Requesting more trials verbally\n- Smiling during reinforcement delivery\n- Quick response latency (within 2\u20133 seconds)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- Steady improvement: accuracy rose from 44% to 95% across 11 sessions.\n- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.\n\n## Pattern Classification\nmastery_progression\n\nAccuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Continue current teaching procedure \u2014 it is effective for this learner.\n- Maintain current stimulus arrangement; begin planning next skill in sequence.\n\n### Consequence strategies\n- Begin thinning reinforcement from CRF to FR-2 as accuracy stabilizes above 85%.\n- Continue differential reinforcement for independent responses.\n\n## Escalation Level\n1 \u2014 Continue monitoring\n\nCurrent programming is effective; no change needed.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 11 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 11 sessions: 69.8%.\n- Trend direction: upward (first session: 44%, last session: 95%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "585482a9ed73dc23", "gold_labels": {"pattern_class": "mastery_progression", "behavior_functions": {}, "escalation_level": 1, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "mastery_progression", "learner_profile": "adolescent", "n_sessions": 11, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "70409"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects items by function (find something you eat with)\nCurriculum: Listener Responding L2\nLearner Level: School-Age Learner\nBaseline: Developing\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function (find something you eat with)\n- Curriculum: Listener Responding L2\n- Learner: School-Age Learner\n- Current mastery: Developing\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the developing stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: phone, book, pencil\nDistractor stimuli: scissors, ball, block, cup\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5bdeaf87728c7703", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L2", "learner_profile": "school_age", "mastery_state": "developing"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function (find something you eat with)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "backstep", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "11906"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-7079\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand\nDate range: Sessions 1\u201310 across 20 days (2026-05-20 to 2026-06-09)\n\nACCELERATION PROGRAMS\n1. identifying community safety signs (Community)\n Method: DTT\n2. requesting preferred items using AAC device (Mand)\n Method: NET\n3. requesting a break during non-preferred activities (Mand)\n Method: NET\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-20 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 9/15 correct (57%); latency 3.7s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d710\n requesting preferred items using AAC device: 6/13 correct (44%); latency 4.0s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d76\n requesting a break during non-preferred activities: 3/9 correct (35%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n Self-injurious behavior (SIB): freq 0\n\nSession 2 \u2014 2026-05-22 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 7/11 correct (68%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d76\n requesting preferred items using AAC device: 5/10 correct (50%); latency 3.8s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n requesting a break during non-preferred activities: 6/15 correct (41%); latency 4.3s; prompts FP\u00d75, PP\u00d70, G\u00d73, I\u00d77\n Self-injurious behavior (SIB): freq 0\n\nSession 3 \u2014 2026-05-24 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 10/13 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d73\n requesting preferred items using AAC device: 8/13 correct (60%); latency 3.4s; prompts FP\u00d74, PP\u00d73, G\u00d73, I\u00d73\n requesting a break during non-preferred activities: 6/13 correct (46%); latency 4.2s; prompts FP\u00d73, PP\u00d74, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 1\n\nSession 4 \u2014 2026-05-26 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 7/8 correct (83%); latency 2.1s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n requesting preferred items using AAC device: 10/14 correct (69%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n requesting a break during non-preferred activities: 6/11 correct (55%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 1\n\nSession 5 \u2014 2026-05-28 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 8/10 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d76\n requesting preferred items using AAC device: 8/11 correct (75%); latency 2.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n requesting a break during non-preferred activities: 9/15 correct (61%); latency 3.3s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d78\n Self-injurious behavior (SIB): freq 0\n\nSession 6 \u2014 2026-05-30 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 9/11 correct (83%); latency 2.3s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d74\n requesting preferred items using AAC device: 12/15 correct (80%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d79\n requesting a break during non-preferred activities: 7/10 correct (67%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d77\n Self-injurious behavior (SIB): freq 0\n\nSession 7 \u2014 2026-06-01 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 7/8 correct (83%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d76\n requesting preferred items using AAC device: 12/15 correct (80%); latency 2.2s; prompts FP\u00d74, PP\u00d71, G\u00d74, I\u00d76\n requesting a break during non-preferred activities: 8/12 correct (71%); latency 2.8s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d76\n Self-injurious behavior (SIB): freq 1\n\nSession 8 \u2014 2026-06-03 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 10/12 correct (83%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n requesting preferred items using AAC device: 10/13 correct (80%); latency 2.5s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n requesting a break during non-preferred activities: 6/9 correct (71%); latency 3.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d77\n Self-injurious behavior (SIB): freq 2\n\nSession 9 \u2014 2026-06-05 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 9/11 correct (83%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n requesting preferred items using AAC device: 8/10 correct (80%); latency 2.6s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d78\n requesting a break during non-preferred activities: 9/13 correct (71%); latency 2.7s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d78\n Self-injurious behavior (SIB): freq 1\n\nSession 10 \u2014 2026-06-07 \u2014 59 min \u2014 1 observer\n identifying community safety signs: 12/14 correct (83%); latency 2.2s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d78\n requesting preferred items using AAC device: 11/14 correct (80%); latency 2.3s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d77\n requesting a break during non-preferred activities: 11/15 correct (71%); latency 2.8s; prompts FP\u00d71, PP\u00d75, G\u00d75, I\u00d74\n Self-injurious behavior (SIB): freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Delayed responding (>5 seconds consistently)\n- Leaving seat without permission"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 57% at return, recovering to 83%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 78.3%.\n- Trend direction: upward (first session: 57%, last session: 83%).\n- Self-injurious behavior (SIB): mean 0.8/session (early-window mean 0.3 -> late-window mean 1.7).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "928d2ad1cc4cb114", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 3, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "35204"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-5994\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Self-Care + Tolerating Denied Access\nDate range: Sessions 1\u201310 across 20 days (2026-03-18 to 2026-04-07)\n\nACCELERATION PROGRAMS\n1. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n2. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n3. ordering in a restaurant (Community)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Pica \u2014 function hypothesized: attention\n2. Fecal smearing (scatolia) \u2014 function hypothesized: automatic\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-18 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 8/10 correct (77%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n wait calmly when access to preferred item is denied: 7/11 correct (68%); latency 2.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n ordering in a restaurant: 8/12 correct (63%); latency 2.9s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d76\n Pica: attempts 2 (0 unsuccessful \u2014 staff retrieved item before ingestion; 2 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 1 (1 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 2 \u2014 2026-03-20 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 9/11 correct (79%); latency 2.5s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 6/10 correct (61%); latency 3.5s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n ordering in a restaurant: 7/10 correct (66%); latency 3.2s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d77\n Pica: attempts 1 (0 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 1 (0 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 3 \u2014 2026-03-22 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 10/13 correct (78%); latency 2.4s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 8/13 correct (61%); latency 3.5s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d77\n ordering in a restaurant: 6/9 correct (67%); latency 3.0s; prompts FP\u00d70, PP\u00d73, G\u00d73, I\u00d73\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 4 \u2014 2026-03-24 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 10/14 correct (74%); latency 2.7s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n wait calmly when access to preferred item is denied: 7/10 correct (66%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d76\n ordering in a restaurant: 9/13 correct (70%); latency 2.6s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d79\n Pica: attempts 3 (2 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 1 (0 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 5 \u2014 2026-03-26 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 5/12 correct (44%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n wait calmly when access to preferred item is denied: 3/9 correct (35%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d75\n ordering in a restaurant: 5/10 correct (46%); latency 3.9s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d710\n Pica: attempts 5 (4 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 3 (2 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 6 \u2014 2026-03-28 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 4/9 correct (45%); latency 4.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n wait calmly when access to preferred item is denied: 4/11 correct (37%); latency 4.5s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d73\n ordering in a restaurant: 6/12 correct (51%); latency 3.7s; prompts FP\u00d74, PP\u00d70, G\u00d70, I\u00d78\n Pica: attempts 5 (3 unsuccessful \u2014 staff retrieved item before ingestion; 2 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 4 (2 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 7 \u2014 2026-03-30 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 5/12 correct (45%); latency 4.0s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n wait calmly when access to preferred item is denied: 4/9 correct (47%); latency 3.6s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n ordering in a restaurant: 7/13 correct (55%); latency 3.6s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d77\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 2 (2 intercepted \u2014 staff redirected before smearing; 0 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 8 \u2014 2026-04-01 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 6/11 correct (52%); latency 3.5s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 8/15 correct (56%); latency 3.6s; prompts FP\u00d74, PP\u00d74, G\u00d75, I\u00d72\n ordering in a restaurant: 7/12 correct (61%); latency 3.4s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d74\n Pica: attempts 2 (2 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 2 (0 intercepted \u2014 staff redirected before smearing; 2 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 9 \u2014 2026-04-03 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 8/13 correct (61%); latency 3.1s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 8/13 correct (59%); latency 3.5s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d76\n ordering in a restaurant: 7/10 correct (68%); latency 2.9s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d73\n Pica: attempts 2 (1 unsuccessful \u2014 staff retrieved item before ingestion; 1 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nSession 10 \u2014 2026-04-05 \u2014 31 min \u2014 1 observer\n dressing (pants, shirt): 5/8 correct (62%); latency 2.9s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n wait calmly when access to preferred item is denied: 10/15 correct (63%); latency 3.2s; prompts FP\u00d71, PP\u00d75, G\u00d73, I\u00d76\n ordering in a restaurant: 10/14 correct (68%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d75\n Pica: attempts 1 (1 unsuccessful \u2014 staff retrieved item before ingestion; 0 successful \u2014 item ingested)\n Fecal smearing (scatolia): attempts 2 (1 intercepted \u2014 staff redirected before smearing; 1 completed \u2014 feces transferred to skin, clothing, or surface)\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Playing with non-task materials\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).\n- Correlation between setting event and performance should be tracked explicitly.\n\n## Pattern Classification\nsetting_event_trigger\n\nA discontinuous performance change at a specific session correlates with an external setting event.\n\n## Behavior Function Hypothesis\nPica: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\nFecal smearing (scatolia): automatic\n Evidence: behavior persists across contexts without clear social mediation.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Identify and document the setting event (sleep, illness, schedule change, medication).\n- Modify programming demand on days when setting event is present.\n- Coordinate with caregivers / medical team on setting-event factors outside the program.\n\n### Consequence strategies\n- Track setting events alongside performance data to quantify their effect.\n\n### Crisis plan\n- If the setting event involves a medical concern (seizure-like activity, injury, illness): stop the program immediately, follow the facility's medical-incident protocol, and refer to the appropriate medical professional. Medical events are NOT behavioral targets and must not be treated as ABA programming issues.\n- If behavior escalates during the setting-event window: use verbal de-escalation and environmental safety first; do not add demand. Only staff currently certified in the facility's crisis-prevention framework may use physical intervention, and only if imminent safety risk warrants it AND the learner's BIP specifically authorizes it (contraindications and BIP restrictions may rule it out).\n- Coordinate with caregivers, the medical team, and the supervising BCBA; document the setting event and the associated behavior for pattern analysis. Review the BIP in light of the setting-event data.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 61.8%.\n- Trend direction: downward (first session: 77%, last session: 62%).\n- Pica: mean 2.4/session (early-window mean 1.3 -> late-window mean 1.7).\n- Fecal smearing (scatolia): mean 2.0/session (early-window mean 1.3 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "e271264ff519079b", "gold_labels": {"pattern_class": "setting_event_trigger", "behavior_functions": {"Pica": "attention", "Fecal smearing (scatolia)": "automatic"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "setting_event_trigger", "learner_profile": "adolescent", "n_sessions": 10, "n_programs": 3, "n_behaviors": 2, "behavior_ids": ["pica", "fecal_smearing"], "behavior_functions": ["attention", "automatic"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "43895"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a NET program for this skill.\n\nTarget: mands for information using \"why\"\nCurriculum: Mand L3\nLearner: Adolescent Learner\nMastery Status: Near mastery\n\nPlease include the MO arrangement, natural opportunity, prompt strategy, natural reinforcer, generalization tactics, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: mands for information using \"why\"\n- Curriculum: Mand L3\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nBriefly withhold or restrict access to the target preferred item/activity for 30\u201360 seconds to establish motivation before the teaching opportunity.\nPrimary natural context: leisure time.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's leisure time routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Expectant look (pause)\nDeliver no prompt \u2014 maintain eye contact and hold the preferred item until the learner responds.\nCurrent prompt level: At the near mastery stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: immediate access to the requested item for 30\u201360 seconds.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use across 3 different adults and 2 different settings within 2 weeks.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "a24dbd9a59e23a0c", "gold_labels": {"method": "net", "domain": "VB-MAPP.mand", "level": "L3", "learner_profile": "adolescent", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "mands for information using \"why\"", "mo_category": "mand_item", "mo_arrangement": "deprivation_preferred_item", "prompt_strategy": "expectant_look", "mastery_criterion": "mc_generalize_multi_therapist"}, "teacher_model": null, "seed_tag": "16366"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: clocking in at start of shift\nCurriculum: AFLS Vocational Skills\nLearner Level: Adolescent Learner\nBaseline: Near mastery\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: clocking in at start of shift\n- Curriculum: AFLS Vocational Skills\n- Learner: Adolescent Learner\n- Current mastery: Near mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Arrive at the workplace by the scheduled time.\n2. Locate the time clock or digital system.\n3. Enter identification (ID card / badge / code).\n4. Confirm the clock-in timestamp.\n5. Notify the supervisor or check the day's task list.\n6. Proceed to the assigned workstation.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the near mastery stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nToken economy \u2014 One token per correctly-completed step; token exchange at chain completion.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "6824002777438d82", "gold_labels": {"method": "task_analysis", "domain": "AFLS.vocational", "skill": "clocking in at start of shift", "program_type": "independence", "learner_profile": "adolescent", "mastery_state": "near", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "clocking in at start of shift", "module": "vocational", "mastery_criterion": "mc_self_initiation", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "6490"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: takes turns during structured activities\nCurriculum: Social Behavior & Social Play L2\nLearner Level: Early Learner\nBaseline: Mastered at current level\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: takes turns during structured activities\n- Curriculum: Social Behavior & Social Play L2\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nArrange a shared activity with one or more peers where the target response is naturally occasioned (turn-taking game, shared snack, collaborative task).\nPrimary natural context: peer play.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's peer play routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Model + immediate fade\nDeliver a vocal/gestural model if no response within the motivated moment; fade to expectant look over sessions.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: turn is delivered; game continues.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use of target response across 3 natural opportunities per day for 3 consecutive days.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "c54c9ff869f1ac4c", "gold_labels": {"method": "net", "domain": "VB-MAPP.social", "level": "L2", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "takes turns during structured activities", "mo_category": "social_turn_taking", "mo_arrangement": "peer_presence", "prompt_strategy": "model_then_fade", "mastery_criterion": "mc_indep_natural_env"}, "teacher_model": null, "seed_tag": "16366"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: sorts items into 2\u20133 categories\nCurriculum: Visual Perceptual Skills & Matching-to-Sample L2\nLearner Level: Early Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: sorts items into 2\u20133 categories\n- Curriculum: Visual Perceptual Skills & Matching-to-Sample L2\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Match\"\nVariations: \"Find the same\"; \"Put with same\"\nPresentation: Present the sample stimulus and an array of comparison stimuli simultaneously.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: block, bottle, towel\nDistractor stimuli: crayon, sock, book, backpack\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "e7ffef93072adc73", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.vp_mts", "level": "L2", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "sorts items into 2\u20133 categories", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "fr2", "error_correction": "backstep", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "54357"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: unloading the dishwasher\nCurriculum: AFLS Home Skills\nLearner Level: Adult Learner\nBaseline: Generalization phase\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: unloading the dishwasher\n- Curriculum: AFLS Home Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Forward chaining\nTeach step 1 to mastery, then 1->2, then 1->2->3, progressively adding later steps.\n\nSteps:\n1. Open the dishwasher once the cycle is complete. <- Learner-taught step (forward chaining starts here)\n2. Unload the top rack first (cups, glasses, small items).\n3. Put cups and glasses away in the correct cabinet.\n4. Unload the utensil basket.\n5. Put utensils away in the drawer.\n6. Unload the bottom rack (plates, bowls, larger items).\n7. Put plates and bowls away in the correct cabinet.\n8. Close the dishwasher door.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nBack-step + re-approach: Return to the previously-mastered step; re-execute that step correctly; then re-attempt the missed step with one prompt level above the last independent level\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nChain self-initiated from the appropriate antecedent (no verbal prompt to begin) across 3 consecutive opportunities.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 2 different settings and with 3 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "f6fc0940ee2e5472", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "unloading the dishwasher", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "forward"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "unloading the dishwasher", "module": "home", "mastery_criterion": "mc_self_initiation", "chain_type": "forward", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "back_to_previous_step"}, "teacher_model": null, "seed_tag": "97806"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-1774\nProfile: Early Learner (dev age 0\u201348 months)\nCurricula: Intraverbal + Mand + Motor Imitation + Tact\nDate range: Sessions 1\u201310 across 20 days (2026-02-07 to 2026-02-27)\n\nACCELERATION PROGRAMS\n1. imitates 3-step motor sequences (Motor Imitation)\n Method: DTT variant\n2. 2-word mand phrases (e.g., \"more cracker\", \"open please\") (Mand)\n Method: NET\n3. tacts colors of objects (Tact)\n Method: DTT\n4. answers WH-questions about familiar topics (Intraverbal)\n Method: NET\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-02-07 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 7/8 correct (83%); latency 2.5s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/13 correct (79%); latency 2.2s; prompts FP\u00d74, PP\u00d72, G\u00d72, I\u00d75\n tacts colors of objects: 12/14 correct (83%); latency 2.3s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d79\n answers WH-questions about familiar topics: 10/13 correct (74%); latency 2.5s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n Self-injurious behavior (SIB): freq 2\n\nSession 2 \u2014 2026-02-09 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 3/8 correct (37%); latency 4.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 3/10 correct (28%); latency 4.4s; prompts FP\u00d70, PP\u00d70, G\u00d71, I\u00d79\n tacts colors of objects: 3/9 correct (31%); latency 4.8s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d73\n answers WH-questions about familiar topics: 3/9 correct (35%); latency 4.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d74\n Self-injurious behavior (SIB): freq 2\n\nSession 3 \u2014 2026-02-11 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 10/12 correct (86%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/12 correct (78%); latency 2.3s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n tacts colors of objects: 10/12 correct (87%); latency 2.1s; prompts FP\u00d74, PP\u00d74, G\u00d72, I\u00d72\n answers WH-questions about familiar topics: 12/15 correct (81%); latency 2.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d79\n Self-injurious behavior (SIB): freq 1\n\nSession 4 \u2014 2026-02-13 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 5/15 correct (34%); latency 4.5s; prompts FP\u00d73, PP\u00d73, G\u00d74, I\u00d75\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/11 correct (22%); latency 4.7s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n tacts colors of objects: 3/9 correct (30%); latency 4.7s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n answers WH-questions about familiar topics: 3/9 correct (28%); latency 5.0s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d77\n Self-injurious behavior (SIB): freq 2\n\nSession 5 \u2014 2026-02-15 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 9/10 correct (87%); latency 2.4s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 9/12 correct (74%); latency 2.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d74\n tacts colors of objects: 9/11 correct (85%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n answers WH-questions about familiar topics: 8/10 correct (78%); latency 2.2s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d76\n Self-injurious behavior (SIB): freq 2\n\nSession 6 \u2014 2026-02-17 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 5/15 correct (31%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d714\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/8 correct (28%); latency 4.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n tacts colors of objects: 4/15 correct (29%); latency 5.0s; prompts FP\u00d74, PP\u00d72, G\u00d70, I\u00d79\n answers WH-questions about familiar topics: 3/8 correct (33%); latency 4.5s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 3\n\nSession 7 \u2014 2026-02-19 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 8/9 correct (89%); latency 1.9s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d78\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 10/13 correct (76%); latency 2.8s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d72\n tacts colors of objects: 10/12 correct (86%); latency 1.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n answers WH-questions about familiar topics: 8/11 correct (75%); latency 2.9s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d76\n Self-injurious behavior (SIB): freq 2\n\nSession 8 \u2014 2026-02-21 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 4/11 correct (33%); latency 4.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d73\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 4/13 correct (29%); latency 4.8s; prompts FP\u00d71, PP\u00d74, G\u00d72, I\u00d76\n tacts colors of objects: 4/13 correct (28%); latency 4.5s; prompts FP\u00d74, PP\u00d70, G\u00d73, I\u00d76\n answers WH-questions about familiar topics: 3/11 correct (30%); latency 4.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d78\n Self-injurious behavior (SIB): freq 3\n\nSession 9 \u2014 2026-02-23 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 13/15 correct (87%); latency 2.3s; prompts FP\u00d75, PP\u00d71, G\u00d72, I\u00d77\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 7/9 correct (73%); latency 2.8s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d75\n tacts colors of objects: 8/9 correct (88%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n answers WH-questions about familiar topics: 7/10 correct (75%); latency 2.8s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n Self-injurious behavior (SIB): freq 2\n\nSession 10 \u2014 2026-02-25 \u2014 44 min \u2014 1 observer\n imitates 3-step motor sequences: 4/12 correct (29%); latency 4.6s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d76\n 2-word mand phrases (e.g., \"more cracker\", \"open please\"): 2/9 correct (23%); latency 4.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d76\n tacts colors of objects: 3/10 correct (33%); latency 4.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n answers WH-questions about familiar topics: 4/15 correct (27%); latency 4.6s; prompts FP\u00d74, PP\u00d70, G\u00d74, I\u00d77\n Self-injurious behavior (SIB): freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Looking away during SD presentation\n- Leaving seat without permission\n- Playing with non-task materials\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.\n- Skill may be under restricted stimulus control.\n\n## Pattern Classification\ngeneralization_failure\n\nAlternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement structured generalization training: vary therapists, settings, and materials systematically.\n- Use multiple-exemplar training \u2014 teach with 3\u20135 different stimulus sets.\n- Program common stimuli across training and generalization settings.\n\n### Consequence strategies\n- Consider natural-environment training (NET) to supplement DTT for this skill.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 59.7%.\n- Trend direction: downward (first session: 83%, last session: 29%).\n- Self-injurious behavior (SIB): mean 2.1/session (early-window mean 1.7 -> late-window mean 2.3).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "c6af2a043c26cf25", "gold_labels": {"pattern_class": "generalization_failure", "behavior_functions": {"Self-injurious behavior (SIB)": "attention"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "generalization_failure", "learner_profile": "early", "n_sessions": 10, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "91831"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-9514\nProfile: Adult Learner (18+ years)\nCurricula: Home Skills + Independent Living + Mand + Tolerating Denied Access + Vocational\nDate range: Sessions 1\u20139 across 18 days (2026-04-23 to 2026-05-11)\n\nACCELERATION PROGRAMS\n1. clocking in at start of shift (Vocational)\n Method: Task Analysis\n2. requesting a break during extended tasks (Mand)\n Method: NET\n3. doing laundry (Home Skills)\n Method: Task Analysis\n4. wait calmly when a preferred activity is denied (Tolerating Denied Access)\n Method: FCT\n5. self-advocacy statements (Independent Living)\n Method: BST\n\nDECELERATION TARGETS\n1. Self-injurious behavior (SIB) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-23 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 9/12 correct (77%); latency 2.4s; prompts FP\u00d74, PP\u00d73, G\u00d74, I\u00d71\n requesting a break during extended tasks: 6/8 correct (73%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n doing laundry: 6/8 correct (76%); latency 2.8s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n wait calmly when a preferred activity is denied: 11/15 correct (71%); latency 2.8s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d78\n self-advocacy statements: 10/14 correct (75%); latency 2.6s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d712\n Self-injurious behavior (SIB): freq 3\n\nSession 2 \u2014 2026-04-25 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 9/15 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d71, G\u00d74, I\u00d78\n requesting a break during extended tasks: 8/12 correct (71%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d77\n doing laundry: 11/15 correct (71%); latency 3.0s; prompts FP\u00d72, PP\u00d70, G\u00d74, I\u00d79\n wait calmly when a preferred activity is denied: 9/15 correct (61%); latency 3.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d711\n self-advocacy statements: 9/12 correct (74%); latency 3.0s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d71\n Self-injurious behavior (SIB): freq 3\n\nSession 3 \u2014 2026-04-27 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 8/14 correct (56%); latency 3.3s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n requesting a break during extended tasks: 7/13 correct (57%); latency 3.7s; prompts FP\u00d70, PP\u00d70, G\u00d74, I\u00d79\n doing laundry: 6/9 correct (65%); latency 3.1s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d74\n wait calmly when a preferred activity is denied: 6/12 correct (53%); latency 3.5s; prompts FP\u00d71, PP\u00d74, G\u00d70, I\u00d77\n self-advocacy statements: 8/12 correct (66%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n Self-injurious behavior (SIB): freq 3\n\nSession 4 \u2014 2026-04-29 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 5/10 correct (49%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d72, I\u00d76\n requesting a break during extended tasks: 7/14 correct (52%); latency 3.6s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d710\n doing laundry: 8/14 correct (55%); latency 3.8s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n wait calmly when a preferred activity is denied: 7/15 correct (48%); latency 3.9s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d79\n self-advocacy statements: 6/10 correct (61%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d75\n Self-injurious behavior (SIB): freq 5\n\nSession 5 \u2014 2026-05-01 \u2014 IOA SESSION \u2014 2 observers\n clocking in at start of shift: 42% (primary); IOA 86% agreement\n requesting a break during extended tasks: 53% (primary); IOA 86% agreement\n doing laundry: 51% (primary); IOA 86% agreement\n wait calmly when a preferred activity is denied: 47% (primary); IOA 86% agreement\n self-advocacy statements: 60% (primary); IOA 86% agreement\n Self-injurious behavior (SIB): freq 5; IOA 86% agreement\n\nSession 6 \u2014 2026-05-03 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 5/14 correct (38%); latency 4.6s; prompts FP\u00d70, PP\u00d74, G\u00d71, I\u00d79\n requesting a break during extended tasks: 6/14 correct (46%); latency 4.0s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d76\n doing laundry: 3/8 correct (42%); latency 3.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n wait calmly when a preferred activity is denied: 4/11 correct (33%); latency 4.6s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d75\n self-advocacy statements: 6/12 correct (54%); latency 3.5s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d73\n Self-injurious behavior (SIB): freq 7\n\nSession 7 \u2014 2026-05-05 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 4/13 correct (27%); latency 4.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n requesting a break during extended tasks: 3/8 correct (42%); latency 4.0s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n doing laundry: 5/14 correct (36%); latency 4.4s; prompts FP\u00d73, PP\u00d70, G\u00d73, I\u00d78\n wait calmly when a preferred activity is denied: 3/10 correct (33%); latency 4.4s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d73\n self-advocacy statements: 6/15 correct (43%); latency 3.9s; prompts FP\u00d75, PP\u00d71, G\u00d74, I\u00d75\n Self-injurious behavior (SIB): freq 5\n ABC (self-injurious behavior (sib)): A = asked to finish dressing task; B = kicked at materials; C = task paused for 3 min\n\nSession 8 \u2014 2026-05-07 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 3/13 correct (20%); latency 5.3s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d74\n requesting a break during extended tasks: 4/15 correct (30%); latency 4.9s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d710\n doing laundry: 5/14 correct (34%); latency 4.7s; prompts FP\u00d74, PP\u00d71, G\u00d73, I\u00d76\n wait calmly when a preferred activity is denied: 4/15 correct (27%); latency 4.7s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d77\n self-advocacy statements: 4/10 correct (45%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d71, I\u00d78\n Self-injurious behavior (SIB): freq 7\n ABC (self-injurious behavior (sib)): A = worksheet task presented; B = refused + pushed materials away; C = task removed for 2 min\n\nSession 9 \u2014 2026-05-09 \u2014 46 min \u2014 1 observer\n clocking in at start of shift: 1/8 correct (15%); latency 5.6s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d73\n requesting a break during extended tasks: 3/14 correct (22%); latency 5.2s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d78\n doing laundry: 3/15 correct (22%); latency 4.8s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d78\n wait calmly when a preferred activity is denied: 1/10 correct (11%); latency 5.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d71\n self-advocacy statements: 3/11 correct (31%); latency 4.3s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n Self-injurious behavior (SIB): freq 8\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Turning away from task or staff\n- Increased response latency (3\u20135\u00d7 baseline)\n- Elopement from the work area\n- Crying during demand presentation"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (77% -> 15%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nSelf-injurious behavior (SIB): escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n4 \u2014 Safety immediate\n\nSafety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.\n\n## Confidence\nhigh\n\nPattern is well-supported: 9 sessions of data, low variance, and IOA evidence confirm reliability.\n\n## Data-Supported Rationale\n- Mean accuracy across 9 sessions: 42.7%.\n- Trend direction: downward (first session: 77%, last session: 15%).\n- Self-injurious behavior (SIB): mean 5.1/session (early-window mean 3.0 -> late-window mean 6.7).\n- IOA data: 86% agreement on IOA session \u2014 adequate."}], "meta": {"task_type": "session_interpretation", "example_id": "23e6ce22da6c09cc", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Self-injurious behavior (SIB)": "escape"}, "escalation_level": 4, "confidence": "high", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "adult", "n_sessions": 9, "n_programs": 5, "n_behaviors": 1, "behavior_ids": ["sib"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": true}, "teacher_model": null, "seed_tag": "54279"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: identifies own name in print\nCurriculum: Reading L1\nLearner Level: Early Learner\nBaseline: Near mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: identifies own name in print\n- Curriculum: Reading L1\n- Learner: Early Learner\n- Current mastery: Near mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What letter is this?\"\nVariations: \"Read this letter\"; \"What does this say?\"\nPresentation: Present the letter or word on a flashcard.\n\n## Prompt Hierarchy\nStrategy: Stimulus Shaping\nSequence: modified stimulus -> intermediate shape -> target stimulus\nCurrent prompt level: At the near mastery stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: jacket, block, crayon\nDistractor stimuli: apple, hat\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a782688f2d5f3e1e", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.reading", "level": "L1", "learner_profile": "early", "mastery_state": "near"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "identifies own name in print", "prompt_hierarchy": "stimulus_shaping", "reinforcement_schedule": "fr2", "error_correction": "transfer_trial", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "58800"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: copies all upper-case letters from model\nCurriculum: Writing L2\nLearner: Early Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: copies all upper-case letters from model\n- Curriculum: Writing L2\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Write the target.\"\nVariations: \"Trace this\"; \"Copy this\"; \"Write {word}\"\nPresentation: Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: Q, N, Z\nDistractor stimuli: D, C, U, K\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "af33fbfd967e3e68", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.writing", "level": "L2", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "copies all upper-case letters from model", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "token_economy", "error_correction": "four_step", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "69141"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2860\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Community + Mand + Reading + Self-Care\nDate range: Sessions 1\u20136 across 12 days (2026-05-23 to 2026-06-04)\n\nACCELERATION PROGRAMS\n1. washing hands independently (Self-Care)\n Method: Task Analysis\n2. indicating 'all done' with a step or activity (Mand)\n Method: NET\n3. sounds out CVC words (Reading)\n Method: DTT\n4. dressing (pants, shirt) (Self-Care)\n Method: Task Analysis\n5. tolerating tooth brushing by caregiver (Self-Care)\n Method: Task Analysis\n6. ordering in a restaurant (Community)\n Method: Task Analysis\n\nDECELERATION TARGETS\n1. Verbal aggression \u2014 function hypothesized: escape\n2. Non-compliance \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-05-23 \u2014 49 min \u2014 1 observer\n washing hands independently: 8/12 correct (68%); latency 2.8s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d77\n indicating 'all done' with a step or activity: 8/12 correct (69%); latency 2.7s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n sounds out CVC words: 9/14 correct (65%); latency 2.9s; prompts FP\u00d73, PP\u00d70, G\u00d74, I\u00d77\n dressing (pants, shirt): 7/12 correct (55%); latency 3.5s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n tolerating tooth brushing by caregiver: 8/13 correct (64%); latency 3.1s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d710\n ordering in a restaurant: 5/9 correct (53%); latency 3.3s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d75\n Verbal aggression: freq 2\n Non-compliance: freq 3\n\nSession 2 \u2014 2026-05-25 \u2014 49 min \u2014 1 observer\n washing hands independently: 7/12 correct (61%); latency 3.2s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d74\n indicating 'all done' with a step or activity: 10/14 correct (74%); latency 2.9s; prompts FP\u00d70, PP\u00d74, G\u00d74, I\u00d76\n sounds out CVC words: 9/14 correct (61%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d71, I\u00d79\n dressing (pants, shirt): 8/14 correct (59%); latency 3.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n tolerating tooth brushing by caregiver: 7/11 correct (63%); latency 3.4s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d711\n ordering in a restaurant: 6/11 correct (55%); latency 3.6s; prompts FP\u00d72, PP\u00d71, G\u00d72, I\u00d76\n Verbal aggression: freq 1\n Non-compliance: freq 3\n\nSession 3 \u2014 2026-05-27 \u2014 49 min \u2014 1 observer\n washing hands independently: 9/13 correct (70%); latency 2.9s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n indicating 'all done' with a step or activity: 9/14 correct (67%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d711\n sounds out CVC words: 5/8 correct (65%); latency 3.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n dressing (pants, shirt): 7/11 correct (63%); latency 3.3s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d76\n tolerating tooth brushing by caregiver: 9/14 correct (63%); latency 3.2s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d78\n ordering in a restaurant: 4/8 correct (52%); latency 3.6s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n Verbal aggression: freq 7\n Non-compliance: freq 6\n\nSession 4 \u2014 2026-05-29 \u2014 49 min \u2014 1 observer\n washing hands independently: 11/15 correct (74%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d713\n indicating 'all done' with a step or activity: 9/12 correct (71%); latency 2.8s; prompts FP\u00d74, PP\u00d74, G\u00d71, I\u00d73\n sounds out CVC words: 8/12 correct (66%); latency 3.3s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n dressing (pants, shirt): 9/15 correct (60%); latency 3.4s; prompts FP\u00d70, PP\u00d71, G\u00d75, I\u00d79\n tolerating tooth brushing by caregiver: 7/10 correct (66%); latency 2.9s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n ordering in a restaurant: 5/10 correct (53%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d76\n Verbal aggression: freq 7\n Non-compliance: freq 7\n\nSession 5 \u2014 2026-05-31 \u2014 49 min \u2014 1 observer\n washing hands independently: 10/14 correct (71%); latency 2.9s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d75\n indicating 'all done' with a step or activity: 11/15 correct (74%); latency 2.7s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d77\n sounds out CVC words: 8/11 correct (68%); latency 2.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n dressing (pants, shirt): 7/11 correct (60%); latency 3.2s; prompts FP\u00d71, PP\u00d72, G\u00d71, I\u00d77\n tolerating tooth brushing by caregiver: 10/15 correct (69%); latency 3.0s; prompts FP\u00d70, PP\u00d74, G\u00d73, I\u00d78\n ordering in a restaurant: 5/10 correct (53%); latency 3.8s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d77\n Verbal aggression: freq 3\n Non-compliance: freq 2\n\nSession 6 \u2014 2026-06-02 \u2014 49 min \u2014 1 observer\n washing hands independently: 10/14 correct (75%); latency 2.7s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d78\n indicating 'all done' with a step or activity: 6/8 correct (69%); latency 3.1s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n sounds out CVC words: 7/10 correct (73%); latency 2.9s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n dressing (pants, shirt): 5/9 correct (58%); latency 3.4s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d74\n tolerating tooth brushing by caregiver: 10/15 correct (68%); latency 2.9s; prompts FP\u00d74, PP\u00d72, G\u00d73, I\u00d76\n ordering in a restaurant: 6/13 correct (48%); latency 3.6s; prompts FP\u00d70, PP\u00d70, G\u00d73, I\u00d710\n Verbal aggression: freq 3\n Non-compliance: freq 2\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Vocal refusal ('no', 'I don't want to')\n- Attempts at property destruction (throwing materials, swiping items off the surface)\n- Crying during demand presentation\n- Increased response latency (3\u20135\u00d7 baseline)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.\n- This is an expected phase of behavior change and typically indicates the intervention is working.\n\n## Pattern Classification\nextinction_burst\n\nMid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.\n\n## Behavior Function Hypothesis\nVerbal aggression: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\nNon-compliance: escape\n Evidence: behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Maintain current antecedent arrangements \u2014 do not add extra demand.\n\n### Replacement behavior (FCT)\n- Ensure replacement response is available and reinforced at high magnitude during the burst.\n\n### Consequence strategies\n- **Maintain current protocol** \u2014 consistency during extinction bursts is critical.\n- Ensure all team members are trained on the procedure.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 6 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 6 sessions: 69.8%.\n- Trend direction: upward (first session: 68%, last session: 75%).\n- Verbal aggression: mean 3.8/session (early-window mean 1.5 -> late-window mean 3.0).\n- Non-compliance: mean 3.8/session (early-window mean 3.0 -> late-window mean 2.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1c709a82dca54bf7", "gold_labels": {"pattern_class": "extinction_burst", "behavior_functions": {"Verbal aggression": "escape", "Non-compliance": "escape"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "extinction_burst", "learner_profile": "adolescent", "n_sessions": 6, "n_programs": 6, "n_behaviors": 2, "behavior_ids": ["verbal_aggression", "non_compliance"], "behavior_functions": ["escape", "escape"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "64993"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: spontaneously comments on events\nCurriculum Reference: Spontaneous Vocal Behavior L1\nLearner Profile: Early Learner\nCurrent Mastery: Mastered at current level\nTeaching Method: NET (Natural Environment Teaching)\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously comments on events\n- Curriculum: Spontaneous Vocal Behavior L1\n- Learner: Early Learner\n- Current mastery: Mastered at current level\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: morning greeting.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's morning greeting routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Time Delay\nWait a brief interval (0->2->4 s) for independent response before delivering a model or verbal prompt.\nCurrent prompt level: At the mastered stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 4 different stimuli or peers.\n- Rotate between 2 different natural settings.\n- Rotate across 2 different adults during the teaching phase.\n\n## Mastery Criteria\nAverage rate of \u2265 3 spontaneous responses per 30-minute session across 3 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "d759e7c8ce006a9e", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L1", "learner_profile": "early", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously comments on events", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "time_delay", "mastery_criterion": "mc_spontaneous_rate"}, "teacher_model": null, "seed_tag": "3768"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a chaining program for the following.\n\nSkill: ordering in a restaurant\nCurriculum: AFLS Community Participation Skills\nLearner Level: Adult Learner\nBaseline: Generalization phase\n\nInclude the full Task Analysis structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: ordering in a restaurant\n- Curriculum: AFLS Community Participation Skills\n- Learner: Adult Learner\n- Current mastery: Generalization phase\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Approach the counter or wait to be seated.\n2. Review the menu.\n3. Decide on a choice.\n4. Place the order verbally (or with an AAC device).\n5. Respond to clarifying questions.\n6. Pay using money or a card.\n7. Wait for the order and collect it.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the generalization stage, deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps.\n\n## Error Correction Procedure\nModel -> prompt -> fade: Demonstrate the correct step; provide physical or gestural prompt as needed; fade prompt across subsequent trials\n\n## Reinforcement\nTerminal reinforcement (mastered) \u2014 Reinforcement delivered only at chain completion; no per-step reinforcement.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed independently with 2 different adults in 2 different settings across 2 consecutive days.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "0f08c6e9cecb1dae", "gold_labels": {"method": "task_analysis", "domain": "AFLS.community", "skill": "ordering in a restaurant", "program_type": "independence", "learner_profile": "adult", "mastery_state": "generalization", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "ordering in a restaurant", "module": "community", "mastery_criterion": "mc_generalization", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "terminal_reinforcement", "error_correction": "model_prompt_fade"}, "teacher_model": null, "seed_tag": "98892"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: selects items by function (What do you drink from?)\nCurriculum: LRFFC L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects items by function (What do you drink from?)\n- Curriculum: LRFFC L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: sock, crayon, apple\nDistractor stimuli: key, ball\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% accuracy across 3 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "3f8c8499b5d36ada", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.lrffc", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects items by function (What do you drink from?)", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_90_3sessions"}, "teacher_model": null, "seed_tag": "36112"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts shapes (circle, square, triangle)\nCurriculum: Tact L2\nLearner: School-Age Learner\nMastery Status: Mastered at current level\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts shapes (circle, square, triangle)\n- Curriculum: Tact L2\n- Learner: School-Age Learner\n- Current mastery: Mastered at current level\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What shape?\"\nVariations: \"What shape is this?\"; \"Tell me the shape\"\nPresentation: Present a clearly-shaped object or shape card and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Time Delay (progressive)\nSequence: 0s delay -> 2s delay -> 4s delay -> 6s delay -> independent\nCurrent prompt level: At the mastered stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 5 (target + 4 distractors)\nTarget stimuli: oval, heart, circle\nDistractor stimuli: triangle, square, star, diamond\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nReturn to previous prompt level that produced success -> successful response -> re-attempt at target prompt level.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "d8241976657e432a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L2", "learner_profile": "school_age", "mastery_state": "mastered"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts shapes (circle, square, triangle)", "prompt_hierarchy": "time_delay", "reinforcement_schedule": "token_economy", "error_correction": "backstep", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "85854"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Here is a multi-session behavioral log for a learner. Analyze it and provide your clinical interpretation using the structured format.\n\nLEARNER PROFILE\nSynthetic ID: SYN-8858\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Social Behavior\nDate range: Sessions 1\u201310 across 20 days (2026-03-19 to 2026-04-08)\n\nACCELERATION PROGRAMS\n1. requesting a break during non-preferred tasks (Mand)\n Method: NET\n2. takes turns during structured activities (Social Behavior)\n Method: PRT\n3. requesting preferred items using a full sentence (Mand)\n Method: NET\n4. reads 10\u201320 sight words (Reading)\n Method: DTT\n5. identifies numerals 1\u201320 (Math)\n Method: DTT\n6. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-19 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 7/9 correct (79%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d74\n takes turns during structured activities: 11/15 correct (76%); latency 2.5s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d713\n requesting preferred items using a full sentence: 13/15 correct (85%); latency 2.4s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d712\n reads 10\u201320 sight words: 10/13 correct (75%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d74\n identifies numerals 1\u201320: 10/12 correct (82%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/15 correct (74%); latency 2.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d711\n\nSession 2 \u2014 2026-03-21 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 7/9 correct (81%); latency 2.5s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d75\n takes turns during structured activities: 9/12 correct (76%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d78\n requesting preferred items using a full sentence: 10/12 correct (80%); latency 2.4s; prompts FP\u00d70, PP\u00d71, G\u00d74, I\u00d77\n reads 10\u201320 sight words: 11/15 correct (73%); latency 2.9s; prompts FP\u00d72, PP\u00d74, G\u00d71, I\u00d78\n identifies numerals 1\u201320: 10/12 correct (86%); latency 2.0s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d711\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (72%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d75\n\nSession 3 \u2014 2026-03-23 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 9/11 correct (82%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d78\n takes turns during structured activities: 9/11 correct (80%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d77\n requesting preferred items using a full sentence: 8/10 correct (83%); latency 2.2s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d75\n reads 10\u201320 sight words: 10/14 correct (74%); latency 2.5s; prompts FP\u00d72, PP\u00d74, G\u00d72, I\u00d76\n identifies numerals 1\u201320: 7/8 correct (84%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (73%); latency 2.5s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d77\n\nSession 4 \u2014 2026-03-25 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 5/11 correct (46%); latency 4.1s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d79\n takes turns during structured activities: 5/13 correct (41%); latency 3.9s; prompts FP\u00d71, PP\u00d70, G\u00d73, I\u00d79\n requesting preferred items using a full sentence: 6/11 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d78\n reads 10\u201320 sight words: 4/10 correct (44%); latency 4.3s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d77\n identifies numerals 1\u201320: 4/8 correct (52%); latency 3.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/12 correct (50%); latency 3.8s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n\nSession 5 \u2014 2026-03-27 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 5/10 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d76\n takes turns during structured activities: 5/11 correct (43%); latency 3.9s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n requesting preferred items using a full sentence: 6/10 correct (59%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n reads 10\u201320 sight words: 5/11 correct (45%); latency 4.1s; prompts FP\u00d73, PP\u00d72, G\u00d71, I\u00d75\n identifies numerals 1\u201320: 5/10 correct (49%); latency 3.7s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/13 correct (45%); latency 4.1s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d711\n\nSession 6 \u2014 2026-03-29 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 6/14 correct (45%); latency 4.3s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d79\n takes turns during structured activities: 7/13 correct (51%); latency 4.0s; prompts FP\u00d70, PP\u00d72, G\u00d74, I\u00d77\n requesting preferred items using a full sentence: 4/8 correct (56%); latency 3.6s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 4/11 correct (41%); latency 4.4s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d72\n identifies numerals 1\u201320: 5/9 correct (52%); latency 3.4s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (51%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d72, I\u00d75\n\nSession 7 \u2014 2026-03-31 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 7/14 correct (47%); latency 3.9s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d77\n takes turns during structured activities: 9/15 correct (57%); latency 3.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d79\n requesting preferred items using a full sentence: 6/11 correct (58%); latency 3.6s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n reads 10\u201320 sight words: 5/12 correct (46%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d75\n identifies numerals 1\u201320: 4/8 correct (55%); latency 3.8s; prompts FP\u00d70, PP\u00d72, G\u00d70, I\u00d76\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 5/9 correct (58%); latency 3.4s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n\nSession 8 \u2014 2026-04-02 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 6/13 correct (49%); latency 3.5s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n takes turns during structured activities: 9/14 correct (66%); latency 2.9s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d78\n requesting preferred items using a full sentence: 6/10 correct (64%); latency 3.3s; prompts FP\u00d73, PP\u00d71, G\u00d71, I\u00d75\n reads 10\u201320 sight words: 7/15 correct (48%); latency 4.1s; prompts FP\u00d71, PP\u00d72, G\u00d75, I\u00d77\n identifies numerals 1\u201320: 9/15 correct (59%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d74, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (64%); latency 3.3s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d75\n\nSession 9 \u2014 2026-04-04 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 7/14 correct (53%); latency 3.6s; prompts FP\u00d74, PP\u00d74, G\u00d74, I\u00d72\n takes turns during structured activities: 8/11 correct (76%); latency 2.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d74\n requesting preferred items using a full sentence: 9/13 correct (68%); latency 3.0s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d75\n reads 10\u201320 sight words: 5/9 correct (53%); latency 3.7s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d72\n identifies numerals 1\u201320: 6/9 correct (66%); latency 2.9s; prompts FP\u00d72, PP\u00d70, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/12 correct (72%); latency 2.7s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d72\n\nSession 10 \u2014 2026-04-06 \u2014 33 min \u2014 1 observer\n requesting a break during non-preferred tasks: 5/8 correct (59%); latency 3.2s; prompts FP\u00d71, PP\u00d71, G\u00d72, I\u00d74\n takes turns during structured activities: 12/15 correct (79%); latency 2.4s; prompts FP\u00d74, PP\u00d70, G\u00d75, I\u00d76\n requesting preferred items using a full sentence: 7/9 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d76\n reads 10\u201320 sight words: 6/11 correct (55%); latency 3.3s; prompts FP\u00d70, PP\u00d73, G\u00d72, I\u00d76\n identifies numerals 1\u201320: 11/15 correct (77%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d75, I\u00d79\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n- Playing with non-task materials\n- Echolalic responding (repeating SD rather than answering)"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.\n- Possible satiation with current reinforcer; preference reassessment indicated.\n\n## Pattern Classification\nmotivating_operation_shift\n\nA mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Conduct preference assessment to identify a currently-valued reinforcer.\n- Program MO-establishing operations (deprivation for target reinforcer) before teaching sessions.\n- Rotate reinforcers to prevent satiation.\n\n### Consequence strategies\n- Adjust reinforcement density to match current MO level.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 10 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 10 sessions: 58.7%.\n- Trend direction: downward (first session: 79%, last session: 59%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "63e50ee5fa3f0235", "gold_labels": {"pattern_class": "motivating_operation_shift", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "motivating_operation_shift", "learner_profile": "school_age", "n_sessions": 10, "n_programs": 6, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "43949"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6177\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Math + Reading + Social Behavior + Tolerating Denied Access\nDate range: Sessions 1\u20138 across 16 days (2026-03-22 to 2026-04-07)\n\nACCELERATION PROGRAMS\n1. reads 10\u201320 sight words (Reading)\n Method: DTT\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n4. takes turns during structured activities (Social Behavior)\n Method: PRT\n5. identifies numerals 1\u201320 (Math)\n Method: DTT\n6. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Tantrum \u2014 function hypothesized: attention\n\nSESSION DATA\n\nSession 1 \u2014 2026-03-22 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 14/15 correct (92%); latency 1.7s; prompts FP\u00d75, PP\u00d74, G\u00d70, I\u00d76\n single-word requests for preferred items: 11/12 correct (89%); latency 2.3s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (90%); latency 2.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n takes turns during structured activities: 13/14 correct (95%); latency 1.8s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n identifies numerals 1\u201320: 11/14 correct (82%); latency 2.1s; prompts FP\u00d73, PP\u00d70, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 11/15 correct (72%); latency 2.9s; prompts FP\u00d75, PP\u00d72, G\u00d73, I\u00d75\n Tantrum: freq 0\n\nSession 2 \u2014 2026-03-24 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 12/13 correct (94%); latency 2.1s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d79\n single-word requests for preferred items: 7/8 correct (92%); latency 2.0s; prompts FP\u00d72, PP\u00d71, G\u00d70, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/13 correct (89%); latency 2.0s; prompts FP\u00d74, PP\u00d71, G\u00d71, I\u00d77\n takes turns during structured activities: 8/9 correct (89%); latency 2.0s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d76\n identifies numerals 1\u201320: 7/9 correct (83%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d73, I\u00d72\n wait calmly when access to preferred item is denied: 8/11 correct (73%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d710\n Tantrum: freq 0\n\nSession 3 \u2014 2026-03-26 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 12/12 correct (98%); latency 1.4s; prompts FP\u00d73, PP\u00d73, G\u00d70, I\u00d76\n single-word requests for preferred items: 15/15 correct (98%); latency 1.7s; prompts FP\u00d71, PP\u00d75, G\u00d74, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 9/10 correct (90%); latency 1.9s; prompts FP\u00d73, PP\u00d71, G\u00d72, I\u00d74\n takes turns during structured activities: 8/8 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d73\n identifies numerals 1\u201320: 12/14 correct (89%); latency 1.7s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n wait calmly when access to preferred item is denied: 6/8 correct (80%); latency 2.4s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d75\n Tantrum: freq 0\n\nSession 4 \u2014 2026-03-28 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 13/14 correct (91%); latency 1.6s; prompts FP\u00d71, PP\u00d74, G\u00d73, I\u00d76\n single-word requests for preferred items: 15/15 correct (98%); latency 1.9s; prompts FP\u00d72, PP\u00d72, G\u00d71, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/8 correct (92%); latency 1.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n takes turns during structured activities: 13/14 correct (95%); latency 1.6s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n identifies numerals 1\u201320: 10/11 correct (89%); latency 2.2s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d74\n wait calmly when access to preferred item is denied: 9/12 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d70, G\u00d70, I\u00d79\n Tantrum: freq 0\n\nSession 5 \u2014 2026-03-30 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 11/11 correct (96%); latency 1.5s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d75\n single-word requests for preferred items: 14/14 correct (98%); latency 1.8s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d710\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/13 correct (90%); latency 2.0s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d76\n takes turns during structured activities: 14/15 correct (92%); latency 1.7s; prompts FP\u00d73, PP\u00d72, G\u00d74, I\u00d76\n identifies numerals 1\u201320: 11/13 correct (86%); latency 2.1s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d76\n wait calmly when access to preferred item is denied: 12/15 correct (78%); latency 2.4s; prompts FP\u00d75, PP\u00d73, G\u00d75, I\u00d72\n Tantrum: freq 0\n\nSession 6 \u2014 2026-04-01 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 10/10 correct (98%); latency 1.7s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d73\n single-word requests for preferred items: 12/12 correct (98%); latency 1.9s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d77\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/13 correct (95%); latency 2.0s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d75\n takes turns during structured activities: 12/12 correct (96%); latency 1.6s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d73\n identifies numerals 1\u201320: 8/10 correct (83%); latency 2.4s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 9/12 correct (79%); latency 2.3s; prompts FP\u00d70, PP\u00d70, G\u00d72, I\u00d710\n Tantrum: freq 0\n\nSession 7 \u2014 2026-04-03 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 12/13 correct (94%); latency 1.9s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d76\n single-word requests for preferred items: 9/9 correct (98%); latency 1.4s; prompts FP\u00d70, PP\u00d73, G\u00d71, I\u00d75\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 11/12 correct (95%); latency 2.0s; prompts FP\u00d74, PP\u00d71, G\u00d72, I\u00d75\n takes turns during structured activities: 14/15 correct (95%); latency 1.8s; prompts FP\u00d73, PP\u00d70, G\u00d75, I\u00d77\n identifies numerals 1\u201320: 8/9 correct (88%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 10/12 correct (81%); latency 2.2s; prompts FP\u00d71, PP\u00d73, G\u00d72, I\u00d76\n Tantrum: freq 0\n\nSession 8 \u2014 2026-04-05 \u2014 60 min \u2014 1 observer\n reads 10\u201320 sight words: 8/8 correct (98%); latency 1.7s; prompts FP\u00d72, PP\u00d72, G\u00d70, I\u00d74\n single-word requests for preferred items: 15/15 correct (98%); latency 1.8s; prompts FP\u00d74, PP\u00d72, G\u00d71, I\u00d78\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/13 correct (94%); latency 1.6s; prompts FP\u00d73, PP\u00d72, G\u00d72, I\u00d76\n takes turns during structured activities: 12/13 correct (95%); latency 2.0s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d77\n identifies numerals 1\u201320: 7/8 correct (85%); latency 2.3s; prompts FP\u00d70, PP\u00d72, G\u00d72, I\u00d74\n wait calmly when access to preferred item is denied: 6/8 correct (80%); latency 2.3s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d72\n Tantrum: freq 0\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Delayed responding (>5 seconds consistently)\n- Looking away during SD presentation\n- Leaving seat without permission\n- Playing with non-task materials"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Prompt dependency**: high accuracy maintained (92%\u201398%) but primarily at prompted levels.\n- Independent responding remains low despite extended training.\n\n## Pattern Classification\nprompt_dependency\n\nAccuracy stays high only at prompted levels; independent responding remains low despite extended training.\n\n## Behavior Function Hypothesis\nTantrum: attention\n Evidence: behavior occurs during low-attention periods and results in social attention contingently.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Switch to a time-delay procedure (0s -> 2s -> 4s -> 6s) to encourage independent responding.\n- Avoid inadvertent prompt delivery (therapist gaze, body positioning) that may function as a prompt.\n\n### Consequence strategies\n- Ensure differential reinforcement \u2014 independent responses receive higher-quality reinforcement than prompted responses.\n- Consider stimulus fading instead of response prompting for this skill.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 95.0%.\n- Trend direction: upward (first session: 92%, last session: 98%).\n- Tantrum: mean 0.0/session (early-window mean 0.0 -> late-window mean 0.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "1a44805f5a5ad3da", "gold_labels": {"pattern_class": "prompt_dependency", "behavior_functions": {"Tantrum": "attention"}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "prompt_dependency", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 6, "n_behaviors": 1, "behavior_ids": ["tantrum"], "behavior_functions": ["attention"], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "62567"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts common actions (running, eating, sleeping)\nCurriculum: Tact L1\nLearner: Early Learner\nMastery Status: Emerging\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common actions (running, eating, sleeping)\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is he doing?\"\nVariations: \"What action?\"; \"What is she doing?\"; \"What is happening here?\"\nPresentation: Present an action picture, video clip, or live demonstration.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: pouring, jumping, eating\nDistractor stimuli: brushing, singing\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n80% accuracy at independent prompt level across 3 sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "8cc5cbf37208c89a", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common actions (running, eating, sleeping)", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "crf", "error_correction": "simple_correction", "mastery_criterion": "mc_80_independent"}, "teacher_model": null, "seed_tag": "97982"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Discrete Trial Training program for the following.\n\nSkill: selects correct item from an array of 2\nCurriculum: Listener Responding L1\nLearner Level: Early Learner\nBaseline: Approaching mastery\n\nInclude the full DTT trial structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: selects correct item from an array of 2\n- Curriculum: Listener Responding L1\n- Learner: Early Learner\n- Current mastery: Approaching mastery\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"Touch [target item]\"\nVariations: \"Point to [target item]\"; \"Give me [target item]\"; \"Show me [target item]\"\nPresentation: Present an array of items on the table and deliver a clear verbal SD.\n\n## Prompt Hierarchy\nStrategy: Stimulus Fading\nSequence: exaggerated stimulus -> moderate fade -> minimal fade -> natural stimulus\nCurrent prompt level: At the approaching mastery stage, deliver gestural or positional prompts; fade toward independence.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: glue, jacket, towel\nDistractor stimuli: apple, hat\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\n\"No, watch me\" -> model -> re-present SD -> differential reinforcement for correct response.\n\n## Reinforcement Schedule\nFR-2 (fixed ratio 2) \u2014 Every 2nd correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\n90% first-trial-correct across 5 consecutive sessions.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "63fd5bc3d79cbc3d", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.listener_responding", "level": "L1", "learner_profile": "early", "mastery_state": "approaching"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "selects correct item from an array of 2", "prompt_hierarchy": "stimulus_fading", "reinforcement_schedule": "fr2", "error_correction": "simple_correction", "mastery_criterion": "mc_90_first_trial"}, "teacher_model": null, "seed_tag": "56565"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "I need a DTT program for this skill.\n\nTarget: tacts community helpers and their roles\nCurriculum: Tact L3\nLearner: School-Age Learner\nMastery Status: Generalization phase\n\nPlease include the SD, prompt hierarchy, stimulus array, error correction procedure, reinforcement schedule, and mastery criteria."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts community helpers and their roles\n- Curriculum: Tact L3\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is this?\"\nVariations: \"Tell me what you see\"; \"What's that?\"; \"Name this\"\nPresentation: Present the target stimulus within visual field and ask while pointing to it.\n\n## Prompt Hierarchy\nStrategy: Least-to-Most\nSequence: independent -> gestural -> positional -> partial physical -> full physical\nCurrent prompt level: At the generalization stage, use minimal prompts (positional or verbal hint) only when needed; expect independent responding.\n\n## Stimulus Array\nArray size: field of 6 (target + 5 distractors, mixed categories)\nTarget stimuli: napkin, hat, jacket\nDistractor stimuli: blanket, plate, ball, apple, spoon\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nRe-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nToken economy \u2014 Tokens delivered per correct response; exchanged at session end for a back-up reinforcer appropriate to the learner (for children: candy, cookies, brief leisure-time access, preferred activity, screen time, or a favorite sensory item). Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nFluency criterion: correct responses within 3-second latency at 90% accuracy.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 5 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 2 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "5bf345c665d54b6b", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L3", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts community helpers and their roles", "prompt_hierarchy": "least_to_most", "reinforcement_schedule": "token_economy", "error_correction": "transfer_trial", "mastery_criterion": "mc_fluency"}, "teacher_model": null, "seed_tag": "56565"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Review this session log and produce a structured clinical interpretation.\n\nLEARNER PROFILE\nSynthetic ID: SYN-2501\nProfile: School-Age Learner (6\u201312 years)\nCurricula: Mand + Tolerating Denied Access\nDate range: Sessions 1\u20138 across 16 days (2026-07-07 to 2026-07-23)\n\nACCELERATION PROGRAMS\n1. single-word mands for preferred items (e.g., \"cookie\", \"iPad\") (Mand)\n Method: NET\n2. single-word requests for preferred items (Mand)\n Method: NET\n3. indicating 'all done' when an activity is complete (Mand)\n Method: NET\n4. wait calmly when access to preferred item is denied (Tolerating Denied Access)\n Method: FCT\n\nDECELERATION TARGETS\n1. Toileting accident (urine or bowel) \u2014 function hypothesized: escape\n\nSESSION DATA\n\nSession 1 \u2014 2026-07-07 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/8 correct (79%); latency 2.7s; prompts FP\u00d71, PP\u00d72, G\u00d70, I\u00d75\n single-word requests for preferred items: 10/14 correct (74%); latency 2.8s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d710\n indicating 'all done' when an activity is complete: 11/14 correct (78%); latency 2.6s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d79\n wait calmly when access to preferred item is denied: 6/9 correct (69%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d73, I\u00d73\n Toileting accident (urine or bowel): urine: 2 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 2 \u2014 2026-07-09 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 12/15 correct (80%); latency 2.4s; prompts FP\u00d75, PP\u00d75, G\u00d71, I\u00d74\n single-word requests for preferred items: 11/14 correct (76%); latency 2.6s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d711\n indicating 'all done' when an activity is complete: 9/13 correct (66%); latency 2.8s; prompts FP\u00d72, PP\u00d72, G\u00d74, I\u00d75\n wait calmly when access to preferred item is denied: 10/15 correct (67%); latency 2.8s; prompts FP\u00d75, PP\u00d75, G\u00d73, I\u00d72\n Toileting accident (urine or bowel): urine: 5 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 3 \u2014 2026-07-11 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/9 correct (75%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d72, I\u00d71\n single-word requests for preferred items: 6/9 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d73, G\u00d70, I\u00d74\n indicating 'all done' when an activity is complete: 8/12 correct (63%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n wait calmly when access to preferred item is denied: 5/8 correct (64%); latency 2.9s; prompts FP\u00d70, PP\u00d71, G\u00d70, I\u00d77\n Toileting accident (urine or bowel): urine: 2 in-toilet / 2 accidents; BM: 1 in-toilet / 0 accidents\n\nSession 4 \u2014 2026-07-13 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/10 correct (61%); latency 3.1s; prompts FP\u00d72, PP\u00d72, G\u00d72, I\u00d74\n single-word requests for preferred items: 9/14 correct (61%); latency 3.5s; prompts FP\u00d74, PP\u00d73, G\u00d71, I\u00d76\n indicating 'all done' when an activity is complete: 7/14 correct (48%); latency 3.8s; prompts FP\u00d72, PP\u00d74, G\u00d73, I\u00d75\n wait calmly when access to preferred item is denied: 8/13 correct (58%); latency 3.6s; prompts FP\u00d73, PP\u00d73, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 2 in-toilet / 1 accidents; BM: 0 in-toilet / 1 accidents\n\nSession 5 \u2014 2026-07-15 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 6/9 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d76\n single-word requests for preferred items: 7/15 correct (50%); latency 3.5s; prompts FP\u00d71, PP\u00d75, G\u00d71, I\u00d78\n indicating 'all done' when an activity is complete: 5/13 correct (42%); latency 3.9s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d79\n wait calmly when access to preferred item is denied: 4/9 correct (45%); latency 4.0s; prompts FP\u00d72, PP\u00d70, G\u00d71, I\u00d76\n Toileting accident (urine or bowel): urine: 1 in-toilet / 3 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = during demand sequence at the work table; B = voided (urine) without signaling need; C = staff paused programming, changed clothing, resumed after 10 min\n\nSession 6 \u2014 2026-07-17 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 7/14 correct (51%); latency 3.4s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d78\n single-word requests for preferred items: 7/15 correct (46%); latency 4.2s; prompts FP\u00d71, PP\u00d74, G\u00d74, I\u00d76\n indicating 'all done' when an activity is complete: 5/14 correct (37%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d70, I\u00d713\n wait calmly when access to preferred item is denied: 7/15 correct (46%); latency 3.8s; prompts FP\u00d75, PP\u00d75, G\u00d75, I\u00d70\n Toileting accident (urine or bowel): urine: 3 in-toilet / 4 accidents; BM: 0 in-toilet / 0 accidents\n\nSession 7 \u2014 2026-07-19 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 4/8 correct (47%); latency 3.9s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n single-word requests for preferred items: 5/13 correct (41%); latency 4.4s; prompts FP\u00d70, PP\u00d73, G\u00d70, I\u00d710\n indicating 'all done' when an activity is complete: 4/13 correct (30%); latency 4.5s; prompts FP\u00d73, PP\u00d74, G\u00d70, I\u00d76\n wait calmly when access to preferred item is denied: 3/9 correct (37%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n Toileting accident (urine or bowel): urine: 1 in-toilet / 5 accidents; BM: 0 in-toilet / 0 accidents\n ABC (toileting accident (urine or bowel)): A = end of 90-minute block without scheduled toileting; B = soiled (bowel) at the work table; C = staff initiated clean-up and hygiene routine; program paused\n\nSession 8 \u2014 2026-07-21 \u2014 30 min \u2014 1 observer\n single-word mands for preferred items (e.g., \"cookie\", \"iPad\"): 3/9 correct (38%); latency 4.4s; prompts FP\u00d73, PP\u00d72, G\u00d73, I\u00d71\n single-word requests for preferred items: 5/13 correct (35%); latency 4.3s; prompts FP\u00d72, PP\u00d74, G\u00d70, I\u00d77\n indicating 'all done' when an activity is complete: 3/14 correct (22%); latency 5.1s; prompts FP\u00d71, PP\u00d72, G\u00d72, I\u00d79\n wait calmly when access to preferred item is denied: 4/12 correct (33%); latency 4.6s; prompts FP\u00d70, PP\u00d72, G\u00d73, I\u00d77\n Toileting accident (urine or bowel): urine: 1 in-toilet / 5 accidents; BM: 0 in-toilet / 0 accidents\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Increased response latency (3\u20135\u00d7 baseline)\n- Elopement from the work area\n- Vocal refusal ('no', 'I don't want to')\n- Self-injurious behavior onset during task difficulty (self-biting, hair-pulling, head-hitting)\n\nPlease include: (1) clinical concerns, (2) pattern classification, (3) function hypothesis for any target behaviors, (4) recommendations structured as antecedent / replacement / consequence / crisis, (5) escalation level and confidence, (6) data-supported rationale."}, {"role": "assistant", "content": "## Clinical Concerns\n- **Frustration pattern**: declining accuracy (79% -> 38%) with escape-indicator behaviors.\n- Risk of learned helplessness and escape-maintained behavior if not addressed.\n\n## Pattern Classification\nfrustration_pattern\n\nDeclining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.\n\n## Behavior Function Hypothesis\nToileting accident (urine or bowel): escape\n Evidence: ABC events in the log support the escape function; behavior occurs in the context of demand presentation and terminates contingent on task removal.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- **Immediately** reduce task difficulty \u2014 drop back 1\u20132 prompt levels or present previously-mastered targets.\n- Intersperse easy (mastered) tasks with acquisition targets at a 3:1 or 4:1 ratio.\n- Shorten session length and increase break frequency.\n\n### Replacement behavior (FCT)\n- Teach a replacement response ('break please', 'wait please', 'all done') and reinforce it at the function of the problem behavior.\n- Pair the replacement response with short, predictable escape windows.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum / escape behavior \u2014 this reinforces the escape function.\n- Provide brief break contingent on independent replacement response instead.\n\n### Crisis plan\n- If aggression toward staff, peer aggression, SIB, or property destruction exceeds the learner's safety threshold: first use verbal de-escalation and environmental safety measures (clear the area, move peers to safe distance, remove hazards).\n- If behavior continues to pose imminent risk, follow the facility's tiered crisis-intervention protocol \u2014 only staff currently certified in the facility's training (e.g., Safety-Care, CPI, PMT, or equivalent) may implement physical intervention, and the least-restrictive option appropriate to the risk level is used. Physical intervention is applied only when specifically authorized in this learner's BIP; for learners with contraindications (medical, developmental, trauma history, or explicit BIP restrictions), the plan stops at de-escalation + environmental safety + supervisor contact.\n- Notify the supervising BCBA immediately; pause the skill-acquisition program until the BIP is reviewed. Document the incident (antecedent, topography, duration, staff response, post-incident state) per facility and BACB Ethics Code (2020, Section 3.05) requirements.\n\n## Escalation Level\n3 \u2014 Supervisor review\n\nCombination of performance decline and/or behavior data warrants supervising BCBA review within 24\u201348 hours.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 8 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 8 sessions: 62.0%.\n- Trend direction: downward (first session: 79%, last session: 38%).\n- Toileting accident (urine or bowel): mean 3.1/session (early-window mean 2.0 -> late-window mean 5.0).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "73287ecf45439a6f", "gold_labels": {"pattern_class": "frustration_pattern", "behavior_functions": {"Toileting accident (urine or bowel)": "escape"}, "escalation_level": 3, "confidence": "moderate", "crisis_plan_required": true}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "frustration_pattern", "learner_profile": "school_age", "n_sessions": 8, "n_programs": 4, "n_behaviors": 1, "behavior_ids": ["toileting_accident"], "behavior_functions": ["escape"], "has_abc_data": true, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "91480"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Design a Natural Environment Teaching program for the following.\n\nSkill: spontaneously asks questions about the environment\nCurriculum: Spontaneous Vocal Behavior L2\nLearner Level: School-Age Learner\nBaseline: Generalization phase\n\nInclude the full naturalistic teaching structure."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: spontaneously asks questions about the environment\n- Curriculum: Spontaneous Vocal Behavior L2\n- Learner: School-Age Learner\n- Current mastery: Generalization phase\n\nThis NET program teaches the target skill through naturalistic embedding \u2014 capturing motivated moments within the learner's routine rather than using contrived massed trials.\n\n## Motivating Operation (MO) Arrangement\nCreate a small environmental change (new item, familiar item in new location, surprise event) that invites a verbal comment or question.\nPrimary natural context: activity setup.\n\n## Natural Opportunity\nEmbed the teaching opportunity within the learner's activity setup routine. Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest).\n\n## Prompt Strategy\nStrategy: Verbal fill-in\nDeliver a partial phrase and pause; learner completes the vocal response.\nCurrent prompt level: At the generalization stage, use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response.\n\n## Natural Reinforcer\nNatural reinforcer: adult attends to the comment with matching affect.\nAvoid contrived consequences \u2014 the functional outcome IS the reinforcer.\n\n## Generalization Tactics\n- Program multiple exemplars across 5 different stimuli or peers.\n- Rotate between 3 different natural settings.\n- Rotate across 3 different adults during the teaching phase.\n\n## Mastery Criteria\nIndependent use with a peer in unstructured free play across 2 consecutive sessions.\n\n## Data Collection\nOpportunity-based data recording: each naturally-occasioned trial is scored for independence and topography. IOA scheduled every 4 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe across novel peers, settings, and materials. Maintenance probes weekly for 4 weeks, then monthly. Fade adult-delivered prompts before declaring mastery."}], "meta": {"task_type": "teaching_program", "example_id": "60e830acdbc1123b", "gold_labels": {"method": "net", "domain": "VB-MAPP.spontaneous_vocal", "level": "L2", "learner_profile": "school_age", "mastery_state": "generalization"}, "provenance": {"layer": 1, "area": "net", "template_id": "task1_net_v1", "taxonomy_cells": {"skill_target": "spontaneously asks questions about the environment", "mo_category": "spontaneous_comment", "mo_arrangement": "unexpected_change", "prompt_strategy": "verbal_fill_in", "mastery_criterion": "mc_peer_initiation"}, "teacher_model": null, "seed_tag": "77480"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: tacts common actions (running, eating, sleeping)\nCurriculum Reference: Tact L1\nLearner Profile: Early Learner\nCurrent Mastery: Emerging\nTeaching Method: DTT\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: tacts common actions (running, eating, sleeping)\n- Curriculum: Tact L1\n- Learner: Early Learner\n- Current mastery: Emerging\n\nThis DTT program establishes stimulus control over the target skill through structured discrete trials with a defined prompt hierarchy, stimulus array, and systematic error-correction procedure.\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What is he doing?\"\nVariations: \"What action?\"; \"What is she doing?\"; \"What is happening here?\"\nPresentation: Present an action picture, video clip, or live demonstration.\n\n## Prompt Hierarchy\nStrategy: Most-to-Least (errorless)\nSequence: full physical -> partial physical -> gestural -> positional -> independent\nCurrent prompt level: At the emerging stage, begin at higher prompt levels (full-physical or partial-physical) with systematic fading.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: jumping, eating, clapping\nDistractor stimuli: writing, singing\nRotation: Randomize position of target across trials to prevent position bias.\n\n## Error Correction Procedure\nModel correct response -> lead (do together) -> test (independent) -> distractor -> retest.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) \u2014 Every correct response reinforced. Provide differential reinforcement: higher-quality reinforcement for independent correct responses than for prompted correct responses.\n\n## Mastery Criteria\nDemonstrated use in natural environment across 3 independent instances.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel stimuli, across 3 additional therapists, and in at least 3 natural contexts. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "a3ab30beb7b37983", "gold_labels": {"method": "dtt", "domain": "VB-MAPP.tact", "level": "L1", "learner_profile": "early", "mastery_state": "emerging"}, "provenance": {"layer": 1, "area": "dtt", "template_id": "task1_dtt_v1", "taxonomy_cells": {"skill_target": "tacts common actions (running, eating, sleeping)", "prompt_hierarchy": "most_to_least", "reinforcement_schedule": "crf", "error_correction": "four_step", "mastery_criterion": "mc_natural_env"}, "teacher_model": null, "seed_tag": "56565"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You\nanalyze multi-session behavioral session logs for individuals with autism and\nproduce structured clinical interpretations that identify patterns, hypothesize\nbehavior functions when applicable, and recommend programming adjustments\nstructured along BIP lines (antecedent strategies, replacement behaviors,\nconsequence strategies, crisis plan). Your interpretation also includes an\nescalation level and a confidence expression. Every recommendation is grounded\nin the data provided. Follow BACB ethical guidelines and reference no real\nclient data."}, {"role": "user", "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-6006\nProfile: Adolescent Learner (13\u201317 years)\nCurricula: Mand + Reading\nDate range: Sessions 1\u20137 across 14 days (2026-04-27 to 2026-05-11)\n\nACCELERATION PROGRAMS\n1. requesting a break during non-preferred activities (Mand)\n Method: NET\n2. sounds out CVC words (Reading)\n Method: DTT\n3. indicating 'all done' with a step or activity (Mand)\n Method: NET\n4. requesting help when a task is difficult (Mand)\n Method: NET\n\nSESSION DATA\n\nSession 1 \u2014 2026-04-27 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 7/13 correct (51%); latency 3.7s; prompts FP\u00d70, PP\u00d71, G\u00d73, I\u00d79\n sounds out CVC words: 3/8 correct (42%); latency 4.0s; prompts FP\u00d70, PP\u00d71, G\u00d71, I\u00d76\n indicating 'all done' with a step or activity: 5/11 correct (44%); latency 4.1s; prompts FP\u00d73, PP\u00d71, G\u00d73, I\u00d74\n requesting help when a task is difficult: 4/10 correct (44%); latency 4.2s; prompts FP\u00d71, PP\u00d70, G\u00d72, I\u00d77\n\nSession 2 \u2014 2026-04-29 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 8/13 correct (58%); latency 3.3s; prompts FP\u00d74, PP\u00d74, G\u00d73, I\u00d72\n sounds out CVC words: 5/11 correct (48%); latency 3.8s; prompts FP\u00d73, PP\u00d71, G\u00d70, I\u00d77\n indicating 'all done' with a step or activity: 7/14 correct (51%); latency 4.0s; prompts FP\u00d71, PP\u00d72, G\u00d73, I\u00d78\n requesting help when a task is difficult: 6/12 correct (47%); latency 3.9s; prompts FP\u00d74, PP\u00d70, G\u00d72, I\u00d76\n\nSession 3 \u2014 2026-05-01 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 10/15 correct (69%); latency 2.9s; prompts FP\u00d75, PP\u00d73, G\u00d70, I\u00d77\n sounds out CVC words: 8/14 correct (59%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d71, I\u00d78\n indicating 'all done' with a step or activity: 7/12 correct (60%); latency 3.3s; prompts FP\u00d72, PP\u00d73, G\u00d72, I\u00d75\n requesting help when a task is difficult: 5/9 correct (52%); latency 3.5s; prompts FP\u00d72, PP\u00d73, G\u00d73, I\u00d71\n\nSession 4 \u2014 2026-05-03 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 6/8 correct (76%); latency 2.5s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n sounds out CVC words: 7/11 correct (66%); latency 3.1s; prompts FP\u00d70, PP\u00d72, G\u00d71, I\u00d78\n indicating 'all done' with a step or activity: 6/8 correct (70%); latency 2.7s; prompts FP\u00d72, PP\u00d71, G\u00d71, I\u00d74\n requesting help when a task is difficult: 8/13 correct (60%); latency 3.2s; prompts FP\u00d73, PP\u00d72, G\u00d70, I\u00d78\n\nSession 5 \u2014 2026-05-05 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 9/11 correct (80%); latency 2.2s; prompts FP\u00d73, PP\u00d70, G\u00d71, I\u00d77\n sounds out CVC words: 8/11 correct (75%); latency 2.4s; prompts FP\u00d71, PP\u00d73, G\u00d70, I\u00d77\n indicating 'all done' with a step or activity: 11/14 correct (79%); latency 2.7s; prompts FP\u00d71, PP\u00d70, G\u00d74, I\u00d79\n requesting help when a task is difficult: 5/8 correct (64%); latency 3.1s; prompts FP\u00d71, PP\u00d71, G\u00d71, I\u00d75\n\nSession 6 \u2014 2026-05-07 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 8/10 correct (80%); latency 2.5s; prompts FP\u00d72, PP\u00d72, G\u00d73, I\u00d73\n sounds out CVC words: 7/8 correct (83%); latency 2.2s; prompts FP\u00d70, PP\u00d70, G\u00d70, I\u00d78\n indicating 'all done' with a step or activity: 10/13 correct (79%); latency 2.5s; prompts FP\u00d70, PP\u00d74, G\u00d72, I\u00d77\n requesting help when a task is difficult: 8/12 correct (70%); latency 3.1s; prompts FP\u00d71, PP\u00d73, G\u00d74, I\u00d74\n\nSession 7 \u2014 2026-05-09 \u2014 49 min \u2014 1 observer\n requesting a break during non-preferred activities: 11/14 correct (80%); latency 2.3s; prompts FP\u00d72, PP\u00d73, G\u00d74, I\u00d75\n sounds out CVC words: 12/14 correct (83%); latency 2.3s; prompts FP\u00d72, PP\u00d74, G\u00d74, I\u00d74\n indicating 'all done' with a step or activity: 7/9 correct (79%); latency 2.7s; prompts FP\u00d72, PP\u00d70, G\u00d73, I\u00d74\n requesting help when a task is difficult: 10/14 correct (71%); latency 2.6s; prompts FP\u00d73, PP\u00d73, G\u00d73, I\u00d75\n\nBEHAVIORAL OBSERVATIONS (across sessions)\n- Echolalic responding (repeating SD rather than answering)\n- Looking away during SD presentation\n- Stereotypic behavior during inter-trial intervals"}, {"role": "assistant", "content": "## Clinical Concerns\n- **Skill loss following break**: performance dropped to 51% at return, recovering to 80%.\n- Maintenance schedule should be reviewed and strengthened.\n\n## Pattern Classification\nskill_loss_after_break\n\nSharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Implement booster sessions at the most-recently-effective prompt level.\n- Increase session frequency temporarily until pre-break levels are recovered.\n\n### Consequence strategies\n- Review and strengthen the maintenance schedule to prevent future skill loss.\n- Consider sending practice materials home during extended breaks.\n\n## Escalation Level\n2 \u2014 Adjust next session\n\nProgramming adjustments indicated at next session; no immediate safety concern.\n\n## Confidence\nmoderate\n\nPattern is supported but alternatives cannot be fully excluded: 7 sessions of data; no IOA session included.\n\n## Data-Supported Rationale\n- Mean accuracy across 7 sessions: 70.5%.\n- Trend direction: upward (first session: 51%, last session: 80%).\n- No IOA session in this log; reliability of measurements cannot be independently verified."}], "meta": {"task_type": "session_interpretation", "example_id": "49110f5cf98a72c3", "gold_labels": {"pattern_class": "skill_loss_after_break", "behavior_functions": {}, "escalation_level": 2, "confidence": "moderate", "crisis_plan_required": false}, "provenance": {"layer": 1, "area": "session_interpretation", "template_id": "task2_session_interp_v1", "taxonomy_cells": {"hidden_pattern_id": "skill_loss_after_break", "learner_profile": "adolescent", "n_sessions": 7, "n_programs": 4, "n_behaviors": 0, "behavior_ids": [], "behavior_functions": [], "has_abc_data": false, "has_ioa_session": false}, "teacher_model": null, "seed_tag": "28791"}}} +{"messages": [{"role": "system", "content": "You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help\nBoard Certified Behavior Analysts and staff design teaching programs for\nindividuals with autism. Your responses are clinically accurate, individualized\nto the learner profile, follow BACB ethical guidelines, and reference no real\nclient data. Select the appropriate teaching method (DTT, NET, Task Analysis,\nFCT, BST, PRT) based on the skill target and learner profile."}, {"role": "user", "content": "Generate a teaching program for the following target.\n\nSkill Target: meal cleanup\nCurriculum Reference: AFLS Home Skills\nLearner Profile: Adult Learner\nCurrent Mastery: Approaching mastery\nTeaching Method: Task Analysis / Chaining\n\nProvide the full program structure appropriate to the selected method."}, {"role": "assistant", "content": "## Program Overview\n- Target skill: meal cleanup\n- Curriculum: AFLS Home Skills\n- Learner: Adult Learner\n- Current mastery: Approaching mastery\n\nThis Task Analysis program teaches the target skill as a chain of discrete steps, each taught and reinforced until the learner can complete the full sequence independently.\n\n## Task Analysis\nChain type: Total-task presentation\nPrompt the learner through the entire chain every trial; fade prompts across all steps simultaneously.\n\nSteps:\n1. Stack used dishes on one side of the table.\n2. Carry dishes to the sink or dishwasher.\n3. Rinse food residue into the trash or disposal.\n4. Load dishes into the dishwasher (or wash by hand).\n5. Wipe the table with a damp cloth.\n6. Put away any leftover food into containers.\n7. Place containers in the refrigerator.\n\n## Prompt Strategy Per Step\nStrategy: Least-to-Most\nAllow independent attempt; add progressively more intrusive prompts only if learner does not initiate.\nCurrent prompt level: At the approaching mastery stage, use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere.\n\n## Error Correction Procedure\nTask reset: Pause the chain; reset environmental stimuli; restart the chain from the beginning with full prompts at the previously-missed step\n\n## Reinforcement\nCRF per step (acquisition) \u2014 Continuous reinforcement for each correctly-completed step during acquisition.\nPair the terminal reinforcer with the natural consequence of the completed task (e.g., clean hands, clean dishes, completed routine).\n\n## Mastery Criteria\nFull chain completed within 180s across 3 consecutive sessions with independent responding on all steps.\n\n## Data Collection\nPer-step data: each step is scored for independence / prompt level on every trial. IOA scheduled every 3 sessions minimum; target agreement \u2265 80%.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe the full chain across 3 different settings and with 2 different adults. Maintenance probes weekly for 4 weeks, then monthly."}], "meta": {"task_type": "teaching_program", "example_id": "348caeabcfed6cfc", "gold_labels": {"method": "task_analysis", "domain": "AFLS.home", "skill": "meal cleanup", "program_type": "independence", "learner_profile": "adult", "mastery_state": "approaching", "chain_type": "total_task"}, "provenance": {"layer": 1, "area": "task_analysis", "template_id": "task1_task_analysis_v1", "taxonomy_cells": {"skill_target": "meal cleanup", "module": "home", "mastery_criterion": "mc_fluent_chain", "chain_type": "total_task", "prompt_strategy": "least_to_most", "reinforcement_schedule": "crf_per_step", "error_correction": "task_reset"}, "teacher_model": null, "seed_tag": "6490"}}} diff --git a/docs/curation/LEGEND.md b/docs/curation/LEGEND.md new file mode 100644 index 0000000000000000000000000000000000000000..8b2469fee26ecb308c35e98e4b41b7b31fb2c862 --- /dev/null +++ b/docs/curation/LEGEND.md @@ -0,0 +1,177 @@ +# Session-Log Reading Guide + +Reference for reading Task 2 (session-interpretation) examples. Keep this +open in a side pane while browsing `review.md`. + +--- + +## Session header line + +``` +Session 3 — 2026-10-10 — 45 min — 1 observer +``` + +Or, for inter-observer agreement sessions: + +``` +Session 5 — 2026-10-14 — IOA SESSION — 2 observers +``` + +| Piece | Meaning | +|---|---| +| `Session N` | Nth observation session in this log | +| `2026-10-10` | Synthetic date (always in 2026 range) | +| `45 min` | Session duration | +| `1 observer` / `2 observers` | Single observer (primary) or IOA session | +| `IOA SESSION` | (If present) agreement check; behavior lines in this session show a trailing `IOA X% agreement` | + +--- + +## Skill data line (acquisition programs) + +``` +ordering in a restaurant: 9/13 correct (67%); latency 3.1s; prompts 5 +``` + +| Piece | Meaning | +|---|---| +| `9/13 correct` | Correct trials / total trials | +| `67%` | Accuracy | +| `latency 3.1s` | Mean response latency after SD presentation | +| `prompts 5` | Number of trials on which any prompt was delivered | + +--- + +## Behavior data lines + +Each target behavior gets its own measurement format matched to the +behavior's clinical shape. Generic `freq` lines apply to most behaviors; +behaviors with clinically distinctive shapes have behavior-specific +measurements. + +### Generic frequency behaviors + +``` +Aggression: freq 3 +Elopement: freq 2 +SIB: freq 5 +Property destruction: freq 1 +Non-compliance: freq 4 +Verbal aggression: freq 2 +``` + +`freq N` = N occurrences this session. + +### Tantrum (includes duration) + +``` +Tantrum: freq 2, duration 7m total +``` + +### Stereotypy and mouthing (include partial-interval recording) + +``` +Motor stereotypy: freq 8; PIR 18% +Vocal stereotypy: freq 5; PIR 12% +Mouthing: freq 6; PIR 15% +``` + +`PIR X%` = partial-interval recording: % of intervals in which the behavior +occurred at any point. + +### Pica (attempts vs successful ingestion) + +``` +Pica: attempts 3 (2 unsuccessful — staff retrieved item before ingestion; 1 successful — item ingested) +``` + +Staff often intercept pica attempts; tracking attempts / successful +separately preserves the severity signal that a raw frequency loses. + +### Fecal smearing / scatolia (attempts vs completed smearing) + +``` +Fecal smearing (scatolia): attempts 2 (1 intercepted — staff redirected before smearing; 1 completed — feces transferred to skin, clothing, or surface) +``` + +Same intercepted / completed split as pica — clinically critical because +staff responsiveness directly shapes outcome severity. + +### Toileting (four-count voiding log) + +``` +Toileting accident (urine or bowel): urine: 3 in-toilet / 2 accidents; BM: 0 in-toilet / 1 accidents +``` + +Mirrors a standard clinical voiding log. The deceleration target is +**accidents** (urine + BM), but successful in-toilet voids are tracked +alongside for context: + +- `urine: X in-toilet / Y accidents` — successful urinations vs. urine accidents +- `BM: P in-toilet / Q accidents` — successful bowel movements vs. BM accidents + +### IOA annotation + +On sessions marked `IOA SESSION`, each behavior line ends with `; IOA X% agreement`: + +``` +Aggression: freq 3; IOA 88% agreement +``` + +--- + +## ABC line + +``` +ABC (elopement): A = peer took toy; B = ran from room; C = staff retrieved learner +``` + +- `A` = Antecedent (what happened immediately before the behavior) +- `B` = Behavior (operational description) +- `C` = Consequence (what happened immediately after) + +ABC evidence feeds the behavior-function hypothesis in the assistant's +response (escape / attention / tangible / automatic, per Iwata et al. 1994 +and Hanley, Iwata, & McCord 2003). + +--- + +## Function-hypothesis line (log header) + +Near the top of a log you'll see one line per tracked behavior: + +``` +1. Fecal smearing (scatolia) — function hypothesized: automatic +2. Aggression — function hypothesized: escape +``` + +These are the sampled (gold) functions that the interpretation response +should corroborate with the evidence in the log. + +--- + +## Program list + +``` +Programs tracked this session block: +1. ordering in a restaurant (AFLS Community) — task_analysis +2. mands for a break (VB-MAPP Mand L1) — net +3. tacts colors of objects (VB-MAPP Tact L2) — dtt +``` + +Each entry names the skill target, its curriculum location, and the +teaching method. + +--- + +## Behavioral indicator block (across sessions) + +``` +BEHAVIORAL OBSERVATIONS (across sessions) +- Increased response latency (3–5× baseline) +- Pushing materials away from the work area +- Vocal refusal ("no", "I don't want to") +``` + +These are pattern-specific indicator clusters (frustration, engagement, or +disengagement) sampled from `the `behavioral_indicators` block in configs/session_interpretation/taxonomy.yaml`. diff --git a/docs/curation/README.md b/docs/curation/README.md new file mode 100644 index 0000000000000000000000000000000000000000..5e07f0de858fa969fac139a6e524584e2e1befdf --- /dev/null +++ b/docs/curation/README.md @@ -0,0 +1,93 @@ +# Curation Workflow + +The curation step turns the pipeline-held-out `curation_pool.jsonl` into the +two evaluation files shipped in `data/splits/`: + +- **`test.jsonl`** — evaluation set. The paper's headline metrics come from this set. +- **`sanity.jsonl`** — 20-example smoke-test set for development. + +In TRACE v1 the whole curation pool is promoted as the test corpus (no hand +curation). The **review + compile** scripts are still part of the pipeline +because (a) regenerating the test and sanity splits is deterministic from the +pool, and (b) anyone adapting TRACE to a new clinical domain can reuse the +same workflow with their own taxonomy. + +--- + +## Two scripts + +| Script | What it does | +|---|---| +| `src/prepare_curation.py` | Renders `curation_pool.jsonl` as a browseable Markdown document in `docs/curation/review.md` — one candidate per section, grouped by task × category, with gold labels and provenance visible. | +| `src/compile_curation.py` | Splits the curation pool into `data/splits/test.jsonl` (the remainder) and `data/splits/sanity.jsonl` (20 examples, largest-remainder stratified by category). Deterministic under `--seed`. | + +--- + +## Commands + +```bash +# Regenerate the browseable review document. +uv run python src/prepare_curation.py + +# Compile the test + sanity splits from the pool. +uv run python src/compile_curation.py + +# Override the sanity size (default 20) or seed (default 42). +uv run python src/compile_curation.py --sanity-n 20 --seed 42 +``` + +After `compile_curation.py` runs you'll have: +- `data/splits/test.jsonl` — the evaluation set +- `data/splits/sanity.jsonl` — the smoke-test set + +--- + +## Reading `review.md` + +`review.md` renders each candidate with: +- A heading showing the task type, category, and `example_id`. +- Gold labels inline (method, domain, level, learner profile, mastery state; + or pattern class, behavior functions, escalation, confidence, crisis-plan flag). +- A short provenance line (for session interpretation: number of sessions, + number of behaviors, whether IOA is included, whether ABC is included). +- The full user message (the teaching-program prompt or the session log). +- The full assistant message (the structured response). + +Use `LEGEND.md` as a side-pane reference for the session-log notation in +Task 2 examples. + +--- + +## Flagging issues for re-generation + +If you spot a clinical inaccuracy while browsing `review.md`: + +1. Note the candidate's `example_id` (shown in its heading). +2. Look up the candidate's `meta.provenance.taxonomy_cells` in the JSONL to + identify which taxonomy dimension produced the issue. +3. Edit the relevant YAML under `configs/` (or the renderer in `src/generators/` + if it's a rendering-logic issue) — *never* hand-edit individual JSONL + entries, because the fix should propagate to every example that sampled + the same cells. +4. Regenerate the corpus with `src/generate.py --all`, re-split with + `src/split_data.py`, and re-compile with `src/compile_curation.py`. + +Every clinical-review flag during v1 development landed as a single +targeted taxonomy edit plus a full regeneration — never a hand-edit of +the JSONL. This is the invariant the pipeline relies on: fixes must +propagate through the taxonomy to be systematic across the corpus. + +--- + +## Adapting the workflow to a new clinical domain + +The scripts are domain-agnostic. To port to a different dataset: + +- `prepare_curation.py` just reads `data/splits/curation_pool.jsonl` and + renders it. Works on any JSONL with the same envelope shape. +- `compile_curation.py` stratifies on `category_of(example)` — a small + function near the top of the script. Change its logic to use whichever + field best represents your category (the `method` field for TRACE's + teaching programs; the `pattern_class` field for session interpretation). +- Everything else — largest-remainder allocation, deterministic splitting, + provenance preservation — carries over unchanged. diff --git a/docs/data-statement.md b/docs/data-statement.md new file mode 100644 index 0000000000000000000000000000000000000000..d544123fb5c94bfbdab157d341245c632812dedb --- /dev/null +++ b/docs/data-statement.md @@ -0,0 +1,126 @@ +# Data Statement for TRACE (v1) + +**TRACE** — **T**axonomy-**R**eferenced **A**BA **C**linical **E**xamples + +Template: Bender, E. M., & Friedman, B. (2018). *Data Statements for Natural Language Processing: Toward Mitigating System Bias and Enabling Better Science.* Transactions of the Association for Computational Linguistics. https://aclanthology.org/Q18-1041/ + +--- + +## A. Curation Rationale + +### Why this data, and why this composition? +Applied Behavior Analysis (ABA) is a clinical discipline with high documentation workload — BCBAs produce teaching programs and interpret multi-session behavioral logs continuously across their caseloads. Existing general-purpose language assistants are both unspecialized to ABA's structured conventions (operational definitions, VB-MAPP / AFLS curricula, functional-analysis frameworks) and unvetted against its clinical standards (BACB Ethics Code 2020; ABAI 2010 Position Statement on Restraint). This dataset was curated to support fine-tuning a small on-device language model (Gemma 4 E2B) specifically for **drafting** — i.e., producing first-pass teaching programs and session interpretations that a BCBA then reviews and revises — without any ingestion of real client data. + +Two tasks are represented because they are the two highest-frequency authorial tasks in an ABA clinical team's week: (1) drafting teaching programs for new or evolving skill targets, and (2) interpreting session-by-session behavioral data to adjust programming. They share a taxonomy (learner profiles, mastery states, teaching methods, target behaviors) but diverge in structure (teaching-program responses are single-program documents; session-interpretation responses are diagnostic summaries). + +### Why synthetic? +Real ABA session data contains 45 CFR 164 protected health information (PHI) under HIPAA and is additionally subject to BACB Ethics Code confidentiality rules (section 2.03, section 2.05). Any public dataset drawn from real client data would require individual consent, facility release, and de-identification that cannot reliably preserve the clinical detail required for training. Synthetic generation avoids these constraints by construction: the data never represented a real person in the first place. The trade-off — distributions may not match any one real caseload — is made explicit in the dataset card (section 6.5) and flagged to users. + +--- + +## B. Language Variety + +All text is in **English (en-US)**, written in standard American clinical register. The authorial style follows the conventions of: +- **Peer-reviewed ABA literature** — *Journal of Applied Behavior Analysis* (JABA), *Behavior Analysis in Practice* (BAP). +- **Reference textbook** — Cooper, Heron, & Heward (2020), *Applied Behavior Analysis* (3rd ed.). +- **Professional documentation** — BACB Ethics Code (2020), ABAI Position Statements. + +Within that register, the dataset covers: +- **Teaching-program language** — imperative instructional prose (stimulus control description, prompt hierarchy specification, reinforcement-schedule specification, error-correction procedure, mastery criterion). +- **Session-log language** — telegraphic data-sheet notation (`accuracy 6/10 (60%); freq 2; duration 3m; IOA 88%`) mixed with ABC-entry prose. +- **Clinical-interpretation language** — structured prose matching the assistant template's headings (Clinical Concerns, Pattern Classification, Behavior Function Hypothesis, Programming Recommendations, Crisis Plan when applicable, Confidence). + +There is no dialectal variation, no code-switching, no second-language speaker voice. Generalization to non-US clinical conventions (BCBA-D vs. international RBT frameworks, different curricula) is explicitly out of scope for v1. + +--- + +## C. Speaker Demographic + +**Not applicable.** The dataset has no speakers. All content is programmatically generated from taxonomy configs written by a single author with ABA domain fluency. The written conventions inherit from the sources listed in section B. + +--- + +## D. Annotator Demographic + +**Not applicable in the conventional sense.** No human annotators labeled instances after generation. Gold labels are emitted at generation time from the same taxonomy cells that produced the example — they are by-construction the ground truth of the sampling step. + +**Quasi-annotator role: clinical-accuracy reviewer.** One individual (the dataset author) performed ad-hoc clinical-accuracy review of generated candidates during iteration. This reviewer: + +- Has practitioner-adjacent exposure to ABA clinical settings (not a BCBA; not a BCaBA; not an RBT in active practice). +- Is a native English speaker, fluent in the English clinical-documentation register. +- Performed review by browsing full-text candidate renderings and flagging clinical implausibilities against published sources (CHH 2020; BACB Ethics Code 2020; JABA papers for operational definitions). + +**What this means for the dataset:** +- Clinical-accuracy review was grounded in published sources, not in one reviewer's clinical judgment alone. Where a reviewer uncertainty could not be resolved from a source, the item was retained with a flag rather than decided by individual preference. +- Full BCBA-level clinical validation (multi-reviewer, κ-scored) was not performed on v1. +- Bias from the reviewer's particular exposure (specific facility types, specific learner age ranges, specific behavioral topographies) is possible. The taxonomy's grounding in CHH and JABA is the mitigating control. + +--- + +## E. Speech Situation + +**Not applicable** (no speech). The imagined reader/user for the generated text: +- **Teaching programs** — a BCBA or behavior technician reading a first-pass program draft for clinical review, before it is finalized and implemented with a learner. +- **Session interpretations** — a BCBA or supervising analyst reading a summary of the most recent N sessions of a learner's data, as a decision-support document for programming adjustments. + +The modality is written documentation, asynchronous, read in quiet clinical-office settings. The register is clinical-professional. + +--- + +## F. Text Characteristics + +### Genre +- Clinical documentation (teaching programs, behavioral session interpretations). +- Instructional prose interleaved with structured data-sheet notation. + +### Structural Conventions +- Teaching programs follow a fixed section layout (Program Overview, Stimulus Control, Prompt Hierarchy, Reinforcement, Error Correction, Mastery Criteria, Generalization & Maintenance Plan). Variants exist for DTT, NET, and Task Analysis; Task Analysis has a distinct toleration-program layout. +- Session logs are rendered as pseudo-data-sheet blocks, one block per session, with per-program accuracy lines, per-behavior measurement lines, and optional ABC / IOA entries. +- Session interpretations follow a fixed layout (Clinical Concerns, Pattern Classification, Behavior Function Hypothesis, Programming Recommendations, Crisis Plan, Confidence). + +### Domain-specific notation +- ABA abbreviations used freely: SD (discriminative stimulus), MO (motivating operation), ABC (antecedent-behavior-consequence), IOA (interobserver agreement), PIR (partial-interval recording), CRF (continuous reinforcement), FR-2, VR-3, etc. +- Measurement notation follows the schema in `schema-v1.md section 2.3`. + +### Distribution characteristics +- Target behaviors: sampled per pattern (0–3 per log). +- Accuracy trajectories: driven by pattern-specific generators (ascending for mastery_progression, descending for regression, etc.). +- Behavior frequencies: pattern-trajectory-driven; bounded. +- Severity: per-behavior `typical_severity` flag (low/moderate/high) feeds escalation logic. + +### What the dataset *does not* contain +- Personal names. +- Real identifiers. +- Dates outside 2026-01-01 to 2026-12-31. +- Non-English text. +- Informal register (chat-style, SMS, social media). +- Off-topic content. +- Adversarial or jailbreak examples. +- Real session notes or documentation. + +--- + +## G. Recording Quality + +**Not applicable** (no recording). + +--- + +## H. Other + +### Provenance guarantee +Every example carries full sampling provenance in `meta.provenance.taxonomy_cells`. This is the artifact that makes clinical auditing tractable: when a reviewer flags an implausible example, they can trace it to the exact cells that generated it and propose a taxonomy or compatibility-rule fix. This is a property of the dataset we consider essential and want future versions to preserve. + +### Known biases +- **Caseload-distribution bias.** Pattern frequencies are uniform across the 12 session-interpretation patterns for learnability, whereas real caseloads skew toward mastery_progression. Users training epidemiological models should reweight. +- **Curriculum bias.** VB-MAPP and AFLS are the only curricula covered. Practitioners using ABLLS-R, Essential for Living, PEAK, or other curricula should adapt. +- **Register bias.** All text is US clinical-professional. Non-US or non-clinical register is not represented. +- **Reviewer-exposure bias.** Clinical review was performed by a single reviewer. See section D. +- **Synthetic-distribution bias.** Sampling weights reflect clinical frequency where known but are approximate. + +### Transparency commitments +- Full generator code is public in the repository. +- Full taxonomy configs are public in the repository. +- Dataset regeneration from `(configs, seed)` is deterministic. +- Every version of the dataset is pinned to a specific repository commit hash. +- All citations used to ground taxonomy content are enumerated in `docs/research/citations-to-use.md`. diff --git a/docs/dataset-card.md b/docs/dataset-card.md new file mode 100644 index 0000000000000000000000000000000000000000..b96b0d6bfbc02903cf710329af16367072223821 --- /dev/null +++ b/docs/dataset-card.md @@ -0,0 +1,358 @@ +--- +pretty_name: "TRACE — Taxonomy-Referenced ABA Clinical Examples" +license: cc-by-nc-4.0 +language: + - en +size_categories: + - 1K"}, + {"role": "user", "content": ""}, + {"role": "assistant", "content": ""} + ], + "meta": { + "task_type": "teaching_program" | "session_interpretation", + "example_id": "", + "gold_labels": { ... task-specific labels ... }, + "provenance": { + "layer": 1, + "area": "dtt" | "net" | "task_analysis" | "session_interpretation", + "template_id": "...", + "taxonomy_cells": { ... exact sampled values ... }, + "teacher_model": null, + "seed_tag": "...", + "generated_at": "..." + } + } +} +``` + +Only `messages` is used for training (mask_prompt: true). `meta` is preserved for evaluation and provenance tracking. + +### 3.2 Gold labels per task + +**Teaching program:** +``` +{ + "method": "dtt" | "net" | "task_analysis", + "domain": "VB-MAPP." | "AFLS.", + "level": "L1" | "L2" | "L3" | ..., # VB-MAPP only + "learner_profile": "early" | "school_age" | "adolescent" | "adult", + "mastery_state": "emerging" | "developing" | "approaching" | "near" | "mastered" | "generalization", + "program_type": "independence" | "toleration", # Task Analysis only + "chain_type": "forward" | "backward" | "total_task" # Task Analysis only +} +``` + +**Session interpretation:** +``` +{ + "pattern_class": "", + "behavior_functions": { "": "escape"|"attention"|"tangible"|"automatic"|"unknown" }, + "escalation_level": 1 | 2 | 3 | 4, + "confidence": "high" | "moderate" | "low", + "crisis_plan_required": true | false +} +``` + +### 3.3 Provenance + +Every sampled choice is recorded in `meta.provenance.taxonomy_cells`. For teaching programs this includes the skill target, prompt hierarchy, reinforcement schedule, error correction, and mastery criterion. For session interpretation it includes the pattern, number of sessions, behaviors sampled, functions inferred, and whether IOA and ABC data are present. This enables: + +- **Traceability:** any clinical issue in a generated example can be traced back to the exact taxonomy cell that produced it. +- **Reproducibility:** the dataset can be regenerated deterministically from the configs and seed. +- **Stratification:** splits can be regenerated with different stratification keys without regenerating the corpus. +- **Ablation:** subsets can be constructed by filtering on provenance cells. + +--- + +## 4. How It Was Built + +### 4.1 Architecture + +``` +configs/ +├── shared/ # cross-area primitives (learner profiles, mastery states, prompt types) +├── dtt/ # DTT area: taxonomy + template + compatibility +├── net/ # NET area +├── task_analysis/ # Chaining area (independence + toleration templates) +└── session_interpretation/ # Session interp area: + # taxonomy + compatibility + trajectory_rules + # + recommendations (per-pattern bullets) + template +src/generators/ +├── aba_dtt.py # per-area generator +├── aba_net.py +├── aba_task_analysis.py +└── aba_session_interp.py +src/generate.py # orchestrator +``` + +### 4.2 Generation loop + +For each example: +1. Sample a clinical configuration from the taxonomy, weighted by realistic clinical frequency (level weights, module weights, pattern weights). +2. Apply compatibility rules (e.g., DTT errorless error correction pairs only with most-to-least prompting; certain patterns have bias toward certain behavior functions). +3. Compute template slots — current-prompt-level guidance from mastery state; MO arrangement from skill keywords; mastery criterion; reinforcement schedule. +4. Render `user_content` from a random user-variant. +5. Render `assistant_content` by filling `{slots}` in the assistant template. +6. Stamp `meta.gold_labels` and `meta.provenance`. +7. Write as one JSONL line. + +### 4.3 Grounding + +- **Skill curricula** — VB-MAPP milestones (Sundberg 2008) for verbal-behavior domains; AFLS (Partington & Mueller) for adaptive-living domains. +- **Teaching methods** — DTT (Lovaas 1987; Smith 2001); NET (Hart & Risley 1975 in CHH Ch. 18); Task Analysis / chaining (CHH Ch. 20). +- **Operational definitions of target behaviors** — Cooper/Heron/Heward Ch. 3, 27; key JABA papers (Iwata et al. 1994 for SIB and functional analysis; Carr & Durand 1985 for FCT; Hanley, Iwata, & McCord 2003 for FBA). +- **Session patterns** — derived from CHH Ch. 6–7 (analyzing behavior change) and Stokes & Baer 1977 (generalization). +- **Crisis plans** — BACB Ethics Code (2020) section 3.05; ABAI Position Statement on Restraint and Seclusion (2010). Physical-intervention procedures are left vague on purpose because they vary by training program (Safety-Care, CPI, PMT, TCI) and jurisdiction; the dataset emphasizes verbal de-escalation, environmental safety, BIP authorization, and contraindications. + +### 4.4 Clinical-accuracy pipeline + +Initial generation produced the structural skeleton. The corpus was then iterated via **practitioner-in-the-loop review** — a full-text render of the held-out candidate pool was browsed by a reviewer with ABA practitioner exposure, and each flagged clinical inaccuracy was traced to the responsible taxonomy cell and fixed with a single targeted edit plus a full regeneration. Because every example records its sampling provenance, a single cell-level edit propagates to every example that sampled the affected cells — so flagging one example systematically corrects a class of examples. + +--- + +## 5. Intended Uses + +### 5.1 Direct use +- Instruction-tuning a small language model (recommended: Gemma 4 E2B 4-bit with QLoRA) to draft ABA teaching programs and interpret session logs. +- Evaluation of task-specific competencies using the held-out `test.jsonl` (281 examples). +- Research on taxonomy-driven synthetic data generation for clinical decision support. + +### 5.2 Downstream use +- Research on small-LM drafting assistants in structured clinical-documentation domains. +- Comparison baselines for future ABA-specific LLM work (Kumar et al. 2024 "Personalized-ABA" is the closest direct predecessor; the present dataset extends to structured program generation and session-log interpretation). + +### 5.3 Out-of-scope (do NOT use for) +- Autonomous clinical decisions. +- Writing final Behavior Intervention Plans without BCBA review. +- Training models on or combined with real client records (the pipeline and schema are explicitly designed to avoid this). +- Legal or insurance documentation. +- Medical diagnosis. + +--- + +## 6. Ethics & Risks + +### 6.1 Provenance +Every example is synthetic. No real client data, no real session notes, no real identifiers were used at any step. Learner references use synthetic IDs (`SYN-####`); dates fall in the range 2026-01-01 to 2026-12-31. + +### 6.2 Clinical-risk framing +The dataset is designed around a **draft-and-review** authoring pattern. Assistant responses are structured so that a reviewer can quickly see the method, the stimulus arrangement, the prompt hierarchy, the reinforcement plan, the error-correction procedure, the mastery criterion, and the generalization plan — each as a distinct, scannable section. Session-interpretation responses surface a confidence level (high / moderate / low) and an escalation level (1–4) as structured fields. These are design choices that support auditability; they are not clinical advice, and TRACE's responsibility disclaimer applies. + +### 6.3 Crisis-plan sensitivity +Crisis plans were written against the ABAI 2010 Position Statement on Restraint and Seclusion and BACB Ethics Code 2020 section 3.05. The dataset references facility crisis-prevention frameworks (Safety-Care, CPI, PMT, TCI) only as examples and **deliberately avoids specifying restraint procedures** because those procedures are (a) jurisdiction-dependent, (b) training-certification-gated, and (c) learner-specific (many learners have contraindications). The dataset embeds explicit text in every crisis-plan bullet that physical intervention is used only when specifically authorized in the learner's BIP and only by staff currently certified in the facility's training program. + +### 6.4 Population representation +Learner profiles are intentionally abstract (early / school-age / adolescent / adult). The dataset does not encode demographic categories (race, socioeconomic status, gender identity) and does not attempt to characterize clinical presentations by such categories. This is a deliberate choice for a first release; future versions may add representation if grounded in published demographic work. + +### 6.5 Known limitations +- **English only.** Teaching method terminology, curriculum targets, and session-log conventions are rendered in English. +- **Synthetic distributions.** Pattern frequencies are approximately uniform for learnability; real clinical practice has different frequencies (mastery-progression is far more common than frustration-pattern in a healthy caseload). The dataset is explicitly a teaching set, not an epidemiological sample. +- **VB-MAPP + AFLS only.** Other curricula (ABLLS-R, Essential for Living, PEAK) are not covered. Practitioners using those curricula should adapt. +- **No longitudinal data.** Sessions within a log are temporally ordered but the pipeline does not model real continuity over months or years. +- **Toleration covers hygiene only.** Other toleration programs (e.g., wearing glasses, riding in a car seat) are not represented. +- **Toilet-training acquisition is out of scope.** Accidents are tracked in session logs and bathroom-requesting is taught as a NET mand, but the full Azrin & Foxx 1971 rapid toilet-training acquisition protocol is not included as a task-analysis program. + +### 6.6 License +**Data:** CC BY-NC 4.0 — research and non-commercial use with attribution. **Code:** MIT. + +**Responsibility.** TRACE is a research artifact. It is not a clinical tool, has not been clinically validated, and carries no clinical endorsement. Anyone who chooses to deploy TRACE — or any model derived from it — in a clinical setting does so entirely at their own responsibility and under their facility's own oversight. The authors and Pombo Labs make no representation of clinical suitability and accept no liability for clinical outcomes. + +--- + +## 7. Reproducibility + +The corpus is regenerated deterministically from: +- `configs/` (all YAMLs) +- `configs/generation.yaml` (seed and per-area counts) +- `src/generators/*.py` (generator code) + +```bash +uv run python src/generate.py --all # regenerate 3000 examples +uv run python src/split_data.py # stratified split +uv run python src/prepare_curation.py # browseable review.md +uv run python src/compile_curation.py # test.jsonl + sanity.jsonl +``` + +The dataset version is `v1.0.0`; the matching git tag pins the exact configs and generator code that produced the published JSONL splits. + +--- + +## 8. Citation + +Please cite as: + +> Kahunla, F. (2026). *TRACE: Taxonomy-Grounded Synthetic Data for Teaching Program Generation and Session Interpretation in Applied Behavior Analysis.* Pombo Labs. https://github.com/Pombo-Labs/TRACE + +Machine-readable metadata: `CITATION.cff`. + +--- + +## 9. Appendices + +- **Datasheet** (Gebru et al. 2021 format): `datasheet.md` +- **Data statement** (Bender & Friedman 2018 format): `data-statement.md` +- **Taxonomy reference** (operational definitions + citations): `taxonomy-v1.md` +- **Schema reference** (wire format + slot specifications): `schema-v1.md` diff --git a/docs/datasheet.md b/docs/datasheet.md new file mode 100644 index 0000000000000000000000000000000000000000..a2de8cb7816d847db1539597b8ec3c1278c2a63c --- /dev/null +++ b/docs/datasheet.md @@ -0,0 +1,220 @@ +# Datasheet for TRACE (v1) + +**TRACE** — **T**axonomy-**R**eferenced **A**BA **C**linical **E**xamples + +Template: Gebru, T., Morgenstern, J., Vecchione, B., Wortman Vaughan, J., Wallach, H., Daumé III, H., & Crawford, K. (2021). *Datasheets for Datasets.* Communications of the ACM. https://arxiv.org/abs/1803.09010 + +--- + +## Motivation + +### Why was the dataset created? +To provide the first public, open-access synthetic instruction-tuning dataset for two tasks in Applied Behavior Analysis (ABA): (1) teaching program generation and (2) multi-session behavioral interpretation. Board Certified Behavior Analysts (BCBAs) draft large numbers of these documents weekly; current LLM assistants are both unspecialized to ABA and unvetted against its clinical standards. The dataset supports fine-tuning a small, on-device language model (Gemma 4 E2B at 2B parameters) for drafting support, preserving clinical oversight at the point of deployment. + +### Who funded the creation of the dataset? +No external funding. Authored by Festus Kahunla (Drexel University) and released publicly under the **Pombo Labs** organization (https://github.com/Pombo-Labs). + +### Any other comments? +The dataset's primary novelty is its **taxonomy-driven construction with complete provenance**: every example can be traced back to a specific set of taxonomy cells that produced it. This is a deliberate choice to make clinical validity auditable. + +--- + +## Composition + +### What do the instances represent? +Two instance types: + +1. **Teaching-program instances** (1,800 total) — a user message describes a learner profile and a skill target; an assistant message produces a structured teaching program. Covers three methods: Discrete Trial Training (DTT, 800), Natural Environment Teaching (NET, 500), Task Analysis / chaining (500). + +2. **Session-interpretation instances** (1,200 total) — a user message contains a multi-session behavioral log (5–12 sessions, with per-session accuracies, target behaviors with measurements, and optional ABC and IOA data); an assistant message produces clinical concerns, a pattern classification, a function hypothesis, recommendations, and when applicable a crisis plan. + +### How many instances are there in total? +2,999. + +### Does the dataset contain all possible instances or is it a sample? +The dataset is a **programmatic sample** from a combinatorial space defined by the taxonomy. The space is formally enumerable (domains × levels × learner profiles × mastery states × prompt hierarchies × reinforcement schedules × error corrections × mastery criteria) but is far larger than 2,999 examples. Sampling weights reflect approximate clinical frequency where known (e.g., VB-MAPP Level 1 is sampled more heavily than Level 3 for DTT because early learners dominate the data space in practice). + +### What data does each instance consist of? +See `schema-v1.md` and `dataset-card.md section 3`. Each instance is one JSONL line with two top-level keys: +- `messages` — a three-message chat triple (system, user, assistant) that the fine-tuner trains on. +- `meta` — task type, example ID, gold labels, and full sampling provenance (ignored during training). + +### Is there a label or target associated with each instance? +Yes, stored in `meta.gold_labels`. For teaching programs: method, domain, level, learner profile, mastery state. For session interpretation: pattern class, behavior functions, escalation level, confidence, crisis-plan requirement. + +### Is any information missing from individual instances? +Some instances have optional data by design: +- **Task Analysis / toleration programs** have `shaping_steps` instead of `steps` and no `chain_type` (not applicable). +- **Session logs** carry ABC data on approximately 30% of sessions when behaviors are present; IOA data on approximately 25% of logs; behavioral-indicator clusters per pattern. Absence of these is intentional, not missingness. + +### Are relationships between individual instances made explicit? +Instances are independent by construction. No implicit links between examples. + +### Are there recommended data splits? +Yes. The dataset ships with four split files: + +| Split | Count | Fraction | Purpose | +|---|---:|---:|---| +| `train.jsonl` | 2,549 | 85% | LoRA fine-tuning | +| `valid.jsonl` | 149 | 5% | Validation during training | +| `test.jsonl` | 281 | 9.4% | Held-out evaluation | +| `sanity.jsonl` | 20 | 0.7% | Training smoke-test | + +All splits are stratified by area × category to preserve the corpus distribution. + +### Are there errors, sources of noise, or redundancies? +- **Template repetition.** Assistant responses follow a structured template; lexical patterns recur by design. This is an intentional trade-off for learnability. +- **Sampling drift.** Because the generator samples independently per example, some combinations of cells can land improbable but not impossible (e.g., a "mastered" skill still running an error-correction procedure). Compatibility rules reduce this; ad-hoc practitioner review catches the rest. +- **Clinical accuracy.** Every example passed at least a syntactic quality check. A subset underwent ad-hoc practitioner review during iteration; not every example has been individually inspected. + +### Does the dataset contain data that might be considered confidential? +No. All data is synthetic. No real client information. No real session notes. No real names, dates, or IDs. Synthetic IDs use the `SYN-####` pattern from a fixed range; synthetic dates fall in 2026-01-01 to 2026-12-31. + +### Does the dataset contain data that might be offensive, insulting, threatening, or otherwise cause anxiety? +The session logs describe challenging behavior (aggression, self-injury, property destruction, pica, elopement, fecal smearing, toileting accidents, tantrum). These descriptions are clinical in register and grounded in the operational definitions used in the peer-reviewed JABA literature. They may be distressing to readers unfamiliar with ABA. Target behaviors appear only in the context of a session log being interpreted — they are never endorsed, glorified, or represented outside clinical framing. + +### Does the dataset relate to people? +Indirectly. Instances describe synthetic learners in clinical scenarios. No real people are identifiable. No real demographic distributions are encoded. + +### Does the dataset identify any sub-populations? +Learner profiles are: +- **Early Learner** (≈3–5 years; VB-MAPP L1 or L2 depending on skill) +- **School-Age** (≈6–10 years; VB-MAPP L2–L3, AFLS basic_living and home) +- **Adolescent** (≈11–17 years; AFLS community, vocational) +- **Adult** (≈18+ years; AFLS independent_living primary) + +These are developmentally anchored; they do not encode race, ethnicity, socioeconomic status, or gender identity. No sub-populations are characterized on those axes. + +### Is it possible to identify individuals from the dataset? +No. The dataset is synthetic. + +### Does the dataset contain data that might be considered sensitive? +The subject matter is sensitive — autism-related clinical challenges, problem behavior, restraint-related crisis planning. The data itself is not sensitive because it is synthetic. Care was taken in the crisis-plan content to cite authoritative sources (BACB Ethics Code 2020 section 3.05; ABAI 2010 Position Statement) and to avoid prescribing restraint procedures directly. + +--- + +## Collection Process + +### How was the data associated with each instance acquired? +Programmatic generation. The pipeline is a deterministic function of (a) the YAML taxonomy configs and (b) a random seed. Inputs to the generator come from published curricula (VB-MAPP; AFLS) and textbook / JABA / BACB Ethics Code content that was manually encoded into the taxonomy YAMLs by the authors. **No web scraping. No ingestion of patient records. No use of real session notes.** + +### What mechanisms or procedures were used? +Python scripts under `src/generators/` with taxonomy YAMLs under `configs/`. See `dataset-card.md section 4`. + +### If sampled, what was the sampling strategy? +Weighted random sampling on each taxonomy dimension, with compatibility rules applied afterward to reject clinically inconsistent combinations. Weights are human-specified (in `compatibility.yaml` files) to approximate clinical frequency. + +### Who was involved in the data collection process? +Festus Kahunla (Drexel; Pombo Labs) wrote the generator and the taxonomy, conducted clinical-accuracy review, and authored the dataset docs. + +### Over what timeframe was the data collected? +January 2026 to April 2026. The early months covered taxonomy research, literature grounding, and pipeline design; the corpus itself was generated and iterated in April 2026. + +### Were any ethical review processes conducted? +No IRB required — the dataset is entirely synthetic. The generator architecture is explicitly designed so no real client data can enter the pipeline. Clinical scope (what counts as a reasonable teaching program, what crisis procedures are appropriate) was grounded in published sources (BACB Ethics Code 2020; ABAI 2010 Position Statement; Cooper, Heron, & Heward 2020) rather than personal clinical experience. + +### Does the dataset relate to people? If so, did you collect data from them directly or from a third-party? +The dataset does not relate to any real person. No data was collected from any person. + +### Were the individuals in question notified about data collection? +Not applicable (no real individuals). + +### Did the individuals in question consent to data collection? +Not applicable (no real individuals). + +### If consent was obtained, were they given a mechanism to revoke consent? +Not applicable. + +### Has an analysis of the potential impact of the dataset and its use on data subjects been conducted? +The dataset has no data subjects. The intended model users (BCBAs and ABA technicians) are discussed in section 6.2 of the dataset card (clinical-risk framing) and section 6.3 (crisis-plan sensitivity). + +--- + +## Preprocessing / Cleaning / Labeling + +### Was any preprocessing / cleaning / labeling of the data done? +- **Labeling.** Gold labels are emitted at generation time from the same taxonomy cells that produced the example. No post-hoc annotation. Labels are therefore by-construction correct for the taxonomy; they are only as valid as the taxonomy is (see section 4.3 of the dataset card for grounding). +- **Deduplication.** Example IDs are deterministic hashes of the message content; duplicates (same user + assistant content) would be detected across generations, but in practice duplicates do not occur because every render draws fresh random values for the slots. +- **Filtering.** Compatibility rules reject clinically inconsistent samples before rendering. No post-rendering filter is applied. + +### Was the "raw" data saved in addition to the preprocessed data? +The generator is a deterministic function of (configs, seed) and produces the same data on every run. There is no separate "raw" stage — every artifact is reproducible from source. + +### Is the software used to preprocess / clean / label the data available? +Yes, in the repository: `src/generators/`, `src/prepare_data.py`, `src/split_data.py`, `src/prepare_curation.py`, `src/compile_curation.py`. + +--- + +## Uses + +### Has the dataset been used for any tasks already? +At time of v1 release, the dataset has been used for the associated research paper and for Pombo Labs's on-device small-LM fine-tuning experiments. Downstream fine-tuning + evaluation results will be referenced from this dataset card as they are published. + +### Is there a repository that links to any or all papers or systems that use the dataset? +Repository: https://github.com/Pombo-Labs/TRACE (v1 pinned to the initial release commit). Downstream artifacts — training logs, evaluation results, and any derived models — will be released separately and referenced from this dataset card as they become available. + +### What other tasks could the dataset be used for? +- Evaluation baselines for future ABA-specific LLM work. +- Research on taxonomy-driven synthetic data generation (the pipeline is domain-agnostic; only the configs are ABA-specific). +- Research on provenance-traceable data and auditable clinical AI. + +### Is there anything about the composition of the dataset or the way it was collected and preprocessed / cleaned / labeled that might impact future uses? +- **Pattern frequencies are uniform for learnability.** Real clinical caseloads are skewed toward mastery-progression; users who care about epidemiological plausibility should reweight. +- **The dataset is English-only and US-clinical-context-only.** Users in other jurisdictions or languages will need to adapt. +- **Target-behavior presence in logs is bounded** (0–3 per log by pattern); real logs may have more or fewer. +- **Crisis plans are deliberately vague on physical-intervention procedures** because those procedures are training- and jurisdiction-specific. Users building systems that need to recommend specific restraint procedures will need additional data curated to their jurisdiction. + +### Are there tasks for which the dataset should not be used? +TRACE is a research artifact and is not a clinical tool. Any clinical use is at the user's own risk. See dataset card section 5.3 for the full framing. Summarized: + +- Not for autonomous clinical decisions. +- Not for final BIP writing without BCBA review. +- Not to be combined with real client data during training or inference. +- Not for medical diagnosis, legal documentation, or insurance reimbursement. + +--- + +## Distribution + +### Will the dataset be distributed to third parties outside of the entity on behalf of which it was created? +Yes — intended for open release under a research license (CC BY-NC 4.0 recommended, pending finalization). + +### How will the dataset be distributed? +Public GitHub repository and Hugging Face Hub. + +### When will the dataset be distributed? +2026-04-25 for the v1 release. + +### Will the dataset be distributed under a copyright or intellectual property license? +Data: CC BY-NC 4.0. Code: MIT. TRACE is a research artifact; it has not been clinically validated and is not a clinical tool. Any use in a clinical setting is at the sole responsibility of the user and their facility. + +### Have any third parties imposed IP-based or other restrictions? +No third-party IP is embedded. Curriculum references are to published frameworks (VB-MAPP, AFLS); no proprietary items are reproduced. Citation-based references to the Cooper/Heron/Heward textbook and JABA papers are just that — citations, not text reproductions. + +### Do any export controls or other regulatory restrictions apply? +No. + +--- + +## Maintenance + +### Who will be supporting / hosting / maintaining the dataset? +**Pombo Labs** (https://github.com/Pombo-Labs) is the release maintainer. Festus Kahunla is the primary author and active maintainer of the dataset. + +### How can the owner / curator / manager of the dataset be contacted? +Via the GitHub repository issue tracker. + +### Is there an erratum? +A CHANGELOG will be maintained in the repository for each dataset version. + +### Will the dataset be updated? +Yes, as ongoing research warrants. Future versions will be released as new git tags (semantic versioning) with full CHANGELOG entries. No public roadmap is committed to at v1. + +### If the dataset relates to people, are there applicable limits on retention? +Not applicable (no real people). + +### Will older versions continue to be supported? +Yes. Each dataset version is pinned to a specific commit hash; old versions remain reproducible from the corresponding configs and generator code. + +### If others want to extend / augment / build on / contribute to the dataset, is there a mechanism for them to do so? +Yes. The taxonomy YAMLs are the intended extension surface. Adding a new skill target, a new target behavior, a new mastery criterion, or a new teaching method follows a documented pattern (see `configs/` for structure). Pull requests to the repository are the intended contribution mechanism. diff --git a/docs/references.md b/docs/references.md new file mode 100644 index 0000000000000000000000000000000000000000..eca31790259ac42f4a0818b46e59f030f8bdd446 --- /dev/null +++ b/docs/references.md @@ -0,0 +1,142 @@ +# Citations to use — the focused shortlist + +**Principle:** The paper's References section is lean and earned, not a literature dump. This shortlist is the ~30 papers we actually cite across the paper and the dataset documentation (dataset card, datasheet, data statement, taxonomy reference). + +--- + +## A. ABA × NLP — direct predecessors and positioning (6) + +1. **Kumar, A., et al. (2024). Personalized-ABA.** *NLP4Science @ ACL.* https://aclanthology.org/2024.nlp4science-1.16/ + *How we cite:* direct predecessor; benchmark to beat. + +2. **Cox, D. J., & Jennings, A. M. (2024). Promises and possibilities of AI in behavior analytic services.** *Behavior Analysis in Practice, 17*, 123–136. https://pmc.ncbi.nlm.nih.gov/articles/PMC10890993/ + *How we cite:* motivates the application space, especially "NLP on session notes." + +3. **Jennings, A. M., & Cox, D. J. (2024). Starting the conversation around the ethical use of AI in ABA.** *Behavior Analysis in Practice, 17*, 107–122. https://pmc.ncbi.nlm.nih.gov/articles/PMC10891004/ + *How we cite:* ethics spine — BACB 2.03/2.05, HIPAA, explainability criteria our stack satisfies by construction. + +4. **Peck, S., O'Brien, C., Bourret, J., & Agostinelli, D. (2025). ChatGPT versus clinician responses to questions in ABA.** *JABA.* https://onlinelibrary.wiley.com/doi/10.1002/jaba.70029 + *How we cite:* BCBAs already prefer LLMs blind — raises hallucination stakes. + +5. **Garg, M., Raza, S., Rayana, S., Liu, X., & Sohn, S. (2025). The rise of small language models in healthcare: A comprehensive survey.** arXiv:2504.17119. https://arxiv.org/abs/2504.17119 + *How we cite:* positions our work in SLM-clinical field; ABA is a named white-space in their taxonomy. + +6. **Gao, L. et al. (2025). Generative AI for assessment and treatment of autism spectrum disorders: A scoping review.** *Frontiers in Psychiatry.* https://pmc.ncbi.nlm.nih.gov/articles/PMC12322814/ + *How we cite:* none of the 10 studies surveyed cover BCBA workflow / session logs / on-device — directly positions our contribution. + +## B. Stack precedents — small clinical on-device LMs (3) + +7. **Zhang, T., et al. (2025). Menta: On-device SLM for mental health.** arXiv:2512.02716. https://arxiv.org/abs/2512.02716 + *How we cite:* direct recipe precedent (4B + LoRA r=16 α=32 + 4-bit on iPhone, beats 13B). + +8. **MedGemma Team, Google DeepMind (2025). MedGemma Technical Report.** arXiv:2507.05201. https://arxiv.org/abs/2507.05201 + *How we cite:* Gemma family at 4B is clinically competent — direct stack-choice precedent. + +9. **Dettmers, T., et al. (2023). QLoRA: Efficient Finetuning of Quantized LLMs.** *NeurIPS.* https://arxiv.org/abs/2305.14314 + *How we cite:* foundational method for our QLoRA fine-tuning. + +## C. Data methodology (4) + +10. **Wang, Y., et al. (2023). Self-Instruct.** *ACL.* https://arxiv.org/abs/2212.10560 + *How we cite:* foundational instruction-bootstrapping pipeline; source of ROUGE-L<0.7 dedup. + +11. **Zhang, X., et al. (2023). AlpaCare.** arXiv:2310.14558. https://arxiv.org/abs/2310.14558 + *How we cite:* clinician-seeded stratification recipe — the key quality lever for medical instruction tuning. + +12. **Gekhman, Z., et al. (2024). Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?** *EMNLP.* https://arxiv.org/abs/2405.05904 + *How we cite:* grounds our "teach format, not new facts" design. + +13. **Patel, M., et al. (2025). How to Design, Create, and Evaluate an Instruction-Tuning Dataset in Health Care.** *JMIR.* https://www.jmir.org/2025/1/e70481 + *How we cite:* our validation protocol (4-dim rubric, κ with cooldown) follows this tutorial directly. + +## D. Evaluation methodology (7) + +14. **Singhal, K., et al. (2025). Toward expert-level medical question answering with LLMs (Med-PaLM 2).** *Nature Medicine.* https://www.nature.com/articles/s41591-024-03423-7 + *How we cite:* rubric design (9 axes, pairwise ranking) — the clinical-NLG gold standard. + +15. **Arora, R. K., et al. (2025). HealthBench.** arXiv:2505.08775. https://arxiv.org/abs/2505.08775 + *How we cite:* justifies our 3-point rubric scale over Likert-5/7. + +16. **Kim, S., et al. (2024). Prometheus 2.** arXiv:2405.01535. https://arxiv.org/abs/2405.01535 + *How we cite:* the open, local LLM judge we run alongside GPT-4.1 and Claude. + +17. **Zheng, L., et al. (2024). Judging LLM-as-a-Judge (MT-Bench).** *NeurIPS 2023.* https://arxiv.org/abs/2306.05685 + *How we cite:* sources of position / verbosity / self-enhancement bias; justifies our swap-augmentation and panel diversity. + +18. **Kulkarni et al. (2025). TN-Eval: Behavioral therapy note rubrics.** arXiv:2503.20648. https://arxiv.org/abs/2503.20648 + *How we cite:* closest adjacent rubric; faithfulness warning (therapists preferred hallucinated notes). + +19. **Manakul, P., et al. (2023). SelfCheckGPT.** arXiv:2303.08896. https://arxiv.org/abs/2303.08896 + *How we cite:* our zero-resource hallucination metric. + +20. **Reiter, E. (2018). A Structured Review of the Validity of BLEU.** *Computational Linguistics, 44*(3), 393–401. https://direct.mit.edu/coli/article/44/3/393/ + *How we cite:* justifies NOT using BLEU as a primary metric (negation matters in evaluation claims). + +## E. Statistical grounding (3) + +21. **Cohen, J. (1968). Weighted kappa.** *Psychological Bulletin, 70*(4). + *How we cite:* quadratic-weighted κ for our ordinal escalation-level classifier. + +22. **Chicco, D., & Jurman, G. (2020). Advantages of MCC over F1 and accuracy.** *BMC Genomics.* https://link.springer.com/article/10.1186/s12864-019-6413-7 + *How we cite:* justifies reporting MCC alongside macro-F1 on imbalanced classification heads. + +23. **Guo, C., et al. (2017). On Calibration of Modern Neural Networks.** *ICML.* https://arxiv.org/abs/1706.04599 + *How we cite:* source of ECE + temperature-scaling for our confidence calibration. + +## F. ABA foundation (5) + +24. **Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). *Applied Behavior Analysis* (3rd ed.).** Pearson. ISBN 978-0134752556. + *How we cite:* canonical ABA reference — chapter citations for DTT, task analysis, measurement, IOA, FBA, BIP components. + +25. **Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1982/1994). Toward a functional analysis of self-injury.** *JABA, 27*(2), 197–209. https://doi.org/10.1901/jaba.1994.27-197 + *How we cite:* bedrock of the four-function taxonomy we encode. + +26. **Carr, E. G., & Durand, V. M. (1985). Reducing behavior problems through functional communication training.** *JABA, 18*(2), 111–126. https://doi.org/10.1901/jaba.1985.18-111 + *How we cite:* origin of FCT; replacement-behavior framework. + +27. **Hanley, G. P., Iwata, B. A., & McCord, B. E. (2003). Functional analysis of problem behavior: A review.** *JABA, 36*(2), 147–185. https://doi.org/10.1901/jaba.2003.36-147 + *How we cite:* definitive FBA review, finalization of four-function taxonomy. + +28. **Lovaas, O. I. (1987). Behavioral treatment and normal educational and intellectual functioning in young autistic children.** *JCCP, 55*(1), 3–9. https://doi.org/10.1037/0022-006X.55.1.3 + *How we cite:* DTT / EIBI historical origin. + +## G. Dataset documentation (2) + +29. **Gebru, T., et al. (2021). Datasheets for Datasets.** *CACM.* https://arxiv.org/abs/1803.09010 + *How we cite:* template for our dataset's datasheet appendix. + +30. **Bender, E., & Friedman, B. (2018). Data Statements for NLP.** *TACL.* https://aclanthology.org/Q18-1041/ + *How we cite:* NLP-specific data disclosure — complements Gebru datasheet. + +--- + +## What was intentionally dropped (and why) + +Comprehensive background is in `literature-foundation.md`. The following categories do **not** earn spots in the paper's References: + +- **Target-behavior JABA papers** (Kodak elopement, Piazza pica, Marcus aggression, Rapp & Vollmer stereotypy, etc.) — These ground the dataset's operational definitions and belong in the **dataset card / datasheet**, not the main paper. They'll appear in a supplementary `taxonomy-v1.md`. +- **Curriculum primary sources beyond VB-MAPP** (AFLS, ABLLS-R, Essential for Living, PEAK) — cite only if we actually use them as data sources. Current pipeline uses VB-MAPP only; others stay in background reading. +- **Secondary synthetic-data papers** (phi-1, Orca, LIMA, Baize, MedAlpaca, Clinical Camel, Persona Hub, Evol-Instruct, LAB, DataDreamer) — Self-Instruct + AlpaCare + Gekhman + Patel cover our method grounding. The rest are design inspiration, not required citations. +- **Secondary evaluation papers** (GPTScore, JudgeLM, BARTScore, Med-HALT, MedHallu, HELM critiques, Abacha MEDIQA, G-Eval) — Prometheus 2 + MT-Bench + TN-Eval + SelfCheckGPT + Reiter + Med-PaLM 2 cover our method grounding. +- **Older measurement / IOA JABA papers** (Powell 1975, Harrop & Daniels 1986, Kazdin 1977) — covered by Cooper/Heron/Heward textbook chapters. +- **Teaching-method origins beyond DTT** (Hart & Risley 1975 NET, Koegel 1987 PRT, Parsons 2012 BST, Tiger 2008 FCT review, Slocum & Tiger 2011 chaining) — covered by CHH textbook chapter citations. Only cite these JABA papers individually if our paper makes a specific claim about that teaching method. +- **Neuromnia blog / Meta-AI press** — not citable in an academic venue; mention in passing without reference. +- **Lanovaz & Hranchuk 2021** — different task (visual-inspection binary classification from graphs); only cite if we explicitly contrast against it. Likely drop. +- **Dataset documentation meta-papers** (Bender 2023 Data Statements v2, HF dataset card docs, NeurIPS 2025 D&B CfP, Giuffrè 2023 synthetic health, Pezoulas 2025 privacy, Mozilla/AI-Alliance 2024) — cite only where directly relevant in the datasheet, not in the main paper. + +--- + +## Summary + +| Category | Count | +|---|---| +| ABA × NLP | 6 | +| Stack precedents | 3 | +| Data methodology | 4 | +| Evaluation methodology | 7 | +| Statistical grounding | 3 | +| ABA foundation | 5 | +| Dataset documentation | 2 | +| **Total** | **30** | + +Appropriate for a 4-page ACL paper (typical range 20–40 citations) plus a NeurIPS D&B dataset paper appendix. Add more only when the text actually needs them. diff --git a/docs/schema-v1.md b/docs/schema-v1.md new file mode 100644 index 0000000000000000000000000000000000000000..aeb8255bf9089ef1406237452c5ab62289107444 --- /dev/null +++ b/docs/schema-v1.md @@ -0,0 +1,756 @@ +# TRACE Dataset Schema v1 + +**Purpose.** This document defines the **wire format** for TRACE training examples — the precise shape each JSONL line takes, the structure of user messages and assistant responses, and the metadata fields used for evaluation label extraction. + +**Scope.** Two tasks, each with its own input/output conventions but sharing a common chat-format wrapper. Every taxonomy category from `taxonomy-v1.md` maps to a field or sampling decision here. + +**Status.** Version 1, 2026-04-23. Breaking changes require a version bump. + +**Companion documents.** +- `taxonomy-v1.md` — the controlled vocabulary (what categories exist) +- `schema-v1.md` — this document (how examples are shaped) +- `datasheet.md` (pending) — dataset card +- `src/prepare_data.py` (pending rewrite) — implements this schema + +--- + +## 0. Overview + +Two tasks: + +| Task | Input | Output | MLX-LM role mapping | +|---|---|---|---| +| **1. Teaching Program Generation** | Learner profile + skill target + method + context | Structured teaching program with method-specific fields | user -> assistant | +| **2. Behavioral Session Interpretation** | Multi-session behavioral log (8–12 sessions typical) | Structured clinical interpretation (concerns + pattern + function + recommendations + escalation + confidence + rationale) | user -> assistant | + +Each training example is one JSONL line with: +- `messages` — the chat-format messages the model trains on (system + user + assistant) +- `meta` — evaluation metadata (gold labels, provenance) that is *not* shown to the model during training but is used by `src/evaluate.py` for metric computation + +--- + +## 1. Wire format + +### 1.1 MLX-LM chat format + +MLX-LM's `--data` expects JSONL files where each line has a `"messages"` array. Each message is `{"role": "system"|"user"|"assistant", "content": "..."}`. When `mask_prompt: true`, loss is computed only on assistant tokens. + +### 1.2 TRACE extended format + +```json +{ + "messages": [ + {"role": "system", "content": "..."}, + {"role": "user", "content": "..."}, + {"role": "assistant", "content": "..."} + ], + "meta": { + "task_type": "teaching_program" | "session_interpretation", + "example_id": "", + "gold_labels": { ... task-specific ... }, + "provenance": { + "layer": 1 | 2 | 3, + "template_id": "", + "taxonomy_cells": { ... sampled values ... }, + "teacher_model": "", + "seed": , + "generated_at": "" + } + } +} +``` + +The `meta` field is ignored by MLX-LM during training but preserved for evaluation and provenance tracking. `example_id` is a deterministic hash of the user + assistant content so that duplicates can be detected across generations. + +### 1.3 Splits + +Train / valid / test splits live in `data/splits/`: +- `train.jsonl` — 85% (used for LoRA training) +- `valid.jsonl` — 10% (used by MLX-LM for periodic validation loss) +- `test.jsonl` — 5% (held out; only used for final evaluation) + +Stratified by `meta.task_type` and by `meta.gold_labels.pattern_class` (task 2) so each split is balanced. + +--- + +## 2. Common conventions + +### 2.1 Synthetic identifiers + +Every learner referenced in any example has a synthetic ID: `SYN-####` where `####` is a 4-digit number from a fixed range (1000–9999). Generator uses `random.Random(seed)` to draw these deterministically. **No real-world ID patterns, no initials, no names.** + +### 2.2 Dates + +Synthetic dates always fall in the range `2026-01-01` -> `2026-12-31` (stable, non-leaking). Format: `YYYY-MM-DD`. Session timestamps within a day are not included (unnecessary detail). + +### 2.3 Measurement notation (within session logs) + +| Measurement | Abbreviation / format | +|---|---| +| Accuracy (trial-based) | `X/N correct (PP%)` e.g., `6/10 correct (60%)` | +| Frequency (count) | `freq = N` | +| Rate | `rate N.NN/min` | +| Duration | `Nm` or `N m Ns` | +| Latency | `latency N.Ns` | +| Partial-interval | `PIR PP% of intervals` | +| Whole-interval | `WIR PP% of intervals` | +| Momentary time sampling | `MTS PP%` | +| Episode-based | `N episodes, mean duration Nm Ns` | +| IOA | `IOA PP%` (with "session marked IOA" header) | + +### 2.4 Prompt-level shorthand + +Used inside session logs to describe trial-level prompting distribution: + +- `FP` — full physical +- `PP` — partial physical +- `M` — model +- `G` — gestural +- `V` — verbal +- `Pos` — positional +- `Vis` — visual +- `I` — independent + +Example: `prompts: FP×3, PP×2, G×1, I×4` means 3 full-physical, 2 partial-physical, 1 gestural, 4 independent trials. + +### 2.5 Markdown conventions (assistant output) + +All assistant outputs use GitHub-flavored markdown. Section headers use `##` for top-level fields and `###` for sub-fields. Structured labels (pattern class, escalation, confidence) appear under their own `##` headers and contain a single canonical value in the first paragraph for regex extraction. + +--- + +## 3. Task 1 — Teaching Program Generation + +### 3.1 System prompt (task 1) + +``` +You are an expert ABA (Applied Behavior Analysis) clinical assistant. You help +Board Certified Behavior Analysts and staff design teaching programs for +individuals with autism. Your responses are clinically accurate, individualized +to the learner profile, follow BACB ethical guidelines, and reference no real +client data. Select the appropriate teaching method (DTT, NET, Task Analysis, +FCT, BST, PRT) based on the skill target and learner profile. +``` + +### 3.2 User message format (task 1) + +Seven required fields filled from the taxonomy. The generator samples valid combinations per section 14 of `taxonomy-v1.md`. + +``` +Generate a teaching program for the following target. + +Skill Target: {skill_name} +Curriculum Reference: {vbmapp_domain or afls_module} — {level or none} +Learner Profile: {early | school-age | adolescent | adult} +Current Mastery: {emerging | developing | approaching | near | mastered | generalization | maintenance} +Teaching Method: {dtt | net | task_analysis | fct | bst | prt} +Program Context: {D | R | Both} + +Provide the full program structure appropriate to the selected method. +``` + +Template variations (~5 paraphrases) are sampled per Self-Instruct convention to avoid canonical-form overfitting. + +### 3.3 Assistant message format (method-agnostic sections) + +Every task-1 output contains these top-level sections in order: + +```markdown +## Program Overview +{1-2 sentence summary of what is being taught and to whom} + +## {method-specific sections — see section 3.4} + +## Mastery Criteria +{which of the 7 mastery conventions from taxonomy section 11.1} + +## Data Collection +{what measurement types will be used; when IOA will be scheduled} + +## Generalization & Maintenance Plan +{when to probe across therapists / settings / materials; maintenance schedule} +``` + +### 3.4 Method-specific output variants + +The middle section of the output varies by method. Each has a fixed field structure. + +#### 3.4.1 DTT + +```markdown +## Discriminative Stimulus (SD) +Primary SD: "{sd_text}" +Variations: {list} +Presentation: {how stimulus is presented} + +## Prompt Hierarchy +Strategy: {most-to-least | least-to-most | time-delay | graduated-guidance | stimulus-fading | stimulus-shaping} +Sequence: {ordered prompt-level levels} +Current prompt level: {based on mastery state} + +## Stimulus Array +Array size: {field of N or no-array} +Target stimuli: {list} +Distractor stimuli: {list or N/A} +Rotation: {how position/order is varied} + +## Error Correction Procedure +{one of the 5 procedures from taxonomy section 10} + +## Reinforcement Schedule +{one of the 7 schedules from taxonomy section 9} +``` + +#### 3.4.2 NET + +```markdown +## Motivating Operation (MO) Arrangement +{how the environment is set up to establish value for the target} + +## Natural Opportunity +{when and where in the natural routine the teaching occurs} + +## Prompt Strategy +Strategy: {prompt hierarchy} +Delivery: {how prompts are embedded naturally} + +## Natural Reinforcer +{the functional reinforcer that follows the target behavior} + +## Generalization Tactics +{multiple exemplar training; programming common stimuli} +``` + +#### 3.4.3 Task Analysis / Chaining + +```markdown +## Task Analysis +Chain type: {forward | backward | total-task} +Steps: +1. {step description} +2. {step description} +... +N. {step description} + +## Prompt Strategy Per Step +{prompt level used at each step, fading plan} + +## Error Correction +{procedure for when a step is missed} + +## Reinforcement +Per-step reinforcement: {yes/no, schedule} +Terminal reinforcement: {on chain completion} +``` + +#### 3.4.4 FCT + +```markdown +## Target Behavior (to reduce) +{operational definition, hypothesized function} + +## Replacement Response +Topography: {vocal phrase | sign | AAC icon | card exchange} +Training sequence: {how the replacement is taught} + +## Extinction Plan +{how the problem behavior is placed on extinction} + +## Reinforcement for Replacement +Schedule: {initially CRF; thinning plan} +Magnitude: {quality and duration matched to natural reinforcer for problem behavior} + +## Crisis Plan +{if extinction burst or safety concern, what staff do} +``` + +#### 3.4.5 BST (staff-facing) + +```markdown +## Training Target +{the program or skill the trainee will learn to implement} + +## Training Components +1. Instruction: {what written/verbal instruction is given} +2. Modeling: {trainer demonstrates; video or live} +3. Rehearsal: {trainee practices with feedback} +4. Feedback: {specific, behavior-specific, immediate} + +## Fidelity Checklist +{bulleted items that must be demonstrated, scored yes/no} + +## Mastery Criterion (trainee) +{e.g., 100% fidelity across 2 consecutive role-plays with novel scenarios} +``` + +#### 3.4.6 PRT + +```markdown +## Motivation Arrangement +{how child choice, preferred materials, and interspersal are set up} + +## Teaching Opportunities +{when target is presented; how multiple cues are programmed} + +## Reinforcement of Attempts +{what counts as an attempt; how attempts are reinforced differentially} + +## Natural Reinforcer +{the functional reinforcer for the target response} + +## Generalization +{multiple-exemplar plan; across people / settings} +``` + +### 3.5 Metadata fields (task 1) + +```json +"meta": { + "task_type": "teaching_program", + "example_id": "", + "gold_labels": { + "method": "dtt", + "domain": "VB-MAPP.Tact", + "level": "L2", + "learner_profile": "early", + "mastery_state": "developing", + "context": "Both" + }, + "provenance": { + "layer": 1, + "template_id": "tp-dtt-001", + "taxonomy_cells": { + "skill_target": "Tacts colors of objects", + "prompt_hierarchy": "most-to-least", + "reinforcement_schedule": "CRF", + "error_correction": "transfer_trial" + }, + "teacher_model": null, + "seed": 20260423001, + "generated_at": "2026-04-23T14:22:11Z" + } +} +``` + +### 3.6 Complete task-1 example (abridged) + +```json +{ + "messages": [ + { + "role": "system", + "content": "You are an expert ABA clinical assistant..." + }, + { + "role": "user", + "content": "Generate a teaching program for the following target.\n\nSkill Target: Tacts colors of objects\nCurriculum Reference: VB-MAPP Tact L2\nLearner Profile: early\nCurrent Mastery: developing\nTeaching Method: dtt\nProgram Context: Both\n\nProvide the full program structure appropriate to the selected method." + }, + { + "role": "assistant", + "content": "## Program Overview\nThis program teaches {learner} to tact (label) the color of presented objects...\n\n## Discriminative Stimulus (SD)\nPrimary SD: \"What color?\"\nVariations: \"Tell me the color\", \"What color is this?\"\nPresentation: ...\n\n## Prompt Hierarchy\nStrategy: most-to-least\nSequence: full physical (hand guidance to point) -> partial physical (touch elbow) -> gestural (point to color) -> positional (color nearest learner) -> independent.\nCurrent prompt level: given current 40% accuracy, begin at partial-physical and fade to gestural across sessions.\n\n## Stimulus Array\nArray size: field of 3 (target + 2 distractors)\nTarget stimuli: red, blue, yellow\nDistractor stimuli: 2 non-target colors per trial\nRotation: randomize position of target across trials.\n\n## Error Correction Procedure\nTransfer trial: re-present SD -> provide effective-level prompt -> reinforce prompted response -> distractor trial -> re-present SD independently.\n\n## Reinforcement Schedule\nCRF (continuous reinforcement) during acquisition; fade to FR-2 once accuracy exceeds 80%.\n\n## Mastery Criteria\n80% accuracy across 2 consecutive sessions with at least 10 trials each.\n\n## Data Collection\nTrial-by-trial data recording prompt level and response accuracy. IOA scheduled every 4th session at minimum.\n\n## Generalization & Maintenance Plan\nPost-mastery: probe with novel color stimuli, across 2 additional therapists, and in at least 2 natural contexts. Maintenance probes weekly for 4 weeks, then monthly." + } + ], + "meta": { + "task_type": "teaching_program", + "example_id": "a3f2b8d14e9c7a0f", + "gold_labels": { + "method": "dtt", + "domain": "VB-MAPP.Tact", + "level": "L2", + "learner_profile": "early", + "mastery_state": "developing", + "context": "Both" + }, + "provenance": { + "layer": 1, + "template_id": "tp-dtt-001", + "taxonomy_cells": { + "skill_target": "Tacts colors of objects", + "prompt_hierarchy": "most-to-least", + "reinforcement_schedule": "CRF", + "error_correction": "transfer_trial" + }, + "teacher_model": null, + "seed": 20260423001, + "generated_at": "2026-04-23T14:22:11Z" + } + } +} +``` + +--- + +## 4. Task 2 — Behavioral Session Interpretation + +### 4.1 System prompt (task 2) + +``` +You are an expert ABA clinical assistant. You analyze multi-session behavioral +session logs for individuals with autism and produce structured clinical +interpretations that identify patterns, hypothesize behavior functions when +applicable, and recommend programming adjustments structured along BIP lines +(antecedent strategies, replacement behaviors, consequence strategies, crisis +plan). Your interpretation also includes an escalation level and a confidence +expression. Every recommendation is grounded in the data provided. Follow BACB +ethical guidelines and reference no real client data. +``` + +### 4.2 User message format — the session log + +The session log is a plain-text block with a fixed top-level structure. The generator produces it deterministically from sampled taxonomy values + a hidden pattern label. + +#### 4.2.1 Learner profile block + +``` +LEARNER PROFILE +Synthetic ID: SYN-#### +Profile: {Early Learner | School-Age Learner | Adolescent Learner | Adult Learner} ({chronological age} yr) +Curricula: {VB-MAPP L# | AFLS {module(s)} | combination} +Primary context: {D | R | Both} +Date range: Sessions 1–N across M days ({start date} to {end date}) +``` + +#### 4.2.2 Acceleration programs block + +``` +ACCELERATION PROGRAMS +1. {Skill Target} ({Curriculum Reference}) + Method: {method}, {method-specific brief, e.g., "backward chaining, 8 steps"} + Context: {D | R | Both} +2. ... +``` + +Typically 3–6 acceleration programs per log. + +#### 4.2.3 Deceleration targets block + +``` +DECELERATION TARGETS +1. {Target behavior} — function hypothesized: {escape | attention | tangible | automatic | unknown} +2. ... +``` + +Zero to three deceleration targets per log. When present, each has a hypothesized function used by the gold-label generator; the hypothesis may or may not be the correct answer in the interpretation (some examples deliberately encode ambiguity). + +#### 4.2.4 Per-session data format + +One block per session. + +``` +Session {N} — {YYYY-MM-DD} ({context-tag}) — {duration} min — {# observers} + {Skill 1 name}: {measurement 1}; {measurement 2}; prompts {prompt distribution} + {Skill 2 name}: {measurement} + ... + {Behavior 1 name}: {measurement} + {Behavior 2 name}: {measurement} + {optional: ABC(behavior): A = {antecedent}; B = {behavior description}; C = {consequence}} +``` + +ABC entries appear in ~30% of logs (taxonomy section 6.1), at a rate of approximately 1 ABC event per session that includes behavior occurrence. + +#### 4.2.5 IOA session format + +Approximately 25% of logs include one IOA session (sampled uniformly from the middle third of sessions in the log). + +``` +Session {N} — IOA SESSION — 2 observers + {Skill 1} IOA: {percentage}% agreement + {Behavior 1} IOA: {percentage}% agreement + ... +``` + +Agreements below 80% in generated IOA data are intentional in a minority of examples to test the model's ability to flag reliability concerns. + +#### 4.2.6 Cross-session observations block + +``` +BEHAVIORAL OBSERVATIONS (across sessions) +- {narrative observation 1} +- {narrative observation 2} +- ... +``` + +3–6 bullet observations that summarize trends visible in the session data. Generated deterministically from the hidden pattern label + the behavioral indicator pool from taxonomy section 7. + +### 4.3 Assistant message format — structured interpretation + +Eight top-level sections in fixed order. Sections marked **required** always appear; sections marked **conditional** appear only when behavior data is present in the log. + +```markdown +## Clinical Concerns (required — free-form prose) + +## Pattern Classification (required — structured label + 1–2 sentence evidence) + +## Behavior Function Hypothesis (conditional — only if deceleration targets observed) + +## Programming Recommendations (required — 4 BIP-structured subsections) +### Antecedent strategies +### Replacement behavior (conditional on behavior presence) +### Consequence strategies +### Crisis plan (conditional — escalation level ≥ 3) + +## Escalation Level (required — structured label + brief justification) + +## Confidence (required — structured label + brief justification) + +## Data-Supported Rationale (required — numeric grounding) +``` + +#### 4.3.1 Clinical Concerns + +Free-form prose, 2–5 bullets or short paragraphs. Each concern references specific data from the log (accuracy values, frequencies, trends). + +#### 4.3.2 Pattern Classification + +First paragraph contains exactly one canonical label (or two labels joined by `+` if co-occurring): + +```markdown +## Pattern Classification +{pattern_label} + +{1–2 sentence explanation of why this pattern was identified} +``` + +Pattern labels drawn from taxonomy section 7 (12 patterns): +`mastery_progression | regression | plateau | frustration_pattern | variable_performance | prompt_dependency | rapid_acquisition | generalization_failure | extinction_burst | skill_loss_after_break | motivating_operation_shift | setting_event_trigger` + +Co-occurring example: `regression + frustration_pattern`. + +#### 4.3.3 Behavior Function Hypothesis (conditional) + +Present only if deceleration targets appear in the log. One sub-entry per target behavior. + +```markdown +## Behavior Function Hypothesis +{behavior name}: {escape | attention | tangible | automatic | unknown} + Evidence: {1–3 sentences grounded in log data, referencing ABC events if present} + +{behavior name}: ... +``` + +If the log lacks evidence to distinguish functions, the hypothesis is `unknown` and the interpretation notes what data would disambiguate. + +#### 4.3.4 Programming Recommendations (BIP-structured) + +Four subsections. `### Antecedent strategies` and `### Consequence strategies` are always present; `### Replacement behavior` appears when deceleration targets are present; `### Crisis plan` appears when escalation level ≥ 3. + +```markdown +## Programming Recommendations + +### Antecedent strategies +- {specific, testable recommendation} +- ... + +### Replacement behavior (FCT) +- {teaching target, function-matched} +- {reinforcement plan for replacement} + +### Consequence strategies +- {how staff respond to target behavior} +- {what is reinforced; what is placed on extinction} + +### Crisis plan +- {safety procedures for escalation} +``` + +#### 4.3.5 Escalation Level + +```markdown +## Escalation Level +{1 | 2 | 3 | 4} — {short label} + +{1–2 sentence justification} +``` + +Labels: `1 — Continue monitoring`, `2 — Adjust next session`, `3 — Supervisor review`, `4 — Safety immediate`. + +#### 4.3.6 Confidence + +```markdown +## Confidence +{high | moderate | low} + +{1 sentence explaining the data-quality basis for the confidence level} +``` + +#### 4.3.7 Data-Supported Rationale + +Numeric grounding for all claims made above. Bulleted list of specific accuracy / frequency / duration / trend statistics drawn from the session log. + +### 4.4 Metadata fields (task 2) + +```json +"meta": { + "task_type": "session_interpretation", + "example_id": "", + "gold_labels": { + "pattern_class": "regression+frustration_pattern", + "behavior_functions": { + "Tantrum": "escape", + "Mouthing": "automatic" + }, + "escalation_level": 3, + "confidence": "moderate", + "crisis_plan_required": true + }, + "log_properties": { + "n_sessions": 10, + "n_acceleration_programs": 4, + "n_deceleration_targets": 2, + "has_abc_data": true, + "has_ioa_session": true, + "learner_profile": "adolescent", + "context_distribution": {"D": 0.3, "R": 0.7, "Both": 0.0} + }, + "provenance": { + "layer": 1, + "template_id": "sess-regress-v1", + "pattern_seed": 20260423007, + "seed": 20260423007, + "generated_at": "2026-04-23T14:30:00Z" + } +} +``` + +### 4.5 Complete task-2 example (abridged user content, full assistant output) + +```json +{ + "messages": [ + { + "role": "system", + "content": "You are an expert ABA clinical assistant. You analyze multi-session behavioral session logs..." + }, + { + "role": "user", + "content": "Interpret the following behavioral session log. Provide clinical concerns, pattern classification, behavior function hypothesis (if applicable), programming recommendations (antecedent / replacement / consequence / crisis), escalation level, confidence, and data-supported rationale.\n\nLEARNER PROFILE\nSynthetic ID: SYN-4721\nProfile: Adolescent Learner (16 yr)\nCurricula: AFLS Basic Living + AFLS Home Skills + VB-MAPP L3 (residual)\nPrimary context: Residential (R), partial Day (D)\nDate range: Sessions 1–10 across 14 days (2026-03-01 to 2026-03-14)\n\nACCELERATION PROGRAMS\n1. Self-Care: Washing Hands (AFLS Basic Living)\n Method: Task Analysis, backward chaining, 8 steps\n Context: Both\n2. Requesting with AAC (VB-MAPP Mand L3 adapted)\n Method: NET\n Context: Both\n3. Community Safety Signs (AFLS Community)\n Method: DTT\n Context: D\n4. Tolerating Denied Access (FCT replacement)\n Method: FCT, replacement response = \"wait please\"\n Context: Both\n\nDECELERATION TARGETS\n1. Tantrum — function hypothesized: escape\n2. Mouthing (non-pica) — function hypothesized: automatic\n\nSession 1 — 2026-03-01 (R) — 45 min — 1 observer\n Washing Hands: 3/8 steps independent (38%); prompts FP×3, PP×2, G×1\n AAC Requests: freq = 4 (rate 0.09/min); 2 independent, 2 prompted\n Safety Signs: 6/10 correct (60%); latency 4.2s\n Tolerating Denied Access: 2/3 trials successful\n Tantrum: freq 1, duration 4m\n Mouthing: freq 8; PIR 20%\n ABC (tantrum): A = iPad removed at 2:15pm; B = screaming + drop-to-floor; C = staff redirected to preferred activity after 4 min\n\nSession 2 — 2026-03-02 (R) — 50 min — 1 observer\n Washing Hands: 3/8 (38%); prompts FP×3, PP×2, G×1\n Safety Signs: 5/10 (50%); latency 4.8s\n Tantrum: freq 2, duration 7m\n Mouthing: freq 12; PIR 28%\n\n[... Sessions 3–9 ...]\n\nSession 7 — IOA SESSION — 2 observers\n Washing Hands IOA: 87% agreement\n Tantrum IOA: 91% agreement\n AAC Requests IOA: 95% agreement\n\nSession 10 — 2026-03-14 (R) — 45 min — 1 observer\n Washing Hands: 2/8 (25%); prompts FP×5, PP×2, G×1\n Safety Signs: 4/10 (40%); latency 6.1s\n Tolerating: 0/3 success\n Tantrum: freq 4, duration 12m\n Mouthing: freq 15; PIR 35%\n ABC (tantrum, occurrence 2): A = asked to finish worksheet; B = hit table + scream; C = task removed for 5 min, staff attention\n\nBEHAVIORAL OBSERVATIONS\n- Increasing response latency in Safety Signs\n- Increasing tantrum duration\n- Mouthing rate climbing\n- Worksheet and non-preferred tasks consistently precede tantrums" + }, + { + "role": "assistant", + "content": "## Clinical Concerns\n- Regression pattern in Safety Signs and Washing Hands: accuracy has declined from 60->40% and 38->25% respectively across 10 sessions.\n- Tantrum frequency and duration are increasing, clustered around demand presentation.\n- Mouthing rate is increasing, consistent with automatic function but also temporally correlated with tantrum increases.\n\n## Pattern Classification\nregression + frustration_pattern\n\nConcurrent regression in previously-approaching skills alongside a frustration pattern characterized by escape-indicator behaviors (declining accuracy + increased tantrum duration + demand-triggered occurrences).\n\n## Behavior Function Hypothesis\nTantrum: escape\n Evidence: both documented ABC events show demand antecedent (iPad removal, worksheet task) followed by task/demand removal or redirection. Tantrum duration increases concurrent with demand difficulty.\nMouthing: automatic\n Evidence: persists across contexts, no clear social consequence, occurs during low-demand periods and throughout sessions.\n\n## Programming Recommendations\n\n### Antecedent strategies\n- Reduce worksheet task duration by 50% for next 3 sessions; introduce choice between two acceptable non-preferred tasks before any demand.\n- Pre-teach \"wait please\" response before predictable iPad-removal windows.\n- Schedule preferred activity access immediately before (not after) known demand windows to leverage MO.\n\n### Replacement behavior (FCT)\n- Strengthen FCT training for Tolerating Denied Access — current 2->1->0/3 rate across sessions suggests teaching is not yet effective. Consider 0-second time delay prompting and increased reinforcement magnitude for independent \"wait please.\"\n- Pair \"wait please\" with a brief, predictable delay (10–15s) followed by reinforcement access.\n\n### Consequence strategies\n- Do NOT remove demand contingent on tantrum (currently reinforcing escape function).\n- Provide brief (20s) break contingent on independent \"wait please\" or equivalent communication.\n- Place tantrum on escape extinction: continue presenting the task while ensuring learner and staff safety.\n\n### Crisis plan\n- If tantrum exceeds 10 minutes or escalates to property destruction or aggression, follow facility crisis protocol, document the incident, and notify supervising BCBA within 24 hours.\n\n## Escalation Level\n3 — Supervisor review\n\nRegression co-occurring with increasing problem behavior warrants BCBA review within 24–48 hours before implementing proposed FCT and extinction changes.\n\n## Confidence\nmoderate\n\nFunction hypothesis is well-supported by 2/2 documented ABC events, but single-observer sessions limit IOA coverage to one session out of 10. Recommend one additional IOA session before finalizing the intervention plan.\n\n## Data-Supported Rationale\n- Mean Safety Signs accuracy sessions 1–3: 58%; sessions 8–10: 42% (16 percentage point decrement, 27% relative decline).\n- Mean tantrum frequency sessions 1–3: 1.3/session; sessions 8–10: 3.3/session (2.5× increase). Mean duration 4 min -> 12 min.\n- 2/2 documented ABC events show demand antecedent -> tantrum behavior -> task removal consequence, consistent with escape-maintained hypothesis.\n- Mouthing PIR 20% -> 35% across the same period (75% relative increase).\n- IOA data (session 7) — 87–95% agreement — supports data reliability for tantrum and AAC measures; no IOA for Safety Signs or Mouthing." + } + ], + "meta": { + "task_type": "session_interpretation", + "example_id": "b5c9e2a14f8d7310", + "gold_labels": { + "pattern_class": "regression+frustration_pattern", + "behavior_functions": { + "Tantrum": "escape", + "Mouthing": "automatic" + }, + "escalation_level": 3, + "confidence": "moderate", + "crisis_plan_required": true + }, + "log_properties": { + "n_sessions": 10, + "n_acceleration_programs": 4, + "n_deceleration_targets": 2, + "has_abc_data": true, + "has_ioa_session": true, + "learner_profile": "adolescent", + "context_distribution": {"D": 0.2, "R": 0.8, "Both": 0.0} + }, + "provenance": { + "layer": 1, + "template_id": "sess-regress-frust-v1", + "pattern_seed": 20260423007, + "seed": 20260423007, + "generated_at": "2026-04-23T14:30:00Z" + } + } +} +``` + +--- + +## 5. Validation and parsing + +### 5.1 Required fields + +For every example (both tasks): +- `messages` present with exactly 3 entries: system, user, assistant (in that order) +- `meta.task_type` in `{teaching_program, session_interpretation}` +- `meta.example_id` present and non-empty +- `meta.gold_labels` present with task-specific required keys +- `meta.provenance.seed` present + +### 5.2 Label extraction regex (for evaluator) + +After model generation, labels are extracted from the assistant response using deterministic regex. If extraction fails, the example is scored as a parse failure (counts against the model). + +**Pattern classification (task 2):** +```python +r"##\s*Pattern\s*Classification\s*\n\s*([a-z_+]+)\s*\n" +``` + +**Escalation level (task 2):** +```python +r"##\s*Escalation\s*Level\s*\n\s*([1-4])\s*—" +``` + +**Confidence (task 2):** +```python +r"##\s*Confidence\s*\n\s*(high|moderate|low)\s*\n" +``` + +**Behavior function (task 2, per behavior):** +```python +r"^([A-Z][A-Za-z\s\-\(\)]+):\s*(escape|attention|tangible|automatic|unknown)\s*$" +``` + +**Method (task 1, validated against user input):** +```python +# No regex needed — method is sampled and stored in gold_labels; evaluator checks +# whether the assistant output contains the method-specific sections (section 3.4) +# corresponding to the expected method. +``` + +### 5.3 Schema validity checks (pre-training) + +`src/prepare_data.py` applies these before writing to splits: + +1. All 3 messages present with correct roles +2. System prompt matches canonical form for the task +3. Assistant response contains all required section headers for its task +4. Structured-label sections are regex-extractable +5. No placeholder strings (`{...}`, `TODO`, empty `Provide:` blocks) in assistant content +6. User message contains no real-world identifying patterns (regex scan for name-like tokens) +7. Total length under `max_seq_length` (4096 tokens for training) +8. No duplicate `example_id` across the full dataset (SimHash + exact-match dedup) + +Examples failing any check are written to `data/processed/rejected.jsonl` with a `reason` field. + +--- + +## 6. Extensibility + +### 6.1 Adding new session patterns + +1. Add entry to `taxonomy-v1.md` section 7 with citation. +2. Add generator template in `src/prepare_data.py` with trajectory rules. +3. Add label to regex enum in section 5.2. +4. Add stratification target in split generator. +5. Bump schema version. + +### 6.2 Adding new teaching methods + +1. Add entry to `taxonomy-v1.md` section 1. +2. Add method-specific output variant to `schema-v1.md` section 3.4. +3. Add generator template in `src/prepare_data.py`. +4. Bump schema version. + +### 6.3 Versioning + +Schema versions follow `major.minor`: +- **Minor** — additive changes (new pattern, new method, new optional field). Old data remains valid. +- **Major** — breaking changes (renamed fields, removed sections, changed required structure). Old data requires migration or a new dataset version. + +Each generated example records the schema version it was generated under in `meta.provenance.schema_version`. + +--- + +## 7. Changelog + +- **v1.0 (2026-04-23)** — initial schema. Two tasks defined. Task 1 covers 6 teaching methods (DTT, NET, Task Analysis, FCT, BST, PRT); Task 2 covers 12 session patterns with structured labels for pattern class, behavior function, escalation level, and confidence. diff --git a/docs/taxonomy-v1.md b/docs/taxonomy-v1.md new file mode 100644 index 0000000000000000000000000000000000000000..e6dbe31b73ee5dbf9e363cff61d2a32d4681ab41 --- /dev/null +++ b/docs/taxonomy-v1.md @@ -0,0 +1,608 @@ +# TRACE Dataset Taxonomy v1 + +**Purpose.** This document is the **controlled vocabulary** for the TRACE synthetic dataset. Every category our data-generation pipeline can draw from is defined here, with an operational definition and a canonical citation. If a category is not in this document, the generator will not produce it — and if it is here, its clinical validity is defensible against a peer-reviewed source. + +**Who this serves.** +- The data-generation pipeline (`src/prepare_data.py`) consumes this taxonomy as its source of truth. +- The dataset paper's composition table is a direct rendering of this document. +- Reviewers verifying "no invented clinical categories" can trace each entry -> citation -> CHH chapter / JABA paper / curriculum manual. + +**Status.** Version 1, 2026-04-22. Breaking changes require a bumped version and a changelog entry. + +**Citation conventions.** Throughout this document: +- **CHH** = Cooper, Heron, & Heward (2020). *Applied Behavior Analysis* (3rd ed.). Pearson. +- **JABA** = *Journal of Applied Behavior Analysis* +- **BAP** = *Behavior Analysis in Practice* +- DOIs link to the authoritative source wherever possible. + +--- + +## 0.A Program orientation — Acceleration vs Deceleration + +Every ABA goal falls into one of two program orientations, a distinction that runs through virtually all clinical data systems: + +| Orientation | Purpose | Covers | +|---|---|---| +| **Acceleration** | Behaviors / skills to **increase** | Skill-acquisition programs (VB-MAPP, AFLS targets); adaptive behaviors (toileting, communication including AAC); FCT-trained replacement behaviors; independent coping with denied access | +| **Deceleration** | Behaviors to **decrease** | Target problem behaviors (section 5): SIB, aggression, elopement, tantrum, stereotypy, pica, etc. | + +**Citation.** Standard ABA terminology; see CHH Ch. 22 ("Differential Reinforcement" — acceleration via reinforcement of desired behavior) and Ch. 24 ("Extinction") together with Ch. 25 ("Antecedent Interventions") for deceleration. + +**Dataset usage.** +- section 1 teaching methods DTT, NET, Task Analysis, PRT, Incidental Teaching apply to **acceleration** programs. +- section 1.4 FCT is the acceleration method paired with deceleration — it teaches a replacement behavior while reducing a target behavior. +- section 1.5 BST is meta-level (trains staff on any program). +- Task 2 session logs sample across both orientations per session — several acceleration targets and zero or more deceleration behaviors. Exact counts are pipeline hyperparameters. + +--- + +## 0.B How taxonomy dimensions combine + +Each **teaching program generation** example is produced by sampling one value from each of these dimensions: + +| Dimension | # values | Section | +|---|---|---| +| Teaching method | 6 (optionally 8 if video modeling + script fading approved) | section 1 | +| Skill domain | 16 (VB-MAPP) + 6 (AFLS) = 22 | section 2 | +| Developmental level | 5 (3 VB-MAPP + 2 AFLS) | section 2 | +| Skill target | ~250 | section 2 | +| Mastery state | 7 | section 3 | +| Prompt hierarchy | 6 | section 8 | +| Reinforcement schedule | 7 | section 9 | +| Error correction | 5 | section 10 | +| Mastery criteria | 7 | section 11 | + +Each **behavioral session interpretation** example is produced by sampling: + +| Dimension | # values | Section | +|---|---|---| +| Session pattern | 12 | section 7 | +| Primary skills observed | 3–8 per session | section 2 | +| Target behaviors observed (if any) | 0–3 per session | section 5 | +| Measurement types used | 2–5 per session | section 6 | +| Behavior function hypothesis | 4 + "not applicable" | section 4 | +| Escalation level | 4 | section 12 | +| Confidence level | 3 | section 13 | +| BIP component structure | 4 | section 11 | + +--- + +## 1. Teaching methods + +The first task (Teaching Program Generation) produces content modality-aware. A "mand" goal is not automatically a DTT program — real clinical practice picks the method that fits the skill and learner. The model must reason about method selection. + +### 1.1 Discrete Trial Training (DTT) +- **Operational definition.** A structured teaching format consisting of clearly defined trial units: *discriminative stimulus (SD) -> learner response -> consequence (reinforcement or correction) -> inter-trial interval (ITI)*. Trials are massed in a contrived setting (usually table-based), with explicit prompt hierarchy and error-correction procedure. +- **When to use.** Acquisition of discrete, decontextualized skills (receptive labels, tacts, matching, early imitation) especially early in instruction when stimulus control must be established. +- **Canonical citation.** Lovaas, O. I. (1987). *JCCP, 55*(1), 3–9. https://doi.org/10.1037/0022-006X.55.1.3 — seminal EIBI trial establishing DTT-based programming. Modern definition: Smith, T. (2001). *Focus on Autism, 16*(2), 86–92. +- **Dataset usage.** Default teaching method for early-learner (VB-MAPP Level 1–2) discrete skills. Output includes explicit SD, prompt hierarchy, stimulus array, error-correction, mastery criteria. + +### 1.2 Natural Environment Teaching (NET) +- **Operational definition.** Teaching that occurs within the learner's natural context, capitalizing on child-initiated motivation and ongoing activities. The teaching opportunity is embedded in the environment, not contrived. +- **When to use.** Generalization phase, functional communication, requesting (mand) skills, when motivation is the limiting factor, or when the skill's natural occasion is ecologically predictable (e.g., requesting food during snack time). +- **Canonical citation.** Hart, B. & Risley, T. R. (1975). *JABA, 8*(4), 411–420. https://doi.org/10.1901/jaba.1975.8-411 (origin of incidental teaching). Autism extension: McGee, G. G., Krantz, P. J., & McClannahan, L. E. (1985). *JABA, 18*(1), 17–31. +- **Dataset usage.** Default teaching method for mand targets and generalization-phase goals. Output emphasizes environmental setup / MO arrangement over contrived SD. + +### 1.3 Task Analysis / Chaining +- **Operational definition.** Decomposition of a complex behavior into sequential component steps, followed by teaching each step and chaining them together. Three chain types: + - **Forward chaining** — teach step 1 to mastery, then 1->2, then 1->2->3, etc. + - **Backward chaining** — complete all steps except the last; teach last; move backward. + - **Total-task presentation** — prompt through entire chain every trial; fade prompts across all steps. +- **When to use.** Multi-step skills with a natural sequence: toileting, handwashing, dressing, cooking, social routines. +- **Canonical citation.** CHH Ch. 23 ("Chaining"). Empirical comparison: Slocum, S. K. & Tiger, J. H. (2011). *JABA, 44*(4), 793–805. https://doi.org/10.1901/jaba.2011.44-793 +- **Dataset usage.** Default for self-care (AFLS Basic Living), domestic, community, and vocational domains. Output structure replaces "SD" with "task analysis steps" (numbered list) + chain type + prompt strategy per step. + +### 1.4 Functional Communication Training (FCT) +- **Operational definition.** Procedure for teaching a communication response that serves the *same function* as a target problem behavior, paired with extinction of the problem behavior. Example: a child who tantrums to escape demands is taught to say/sign "break please," which is reinforced; the tantrum is placed on extinction. +- **When to use.** Any target behavior with a clear social function (escape, attention, tangible). First-line strategy before consequence-based interventions. +- **Canonical citation.** Carr, E. G. & Durand, V. M. (1985). *JABA, 18*(2), 111–126. https://doi.org/10.1901/jaba.1985.18-111 (seminal). Modern practical review: Tiger, J. H., Hanley, G. P., & Bruzek, J. (2008). *BAP, 1*(1), 16–23. +- **Dataset usage.** Default teaching method in behavior-intervention-adjacent recommendations (task 2 interpretation output when behavior data is present). Output ties communication response -> hypothesized function -> reinforcement schedule. + +### 1.5 Behavior Skills Training (BST) +- **Operational definition.** Staff/caregiver training procedure with four components in sequence: **(1) Instruction**, **(2) Modeling**, **(3) Rehearsal**, **(4) Feedback**, iterated until the trainee meets mastery criteria on a performance checklist. +- **When to use.** Training RBTs on a new program, training caregivers on a home procedure, preparing staff for a BIP rollout. +- **Canonical citation.** Parsons, M. B., Rollyson, J. H., & Reid, D. H. (2012). *BAP, 5*(2), 2–11. https://doi.org/10.1007/BF03391819 +- **Dataset usage.** Used in a minority of task 1 outputs when the target is a staff-facing training program (rather than a learner-facing teaching program). Output structure emphasizes trainer steps and fidelity checklist. + +### 1.6 Pivotal Response Training (PRT) +- **Operational definition.** Naturalistic teaching targeting "pivotal" behaviors that produce widespread collateral gains: **motivation**, **self-initiation**, **responsivity to multiple cues**, **self-management**. Uses child choice, interspersal of mastered/acquisition targets, natural reinforcers, and reinforcement of attempts (not just correct responses). +- **When to use.** Language acquisition in autism, social initiation, generalization across contexts. +- **Canonical citation.** Koegel, R. L., O'Dell, M. C., & Koegel, L. K. (1987). *JADD, 17*(2), 187–200. https://doi.org/10.1007/BF01495055 (origin of NLP -> PRT). JABA-indexed: Laski, K. E., Charlop, M. H., & Schreibman, L. (1988). *JABA, 21*(4), 391–400. +- **Dataset usage.** Used for a minority of early-language targets in naturalistic conditions. Output emphasizes motivation arrangement, child-choice, attempt-reinforcement. + +### Method-selection heuristics (encoded as prompt guidance, not hard rules) + +| Skill context | Default method | Alternative | +|---|---|---| +| Early discrete skills (tacts, receptive labels, matching) | DTT | PRT (if motivation is limiting) | +| Mand / requesting | NET | FCT (if replacing problem behavior) | +| Multi-step routines (dressing, handwashing) | Task Analysis | — | +| Problem-behavior reduction | FCT | — | +| Generalization phase | NET | PRT | +| Staff-facing training | BST | — | +| Social-initiation skills | PRT | NET | + +--- + +## 2. Skill curriculum (Acceleration targets) + +All skill domains in this section are **acceleration** targets — goals we want the learner to perform MORE often / more independently. See section 0.A. + +Two overlapping curricula are encoded: **VB-MAPP** (early learners, 0–48 months developmental age) and **AFLS** (adolescents / adults and functional living). Together they cover the full learner age span TRACE should handle. + +### 2.1 VB-MAPP (Verbal Behavior Milestones Assessment and Placement Program) + +**Primary citation.** Sundberg, M. L. (2008). *Verbal Behavior Milestones Assessment and Placement Program*. AVB Press. Validation study: Barnes, C. S., Mellor, J. R., & Rehfeldt, R. A. (2014). *Analysis of Verbal Behavior, 30*(1), 36–47. https://doi.org/10.1007/s40616-013-0004-5 + +**Developmental levels:** +- **Level 1** — 0–18 months typical developmental age +- **Level 2** — 18–30 months +- **Level 3** — 30–48 months + +**16 skill domains**, each expanded to 5 representative targets per developmental level (Sundberg 2008): + +**1. Mand** — requesting. +- L1: single-word mands for preferred items; mands for missing items needed to complete an activity; mands for actions; mands for help; mands using 2-word phrases. +- L2: mands for information using "what"; mands for information using "where"; mands using adjectives (big, little, more); mands for attention from peers; mands for others to stop an action. +- L3: mands for information using "why"; mands for information using "when"; mands using complete sentences with correct grammar; mands for future events or items not present; mands using polite social conventions. + +**2. Tact** — labeling / naming. +- L1: common objects (ball, cup, shoe); familiar people by name; common actions (running, eating, sleeping); body parts (nose, eyes, mouth); common animals (dog, cat, bird). +- L2: colors of objects; shapes (circle, square, triangle); adjectives (big/little, hot/cold); prepositions (in, on, under); emotions in self and others. +- L3: community helpers and their roles; categories (fruits, vehicles, clothing); features of objects (color, shape, function); past-tense events; abstract concepts (same/different, first/last). + +**3. Echoic** — vocal imitation. +- L1: single vowel sounds; single consonant-vowel combinations; 1–2 syllable words; animal sounds; familiar words on request. +- L2: 2-word phrases; 3-word phrases; blended consonants; novel words with correct articulation; short sentences. +- L3: multi-syllable words accurately; full sentences with correct prosody; unfamiliar words from context; phrases with correct intonation patterns; complex instructions verbatim. + +**4. Listener Responding** — receptive language. +- L1: 1-step motor instructions (sit down, stand up); select correct item from an array of 2; point to named body parts; point to named common objects; follow instructions involving objects. +- L2: select correct item from an array of 4–6; follow 2-step instructions; select items by feature (find something red); select items by function (find something you eat with); select items by class (find an animal). +- L3: follow multi-step instructions with qualifiers; select items by multiple features simultaneously; follow conditional instructions (if X, then Y); respond to questions about stories read aloud; follow instructions involving temporal concepts (before, after). + +**5. Visual Perceptual Skills & Matching-to-Sample (VP-MTS).** +- L1: match identical objects; match identical pictures; match colors (identical chips); match shapes (identical blocks); complete simple 3–4 piece puzzles. +- L2: match non-identical items by category; sort items into 2–3 categories; match quantities (1–5); match upper-case letters; complete 8–12 piece puzzles. +- L3: match associated items (sock-shoe, cup-plate); sort by multiple attributes simultaneously; match upper to lower case letters; sequence 3–4 step picture sequences; reproduce block designs from model. + +**6. Motor Imitation.** +- L1: gross motor actions (clap hands, stomp feet); actions with objects (bang drum, push car); 2-step motor sequences; fine motor actions (pinch, point); facial movements (open mouth, stick out tongue). +- L2: 3-step motor sequences; novel motor actions on first attempt; actions in songs and finger plays; actions involving bilateral coordination; motor actions after a delay. +- L3: complex multi-step sequences; motor actions from video models; handwriting strokes and letter forms; craft and art activities; sports-related motor sequences. + +**7. Independent Play.** +- L1: cause-and-effect toys independently; sensory toys for sustained periods; construction toys (blocks, Duplo); vehicles (pushing, rolling); electronic learning toys. +- L2: simple pretend play sequences; art materials (coloring, playdough); age-appropriate puzzles independently; books independently; simple board or card games. +- L3: elaborate pretend play with storylines; creative construction activities; rule-based games independently; independently select and transition between activities; hobby / special-interest activities productively. + +**8. Social Behavior & Social Play.** +- L1: eye contact during interactions; responds to own name; social games (peek-a-boo, tickle); shows items to others spontaneously; tolerates proximity of peers during parallel play. +- L2: initiates social interactions with peers; takes turns during structured activities; shares materials when prompted; cooperative play with one peer; responds to peer initiations appropriately. +- L3: maintains reciprocal conversations with peers; demonstrates empathy and perspective-taking; negotiates and compromises during group activities; joins ongoing peer activities appropriately; maintains friendships over time. + +**9. Intraverbal** — fill-ins, WH answers without visual prompt. +- L1: fills in words in songs and nursery rhymes; answers "what's your name"; fills in missing words in familiar phrases; answers simple what-questions about visible items; completes verbal routines (ready, set, ___). +- L2: answers WH-questions about familiar topics; describes function of common objects; names items in categories when given the category; answers social questions (How are you?); describes recent events in sequence. +- L3: answers why- and how-questions; provides definitions of words; answers hypothetical questions; engages in multi-turn conversations on a topic; makes inferences from given information. + +**10. LRFFC** — Listener Responding by Feature, Function, Class. +- L1: select items by function (What do you drink from?); select items by feature (What is round?); select items by class (Find the animal); select by single function from array of 3; select by single visible feature from array of 3. +- L2: select items by function from array of 5–8; select items by feature from array of 5–8; select items by class from array of 5–8; select items by multiple features (round and red); select items by function when item not visible. +- L3: select items given 2+ features/functions simultaneously; select items by class with exclusion (animal but not a pet); select items by abstract features (something needed when cold); select items by comparison (which is heavier); select items by negative features (not round, not food). + +**11. Reading.** +- L1: match letters to identical letters; identify own name in print; identify 5–10 upper-case letters; match words to identical words; track print left to right. +- L2: identify all upper-case letters; identify all lower-case letters; read 10–20 sight words; sound out CVC words (cat, dog, run); read simple 2–3 word phrases. +- L3: read sentences with comprehension; read short passages and answer questions; phonetic decoding for novel words; read and follow written instructions; read grade-level text with fluency. + +**12. Writing.** +- L1: trace horizontal and vertical lines; trace basic shapes (circle, cross); copy basic shapes from model; trace letters of own name; write own name from model. +- L2: write own name independently; copy all upper-case letters from model; write upper-case letters from dictation; copy simple words from model; write numbers 1–10. +- L3: write words from dictation; write simple sentences with spacing; write answers to questions; write short compositions (3–5 sentences); use basic punctuation and capitalization. + +**13. Math.** +- L1: rote count to 10; count objects 1–5 with 1:1 correspondence; identify numerals 1–5; match quantities to numerals 1–5; identify basic shapes in math context. +- L2: count objects 1–20 with 1:1 correspondence; identify numerals 1–20; compare quantities (more/less/same); solve single-digit addition with manipulatives; identify coins by name. +- L3: add single-digit numbers without manipulatives; subtract single-digit numbers; identify place value (ones, tens); tell time to the hour and half-hour; solve simple word problems. + +**14. Group & Classroom Skills.** +- L1: sits in designated area for 2–3 minutes; attends to teacher during group instruction; follows group instructions (everyone stand up); transitions between activities with prompts; waits in line briefly. +- L2: sits in group for 10–15 minutes; raises hand to request or respond; follows classroom routines independently; works independently at desk for 5–10 minutes; transitions between activities independently. +- L3: participates in group discussions; follows multi-step classroom instructions; works independently for 15+ minutes; self-monitors behavior using checklist; completes and turns in assignments independently. + +**15. Linguistic Structure** — grammar. +- L1: single words to communicate; combines 2 words (agent + action or action + object); basic noun-verb combinations; simple negation (no, not); basic pronouns (I, me, my). +- L2: 3–4 word sentences; regular plurals (-s); present progressive (-ing); prepositions in sentences; regular past tense (-ed). +- L3: complex sentences with conjunctions; irregular past tense correctly; pronouns with correct referents; questions with correct word order; passive voice and embedded clauses. + +**16. Spontaneous Vocal Behavior.** +- L1: spontaneous vocalizations during preferred activities; spontaneous naming of items; spontaneous requests without prompts; spontaneous greetings to familiar people; spontaneous comments on events. +- L2: spontaneous descriptions of ongoing activities; spontaneous questions about the environment; spontaneous reporting of past events; spontaneous use of social phrases; spontaneous initiation of conversation with peers. +- L3: spontaneous storytelling / narratives; spontaneous relevant comments in conversations; spontaneous adjustment of language to different listeners; spontaneous use of humor appropriately; spontaneous provision of explanations and reasons. + +### 2.2 AFLS (Assessment of Functional Living Skills) + +**Primary citation.** Partington, J. W., & Mueller, M. M. (2012). *Assessment of Functional Living Skills*. Behavior Analysts, Inc. / Stimulus Publications. + +**Developmental scope.** Late-childhood through adult; skills are functional rather than milestone-bound. + +**6 skill modules** (we encode a representative subset of each): + +1. **Basic Living Skills** — toileting, eating, dressing, hygiene, bathing (chaining-intensive). +2. **Home Skills** — meal preparation, cleaning, laundry, bed-making. +3. **Community Participation Skills** — safety signs, money handling, transportation, shopping, restaurants. +4. **School Skills** — classroom routines, assignments, social-academic interaction. +5. **Vocational Skills** — task completion, workplace etiquette, punctuality. +6. **Independent Living Skills** — budgeting, appointments, medication, self-advocacy. + +**Dataset usage.** AFLS targets appear in task 1 examples for older learners (labeled "Adolescent" or "Adult" learner profile) and in task 2 session logs for those learner profiles. Teaching method defaults to **Task Analysis / Chaining**. + +### 2.3 Learner-profile age bands + +| Profile label | Age band | Curricula | +|---|---|---| +| Early Learner | developmental age 0–48 mo | VB-MAPP | +| School-Age Learner | 6–12 yr chronological, varying developmental | VB-MAPP L2–3 + AFLS School | +| Adolescent Learner | 13–17 yr chronological, moderate-severe support needs | AFLS + VB-MAPP L3 for residual gaps | +| Adult Learner | 18+ yr in residential / day program | AFLS + Community + Vocational | + +**Dataset usage.** Learner-profile label is a sampling dimension. + +**On distribution.** No research source prescribes an ABA-population age-band distribution. The pipeline config sets a balanced default across the four profiles (rather than skewing toward any band) so the model trains on the full lifespan. Real-world caseload distributions vary by facility type (early-intervention centers skew Early; residential programs skew Adult). The exact ratio is a *pipeline hyperparameter*, not a clinical truth. + +### 2.4 Gaps explicitly not covered in v1 + +- **Essential for Living** (McGreevy, Fry, & Cornwall 2014) — skills for adults with severe support needs. Partially covered by AFLS; full EFL integration is not included in v1. +- **ABLLS-R** (Partington 2010) — heavily overlaps VB-MAPP for early learners; not separately encoded. +- **PEAK** (Dixon 2014+) — relational frame / advanced cognition; out of scope for baseline dataset. + +--- + +## 3. Mastery states + +Seven-state taxonomy describing where a skill is in its learning trajectory. Values are slot values in task 1 user prompts and targets in task 2 interpretation. + +| State | Operational criterion | Clinical meaning | +|---|---|---| +| **Emerging** | < 30% accuracy across recent sessions | Skill not yet acquired; high prompt levels needed | +| **Developing** | 30–50% accuracy, inconsistent | Responding established; stimulus control unstable | +| **Approaching mastery** | 50–70% accuracy, prompt fading in progress | Moving toward independence | +| **Near mastery** | 70–85% accuracy, occasional errors | Stimulus control solid, complex stimuli still variable | +| **Mastered (current level)** | ≥ 85% accuracy across 3 consecutive sessions | Meets mastery criterion for current step | +| **Generalization phase** | Mastered in training but not in novel settings / with novel materials / across therapists | Ready for NET-style generalization probes | +| **Maintenance** | Previously mastered; periodic probes to check retention | Maintenance schedule (weekly->monthly) | + +**Citation.** Mastery-criterion conventions follow CHH Ch. 26 ("Generalization and Maintenance of Behavior Change") and the programmatic mastery-criteria convention in CHH Ch. 28. + +**Dataset usage.** Mastery state is a primary slot in every task 1 prompt. In task 2, the narrative "a previously mastered skill has regressed" vs "a developing skill has plateaued" is the basis for pattern detection (section 7). + +--- + +## 4. Behavior functions + +The **four-function taxonomy** of problem behavior is the spine of all behavior reasoning in ABA. Iwata's original functional analysis identified three (attention, escape, sensory); Hanley, Iwata & McCord formalized the fourth (tangible) as standard. + +| Function | Reinforcer | Example context | Hypothesis test | +|---|---|---|---| +| **Escape** (negative reinforcement) | Termination / avoidance of an aversive stimulus (demand, task, interaction) | Child hits when presented with math worksheet -> demand is removed | Elevated rate in demand condition vs control | +| **Attention** (positive reinforcement, social) | Attention from others (adult or peer) | Child yells when parent on phone -> parent attends | Elevated rate in attention condition vs control | +| **Tangible** (positive reinforcement, material) | Access to an item or activity | Child tantrums when iPad is taken away -> iPad is returned | Elevated rate when preferred item is restricted | +| **Automatic** (non-social / sensory) | Self-produced reinforcement (proprioceptive, auditory, visual, gustatory) | Repetitive hand-flapping persists during alone condition | Elevated rate in alone condition; persists across conditions | + +**Canonical citations.** +- Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1982/1994). *JABA, 27*(2), 197–209. https://doi.org/10.1901/jaba.1994.27-197 — THE foundational functional-analysis paper. +- Hanley, G. P., Iwata, B. A., & McCord, B. E. (2003). *JABA, 36*(2), 147–185. https://doi.org/10.1901/jaba.2003.36-147 — definitive review; formalizes four-function taxonomy including tangible. +- Pre-FA conceptual roots: Carr, E. G. (1977). *Psychological Bulletin, 84*(4), 800–816. +- CHH Ch. 27 ("Functional Behavior Assessment"). + +**Clinical principle.** Same topography can serve different functions across individuals. A "tantrum" in one client may be escape-maintained; in another, attention-maintained. **Interventions must match function**, not topography. + +**Dataset usage.** +- Task 2 interpretation output includes **Function Hypothesis** field when behavior data is present — one of the four functions or "Not Applicable" (skill-only sessions). +- When present, the hypothesis cites the evidence in the session log (e.g., "elevated rate during demand presentation -> escape-maintained"). + +--- + +## 5. Target behaviors — Deceleration targets (with operational definitions) + +All behaviors in this section are **deceleration** targets — behaviors we want to reduce. See section 0.A. + +Eighteen challenging behaviors with JABA-grounded operational definitions suitable for staff scoring. Behaviors without a canonical operational definition in the literature (e.g., disrobing, grabbing/snatching) are deliberately excluded from v1 rather than invented. + +| # | Behavior | Operational definition | Citation | +|---|---|---|---| +| 1 | **Self-injurious behavior (SIB)** | Any response that produces tissue damage or has the potential to produce it, including head-hitting, self-biting, face-slapping, head-banging against objects, skin-picking, self-pinching, hair-pulling directed at self. | Iwata et al. 1982/1994 | +| 2 | **Head-directed SIB** (subtype) | Contact between the individual's hand and head, OR between the head and a stationary object, with audible impact. | Iwata et al. 1982/1994 | +| 3 | **Aggression** | Attempted or completed forceful contact directed toward another person: hitting, kicking, biting, hair-pulling, scratching, pinching, or throwing objects AT another person. | Marcus, B. A. et al. (2001). *Behavior Modification, 25*(2), 189–213. https://doi.org/10.1177/0145445501252002 | +| 4 | **Property destruction** | Hitting or kicking furniture/walls; throwing objects not intended to be thrown; tearing clothing, books, or materials; swiping items off surfaces; overturning furniture. | Piazza, C. C., Bowman, L. G., Contrucci, S. A., et al. (1999). *JABA, 32*(4), 437–449. https://doi.org/10.1901/jaba.1999.32-437 | +| 5 | **Elopement** | Full body (or pre-specified anatomical threshold) crossing a designated boundary (e.g., classroom door, yard fence) without adult approval. | Kodak, T., Grow, L., & Northup, J. (2004). *JABA, 37*(2), 229–232. https://doi.org/10.1901/jaba.2004.37-229; Lang, R. et al. (2010). *JABA, 43*(1), 113–118. | +| 6 | **Pica** | Placement of any inedible (non-food) item past the plane of the lips, including mouthing and ingestion of objects (paper, dirt, cigarette butts, fabric, small toys, hair). | Piazza, C. C., Fisher, W. W., Hanley, G. P., et al. (1998). *JABA, 31*(2), 165–189. https://doi.org/10.1901/jaba.1998.31-165 | +| 7 | **Motor stereotypy** | Repetitive, non-functional motor movements (hand-flapping, body-rocking, finger-flicking, spinning) occurring independent of social context and serving no apparent instrumental purpose. | Rapp, J. T. & Vollmer, T. R. (2005). *RIDD, 26*(6), 527–547. https://doi.org/10.1016/j.ridd.2004.11.005 | +| 8 | **Vocal stereotypy** | Non-contextual, non-communicative vocalization: repetitive sounds, words, phrases, humming, or echolalia outside appropriate conversational context. | Ahearn, W. H., Clark, K. M., MacDonald, R. P. F., & Chung, B. I. (2007). *JABA, 40*(2), 263–275. https://doi.org/10.1901/jaba.2007.30-06 | +| 9 | **Tantrum** | Co-occurring cluster of two or more of: crying/screaming, dropping to floor, kicking, hitting, throwing objects, lasting ≥ 3 s. Scored from onset of first component topography to 5 s without any component behavior. | Kurtz, P. F. et al. (2003). *JABA, 36*(2), 205–219. https://doi.org/10.1901/jaba.2003.36-205; topography adapted from Carr & Durand 1985. | +| 10 | **Non-compliance** | Failure to initiate a requested response within 5 s of an instructional prompt (vocal, model, or physical), OR active refusal (verbal "no," walking away). | Wilder, D. A. et al. (2012). *JABA, 45*(1), 121–126. https://doi.org/10.1901/jaba.2012.45-121 | +| 11 | **Mouthing (non-pica)** | Placement of the hand, fingers, saliva, or non-food object into the mouth past the lip plane, excluding eating/drinking during scheduled meals. | Piazza, C. C., Hanley, G. P., & Fisher, W. W. (1996). *JABA, 29*(4), 437–449. https://doi.org/10.1901/jaba.1996.29-437 | +| 12 | **Throwing** | Propelling an object through the air with force (excluding task-relevant throwing such as during a game). | Property destruction literature, specifically Piazza et al. 1999 — throwing is often coded as a subtype. | +| 13 | **Food refusal / selectivity** | Rejection of presented food by turning the head, closing the mouth, batting the utensil away, or expelling food once placed in the mouth. | Piazza, C. C. et al. (2003). *JABA, 36*(2), 187–204 — feeding-disorder literature. | +| 14 | **Rumination / regurgitation** | Voluntary regurgitation of previously ingested food into the mouth, followed by re-chewing, re-swallowing, or expulsion. Scored per regurgitation episode. | Lyons, E. A., Rue, H. C., Luiselli, J. K., & DiGennaro, F. D. (2007). *JABA, 40*(4), 743–747. https://doi.org/10.1901/jaba.2007.743-747. See also Kahng et al. 2003. | +| 15 | **Sleep resistance / disruption** | Non-compliance with bedtime routine (refusing to go to bed, leaving bed, repeated requests) OR night waking ≥ 5 min at a time. Scored as instances or total disrupted minutes per night. | Friman, P. C., Hoff, K. E., Schnoes, C., Freeman, K. A., Woods, D. W., & Blum, N. (1999). *JABA, 32*(4), 505–508. https://doi.org/10.1901/jaba.1999.32-505 | +| 16 | **Verbal aggression / threats** | Yelling, cursing, name-calling, or making verbal threats directed toward another person (e.g., "I'll hit you," "I hate you"). | Kelley, M. E., Lerman, D. C., & Van Camp, C. M. (2002). *JABA, 35*(1), 59–63. See also Hagopian & Boelter 2005. | +| 17 | **Food stealing** | Taking food items not offered to the individual — from another person's plate, a storage area, or during restricted-access periods — and/or placing such items in the mouth outside designated meal/snack times. | Maglieri, K. A., DeLeon, I. G., Rodriguez-Catter, V., & Sevin, B. M. (2000). Adjunctive delivery of noncontingent reinforcement to treat food stealing during sessions of DRO. *JABA, 33*(4), 615–618. https://doi.org/10.1901/jaba.2000.33-615 | +| 18 | **Inappropriate sexual behavior (ISB)** | Engagement in sexual self-stimulation (genital contact, rhythmic self-stimulatory movements) in public or semi-public contexts where such behavior is socially or institutionally inappropriate. Scored per episode with onset/offset criteria specified in the program. | Davis, T. N., Machalicek, W., Scalzo, R., Kobylecky, A., Campbell, V., Pinkelman, S., Chan, J., & Sigafoos, J. (2016). A review and treatment selection model for individuals with developmental disabilities who engage in inappropriate sexual behavior. *BAP, 9*(4), 389–402. https://doi.org/10.1007/s40617-016-0149-5. See also Fyffe, C. E., Kahng, S., Fittro, E., & Russell, D. (2004). *JABA, 37*(3), 401–404. | + +**Dataset usage.** +- Task 2 session logs include 0–3 target behaviors per log. Sampling weight biased toward the common behaviors (tantrum, non-compliance, aggression, stereotypy, elopement). +- Each target behavior in a log has an associated function hypothesis (section 4) that drives the interpretation output. + +--- + +## 6. Measurement types + +Real session logs mix measurement types; task 2 inputs must reflect this heterogeneity. + +| Type | Operational definition | Use case | Citation | +|---|---|---|---| +| **Frequency** (event recording) | Raw count of occurrences of a discrete, quickly-completing behavior during a session. "12 requests this session." | Discrete, countable behaviors with clear onset/offset. | CHH Ch. 4 | +| **Rate** | Frequency normalized to time: count ÷ session duration. "0.4 requests per minute." | Compare sessions of different durations. | CHH Ch. 4 | +| **Duration** | Total elapsed time a behavior lasts (sum across instances). "Tantrum total duration: 8 min." | Behaviors with meaningful length (tantrums, on-task engagement, stereotypy). | CHH Ch. 4 | +| **Latency** | Time from SD to initiation of response. "Mean latency: 3.2 s." | Responsivity / prompt dependency. | CHH Ch. 4 | +| **Partial-interval recording (PIR)** | Divide session into intervals (e.g., 10 s); mark each interval if behavior occurred *at any point*. Overestimates prevalence. | Hard-to-count behaviors (stereotypy). | Powell, Martindale & Kulp 1975; Harrop & Daniels 1986 | +| **Whole-interval recording** | Mark interval only if behavior occurred for the *entire* interval. Underestimates. | Behaviors to increase (on-task engagement). | Cooper et al. Ch. 4 | +| **Momentary time sampling (MTS)** | Check once at the end of each interval whether behavior is occurring. Efficient for groups. Approximates duration. | Multi-client settings; duration estimation. | Powell et al. 1975 | + +**Inter-Observer Agreement (IOA).** Two observers independently record the same session; agreement expressed as percent or κ. **Minimum acceptable: ≥ 80%** across 33% of sessions for scientific validity. +- Citation: CHH Ch. 5 ("Improving and Assessing the Quality of Behavioral Measurement"); Kazdin, A. E. (1977). *JABA, 10*(1), 141–150. https://doi.org/10.1901/jaba.1977.10-141 + +**Dataset usage.** +- Every task 2 session log includes at least one **Primary Measurement Type** per target skill or behavior. +- ~25% of logs include an **IOA subset** (percentage agreement reported) to test the model's ability to interpret reliability. +- Accuracy % (trial-based correct/total) is the dominant measurement for skill-acquisition programs — this stays. + +### 6.1 Antecedent-Behavior-Consequence (ABC) data + +ABC recording supplements behavior frequency counts with the immediate environmental context around each occurrence. A subset of task 2 session logs (~30%) include ABC entries for target behaviors, enabling function-hypothesis reasoning beyond what frequency data alone supports. + +- **Antecedent** — what was happening immediately before (≤ 10 s) the behavior occurred. +- **Behavior** — the target behavior as operationally defined (section 5). +- **Consequence** — what happened immediately after (≤ 10 s), including any staff response. + +**Citation.** Bijou, S. W., Peterson, R. F., & Ault, M. H. (1968). A method to integrate descriptive and experimental field studies at the level of data and empirical concepts. *JABA, 1*(2), 175–191. https://doi.org/10.1901/jaba.1968.1-175. See also CHH Ch. 27 ("Functional Behavior Assessment"). + +The remaining ~70% of logs include frequency-only behavior data, giving the model exposure to both enriched and minimal log formats. + +### 6.2 Program execution context + +Each acceleration and deceleration program is associated with one or more settings in which it runs: + +| Context tag | Setting | +|---|---| +| D | Day (school or day habilitation) | +| R | Residential (home, group home, dormitory) | +| Both | Runs across settings | + +**Dataset usage.** Each synthetic task 1 program and task 2 session log carries a context tag. Default sampling ~40% D, ~30% R, ~30% Both. No research source prescribes exact proportions; this is a pipeline hyperparameter. + +--- + +## 7. Session patterns (for Task 2) + +Behavioral session interpretation requires recognizing **patterns across sessions**. Fourteen clinically meaningful patterns are encoded. + +| # | Pattern | Signature | Concern level | Extends prepare_data.py? | +|---|---|---|---|---| +| 1 | **Mastery progression** | Ascending accuracy, mastery criterion within reach | None | Yes | +| 2 | **Regression** | Previously mastered skill shows declining accuracy | High | Yes | +| 3 | **Plateau** | Flat accuracy below mastery for ≥ 5 sessions | Moderate | Yes | +| 4 | **Frustration pattern** | Declining accuracy + escape-function behavioral indicators | High | Yes | +| 5 | **Variable performance** | High session-to-session SD, no clear trend | Moderate | Yes | +| 6 | **Prompt dependency** | High prompted accuracy, low independent accuracy, prolonged | Moderate | Yes | +| 7 | **Rapid acquisition** | Accelerated mastery beyond expected timeline | None (good) | Yes | +| 8 | **Generalization failure** | Mastered in training, low accuracy in novel settings/stimuli | Moderate | Yes | +| 9 | **Extinction burst** | Temporary spike in problem behavior during reduction procedure | Expected | Yes | +| 10 | **Skill loss after break** | Performance drop after absence, recovering | Moderate | Yes | +| 11 | **Motivating operation shift** | Responding drops when MO changes (e.g., satiation of reinforcer); recovers when MO restored | Moderate | No — new in v1, grounded in Michael 1993 / CHH Ch.16 | +| 12 | **Setting event trigger** | Accuracy or behavior changes correlated with an external setting event (illness, sleep disruption, schedule change) | Moderate | No — new in v1, grounded in Smith & Iwata 1997 / Bijou & Baer 1961 | + +**Citations.** +- Patterns 1–10 draw from standard single-case-design interpretation literature (CHH Ch. 6, 7). Specific ones: + - Regression / skill loss: CHH Ch. 26 (maintenance). + - Extinction burst: CHH Ch. 24. + - Prompt dependency: Time-delay literature (Touchette, Touchette 1971; CHH Ch. 21 prompting). + - Generalization failure: Stokes & Baer 1977 "An implicit technology of generalization" *JABA, 10*(2); CHH Ch. 26. +- Patterns 11–12 are TRACE additions with specific literature grounding: + - MO shift: CHH Ch. 16 ("Motivating Operations"); Michael, J. (1993). *Journal of the Experimental Analysis of Behavior, 59*(3), 533–552. + - Setting event: Bijou & Baer 1961; Smith, R. G. & Iwata, B. A. (1997). *JABA, 30*(2), 343–375. + +**Honest framing.** This 12-pattern taxonomy is our *operationalization* for the dataset — no single paper proposes these as a canonical pattern set. Each underlying concept is well-grounded (citations above); the clustering into 12 discrete labels is a design choice that makes session interpretation a tractable classification task. + +**Dataset usage.** Each task 2 example is generated from one of these 12 patterns as its hidden label. The interpretation output includes the pattern classification as a structured field (section 12). + +--- + +## 8. Prompt hierarchies + +Six prompt-fading strategies used in DTT and related methods. + +| # | Strategy | Sequence | When to use | +|---|---|---|---| +| 1 | **Most-to-Least (errorless)** | Full physical -> partial physical -> gestural -> positional -> independent | Acquisition, early learner, safety-critical skills | +| 2 | **Least-to-Most** | Independent -> gestural -> positional -> partial physical -> full physical | Learner already has partial repertoire; promotes independence | +| 3 | **Graduated Guidance** | Hand-over-hand with fading pressure -> shadow -> independent | Motor skills, self-care chaining | +| 4 | **Time Delay** (progressive) | 0 s delay -> 2 s -> 4 s -> 6 s -> learner responds independently | Prompt-dependency prevention; ideal for fading | +| 5 | **Stimulus Fading** | Exaggerated stimulus -> gradually reduce salience -> natural stimulus | Receptive discrimination, early reading | +| 6 | **Stimulus Shaping** | Modified stimulus -> gradually reshape to target -> natural stimulus | Complex visual discriminations | + +**Citations.** CHH Ch. 21 ("Imitation, Shaping, and Prompting") and Ch. 17 ("Stimulus Control"). Time delay: Touchette & Howard 1984 *JABA*. + +**Dataset usage.** One of the six sampled per task 1 DTT / Task-Analysis example. Output describes the sequence and current prompt level based on mastery state (section 3). + +--- + +## 9. Reinforcement schedules + +Seven reinforcement arrangements — intentional kept to the commonly-used set for cleanliness. + +| # | Schedule | Description | Typical use | +|---|---|---|---| +| 1 | **CRF** (continuous reinforcement) | Every correct response reinforced | Acquisition phase | +| 2 | **FR-2** (fixed ratio 2) | Every 2nd correct response reinforced | Thinning as accuracy stabilizes | +| 3 | **VR-3** (variable ratio 3) | Average of every 3rd correct response | Maintenance, resistance to extinction | +| 4 | **DRO** (differential reinforcement of other behavior) | Reinforce any behavior *other than* the problem behavior during an interval | Behavior reduction | +| 5 | **DRA** (differential reinforcement of alternative) | Reinforce a specific alternative behavior, extinguish problem behavior | FCT pairing | +| 6 | **DRI** (differential reinforcement of incompatible) | Reinforce a behavior physically incompatible with the problem behavior | Stereotypy reduction | +| 7 | **Token economy** | Tokens delivered per correct response, exchanged later for a back-up reinforcer | Group / classroom settings | + +**Citations.** CHH Ch. 13 ("Schedules of Reinforcement"), Ch. 22 ("Differential Reinforcement"). + +**Dataset usage.** One schedule sampled per task 1 example. Task 2 recommendations may suggest schedule changes (e.g., "thin from CRF to FR-2 as accuracy exceeds 85%"). + +--- + +## 10. Error correction procedures + +Five error-correction procedures used in DTT and related formats. + +| # | Procedure | Steps | +|---|---|---| +| 1 | **Transfer trial** | Represent SD -> prompt at effective level -> reinforce prompted response -> distractor trial -> re-present SD independently | +| 2 | **4-step** | Model -> lead (do together) -> test (independent) -> distractor -> retest | +| 3 | **Backstep** | Return to previous prompt level that produced success -> successful response -> re-attempt at target prompt level | +| 4 | **Simple correction** | "No, watch me" -> model -> re-present SD -> differential reinforcement | +| 5 | **Error-free** | Prevent errors by using high prompts from start; fade gradually (paired with most-to-least) | + +**Citations.** CHH Ch. 21. Specific: Heward, W. L. (1994) error-correction comparison. + +**Dataset usage.** One procedure sampled per task 1 DTT example. + +--- + +## 11. Mastery criteria options + BIP components + +### 11.1 Mastery criteria + +Seven mastery criteria conventions common in ABA practice. + +1. 80% accuracy across 2 consecutive sessions with ≥ 10 trials each. +2. 90% accuracy across 3 consecutive sessions. +3. 80% accuracy with 2 different therapists and 2 different settings (generalization-inclusive). +4. 80% accuracy at independent prompt level across 3 sessions. +5. 90% first-trial-correct across 5 consecutive sessions. +6. Fluency criterion: correct responses within 3-second latency at 90% accuracy. +7. Demonstrated use in natural environment × 3 independent instances. + +**Citation.** CHH Ch. 28 ("Developing Behavior-Change Programs"). + +### 11.2 BIP components (for task 2 recommendations) + +When a target behavior is present in a session log, the task 2 recommendation field is structured along four BIP dimensions. + +| Component | Purpose | Example content | +|---|---|---| +| **Antecedent strategies** | Prevent the behavior before it occurs by modifying the environment | Shorten tasks; offer choice; pre-teach coping; non-contingent attention; reduce demand difficulty; provide warnings before transitions | +| **Replacement behavior (teach)** | A skill that produces the same function more appropriately; pair with FCT | Teach "break please" for escape-function tantrums; teach "look at me" for attention-function disruption | +| **Consequence strategies** | Staff response after the behavior | Do not reinforce (extinction); redirection; blocking; DRA/DRI; differential reinforcement of replacement behavior | +| **Crisis plan** | Safety procedure if behavior escalates | Staff call-tree; environmental safety (clear hazards); agreed restraint/escape procedures per facility policy; post-incident debrief | + +**Citations.** +- Antecedent strategies: CHH Ch. 25 ("Antecedent Interventions"); Smith, R. G. & Iwata, B. A. (1997). *JABA, 30*(2), 343–375. +- Replacement behaviors: Carr & Durand 1985 (FCT). +- Consequence strategies: CHH Ch. 22 (differential reinforcement), Ch. 24 (extinction). +- Crisis plan: no canonical JABA paper — cite BACB 5th ed. Task List G-15 / H-8 and the BACB *Ethics Code for Behavior Analysts* (2020). Flagged in section 15. + +--- + +## 12. Escalation level (structured output for Task 2) + +Every task 2 interpretation emits an **escalation level** as a classification label. This is the safety-critical head. + +| Level | Meaning | Trigger examples | +|---|---|---| +| **1. Continue monitoring** | Current programming is effective; no changes | Mastery progression; rapid acquisition | +| **2. Adjust next session** | Implement a specific programming change at next session | Plateau -> change prompt strategy; prompt dependency -> introduce time delay | +| **3. Supervisor review** | Elevate to supervising BCBA within 24–48 hr | Regression; new emergence of problem behavior; unclear function | +| **4. Safety — immediate** | Stop program; contact supervisor and/or family immediately; consider crisis-plan activation | SIB escalation; aggression resulting in injury; high-rate elopement; extinction burst exceeding safety threshold | + +**Clinical principle.** Escalation level is **ordinal** — mistaking level 4 for level 1 is catastrophically worse than mistaking level 2 for level 1. Evaluation uses quadratic-weighted κ (section 13) to penalize distant mistakes accordingly. + +**Dataset usage.** Every task 2 example has a gold-standard escalation level derived deterministically from the pattern (section 7), the target behaviors (section 5), and severity markers. + +**On distribution.** No research source prescribes a specific class balance for escalation labels. The pipeline config (not this taxonomy) sets a default that keeps the minority class (level 4 "safety-immediate") present for training representation while reflecting that most sessions do not require immediate safety action. The exact ratio is a *pipeline hyperparameter*, not a clinical truth; see `configs/data-generation.yaml` once implemented. + +**Framing as our contribution.** No canonical source specifies this 4-level escalation ordinal. It is TRACE's operationalization, designed to (a) make safety a first-class output and (b) support quadratic-weighted κ evaluation. We cite it as a project design choice, not a clinical taxonomy. + +--- + +## 13. Confidence level (structured output for Task 2) + +Three-level confidence expression acknowledging uncertainty. + +| Level | Meaning | Example justifications | +|---|---|---| +| **High** | Pattern and recommendation well-supported by the data in this log | Multiple sessions of clear data; IOA present and ≥ 80%; single clear pattern | +| **Moderate** | Data supports the hypothesis but alternatives cannot be fully ruled out | No IOA; few sessions; multiple partially-matching patterns | +| **Low** | Insufficient data to rule in a specific pattern; recommend data collection before programming changes | < 3 sessions; no IOA; high variability; contradictory signals | + +**Citations.** Calibrated uncertainty expression follows general clinical-LM principles (Med-PaLM 2, Singhal et al. 2025). No single ABA-specific source; flagged as project innovation. + +**Dataset usage.** Confidence is sampled based on objective data-quality features of the log (number of sessions, IOA presence, variance). Evaluation computes calibration metrics (ECE, Brier — section 13 of `literature-foundation.md`). + +--- + +## 14. Coverage matrix (dimension interactions) + +Not every combination is valid. The pipeline enforces these constraints: + +| Teaching method | Compatible skill domains | Compatible learner profiles | +|---|---|---| +| DTT | All discrete VB-MAPP domains (Mand, Tact, Listener Responding, Echoic, Motor Imitation, VP/MTS, LRFFC, Intraverbal, Math, Reading early) | Early, School-Age | +| NET | Mand, Tact, Social Behavior, Spontaneous Vocal Behavior, Independent Play | All | +| Task Analysis | AFLS modules, Writing (multi-step), complex Math, domestic/vocational | School-Age, Adolescent, Adult | +| FCT | Triggered by target behavior in session log, not by skill domain | All | +| BST | N/A — staff-facing | Training-program variant | +| PRT | Mand, Tact, Social Behavior, Spontaneous Vocal Behavior | Early, School-Age | + +| Target behavior | Plausible functions | Typical severity / escalation | +|---|---|---| +| SIB | All four; most often automatic or escape | L3 – L4 | +| Aggression | Most often escape or tangible | L3 – L4 | +| Elopement | Escape, tangible, or automatic | L2 – L4 (depends on environmental risk) | +| Tantrum | Most often escape or tangible | L2 – L3 | +| Non-compliance | Primarily escape | L1 – L2 | +| Stereotypy | Primarily automatic | L1 – L2 (unless interfering with learning) | +| Pica | Automatic (oral) | L3 – L4 (safety) | + +**Dataset usage.** Generators sample valid combinations only. Invalid combinations (e.g., DTT for a multi-step handwashing skill) are not produced, or are produced deliberately as negative-training examples (rare; labeled). + +--- + +## 15. Gaps (explicitly acknowledged) + +The following areas in the taxonomy have weaker citation grounding. We cite what exists, note the limitation in the dataset card, and flag them as future-work extensions. + +1. **Crisis plan components** — no canonical JABA paper. Cited: BACB 5th ed. Task List G-15 / H-8 and the 2020 Ethics Code. Acceptable because BIP crisis planning is a BACB-regulated practice area. +2. **Patterns 11–12** (MO shift, setting event) — session-level operationalization is TRACE's contribution; underlying concepts grounded in Michael 1993, Smith & Iwata 1997, Bijou & Baer 1961. +3. **Disrobing, grabbing/snatching, cascading drift, comorbid skill-behavior** — considered for v1 but excluded because no canonical operational definition exists. We do not invent what the literature has not defined. +4. **Essential for Living / PEAK / ABLLS-R** — not encoded in v1. VB-MAPP + AFLS provide sufficient coverage for baseline dataset. +5. **Severity index within target behaviors** — no encoded numeric severity scale. v1 treats severity as implicit in the behavior type + escalation level. +6. **Multi-function (mixed-function) behaviors** — v1 assigns a single function hypothesis; real clinical reality includes mixed functions. Flagged as a known limitation. +7. **Cultural / linguistic variation** — v1 is English-only, US/North-American ABA convention. Flagged in dataset card limitations. +8. **Escalation ordinal (4 levels) and confidence ordinal (3 levels)** — TRACE design choices, not citation-grounded taxonomies. Framed as project contributions; their clinical validity will be evaluated empirically (quadratic-weighted κ vs BCBA raters). +9. **Class distributions** (escalation levels, learner profiles, pattern frequencies) — no research source prescribes exact proportions. These are pipeline hyperparameters tuned for training balance, not clinical truths. + +--- + +## 16. Changelog + +- **v1.0 (2026-04-22)** — initial taxonomy, drawn from VB-MAPP (Sundberg 2008), AFLS (Partington & Mueller 2012), Cooper/Heron/Heward (2020), Iwata (1982/1994), Hanley/Iwata/McCord (2003), Carr & Durand (1985), and supporting JABA operational-definition literature. + +--- + +## 17. How this taxonomy is consumed by the pipeline + +- `src/prepare_data.py` imports taxonomy as Python constants generated from this document. +- Each training example is produced by (a) sampling a valid combination of taxonomy values, (b) running the relevant generation template, (c) applying quality + safety filters. +- A provenance record is written per example: taxonomy slot values, template ID, teacher model, seed, filter thresholds, timestamp. +- If this taxonomy changes, the pipeline version and dataset version both bump; old data is preserved as a previous version on HuggingFace. + +See `schema-v1.md` (next) for the precise JSON structure of training examples. diff --git a/pyproject.toml b/pyproject.toml new file mode 100644 index 0000000000000000000000000000000000000000..25c1d301d52a512b90ebb0cfeba22496f22ccab7 --- /dev/null +++ b/pyproject.toml @@ -0,0 +1,54 @@ +[project] +name = "trace" +version = "1.0.0" +description = "TRACE — Taxonomy-Referenced ABA Clinical Examples. Synthetic dataset + generator for Applied Behavior Analysis teaching-program generation and session interpretation." +readme = "README.md" +requires-python = ">=3.10" +license = { text = "MIT" } +authors = [ + { name = "Festus Kahunla" }, +] +maintainers = [ + { name = "Pombo Labs" }, +] +keywords = [ + "applied-behavior-analysis", + "aba", + "autism", + "clinical-nlp", + "synthetic-data", + "instruction-tuning", + "taxonomy", +] +classifiers = [ + "Development Status :: 4 - Beta", + "Intended Audience :: Science/Research", + "License :: OSI Approved :: MIT License", + "Programming Language :: Python :: 3", + "Programming Language :: Python :: 3.10", + "Programming Language :: Python :: 3.11", + "Programming Language :: Python :: 3.12", + "Topic :: Scientific/Engineering :: Artificial Intelligence", +] + +dependencies = [ + "pyyaml>=6.0", +] + +[project.optional-dependencies] +dataset-loading = [ + "datasets>=2.18.0", +] + +[project.urls] +Homepage = "https://github.com/Pombo-Labs/TRACE" +Repository = "https://github.com/Pombo-Labs/TRACE" +Documentation = "https://github.com/Pombo-Labs/TRACE#readme" +Issues = "https://github.com/Pombo-Labs/TRACE/issues" + +[build-system] +requires = ["hatchling"] +build-backend = "hatchling.build" + +[tool.hatch.build.targets.wheel] +packages = ["src"] diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..13d742acb2c98f3e433febc6d074068f4e71c277 --- /dev/null +++ b/requirements.txt @@ -0,0 +1,9 @@ +# Runtime dependencies for the TRACE generator + dataset loading. +# +# Core (required to regenerate the corpus from configs): +pyyaml>=6.0 + +# Optional — only needed if you load TRACE via the Hugging Face datasets +# library (`load_dataset("PomboLabs/TRACE")`). Skip if you only consume the +# raw JSONL files under data/splits/. +datasets>=2.18.0 diff --git a/src/__init__.py b/src/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/src/compile_curation.py b/src/compile_curation.py new file mode 100644 index 0000000000000000000000000000000000000000..a4f98bb3f594ad790734a36cd9c7aa51114ff4f4 --- /dev/null +++ b/src/compile_curation.py @@ -0,0 +1,127 @@ +"""Compile test.jsonl + sanity.jsonl from curation_pool.jsonl. + +No hand-curation: the whole curation pool is promoted as the +evaluation corpus. A small stratified slice is carved out for +sanity.jsonl (training smoke-test); the remainder becomes test.jsonl. + +Stratification is by category (method for teaching programs, +pattern_class for session interpretation) so both splits reflect the +pool's distribution. + +Reads: + data/splits/curation_pool.jsonl + +Writes: + data/splits/test.jsonl + data/splits/sanity.jsonl + +Usage: + uv run python src/compile_curation.py + uv run python src/compile_curation.py --sanity-n 20 --seed 42 +""" + +import argparse +import json +import random +import sys +from collections import Counter, defaultdict +from pathlib import Path + +REPO_ROOT = Path(__file__).resolve().parent.parent +POOL_PATH = REPO_ROOT / "data" / "splits" / "curation_pool.jsonl" +TEST_JSONL = REPO_ROOT / "data" / "splits" / "test.jsonl" +SANITY_JSONL = REPO_ROOT / "data" / "splits" / "sanity.jsonl" + + +def category_of(example: dict) -> str: + gl = example["meta"]["gold_labels"] + if example["meta"]["task_type"] == "teaching_program": + return gl.get("method", "?") + return gl.get("pattern_class", "?") + + +def stratified_sanity_sample( + pool: list[dict], sanity_n: int, rng: random.Random +) -> tuple[list[dict], list[dict]]: + """Carve sanity_n stratified examples out of the pool; return (test, sanity).""" + by_cat: dict[str, list[dict]] = defaultdict(list) + for ex in pool: + by_cat[category_of(ex)].append(ex) + + # Largest-remainder method: allocate sanity slots proportional to category size. + total = len(pool) + raw_quotas = {cat: sanity_n * len(exs) / total for cat, exs in by_cat.items()} + floor_quotas = {cat: int(q) for cat, q in raw_quotas.items()} + remainder = sanity_n - sum(floor_quotas.values()) + # Distribute remaining slots by largest fractional part. + fractional = sorted( + raw_quotas.items(), key=lambda kv: kv[1] - int(kv[1]), reverse=True + ) + for cat, _ in fractional[:remainder]: + floor_quotas[cat] += 1 + + sanity: list[dict] = [] + for cat, quota in floor_quotas.items(): + if quota <= 0: + continue + picks = rng.sample(by_cat[cat], min(quota, len(by_cat[cat]))) + sanity.extend(picks) + + sanity_ids = {ex["meta"]["example_id"] for ex in sanity} + test = [ex for ex in pool if ex["meta"]["example_id"] not in sanity_ids] + return test, sanity + + +def tag_split(example: dict, split_name: str) -> dict: + """Deep-copy and stamp the curation split label into meta.""" + out = json.loads(json.dumps(example)) + out["meta"]["curation"] = {"target_split": split_name} + return out + + +def main() -> int: + parser = argparse.ArgumentParser() + parser.add_argument("--sanity-n", type=int, default=20) + parser.add_argument("--seed", type=int, default=42) + args = parser.parse_args() + + if not POOL_PATH.exists(): + print(f"Curation pool missing: {POOL_PATH}", file=sys.stderr) + print("Run `uv run python src/split_data.py` first.", file=sys.stderr) + return 1 + + pool: list[dict] = [] + with open(POOL_PATH) as f: + for line in f: + pool.append(json.loads(line)) + + if args.sanity_n >= len(pool): + print( + f"sanity_n ({args.sanity_n}) must be smaller than pool size ({len(pool)}).", + file=sys.stderr, + ) + return 1 + + rng = random.Random(args.seed) + test, sanity = stratified_sanity_sample(pool, args.sanity_n, rng) + + TEST_JSONL.parent.mkdir(parents=True, exist_ok=True) + with open(TEST_JSONL, "w") as f: + for ex in test: + f.write(json.dumps(tag_split(ex, "test")) + "\n") + with open(SANITY_JSONL, "w") as f: + for ex in sanity: + f.write(json.dumps(tag_split(ex, "sanity")) + "\n") + + def fmt_dist(rows: list[dict]) -> str: + c = Counter(category_of(r) for r in rows) + return ", ".join(f"{k}={v}" for k, v in sorted(c.items())) + + print(f"Wrote:") + print(f" {TEST_JSONL} {len(test)} examples ({fmt_dist(test)})") + print(f" {SANITY_JSONL} {len(sanity)} examples ({fmt_dist(sanity)})") + return 0 + + +if __name__ == "__main__": + raise SystemExit(main()) diff --git a/src/generate.py b/src/generate.py new file mode 100644 index 0000000000000000000000000000000000000000..bea90f5a4127aaf5f03de0fe254286e8d888dea9 --- /dev/null +++ b/src/generate.py @@ -0,0 +1,124 @@ +"""TRACE synthetic data generator — orchestrator. + +Routes to per-area generators. Each area is self-contained under configs// +with its own taxonomy, template, and compatibility rules. + +Usage: + # Generate one area (count from generation.yaml) + uv run python src/generate.py --area dtt + + # Generate one area with explicit count + uv run python src/generate.py --area dtt --n 50 + + # Generate all enabled areas + uv run python src/generate.py --all + + # Write to a specific file (single-area mode only) + uv run python src/generate.py --area dtt --n 10 --out data/samples/dtt-10.jsonl +""" + +import argparse +import hashlib +import json +import random +import sys +from pathlib import Path + +# Path setup so we can run as a script +REPO_ROOT = Path(__file__).resolve().parent.parent +sys.path.insert(0, str(REPO_ROOT)) + +from src.generators.aba_dtt import DTTGenerator +from src.generators.aba_net import NETGenerator +from src.generators.aba_session_interp import SessionInterpretationGenerator +from src.generators.aba_task_analysis import TaskAnalysisGenerator +from src.generators.base import load_yaml + +CONFIG_DIR = REPO_ROOT / "configs" + +# Area registry — add new generators here as they come online. +GENERATORS = { + "dtt": DTTGenerator, + "net": NETGenerator, + "task_analysis": TaskAnalysisGenerator, + "session_interpretation": SessionInterpretationGenerator, + # "fct": FCTGenerator, + # "bst": BSTGenerator, + # "prt": PRTGenerator, +} + + +def deterministic_area_offset(area: str) -> int: + """Stable per-area seed offset. + + Using Python's built-in ``hash()`` here would be non-deterministic across + processes (PYTHONHASHSEED randomization), which would make the dataset + irreproducible run-to-run. SHA-256 is stable across processes and Python + versions. + """ + return int(hashlib.sha256(area.encode()).hexdigest(), 16) % 10000 + + +def generate_area(area: str, n: int, seed: int, out_stream) -> int: + """Generate `n` examples for `area`; write to `out_stream` as JSONL. Returns count written.""" + if area not in GENERATORS: + raise ValueError(f"Generator for area '{area}' not registered. Available: {list(GENERATORS)}") + GenCls = GENERATORS[area] + generator = GenCls(CONFIG_DIR) + rng = random.Random(seed) + for _ in range(n): + example = generator.render_example(rng) + out_stream.write(json.dumps(example) + "\n") + return n + + +def main(): + parser = argparse.ArgumentParser(description="TRACE per-area generator orchestrator.") + parser.add_argument("--area", default=None, help="Generate only this area.") + parser.add_argument("--all", action="store_true", help="Generate all enabled areas from generation.yaml.") + parser.add_argument("--n", type=int, default=None, help="Override example count.") + parser.add_argument("--seed", type=int, default=None, help="Override global seed.") + parser.add_argument("--out", type=str, default=None, help="Output JSONL file (single-area mode).") + args = parser.parse_args() + + gen_config = load_yaml(CONFIG_DIR / "generation.yaml") + seed = args.seed if args.seed is not None else gen_config["seed"] + + if not args.area and not args.all: + parser.error("Must specify --area or --all.") + + if args.all: + # Multi-area: write one file per area under output_dir + output_dir = REPO_ROOT / gen_config["output_dir"] + output_dir.mkdir(parents=True, exist_ok=True) + total = 0 + for area, area_cfg in gen_config["areas"].items(): + if not area_cfg.get("enabled"): + print(f" - {area}: disabled, skipping") + continue + if area not in GENERATORS: + print(f" - {area}: generator not yet built, skipping") + continue + n = area_cfg["n"] + out_path = output_dir / f"{area}.jsonl" + with open(out_path, "w") as f: + generate_area(area, n, seed + deterministic_area_offset(area), f) + print(f" - {area}: {n} examples -> {out_path}") + total += n + print(f"\nTotal generated: {total} examples.") + else: + # Single-area mode + area = args.area + n = args.n if args.n is not None else gen_config["areas"][area]["n"] + out_stream = open(args.out, "w") if args.out else sys.stdout + try: + generate_area(area, n, seed, out_stream) + finally: + if args.out: + out_stream.close() + if args.out: + print(f"Wrote {n} examples to {args.out}") + + +if __name__ == "__main__": + main() diff --git a/src/generators/__init__.py b/src/generators/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3624f17878b5225d7cac7b686494f2039d49f300 --- /dev/null +++ b/src/generators/__init__.py @@ -0,0 +1 @@ +"""Generator modules for TRACE synthetic data pipeline.""" diff --git a/src/generators/aba_dtt.py b/src/generators/aba_dtt.py new file mode 100644 index 0000000000000000000000000000000000000000..6f3376121512f34062360a212e9cbed8de94f370 --- /dev/null +++ b/src/generators/aba_dtt.py @@ -0,0 +1,293 @@ +"""DTT area generator. + +Loads configs/dtt/ + configs/shared/ and produces JSONL examples with +clinically-consistent combinations sampled from the DTT taxonomy. +""" + +import random +from pathlib import Path + +from .base import ( + article, + load_area, + load_shared, + render_prompt_sequence, + sample_stimuli, + strip_trailing_period, + make_example_envelope, +) + +# Area identifier + task type +AREA = "dtt" +TASK_TYPE = "teaching_program" + + +# SD (Discriminative Stimulus) generation — clinically nuanced + + +def generate_sd(skill_target: str, domain_id: str, rng: random.Random) -> dict: + """Generate primary SD, variations, and presentation description. + + Sub-domain aware: Tact-of-colors gets "What color?", not "What is this?". + """ + skill = skill_target.lower() + + if domain_id == "tact": + if "color" in skill: + return _sd("What color?", '"What color is this?"; "Tell me the color"', + "Present a single-color object and ask while pointing to it.") + if "shape" in skill: + return _sd("What shape?", '"What shape is this?"; "Tell me the shape"', + "Present a clearly-shaped object or shape card and ask while pointing to it.") + if "action" in skill: + return _sd("What is he doing?", '"What action?"; "What is she doing?"; "What is happening here?"', + "Present an action picture, video clip, or live demonstration.") + if "emotion" in skill or "feeling" in skill: + return _sd("How does he feel?", '"What emotion?"; "How is she feeling?"', + "Present a photograph or emoji card depicting the target emotion.") + if "body part" in skill: + return _sd("What is this?", '"Tell me"; "Name this body part"', + "Point to the target body part (on self, learner, or doll) and ask.") + if "animal" in skill: + return _sd("What animal is this?", '"What is this?"; "Name this animal"', + "Present an animal picture card and ask while pointing.") + if "preposition" in skill: + return _sd("Where is it?", '"Where is the {object}?"; "Tell me where"', + "Arrange two objects in a target spatial relation and ask about the position.") + if "adjective" in skill: + return _sd("What is it like?", '"Describe it"; "Is it big or little?"', + "Present two contrasting items and prompt a comparative description.") + if "past" in skill or "event" in skill: + return _sd("What happened?", '"What did she do?"; "Tell me what you did"', + "Reference a recently completed action or event familiar to the learner.") + if "categor" in skill: + return _sd("What category?", '"What group does this belong to?"; "What kind of thing?"', + "Present a clearly categorizable item and prompt the category label.") + return _sd("What is this?", '"Tell me what you see"; "What\'s that?"; "Name this"', + "Present the target stimulus within visual field and ask while pointing to it.") + + if domain_id == "vp_mts": + return _sd("Match", '"Find the same"; "Put with same"', + "Present the sample stimulus and an array of comparison stimuli simultaneously.") + + if domain_id in {"listener_responding", "lrffc"}: + return _sd("Touch [target item]", + '"Point to [target item]"; "Give me [target item]"; "Show me [target item]"', + "Present an array of items on the table and deliver a clear verbal SD.") + + if domain_id == "reading": + if "sight word" in skill or "word" in skill: + return _sd("What word?", '"Read this word"; "What does it say?"', + "Present the sight-word card and deliver the SD.") + if "letter" in skill: + return _sd("What letter is this?", '"Read this letter"; "What does this say?"', + "Present the letter flashcard.") + if "sentence" in skill or "passage" in skill: + return _sd("Read this.", '"Read it aloud"; "What does this sentence say?"', + "Present the written sentence or passage at an appropriate reading level.") + return _sd("What letter is this?", '"Read this letter"; "What does this say?"', + "Present the letter or word on a flashcard.") + + if domain_id == "writing": + return _sd("Write the target.", '"Trace this"; "Copy this"; "Write {word}"', + "Provide pencil and paper; deliver a clear verbal instruction naming the target letter or word.") + + if domain_id == "math": + if "numeral" in skill or "identif" in skill: + return _sd("What number is this?", '"Which number?"; "Tell me the number"', + "Present the numeral card.") + if "count" in skill: + return _sd("How many?", '"Count these"; "How many are there?"', + "Present the set of objects to count.") + if "add" in skill or "subtract" in skill: + return _sd("Solve this.", '"What is the answer?"; "How much?"', + "Present a written or spoken arithmetic problem at target level.") + return _sd("How many?", '"Count these"; "What number is this?"', + "Present the numeral card or a set of objects to count.") + + return _sd("Show me", '"Do this"; "Your turn"', + "Present relevant stimuli and deliver a clear verbal SD.") + + +def _sd(primary: str, variations: str, presentation: str) -> dict: + return {"primary_sd": primary, "sd_variations": variations, "sd_presentation": presentation} + + +def current_prompt_guidance(mastery_state_id: str) -> str: + """Translate mastery state into current-prompt-level guidance.""" + if mastery_state_id in {"emerging", "developing"}: + return "begin at higher prompt levels (full-physical or partial-physical) with systematic fading." + if mastery_state_id == "approaching": + return "deliver gestural or positional prompts; fade toward independence." + if mastery_state_id in {"near", "mastered", "generalization"}: + return "use minimal prompts (positional or verbal hint) only when needed; expect independent responding." + return "prompt level calibrated to current performance." + + +# DTTGenerator class + + +class DTTGenerator: + """Generates teaching-program examples for the DTT area. + + Loads its self-contained taxonomy + template + compatibility rules from + configs/dtt/ and uses shared primitives from configs/shared/. + """ + + def __init__(self, config_dir: Path): + self.config_dir = Path(config_dir) + self.shared = load_shared(self.config_dir) + self.area = load_area(self.config_dir, AREA) + self.template = self.area["template"] + self.taxonomy = self.area["taxonomy"] + self.compat = self.area["compatibility"] + self.prompt_types = self.shared["prompt_types"]["prompt_types"] + + # sampling + + def sample_combination(self, rng: random.Random) -> dict: + """Sample a clinically-valid DTT combo.""" + # Sample level with weights from compatibility config + level_weights = self.compat["level_sampling_weights"] + levels = list(level_weights.keys()) + weights = [level_weights[k] for k in levels] + level_id = rng.choices(levels, weights=weights)[0] + + # Sample domain — filter to only those that define this level (some + # domains like Writing don't have L3 in the DTT taxonomy because + # long-form writing is Task Analysis territory). + all_domains = self.taxonomy["skill_domains"]["vbmapp"]["domains"] + domains_with_level = [d for d in all_domains if level_id in d] + if not domains_with_level: + # Fall back: pick from all domains and re-sample a level they support + domain = rng.choice(all_domains) + available_levels = [lv for lv in levels if lv in domain] + level_id = rng.choice(available_levels) + else: + domain = rng.choice(domains_with_level) + skill_target = rng.choice(domain[level_id]) + + # Learner profile filtered by level + profiles = self.shared["learner_profiles"]["profiles"] + allowed_profile_ids = set(self.compat["level_to_learner_profiles"][level_id]) + profile = rng.choice([p for p in profiles if p["id"] in allowed_profile_ids]) + + mastery_state = rng.choice(self.shared["mastery_states"]["states"]) + + # Prompt hierarchy then error-correction (filtered for errorless compat) + prompt_hierarchy = rng.choice(self.taxonomy["prompt_hierarchies"]) + all_ecs = self.taxonomy["error_corrections"] + required_hierarchy = self.compat.get("errorless_requires_hierarchy") + if prompt_hierarchy["id"] != required_hierarchy: + valid_ecs = [ec for ec in all_ecs if ec["id"] != "errorless"] + else: + valid_ecs = all_ecs + error_correction = rng.choice(valid_ecs) + + # Reinforcement schedule filtered by mastery state + all_schedules = self.taxonomy["reinforcement_schedules"] + allowed_schedule_ids = set(self.compat["mastery_to_reinforcement"][mastery_state["id"]]) + valid_schedules = [s for s in all_schedules if s["id"] in allowed_schedule_ids] + if not valid_schedules: + valid_schedules = [s for s in all_schedules if s["id"] == "crf"] + reinforcement_schedule = rng.choice(valid_schedules) + + mastery_criterion = rng.choice(self.taxonomy["mastery_criteria"]) + + return { + "skill_target": skill_target, + "domain_id": domain["id"], + "domain_name": domain["name"], + "level_id": level_id, + "learner_profile": profile, + "mastery_state": mastery_state, + "prompt_hierarchy": prompt_hierarchy, + "reinforcement_schedule": reinforcement_schedule, + "error_correction": error_correction, + "mastery_criterion": mastery_criterion, + } + + # slot computation + + def compute_slots(self, combo: dict, rng: random.Random) -> dict: + array_size_info = self.compat["array_size_by_level"][combo["level_id"]] + array_size_n = array_size_info["n"] + array_size_text = array_size_info["text"] + + sd = generate_sd(combo["skill_target"], combo["domain_id"], rng) + stim = sample_stimuli(combo["skill_target"], array_size_n, rng) + + if stim["distractors"]: + distractor_block = f"Distractor stimuli: {', '.join(stim['distractors'])}" + else: + distractor_block = "Distractor stimuli: N/A (no-array SD format)" + + mastery_state = combo["mastery_state"] + + return { + "skill_target": combo["skill_target"], + "curriculum_ref": f"{combo['domain_name']} {combo['level_id']}", + "learner_profile_name": combo["learner_profile"]["name"], + "learner_profile_article": article(combo["learner_profile"]["name"]), + "mastery_state_name": mastery_state["name"], + "mastery_state_short": mastery_state.get("short", mastery_state["name"].lower()), + "primary_sd": sd["primary_sd"], + "sd_variations": sd["sd_variations"], + "sd_presentation": sd["sd_presentation"], + "prompt_hierarchy_name": combo["prompt_hierarchy"]["name"], + "prompt_sequence": render_prompt_sequence( + combo["prompt_hierarchy"]["sequence"], self.prompt_types + ), + "current_prompt_guidance": current_prompt_guidance(mastery_state["id"]), + "array_size": array_size_text, + "target_stimuli": ", ".join(stim["targets"]), + "distractor_block": distractor_block, + "error_correction_steps": combo["error_correction"]["steps"].strip(), + "reinforcement_schedule_name": combo["reinforcement_schedule"]["name"], + "reinforcement_description": strip_trailing_period(combo["reinforcement_schedule"]["description"]), + "mastery_criterion_text": combo["mastery_criterion"]["text"], + "ioa_frequency": rng.choice([3, 4, 5]), + "n_generalization_therapists": rng.choice([2, 3]), + "n_generalization_settings": rng.choice([2, 3]), + } + + # rendering + + def render_example(self, rng: random.Random) -> dict: + combo = self.sample_combination(rng) + slots = self.compute_slots(combo, rng) + + user_variant = rng.choice(self.template["user_variants"]) + user_content = user_variant.format(**slots) + assistant_content = self.template["assistant_template"].format(**slots) + + gold_labels = { + "method": AREA, + "domain": f"VB-MAPP.{combo['domain_id']}", + "level": combo["level_id"], + "learner_profile": combo["learner_profile"]["id"], + "mastery_state": combo["mastery_state"]["id"], + } + provenance = { + "layer": 1, + "area": AREA, + "template_id": self.template["template_id"], + "taxonomy_cells": { + "skill_target": combo["skill_target"], + "prompt_hierarchy": combo["prompt_hierarchy"]["id"], + "reinforcement_schedule": combo["reinforcement_schedule"]["id"], + "error_correction": combo["error_correction"]["id"], + "mastery_criterion": combo["mastery_criterion"]["id"], + }, + "teacher_model": None, + "seed_tag": str(rng.getstate()[1][0] % 100000), + } + + return make_example_envelope( + system_content=self.template["system_prompt"], + user_content=user_content, + assistant_content=assistant_content, + task_type=TASK_TYPE, + gold_labels=gold_labels, + provenance=provenance, + ) diff --git a/src/generators/aba_net.py b/src/generators/aba_net.py new file mode 100644 index 0000000000000000000000000000000000000000..cbcfc3d88df4e51ade52f950adec69033dd387de --- /dev/null +++ b/src/generators/aba_net.py @@ -0,0 +1,216 @@ +"""NET area generator. + +Loads configs/net/ + configs/shared/ and produces naturalistic teaching +program examples. Mand / Social / Spontaneous Vocal / some Intraverbal skills. +""" + +import random +from pathlib import Path + +from .base import ( + load_area, + load_shared, + make_example_envelope, +) + +AREA = "net" +TASK_TYPE = "teaching_program" + + +def pick_mo_category(skill_target: str, domain_id: str, compat: dict) -> str: + """Match skill text against domain + keyword rules to pick an MO category. + + Domain is checked first to avoid false matches from shared keywords + (e.g., 'preferred' in both mand skills and spontaneous-vocal skills). + """ + s = skill_target.lower() + + if domain_id == "mand": + if "bathroom" in s or "potty" in s or "toilet" in s or "restroom" in s: + return "mand_bathroom" + if "break" in s: + return "mand_break" + if "all done" in s or "completion" in s or "finished" in s: + return "mand_completion" + if "missing" in s: + return "mand_missing" + if "help" in s: + return "mand_help" + if '"what"' in s or "what" in s.split(): + return "mand_info_what" + if '"where"' in s or "where" in s.split(): + return "mand_info_where" + if "attention" in s or "peers" in s: + return "mand_attention" + if "action" in s: + return "mand_action" + return "mand_item" + + if domain_id == "social": + if "initiate" in s or "initiates" in s: + return "social_initiation" + if "turn" in s: + return "social_turn_taking" + if "reciprocal" in s or "conversation" in s: + return "intraverbal_multi_turn" + return "social_initiation" + + if domain_id == "spontaneous_vocal": + if "greet" in s: + return "spontaneous_greeting" + return "spontaneous_comment" + + if domain_id == "intraverbal": + if "multi-turn" in s or "conversation" in s: + return "intraverbal_multi_turn" + if "routine" in s: + return "intraverbal_routine" + return "intraverbal_multi_turn" + + return compat["default_mo_category"] + + +def pick_mo_arrangement(mo_category: str, mo_arrangements: list, rng: random.Random) -> dict: + """Find an MO arrangement matching the category, falling back to random.""" + matches = [m for m in mo_arrangements if mo_category in m["applies_to"]] + if matches: + return rng.choice(matches) + return rng.choice(mo_arrangements) + + +def current_prompt_guidance(mastery_state_id: str) -> str: + """Guidance for NET prompt level based on mastery state.""" + if mastery_state_id in {"emerging", "developing"}: + return "begin with a model prompt delivered during the motivated moment; fade to expectant look across sessions." + if mastery_state_id == "approaching": + return "use time delay (0 s -> 2 s -> 4 s) before delivering any model prompt; reinforce independent responses differentially." + if mastery_state_id in {"near", "mastered", "generalization"}: + return "use expectant look only; avoid direct prompts. The MO itself should be sufficient to occasion the response." + return "prompt level calibrated to current performance." + + +class NETGenerator: + def __init__(self, config_dir: Path): + self.config_dir = Path(config_dir) + self.shared = load_shared(self.config_dir) + self.area = load_area(self.config_dir, AREA) + self.template = self.area["template"] + self.taxonomy = self.area["taxonomy"] + self.compat = self.area["compatibility"] + + def sample_combination(self, rng: random.Random) -> dict: + # Level with weights + level_weights = self.compat["level_sampling_weights"] + levels = list(level_weights.keys()) + weights = [level_weights[k] for k in levels] + level_id = rng.choices(levels, weights=weights)[0] + + # Sample domain with a level match + all_domains = self.taxonomy["skill_domains"]["vbmapp"]["domains"] + domains_with_level = [d for d in all_domains if level_id in d] + if not domains_with_level: + domain = rng.choice(all_domains) + available_levels = [lv for lv in levels if lv in domain] + level_id = rng.choice(available_levels) + else: + domain = rng.choice(domains_with_level) + skill_target = rng.choice(domain[level_id]) + + # Learner profile + profiles = self.shared["learner_profiles"]["profiles"] + allowed = set(self.compat["level_to_learner_profiles"][level_id]) + profile = rng.choice([p for p in profiles if p["id"] in allowed]) + + mastery_state = rng.choice(self.shared["mastery_states"]["states"]) + + mo_category = pick_mo_category(skill_target, domain["id"], self.compat) + mo_arrangement = pick_mo_arrangement(mo_category, self.taxonomy["mo_arrangements"], rng) + natural_context = rng.choice( + mo_arrangement.get("context_examples") or self.taxonomy["natural_contexts"] + ) + prompt_strategy = rng.choice(self.taxonomy["prompt_strategies"]) + mastery_criterion = rng.choice(self.taxonomy["mastery_criteria"]) + + return { + "skill_target": skill_target, + "domain_id": domain["id"], + "domain_name": domain["name"], + "level_id": level_id, + "learner_profile": profile, + "mastery_state": mastery_state, + "mo_category": mo_category, + "mo_arrangement": mo_arrangement, + "natural_context": natural_context, + "prompt_strategy": prompt_strategy, + "mastery_criterion": mastery_criterion, + } + + def compute_slots(self, combo: dict, rng: random.Random) -> dict: + mastery_state = combo["mastery_state"] + reinforcer_map = self.taxonomy["natural_reinforcer_examples"] + natural_reinforcer = reinforcer_map.get(combo["mo_category"], reinforcer_map["default"]) + + # Build a natural-opportunity description from the MO arrangement + natural context + opportunity_text = ( + f"Embed the teaching opportunity within the learner's {combo['natural_context']} routine. " + f"Deliver the opportunity when the MO is evident (learner oriented to the relevant stimulus, approaching the materials, or demonstrating interest)." + ) + + return { + "skill_target": combo["skill_target"], + "curriculum_ref": f"{combo['domain_name']} {combo['level_id']}", + "learner_profile_name": combo["learner_profile"]["name"], + "mastery_state_name": mastery_state["name"], + "mastery_state_short": mastery_state.get("short", mastery_state["name"].lower()), + "mo_arrangement_text": combo["mo_arrangement"]["text"], + "primary_natural_context": combo["natural_context"], + "natural_opportunity_text": opportunity_text, + "prompt_strategy_name": combo["prompt_strategy"]["name"], + "prompt_strategy_description": combo["prompt_strategy"]["description"], + "current_prompt_guidance": current_prompt_guidance(mastery_state["id"]), + "natural_reinforcer_text": f"Natural reinforcer: {natural_reinforcer}", + "n_exemplars": rng.choice([3, 4, 5]), + "n_settings": rng.choice([2, 3]), + "n_therapists": rng.choice([2, 3]), + "mastery_criterion_text": combo["mastery_criterion"]["text"], + "ioa_frequency": rng.choice([3, 4, 5]), + } + + def render_example(self, rng: random.Random) -> dict: + combo = self.sample_combination(rng) + slots = self.compute_slots(combo, rng) + + user_variant = rng.choice(self.template["user_variants"]) + user_content = user_variant.format(**slots) + assistant_content = self.template["assistant_template"].format(**slots) + + gold_labels = { + "method": AREA, + "domain": f"VB-MAPP.{combo['domain_id']}", + "level": combo["level_id"], + "learner_profile": combo["learner_profile"]["id"], + "mastery_state": combo["mastery_state"]["id"], + } + provenance = { + "layer": 1, + "area": AREA, + "template_id": self.template["template_id"], + "taxonomy_cells": { + "skill_target": combo["skill_target"], + "mo_category": combo["mo_category"], + "mo_arrangement": combo["mo_arrangement"]["id"], + "prompt_strategy": combo["prompt_strategy"]["id"], + "mastery_criterion": combo["mastery_criterion"]["id"], + }, + "teacher_model": None, + "seed_tag": str(rng.getstate()[1][0] % 100000), + } + + return make_example_envelope( + system_content=self.template["system_prompt"], + user_content=user_content, + assistant_content=assistant_content, + task_type=TASK_TYPE, + gold_labels=gold_labels, + provenance=provenance, + ) diff --git a/src/generators/aba_session_interp.py b/src/generators/aba_session_interp.py new file mode 100644 index 0000000000000000000000000000000000000000..3f9efbef4b8b2a1fb7509e846fe46b2be4495a70 --- /dev/null +++ b/src/generators/aba_session_interp.py @@ -0,0 +1,931 @@ +"""Session Interpretation generator (Task 2). + +Generates multi-session behavioral logs and matching structured interpretations. +The log is produced deterministically from a sampled hidden pattern so the +interpretation's gold labels are known by construction. +""" + +import random +from datetime import date, timedelta +from pathlib import Path + +from .base import ( + load_area, + load_shared, + load_yaml, + strip_trailing_period, + make_example_envelope, +) + +AREA = "session_interpretation" +TASK_TYPE = "session_interpretation" + + +# Trajectory math — generate per-session accuracy values following a pattern rule + + +def _clip(x: float, lo: float = 0.0, hi: float = 1.0) -> float: + return max(lo, min(hi, x)) + + +def generate_accuracy_trajectory(pattern_id: str, rule: dict, n_sessions: int, rng: random.Random) -> list: + """Return a list of accuracy values for n_sessions based on pattern rule.""" + acc = rule["accuracy"] + + if pattern_id == "generalization_failure": + alt = acc["alternating"] + tr_lo, tr_hi = alt["training_range"] + nov_lo, nov_hi = alt["novel_range"] + noise = alt["noise"] + vals = [] + training = rng.uniform(tr_lo, tr_hi) + novel = rng.uniform(nov_lo, nov_hi) + for i in range(n_sessions): + base = training if i % 2 == 0 else novel + vals.append(_clip(base + rng.uniform(-noise, noise))) + return vals + + if pattern_id == "skill_loss_after_break": + base = rng.uniform(*acc["base_range"]) + drop = rng.uniform(*acc["initial_drop"]) + rec = rng.uniform(*acc["recovery_per_session_range"]) + noise = acc["noise"] + vals = [_clip(base - drop + rng.uniform(-noise, noise))] + for _ in range(1, n_sessions): + vals.append(_clip(vals[-1] + rec + rng.uniform(-noise, noise), 0.1, base)) + return vals + + if pattern_id == "motivating_operation_shift": + base = rng.uniform(*acc["base_range"]) + dip_len = rng.randint(*acc["mo_dip_sessions"]) + dip_mag = rng.uniform(*acc["mo_dip_magnitude"]) + rec = rng.uniform(*acc["recovery_per_session_range"]) + noise = acc["noise"] + dip_start = max(2, n_sessions // 3) + vals = [] + for i in range(n_sessions): + in_dip = dip_start <= i < dip_start + dip_len + if in_dip: + v = base - dip_mag + rng.uniform(-noise, noise) + elif i >= dip_start + dip_len: + prev = vals[-1] + v = min(base, prev + rec + rng.uniform(-noise, noise)) + else: + v = base + rng.uniform(-noise, noise) + vals.append(_clip(v, 0.1, 0.98)) + return vals + + if pattern_id == "setting_event_trigger": + base = rng.uniform(*acc["base_range"]) + se_idx = max(1, int(acc["setting_event_session_fraction"] * n_sessions)) + mag = rng.uniform(*acc["setting_event_magnitude"]) + rec = rng.uniform(*acc["recovery_per_session_range"]) + noise = acc["noise"] + vals = [] + for i in range(n_sessions): + if i < se_idx: + v = base + rng.uniform(-noise, noise) + elif i == se_idx: + v = base - mag + rng.uniform(-noise, noise) + else: + prev = vals[-1] + v = min(base, prev + rec + rng.uniform(-noise, noise)) + vals.append(_clip(v, 0.1, 0.98)) + return vals + + if pattern_id == "extinction_burst": + base = rng.uniform(*acc["base_range"]) + delta = rng.uniform(*acc["delta_per_session_range"]) + noise = acc["noise"] + vals = [_clip(base + i * delta + rng.uniform(-noise, noise)) for i in range(n_sessions)] + return vals + + # Default: linear delta from base + base = rng.uniform(*acc["base_range"]) + delta = rng.uniform(*acc["delta_per_session_range"]) + noise = acc["noise"] + cap = acc.get("cap", 0.98) + floor_v = acc.get("floor", 0.05) + return [_clip(base + i * delta + rng.uniform(-noise, noise), floor_v, cap) for i in range(n_sessions)] + + +def generate_behavior_frequencies( + traj_name: str, library: dict, n_sessions: int, rule: dict, rng: random.Random +) -> list: + """Return a list of integer behavior frequencies per session.""" + cfg = library[traj_name] + if traj_name == "temporary_spike": + base_lo, base_hi = cfg["base_freq_range"] + mult_range = cfg["spike_multiplier_range"] + noise = cfg["noise"] + burst_start = max(2, int(rule.get("burst_start_fraction", 0.33) * n_sessions)) + burst_len = rule.get("burst_duration_sessions", 2) + base = rng.uniform(base_lo, base_hi) + mult = rng.uniform(*mult_range) + vals = [] + for i in range(n_sessions): + if burst_start <= i < burst_start + burst_len: + v = base * mult + rng.uniform(-noise, noise) + else: + v = base + rng.uniform(-noise, noise) + vals.append(max(0, round(v))) + return vals + if traj_name == "coincident_spike": + # Spike coincides with setting event index + base_lo, base_hi = cfg["base_freq_range"] + mult_range = cfg["spike_multiplier_range"] + noise = cfg["noise"] + se_idx = max(1, int(rule["accuracy"]["setting_event_session_fraction"] * n_sessions)) + base = rng.uniform(base_lo, base_hi) + mult = rng.uniform(*mult_range) + vals = [] + for i in range(n_sessions): + if i in {se_idx, se_idx + 1}: + v = base * mult + rng.uniform(-noise, noise) + else: + v = base + rng.uniform(-noise, noise) + vals.append(max(0, round(v))) + return vals + # default: base_freq_range + cumulative delta + noise + base_lo, base_hi = cfg["base_freq_range"] + delta = cfg.get("delta_per_session", 0.0) + noise = cfg.get("noise", 1.0) + base = rng.uniform(base_lo, base_hi) + return [max(0, round(base + i * delta + rng.uniform(-noise, noise))) for i in range(n_sessions)] + + +# Stimulus / profile sampling helpers + + +SKILL_POOL_BY_CATEGORY = { + "early_language": [ + ("Mand", "single-word mands for preferred items (e.g., \"cookie\", \"iPad\")", "NET"), + ("Mand", "mands for missing items needed to complete an activity", "NET"), + ("Mand", "2-word mand phrases (e.g., \"more cracker\", \"open please\")", "NET"), + ("Mand", "requesting a break ('break please' or AAC 'I need a break')", "NET"), + ("Mand", "indicating completion ('all done', 'finished')", "NET"), + ("Tact", "tacts common objects", "DTT"), + ("Tact", "tacts colors of objects", "DTT"), + ("Listener Responding", "follows 1-step motor instructions", "DTT"), + ("Listener Responding", "selects items by feature", "DTT"), + ("Echoic", "echoes 2-word phrases", "DTT variant"), + ("Motor Imitation", "imitates 3-step motor sequences", "DTT variant"), + ("Intraverbal", "answers WH-questions about familiar topics", "NET"), + ("Tolerating Denied Access", "wait calmly when access to preferred item is denied", "FCT"), + ], + "school_age": [ + ("Reading", "reads 10–20 sight words", "DTT"), + ("Reading", "sounds out CVC words", "DTT"), + ("Writing", "copies simple words from model", "DTT"), + ("Math", "identifies numerals 1–20", "DTT"), + ("Math", "counts objects 1–20", "DTT"), + ("Mand", "requesting preferred items using AAC device (e.g., \"I want cookie please\")", "NET"), + ("Mand", "single-word requests for preferred items", "NET"), + ("Mand", "requesting preferred items using a full sentence", "NET"), + ("Mand", "requesting a break during non-preferred tasks", "NET"), + ("Mand", "indicating 'all done' when an activity is complete", "NET"), + ("Tolerating Denied Access", "wait calmly when access to preferred item is denied", "FCT"), + ("Social Behavior", "takes turns during structured activities", "PRT"), + ], + "adolescent": [ + ("Self-Care", "washing hands independently", "Task Analysis"), + ("Self-Care", "dressing (pants, shirt)", "Task Analysis"), + ("Self-Care", "brushing teeth independently", "Task Analysis"), + ("Self-Care", "tolerating tooth brushing by caregiver", "Task Analysis"), + ("Self-Care", "tolerating haircut", "Task Analysis"), + ("Home Skills", "making the bed", "Task Analysis"), + ("Home Skills", "meal cleanup after eating", "Task Analysis"), + ("Community", "identifying community safety signs", "DTT"), + ("Community", "ordering in a restaurant", "Task Analysis"), + ("Mand", "requesting preferred items using AAC device", "NET"), + ("Mand", "requesting help when a task is difficult", "NET"), + ("Mand", "requesting a break during non-preferred activities", "NET"), + ("Mand", "indicating 'all done' with a step or activity", "NET"), + ("Tolerating Denied Access", "wait calmly when access to preferred item is denied", "FCT"), + ], + "adult": [ + ("Independent Living", "scheduling medical appointments", "Task Analysis"), + ("Independent Living", "managing a budget", "Task Analysis"), + ("Independent Living", "self-advocacy statements", "BST"), + ("Vocational", "clocking in at start of shift", "Task Analysis"), + ("Vocational", "completing assigned tasks in order", "Task Analysis"), + ("Home Skills", "doing laundry", "Task Analysis"), + ("Home Skills", "simple meal preparation", "Task Analysis"), + ("Community", "using public transportation", "Task Analysis"), + ("Mand", "requesting assistance using AAC device", "NET"), + ("Mand", "requesting a break during extended tasks", "NET"), + ("Mand", "indicating 'all done' at task completion", "NET"), + ("Tolerating Denied Access", "wait calmly when a preferred activity is denied", "FCT"), + ], +} + + +def sample_programs(learner_profile_id: str, n: int, rng: random.Random) -> list: + """Sample a realistic set of n acceleration programs for the learner profile.""" + category_map = { + "early": "early_language", + "school_age": "school_age", + "adolescent": "adolescent", + "adult": "adult", + } + primary = category_map.get(learner_profile_id, "school_age") + pool = list(SKILL_POOL_BY_CATEGORY[primary]) + # Mix in one program from an adjacent category for realism + adjacent = {"early": "school_age", "school_age": "early_language", "adolescent": "school_age", "adult": "adolescent"} + pool += SKILL_POOL_BY_CATEGORY[adjacent.get(learner_profile_id, "school_age")][:3] + rng.shuffle(pool) + return pool[:n] + + +# Session log rendering + + +def _render_measurements_for_program( + skill_domain: str, accuracy: float, rng: random.Random +) -> str: + """Return a one-line measurement string for one program in one session.""" + trials = rng.randint(8, 15) + correct = round(accuracy * trials) + fp = rng.randint(0, max(0, trials // 3)) + pp = rng.randint(0, max(0, trials // 3)) + g = rng.randint(0, max(0, trials // 3)) + indep = max(0, trials - fp - pp - g) + prompts = f"FP×{fp}, PP×{pp}, G×{g}, I×{indep}" + latency = 1.5 + (1.0 - accuracy) * 4.5 + rng.uniform(-0.3, 0.3) + return f"{correct}/{trials} correct ({round(accuracy*100)}%); latency {latency:.1f}s; prompts {prompts}" + + +def _render_behavior_measurement(behavior_id: str, freq: int, rng: random.Random) -> str: + """Return a measurement string for one target behavior in one session.""" + if behavior_id in {"tantrum"}: + if freq == 0: + return f"freq 0" + duration_total = freq * rng.randint(2, 5) + return f"freq {freq}, duration {duration_total}m total" + if behavior_id in {"motor_stereotypy", "vocal_stereotypy", "mouthing"}: + pir_pct = min(60, freq * rng.randint(3, 5) + rng.randint(0, 10)) + return f"freq {freq}; PIR {pir_pct}%" + if behavior_id == "pica": + # Pica is measured as attempts (total) and successful ingestions (subset). + # Most pica events are caught by staff before ingestion — successful ≤ attempts. + if freq == 0: + return "attempts 0" + attempts = freq + # Successful ingestion is stochastic but bounded; most are intercepted. + successful = 0 + for _ in range(attempts): + if rng.random() < 0.25: + successful += 1 + unsuccessful = attempts - successful + return ( + f"attempts {attempts} ({unsuccessful} unsuccessful — staff retrieved item before ingestion; " + f"{successful} successful — item ingested)" + ) + if behavior_id == "fecal_smearing": + # Scatolia is measured as attempts (reaching for feces) and completed smearing (subset). + # Staff often intercept before smearing completes — completed ≤ attempts. + if freq == 0: + return "attempts 0" + attempts = freq + completed = 0 + for _ in range(attempts): + if rng.random() < 0.35: + completed += 1 + intercepted = attempts - completed + return ( + f"attempts {attempts} ({intercepted} intercepted — staff redirected before smearing; " + f"{completed} completed — feces transferred to skin, clothing, or surface)" + ) + if behavior_id == "toileting_accident": + # Toileting data tracks both successful voids (in toilet) and accidents. + # BM accidents scale with overall accident load but stay physiologically + # bounded — typical session window is 0–1 BMs; some learners with GI + # clustering (IBS, diet, medication-related) can reach 2. + if freq == 0: + bm_accidents = 0 + elif freq == 1: + bm_accidents = 1 if rng.random() < 0.25 else 0 + elif freq <= 4: + bm_accidents = rng.choices([0, 1], weights=[0.65, 0.35])[0] + else: + bm_accidents = rng.choices([0, 1, 2], weights=[0.50, 0.35, 0.15])[0] + urine_accidents = max(0, freq - bm_accidents) + + # Successful voids scale inversely with accident load — a session full + # of accidents represents missed toileting opportunities. + if freq == 0: + urine_success = rng.randint(3, 6) + elif freq <= 2: + urine_success = rng.randint(2, 5) + elif freq <= 4: + urine_success = rng.randint(1, 3) + else: + urine_success = rng.randint(0, 2) + + # BM successes: typically 0 per session window, sometimes 1. Learners + # with BM accidents can still have an earlier successful BM in the toilet + # (e.g., scheduled morning sit worked, later opportunities missed). + if bm_accidents > 0: + bm_success = 1 if rng.random() < 0.15 else 0 + else: + bm_success = rng.choices([0, 1], weights=[0.75, 0.25])[0] + + return ( + f"urine: {urine_success} in-toilet / {urine_accidents} accidents; " + f"BM: {bm_success} in-toilet / {bm_accidents} accidents" + ) + if behavior_id in {"elopement"}: + return f"freq {freq}" + return f"freq {freq}" + + +ABC_TEMPLATES = { + "escape": [ + ("worksheet task presented", "refused + pushed materials away", "task removed for 2 min"), + ("demand to complete math problem", "threw materials and vocalized 'no'", "staff removed demand, redirected"), + ("transition to non-preferred activity", "dropped to floor, cried", "transition delayed, preferred item offered"), + ("asked to finish dressing task", "kicked at materials", "task paused for 3 min"), + ], + "attention": [ + ("adult attention diverted to another learner", "vocalized loudly + approached staff", "staff turned to address behavior"), + ("staff engaged in paperwork nearby", "yelled name + tapped staff", "staff attended and redirected"), + ], + "tangible": [ + ("preferred iPad removed at end of break", "screamed + reached for device", "staff redirected to alternative"), + ("peer took preferred toy", "grabbed toy back + cried", "staff mediated and returned item"), + ], + "automatic": [ + ("quiet unstructured moment between trials", "engaged in hand-flapping", "no social consequence; self-terminated after 30s"), + ("waiting in line transition", "mouthed own hand", "staff redirected to task materials"), + ], +} + + +# Behavior-specific ABC overrides — used in preference to the function-keyed +# templates above when the behavior has distinctive antecedents / topography. +BEHAVIOR_ABC_TEMPLATES = { + "fecal_smearing": [ + ("quiet moment in bathroom after bowel movement", "reached into diaper, retrieved feces", "staff redirected hands and initiated clean-up"), + ("unstructured downtime, staff out of direct line of sight", "reached behind waistband of pants", "staff intercepted and guided to bathroom"), + ("post-toileting, before hygiene steps completed", "transferred feces to arm and wall surface", "staff blocked further smearing, cleaned skin and surface"), + ("lying down for rest period", "reached into pull-up, brought hand to face", "staff intercepted before oral contact; hygiene routine initiated"), + ], + "toileting_accident": [ + ("during demand sequence at the work table", "voided (urine) without signaling need", "staff paused programming, changed clothing, resumed after 10 min"), + ("transition from preferred activity to non-preferred task", "voided (urine) en route to the work area", "staff redirected to bathroom, changed clothing"), + ("end of 90-minute block without scheduled toileting", "soiled (bowel) at the work table", "staff initiated clean-up and hygiene routine; program paused"), + ("during group activity without requesting bathroom", "voided (urine) in seat", "staff removed learner from group, changed clothing, re-entered after clean-up"), + ], +} + + +def _render_abc_entry(behavior_id: str, behavior_name: str, function: str, rng: random.Random) -> str: + templates = BEHAVIOR_ABC_TEMPLATES.get(behavior_id) or ABC_TEMPLATES.get(function) or ABC_TEMPLATES["escape"] + a, b, c = rng.choice(templates) + return f"ABC ({behavior_name.lower()}): A = {a}; B = {b}; C = {c}" + + +def render_session_log( + *, + synthetic_id: int, + learner_profile: dict, + curricula_line: str, + programs: list, + accuracies_per_program: list, + behaviors: list, + behavior_frequencies_per_behavior: list, + behavior_functions: list, + n_sessions: int, + start_date: date, + duration_per_session: int, + ioa_session_idx: int | None, + ioa_agreement: float | None, + include_abc: bool, + abc_sessions: list, + behavioral_indicators_cluster: list, + rng: random.Random, +) -> str: + """Render the full multi-session log as text.""" + lines = [] + lines.append("LEARNER PROFILE") + lines.append(f"Synthetic ID: SYN-{synthetic_id}") + profile_text = f"{learner_profile['name']}" + if "chronological_age" in learner_profile: + profile_text += f" ({learner_profile['chronological_age']})" + elif "developmental_age" in learner_profile: + profile_text += f" (dev age {learner_profile['developmental_age']})" + lines.append(f"Profile: {profile_text}") + lines.append(f"Curricula: {curricula_line}") + end_date = start_date + timedelta(days=n_sessions * 2) + lines.append(f"Date range: Sessions 1–{n_sessions} across {(end_date - start_date).days} days ({start_date.isoformat()} to {end_date.isoformat()})") + lines.append("") + + lines.append("ACCELERATION PROGRAMS") + for i, (domain, skill, method) in enumerate(programs, 1): + lines.append(f"{i}. {skill} ({domain})") + lines.append(f" Method: {method}") + lines.append("") + + if behaviors: + lines.append("DECELERATION TARGETS") + for i, (beh, fn) in enumerate(zip(behaviors, behavior_functions), 1): + lines.append(f"{i}. {beh['name']} — function hypothesized: {fn}") + lines.append("") + + lines.append("SESSION DATA") + for s_idx in range(n_sessions): + session_num = s_idx + 1 + d = (start_date + timedelta(days=s_idx * 2)).isoformat() + if ioa_session_idx == s_idx: + lines.append(f"") + lines.append(f"Session {session_num} — {d} — IOA SESSION — 2 observers") + for i, (domain, skill, method) in enumerate(programs): + acc = accuracies_per_program[i][s_idx] + lines.append(f" {skill}: {round(acc*100)}% (primary); IOA {round(ioa_agreement*100)}% agreement") + for i, beh in enumerate(behaviors): + freq = behavior_frequencies_per_behavior[i][s_idx] + ms = _render_behavior_measurement(beh["id"], freq, rng) + lines.append(f" {beh['name']}: {ms}; IOA {round(ioa_agreement*100)}% agreement") + else: + lines.append(f"") + lines.append(f"Session {session_num} — {d} — {duration_per_session} min — 1 observer") + for i, (domain, skill, method) in enumerate(programs): + acc = accuracies_per_program[i][s_idx] + ms = _render_measurements_for_program(domain, acc, rng) + lines.append(f" {skill}: {ms}") + for i, beh in enumerate(behaviors): + freq = behavior_frequencies_per_behavior[i][s_idx] + ms = _render_behavior_measurement(beh["id"], freq, rng) + lines.append(f" {beh['name']}: {ms}") + # ABC entry for this session? + if include_abc and s_idx in abc_sessions and behaviors: + beh_idx = rng.randrange(len(behaviors)) + fn = behavior_functions[beh_idx] + if behavior_frequencies_per_behavior[beh_idx][s_idx] > 0: + beh = behaviors[beh_idx] + lines.append(f" {_render_abc_entry(beh['id'], beh['name'], fn, rng)}") + + if behavioral_indicators_cluster: + lines.append("") + lines.append("BEHAVIORAL OBSERVATIONS (across sessions)") + for ind in behavioral_indicators_cluster: + lines.append(f"- {ind}") + + return "\n".join(lines) + + +# Interpretation section assembly + + +def assemble_clinical_concerns( + pattern_id: str, + pattern: dict, + accuracies_first_last: tuple, + behaviors: list, + behavior_frequencies_per_behavior: list, + n_sessions: int, +) -> str: + """Produce the clinical-concerns bullets.""" + first, last = accuracies_first_last + bullets = [] + if pattern_id == "mastery_progression": + bullets.append(f"- Steady improvement: accuracy rose from {round(first*100)}% to {round(last*100)}% across {n_sessions} sessions.") + bullets.append("- No clinical concerns at this time; mastery criteria should be reachable within a few additional sessions.") + elif pattern_id == "regression": + bullets.append(f"- **Regression identified**: accuracy has declined from {round(first*100)}% to {round(last*100)}% across {n_sessions} sessions.") + if behaviors: + bullets.append("- Accompanying behavior data suggests the decline may be function-related.") + elif pattern_id == "plateau": + bullets.append(f"- **Plateau**: accuracy has remained near {round(((first+last)/2)*100)}% for {n_sessions} sessions without meaningful improvement.") + bullets.append("- Performance is below mastery criteria; current teaching procedure may not be effective for this learner.") + elif pattern_id == "frustration_pattern": + bullets.append(f"- **Frustration pattern**: declining accuracy ({round(first*100)}% -> {round(last*100)}%) with escape-indicator behaviors.") + bullets.append("- Risk of learned helplessness and escape-maintained behavior if not addressed.") + elif pattern_id == "variable_performance": + avg = (first + last) / 2 + bullets.append(f"- **Variable performance**: accuracy fluctuating around {round(avg*100)}% with no clear trend.") + bullets.append("- Inconsistency may indicate stimulus-control issues or motivational variability across sessions.") + elif pattern_id == "prompt_dependency": + bullets.append(f"- **Prompt dependency**: high accuracy maintained ({round(first*100)}%–{round(last*100)}%) but primarily at prompted levels.") + bullets.append("- Independent responding remains low despite extended training.") + elif pattern_id == "rapid_acquisition": + bullets.append(f"- **Rapid acquisition**: accelerated mastery from {round(first*100)}% to {round(last*100)}% across {n_sessions} sessions.") + bullets.append("- Acquisition has exceeded expected timeline.") + elif pattern_id == "generalization_failure": + bullets.append("- **Generalization failure**: strong performance in training context but failure to generalize to novel conditions.") + bullets.append("- Skill may be under restricted stimulus control.") + elif pattern_id == "extinction_burst": + bullets.append("- **Extinction burst** observed: temporary increase in problem behavior consistent with an extinction procedure.") + bullets.append("- This is an expected phase of behavior change and typically indicates the intervention is working.") + elif pattern_id == "skill_loss_after_break": + bullets.append(f"- **Skill loss following break**: performance dropped to {round(first*100)}% at return, recovering to {round(last*100)}%.") + bullets.append("- Maintenance schedule should be reviewed and strengthened.") + elif pattern_id == "motivating_operation_shift": + bullets.append("- **Motivating-operation shift**: mid-log dip in responding followed by recovery suggests reinforcer-value change.") + bullets.append("- Possible satiation with current reinforcer; preference reassessment indicated.") + elif pattern_id == "setting_event_trigger": + bullets.append("- **Setting event pattern**: observable change in performance correlated with an external event (illness, sleep, schedule).") + bullets.append("- Correlation between setting event and performance should be tracked explicitly.") + return "\n".join(bullets) + + +def assemble_function_hypothesis_section( + behaviors: list, behavior_functions: list, abc_events_summary: str +) -> str: + """Produce the function hypothesis section, or empty string if no behaviors.""" + if not behaviors: + return "" + lines = ["", "## Behavior Function Hypothesis"] + for beh, fn in zip(behaviors, behavior_functions): + lines.append(f"{beh['name']}: {fn}") + if fn != "unknown": + ev_bits = [] + if abc_events_summary: + ev_bits.append(f"ABC events in the log support the {fn} function") + # Function-specific evidence without showing the operational definition + func_evidence = { + "escape": "behavior occurs in the context of demand presentation and terminates contingent on task removal", + "attention": "behavior occurs during low-attention periods and results in social attention contingently", + "tangible": "behavior occurs when preferred items are restricted and is followed by access to the item", + "automatic": "behavior persists across contexts without clear social mediation", + } + if fn in func_evidence: + ev_bits.append(func_evidence[fn]) + lines.append(f" Evidence: " + "; ".join(ev_bits) + ".") + else: + lines.append(" Evidence: insufficient data to disambiguate functions; recommend FBA / preference assessment before programming changes.") + lines.append("") + return "\n".join(lines) + + +def assemble_replacement_section(behaviors: list, behavior_functions: list, recs_pool: list) -> str: + """Produce the replacement behavior (FCT) section — only if behaviors present.""" + if not behaviors or not recs_pool: + return "" + lines = ["", "### Replacement behavior (FCT)"] + for r in recs_pool[:3]: + lines.append(f"- {r}") + return "\n".join(lines) + "\n" + + +def assemble_crisis_section(recs_pool: list, escalation_level: int) -> str: + """Produce the crisis plan section — only if escalation >= 3 and pool has items.""" + if escalation_level < 3 or not recs_pool: + return "" + lines = ["", "### Crisis plan"] + for r in recs_pool[:3]: + lines.append(f"- {r}") + return "\n".join(lines) + "\n" + + +def assemble_rationale( + accuracies_first_last: tuple, + mean_accuracy: float, + n_sessions: int, + behaviors: list, + behavior_frequencies_per_behavior: list, + ioa_present: bool, + ioa_agreement: float | None, +) -> str: + """Produce the data-supported rationale bullets.""" + first, last = accuracies_first_last + bullets = [] + bullets.append(f"- Mean accuracy across {n_sessions} sessions: {round(mean_accuracy*100, 1)}%.") + direction = "upward" if last > first else "downward" if last < first else "flat" + bullets.append(f"- Trend direction: {direction} (first session: {round(first*100)}%, last session: {round(last*100)}%).") + if behaviors: + for beh, freqs in zip(behaviors, behavior_frequencies_per_behavior): + mean_f = sum(freqs) / len(freqs) + first_third = sum(freqs[:max(1, len(freqs)//3)]) / max(1, len(freqs)//3) + last_third = sum(freqs[-max(1, len(freqs)//3):]) / max(1, len(freqs)//3) + bullets.append(f"- {beh['name']}: mean {mean_f:.1f}/session (early-window mean {first_third:.1f} -> late-window mean {last_third:.1f}).") + if ioa_present and ioa_agreement is not None: + bullets.append(f"- IOA data: {round(ioa_agreement*100)}% agreement on IOA session — {'adequate' if ioa_agreement >= 0.80 else 'below threshold'}.") + else: + bullets.append("- No IOA session in this log; reliability of measurements cannot be independently verified.") + return "\n".join(bullets) + + +# Escalation + confidence inference + + +def infer_escalation( + pattern: dict, + behaviors: list, + behavior_frequencies_per_behavior: list, + rules: dict, +) -> int: + """Compute escalation level from pattern + behavior severity.""" + level = int(pattern["default_escalation"]) + severe_ids = set(rules.get("severe_behavior_ids", [])) + if any(b["id"] in severe_ids for b in behaviors): + level = min(4, level + rules.get("severe_behavior_escalation_bump", 1)) + # Safety-immediate threshold + safety_threshold_ids = set(rules.get("safety_immediate_threshold_behaviors", [])) + threshold_rate = rules.get("safety_immediate_sib_rate_per_session", 8) + for beh, freqs in zip(behaviors, behavior_frequencies_per_behavior): + if beh["id"] in safety_threshold_ids: + if len(freqs) >= 2 and all(f >= threshold_rate for f in freqs[-2:]): + level = 4 + return level + + +def infer_confidence( + n_sessions: int, + ioa_present: bool, + accuracies: list, + rules: dict, +) -> str: + """Compute confidence level from data-quality features. + + Rules: + - low: very short log, OR short log with no IOA and non-trivial variance + - moderate: default when criteria for high are not met + - high: long log with IOA evidence and low variance + """ + mean = sum(accuracies) / len(accuracies) + sd = (sum((a - mean) ** 2 for a in accuracies) / len(accuracies)) ** 0.5 + + # Explicit low triggers + if n_sessions < rules["min_sessions_for_moderate"]: + return "low" + if n_sessions < 7 and not ioa_present and sd > 0.15: + return "low" + if sd > 0.35: + return "low" + + # High requires long log + IOA + modest variance + if ( + n_sessions >= rules["min_sessions_for_high"] + and (not rules["ioa_required_for_high"] or ioa_present) + and sd <= rules["variance_threshold_low"] + ): + return "high" + + return "moderate" + + +# Main generator class + + +class SessionInterpretationGenerator: + def __init__(self, config_dir: Path): + self.config_dir = Path(config_dir) + self.shared = load_shared(self.config_dir) + self.area = load_area(self.config_dir, AREA) + self.template = self.area["template"] + self.taxonomy = self.area["taxonomy"] + self.compat = self.area["compatibility"] + self.trajectory_rules = load_yaml(self.config_dir / AREA / "trajectory_rules.yaml") + self.recommendations = load_yaml(self.config_dir / AREA / "recommendations.yaml") + + def _pattern_by_id(self, pid: str) -> dict: + return next(p for p in self.taxonomy["patterns"] if p["id"] == pid) + + def _behavior_by_id(self, bid: str) -> dict: + return next(b for b in self.taxonomy["target_behaviors"] if b["id"] == bid) + + def _escalation_label(self, level: int) -> str: + return next(e["label"] for e in self.taxonomy["escalation_levels"] if e["id"] == level) + + def render_example(self, rng: random.Random) -> dict: + # Sample hidden pattern + pattern = rng.choice(self.taxonomy["patterns"]) + pattern_id = pattern["id"] + + # Log length + lo, hi = self.compat["log_length_by_pattern"][pattern_id] + n_sessions = rng.randint(lo, hi) + + # Learner profile (any — patterns are profile-agnostic) + profile = rng.choice(self.shared["learner_profiles"]["profiles"]) + + # Acceleration programs + n_programs = rng.randint(*self.compat["programs_per_log_range"]) + programs = sample_programs(profile["id"], n_programs, rng) + curricula_names = sorted(set(p[0] for p in programs)) + curricula_line = " + ".join(curricula_names) + + # Trajectory per program (all programs follow the pattern's trajectory loosely) + rule = self.trajectory_rules["rules"][pattern_id] + accuracies_per_program = [ + generate_accuracy_trajectory(pattern_id, rule, n_sessions, rng) + for _ in programs + ] + + # Target behaviors (if any) + n_beh_lo, n_beh_hi = self.compat["pattern_behavior_count_ranges"][pattern_id] + n_behaviors = rng.randint(n_beh_lo, n_beh_hi) + behaviors = [] + behavior_functions = [] + if n_behaviors > 0: + function_bias = self.compat["pattern_function_bias"].get(pattern_id, []) + candidate_behs = [ + b for b in self.taxonomy["target_behaviors"] + if any(f in b["plausible_functions"] for f in function_bias) or not function_bias + ] + behaviors = rng.sample(candidate_behs, min(n_behaviors, len(candidate_behs))) + for beh in behaviors: + if function_bias: + valid = [f for f in function_bias if f in beh["plausible_functions"]] + behavior_functions.append(rng.choice(valid) if valid else rng.choice(beh["plausible_functions"])) + else: + behavior_functions.append(rng.choice(beh["plausible_functions"])) + + # Behavior frequencies per session + behavior_freqs = [] + if behaviors: + traj_name = self.trajectory_rules["rules"][pattern_id]["behavior_trajectory"] + for _ in behaviors: + freqs = generate_behavior_frequencies( + traj_name, self.trajectory_rules["behavior_trajectory_library"], + n_sessions, rule, rng, + ) + behavior_freqs.append(freqs) + + # IOA decision + include_ioa = rng.random() < self.compat["ioa_inclusion_probability"] + ioa_session_idx = None + ioa_agreement = None + if include_ioa and n_sessions >= 4: + ioa_session_idx = rng.randint(n_sessions // 3, 2 * n_sessions // 3) + if rng.random() < self.compat["ioa_low_probability"]: + ioa_agreement = rng.uniform(*self.compat["ioa_low_range"]) + else: + ioa_agreement = rng.uniform(*self.compat["ioa_agreement_range"]) + + # ABC decision + include_abc = bool(behaviors) and rng.random() < self.compat["abc_inclusion_probability"] + abc_sessions = [] + if include_abc: + candidate_sessions = [i for i in range(n_sessions) if any(bf[i] > 0 for bf in behavior_freqs)] + n_abc = min(rng.randint(1, 3), len(candidate_sessions)) + abc_sessions = rng.sample(candidate_sessions, n_abc) if candidate_sessions else [] + + # Behavioral indicators + cluster = self.compat["pattern_behavioral_indicator_cluster"].get(pattern_id) + indicators = [] + if cluster: + pool = self.taxonomy["behavioral_indicators"][cluster] + indicators = rng.sample(pool, min(rng.randint(2, 4), len(pool))) + + # Log dates + start_date = date(2026, rng.randint(1, 10), rng.randint(1, 28)) + duration_per_session = rng.randint(*self.compat["session_duration_minutes_range"]) + + # Synthetic ID + synthetic_id = rng.randint(1000, 9999) + + # Render the user's session log + session_log = render_session_log( + synthetic_id=synthetic_id, + learner_profile=profile, + curricula_line=curricula_line, + programs=programs, + accuracies_per_program=accuracies_per_program, + behaviors=behaviors, + behavior_frequencies_per_behavior=behavior_freqs, + behavior_functions=behavior_functions, + n_sessions=n_sessions, + start_date=start_date, + duration_per_session=duration_per_session, + ioa_session_idx=ioa_session_idx, + ioa_agreement=ioa_agreement, + include_abc=include_abc, + abc_sessions=abc_sessions, + behavioral_indicators_cluster=indicators, + rng=rng, + ) + + # Compute gold labels + escalation_level = infer_escalation( + pattern, behaviors, behavior_freqs, self.compat["escalation_rules"] + ) + # Flatten accuracies across all programs for variance check + flat_accs = [v for traj in accuracies_per_program for v in traj] + confidence_level = infer_confidence( + n_sessions, ioa_session_idx is not None, flat_accs, self.compat["confidence_rules"] + ) + + # Primary-program accuracies for concerns/rationale (take first program) + primary_accs = accuracies_per_program[0] + mean_primary = sum(primary_accs) / len(primary_accs) + + # Assemble the assistant message sections + clinical_concerns_bullets = assemble_clinical_concerns( + pattern_id, pattern, (primary_accs[0], primary_accs[-1]), + behaviors, behavior_freqs, n_sessions, + ) + + pattern_explanations = { + "mastery_progression": "Accuracy is ascending cleanly across sessions with no interfering behaviors, consistent with a mastery-progression trajectory.", + "regression": "Previously-established responding has declined across recent sessions, warranting review of environmental and reinforcement variables.", + "plateau": "Responding has stabilized below mastery criterion without meaningful movement, indicating the current procedure is limiting further acquisition.", + "frustration_pattern": "Declining accuracy co-occurs with escape-indicator behaviors, consistent with frustration and escape-maintained responding to demand.", + "variable_performance": "High session-to-session variability with no clear trend suggests environmental or motivational inconsistency rather than a single programmatic issue.", + "prompt_dependency": "Accuracy stays high only at prompted levels; independent responding remains low despite extended training.", + "rapid_acquisition": "Accuracy ascended steeply beyond expected timeline for this learner's profile.", + "generalization_failure": "Alternation between strong training performance and poor novel-condition performance indicates restricted stimulus control.", + "extinction_burst": "Mid-log spike in problem-behavior frequency followed by recovery is consistent with a planned extinction procedure.", + "skill_loss_after_break": "Sharp drop at the start of the log followed by gradual recovery indicates skill loss following an absence.", + "motivating_operation_shift": "A mid-log dip followed by recovery maps onto a motivating-operation change rather than a skill issue.", + "setting_event_trigger": "A discontinuous performance change at a specific session correlates with an external setting event.", + } + + function_section = assemble_function_hypothesis_section( + behaviors, behavior_functions, + "ABC events present" if include_abc and abc_sessions else "" + ) + + pattern_recs = self.recommendations["patterns"][pattern_id] + antecedent_pool = pattern_recs["antecedent"] or ["Continue monitoring; reassess in 2 weeks."] + consequence_pool = pattern_recs["consequence"] or ["Maintain current consequences."] + ant_bullets = "\n".join(f"- {r}" for r in antecedent_pool[:4]) + con_bullets = "\n".join(f"- {r}" for r in consequence_pool[:3]) + + rep_section = assemble_replacement_section( + behaviors, behavior_functions, pattern_recs.get("replacement", []) + ) + crisis_section = assemble_crisis_section( + pattern_recs.get("crisis", []), escalation_level + ) + + escalation_justifications = { + 1: "Current programming is effective; no change needed.", + 2: "Programming adjustments indicated at next session; no immediate safety concern.", + 3: "Combination of performance decline and/or behavior data warrants supervising BCBA review within 24–48 hours.", + 4: "Safety-critical signals require immediate cessation of current program and direct BCBA contact; consider crisis-plan activation.", + } + + confidence_justifications = { + "high": f"Pattern is well-supported: {n_sessions} sessions of data, low variance, and IOA evidence confirm reliability.", + "moderate": f"Pattern is supported but alternatives cannot be fully excluded: {n_sessions} sessions of data" + (" with IOA evidence." if ioa_session_idx is not None else "; no IOA session included."), + "low": f"Insufficient data: only {n_sessions} sessions, high variance, and/or no IOA evidence. Recommend additional data collection before programming changes.", + } + + rationale_bullets = assemble_rationale( + (primary_accs[0], primary_accs[-1]), + mean_primary, + n_sessions, + behaviors, + behavior_freqs, + ioa_session_idx is not None, + ioa_agreement, + ) + + # Fill template + user_variant = rng.choice(self.template["user_variants"]) + user_content = user_variant.format(session_log=session_log) + + assistant_content = self.template["assistant_template"].format( + clinical_concerns_bullets=clinical_concerns_bullets, + pattern_class=pattern_id, + pattern_explanation=pattern_explanations[pattern_id], + function_hypothesis_section=function_section, + antecedent_bullets=ant_bullets, + replacement_section=rep_section, + consequence_bullets=con_bullets, + crisis_section=crisis_section, + escalation_level=escalation_level, + escalation_label=self._escalation_label(escalation_level), + escalation_justification=escalation_justifications[escalation_level], + confidence_level=confidence_level, + confidence_justification=confidence_justifications[confidence_level], + rationale_bullets=rationale_bullets, + ) + + # Metadata + gold_labels = { + "pattern_class": pattern_id, + "behavior_functions": {b["name"]: f for b, f in zip(behaviors, behavior_functions)}, + "escalation_level": escalation_level, + "confidence": confidence_level, + "crisis_plan_required": escalation_level >= 3, + } + provenance = { + "layer": 1, + "area": AREA, + "template_id": self.template["template_id"], + "taxonomy_cells": { + "hidden_pattern_id": pattern_id, + "learner_profile": profile["id"], + "n_sessions": n_sessions, + "n_programs": len(programs), + "n_behaviors": len(behaviors), + "behavior_ids": [b["id"] for b in behaviors], + "behavior_functions": list(behavior_functions), + "has_abc_data": include_abc and bool(abc_sessions), + "has_ioa_session": ioa_session_idx is not None, + }, + "teacher_model": None, + "seed_tag": str(rng.getstate()[1][0] % 100000), + } + + return make_example_envelope( + system_content=self.template["system_prompt"], + user_content=user_content, + assistant_content=assistant_content, + task_type=TASK_TYPE, + gold_labels=gold_labels, + provenance=provenance, + ) diff --git a/src/generators/aba_task_analysis.py b/src/generators/aba_task_analysis.py new file mode 100644 index 0000000000000000000000000000000000000000..50ba56dc9217fecc019123ebbc0447e6c6727c38 --- /dev/null +++ b/src/generators/aba_task_analysis.py @@ -0,0 +1,277 @@ +"""Task Analysis / Chaining area generator. + +Loads configs/task_analysis/ + configs/shared/ and produces chaining +teaching program examples for AFLS modules (Basic Living, Home, Community, +Vocational, Independent Living). +""" + +import random +from pathlib import Path + +from .base import ( + load_area, + load_shared, + load_yaml, + make_example_envelope, +) + +AREA = "task_analysis" +TASK_TYPE = "teaching_program" + + +# Mastery state -> which shaping step the learner is currently at (as index range +# within the shaping_steps list). 0-indexed, end-exclusive. +TOLERATION_MASTERY_STEP_RANGES = { + "emerging": (0, 2), + "developing": (1, 3), + "approaching": (2, 4), + "near": (3, 5), + "mastered": (4, 6), + "generalization": (5, 7), +} + + +def current_shaping_step_guidance(mastery_state_id: str, shaping_steps: list) -> str: + """Describe where the learner is currently sitting in the shaping progression.""" + n = len(shaping_steps) + lo, hi = TOLERATION_MASTERY_STEP_RANGES.get(mastery_state_id, (0, 2)) + lo = max(0, min(lo, n - 1)) + hi = max(lo + 1, min(hi, n)) + idx = min(lo, n - 1) + return ( + f"begin at shaping step {idx + 1} (\"{shaping_steps[idx]}\") and advance through " + f"step {hi} across sessions as tolerance stabilizes." + ) + + +def render_shaping_steps(shaping_steps: list) -> str: + """Render numbered shaping progression.""" + return "\n".join(f"{i}. {step}" for i, step in enumerate(shaping_steps, 1)) + + +def current_prompt_guidance(mastery_state_id: str, prompt_strategy_id: str) -> str: + """Guidance for chaining prompt level at the current mastery state.""" + if mastery_state_id in {"emerging", "developing"}: + if prompt_strategy_id == "graduated_guidance": + return "provide hand-over-hand guidance at each step; reduce physical pressure across sessions." + return "use full-physical or partial-physical prompts at each step; fade systematically." + if mastery_state_id == "approaching": + return "use partial-physical or gestural prompts only at steps where errors occur; allow independence elsewhere." + if mastery_state_id in {"near", "mastered", "generalization"}: + return "deliver the least intrusive prompt that prevents an error; expect independence on previously-mastered steps." + return "prompt level calibrated to current performance." + + +def render_steps(step_list: list, chain_type_id: str) -> str: + """Render numbered step list, with callout of the learner-taught step for backward chaining.""" + lines = [] + for i, step in enumerate(step_list, 1): + if chain_type_id == "backward" and i == len(step_list): + lines.append(f"{i}. {step} <- Learner-taught step (backward chaining starts here)") + elif chain_type_id == "forward" and i == 1: + lines.append(f"{i}. {step} <- Learner-taught step (forward chaining starts here)") + else: + lines.append(f"{i}. {step}") + return "\n".join(lines) + + +def render_criterion_text(criterion: dict, rng: random.Random) -> str: + """Fill any slots in the mastery-criterion text (e.g., fluency seconds).""" + text = criterion["text"] + if "{fluency_seconds}" in text: + text = text.replace("{fluency_seconds}", str(rng.choice([60, 90, 120, 180]))) + return text + + +class TaskAnalysisGenerator: + def __init__(self, config_dir: Path): + self.config_dir = Path(config_dir) + self.shared = load_shared(self.config_dir) + self.area = load_area(self.config_dir, AREA) + self.template = self.area["template"] # chain-based (independence) + self.template_toleration = load_yaml( + self.config_dir / AREA / "template_toleration.yaml" + ) + self.taxonomy = self.area["taxonomy"] + self.compat = self.area["compatibility"] + + def sample_combination(self, rng: random.Random) -> dict: + # Weighted module selection + module_weights = self.compat["module_sampling_weights"] + modules = self.taxonomy["afls"]["modules"] + module_ids = [m["id"] for m in modules] + weights = [module_weights[mid] for mid in module_ids] + module = rng.choices(modules, weights=weights)[0] + + # Sample a skill from the module + skill_entry = rng.choice(module["skills"]) + skill_target = skill_entry["name"] + is_toleration = skill_entry.get("program_type") == "toleration" + + # Learner profile — intersect module's allowed with the (shared) learner list + profiles = self.shared["learner_profiles"]["profiles"] + module_allowed = set(module["learner_profiles"]) + profile = rng.choice([p for p in profiles if p["id"] in module_allowed]) + + mastery_state = rng.choice(self.shared["mastery_states"]["states"]) + + # Filter mastery criteria by program type (independence vs toleration) + program_type = "toleration" if is_toleration else "independence" + eligible_criteria = [ + c for c in self.taxonomy["mastery_criteria"] + if c.get("applies_to", "independence") == program_type + ] + if not eligible_criteria: + eligible_criteria = self.taxonomy["mastery_criteria"] + mastery_criterion = rng.choice(eligible_criteria) + + if is_toleration: + # Toleration programs: duration-based shaping, no chain / prompt / EC fields needed. + return { + "skill_target": skill_target, + "program_type": "toleration", + "shaping_steps": skill_entry["shaping_steps"], + "target_activity": skill_entry["target_activity"], + "end_goal_description": skill_entry["end_goal_description"], + "module_id": module["id"], + "module_name": module["name"], + "learner_profile": profile, + "mastery_state": mastery_state, + "mastery_criterion": mastery_criterion, + } + + # Independence (chain-based) program — existing path + steps = skill_entry["steps"] + preferred_chain = skill_entry.get("chain_type_preferred", "total_task") + + if rng.random() < self.compat["use_preferred_chain_type_probability"]: + chain_type = next(c for c in self.taxonomy["chain_types"] if c["id"] == preferred_chain) + else: + chain_type = rng.choice(self.taxonomy["chain_types"]) + + allowed_strategy_ids = set(self.compat["mastery_to_prompt_strategies"][mastery_state["id"]]) + valid_strategies = [s for s in self.taxonomy["prompt_hierarchies"] if s["id"] in allowed_strategy_ids] + if not valid_strategies: + valid_strategies = self.taxonomy["prompt_hierarchies"] + prompt_strategy = rng.choice(valid_strategies) + + allowed_schedule_ids = set(self.compat["mastery_to_reinforcement"][mastery_state["id"]]) + valid_schedules = [s for s in self.taxonomy["reinforcement_schedules"] if s["id"] in allowed_schedule_ids] + if not valid_schedules: + valid_schedules = [s for s in self.taxonomy["reinforcement_schedules"] if s["id"] == "crf_per_step"] + reinforcement_schedule = rng.choice(valid_schedules) + + error_correction = rng.choice(self.taxonomy["error_corrections"]) + + return { + "skill_target": skill_target, + "program_type": "independence", + "steps": steps, + "module_id": module["id"], + "module_name": module["name"], + "chain_type": chain_type, + "learner_profile": profile, + "mastery_state": mastery_state, + "prompt_strategy": prompt_strategy, + "reinforcement_schedule": reinforcement_schedule, + "error_correction": error_correction, + "mastery_criterion": mastery_criterion, + } + + def compute_slots(self, combo: dict, rng: random.Random) -> dict: + mastery_state = combo["mastery_state"] + common = { + "skill_target": combo["skill_target"], + "curriculum_ref": f"AFLS {combo['module_name']}", + "learner_profile_name": combo["learner_profile"]["name"], + "mastery_state_name": mastery_state["name"], + "mastery_state_short": mastery_state.get("short", mastery_state["name"].lower()), + "mastery_criterion_text": render_criterion_text(combo["mastery_criterion"], rng), + "ioa_frequency": rng.choice([3, 4, 5]), + "n_generalization_settings": rng.choice([2, 3]), + "n_generalization_therapists": rng.choice([2, 3]), + } + + if combo["program_type"] == "toleration": + return { + **common, + "target_activity": combo["target_activity"], + "end_goal_description": combo["end_goal_description"], + "shaping_step_list": render_shaping_steps(combo["shaping_steps"]), + "current_step_guidance": current_shaping_step_guidance( + mastery_state["id"], combo["shaping_steps"] + ), + } + + # Independence (chain-based) + ec_steps_clean = combo["error_correction"]["steps"].strip() + return { + **common, + "chain_type_name": combo["chain_type"]["name"], + "chain_type_description": combo["chain_type"]["description"], + "step_list": render_steps(combo["steps"], combo["chain_type"]["id"]), + "prompt_strategy_name": combo["prompt_strategy"]["name"], + "prompt_strategy_description": combo["prompt_strategy"]["description"], + "current_prompt_guidance": current_prompt_guidance(mastery_state["id"], combo["prompt_strategy"]["id"]), + "error_correction_name": combo["error_correction"]["name"], + "error_correction_steps": ec_steps_clean.rstrip("."), + "reinforcement_schedule_name": combo["reinforcement_schedule"]["name"], + "reinforcement_description": combo["reinforcement_schedule"]["description"].rstrip("."), + } + + def render_example(self, rng: random.Random) -> dict: + combo = self.sample_combination(rng) + slots = self.compute_slots(combo, rng) + + template = ( + self.template_toleration + if combo["program_type"] == "toleration" + else self.template + ) + user_variant = rng.choice(template["user_variants"]) + user_content = user_variant.format(**slots) + assistant_content = template["assistant_template"].format(**slots) + + gold_labels = { + "method": AREA, + "domain": f"AFLS.{combo['module_id']}", + "skill": combo["skill_target"], + "program_type": combo["program_type"], + "learner_profile": combo["learner_profile"]["id"], + "mastery_state": combo["mastery_state"]["id"], + } + if combo["program_type"] == "independence": + gold_labels["chain_type"] = combo["chain_type"]["id"] + + taxonomy_cells = { + "skill_target": combo["skill_target"], + "module": combo["module_id"], + "mastery_criterion": combo["mastery_criterion"]["id"], + } + if combo["program_type"] == "independence": + taxonomy_cells.update({ + "chain_type": combo["chain_type"]["id"], + "prompt_strategy": combo["prompt_strategy"]["id"], + "reinforcement_schedule": combo["reinforcement_schedule"]["id"], + "error_correction": combo["error_correction"]["id"], + }) + else: + taxonomy_cells["shaping_n_steps"] = len(combo["shaping_steps"]) + + provenance = { + "layer": 1, + "area": AREA, + "template_id": template["template_id"], + "taxonomy_cells": taxonomy_cells, + "teacher_model": None, + "seed_tag": str(rng.getstate()[1][0] % 100000), + } + + return make_example_envelope( + system_content=template["system_prompt"], + user_content=user_content, + assistant_content=assistant_content, + task_type=TASK_TYPE, + gold_labels=gold_labels, + provenance=provenance, + ) diff --git a/src/generators/base.py b/src/generators/base.py new file mode 100644 index 0000000000000000000000000000000000000000..3b4aa8682a68d6752ff9178088956b5ab748cf12 --- /dev/null +++ b/src/generators/base.py @@ -0,0 +1,214 @@ +"""Shared base utilities for per-area generators. + +This module is domain-agnostic. Area-specific logic lives in +`aba_.py` modules. +""" + +import random +import re +from pathlib import Path + +import yaml + +# ============================================================ +# YAML loading +# ============================================================ + + +def load_yaml(path: Path) -> dict: + """Load a single YAML file.""" + with open(path) as f: + return yaml.safe_load(f) + + +def load_shared(config_dir: Path) -> dict: + """Load the shared primitives used across all areas. + + Returns a dict keyed by primitive name: + learner_profiles, mastery_states, prompt_types + """ + shared_dir = Path(config_dir) / "shared" + return { + "learner_profiles": load_yaml(shared_dir / "learner_profiles.yaml"), + "mastery_states": load_yaml(shared_dir / "mastery_states.yaml"), + "prompt_types": load_yaml(shared_dir / "prompt_types.yaml"), + } + + +def load_area(config_dir: Path, area: str) -> dict: + """Load an area's self-contained config: taxonomy + template + compatibility.""" + area_dir = Path(config_dir) / area + if not area_dir.is_dir(): + raise FileNotFoundError(f"Area config directory not found: {area_dir}") + return { + "taxonomy": load_yaml(area_dir / "taxonomy.yaml"), + "template": load_yaml(area_dir / "template.yaml"), + "compatibility": load_yaml(area_dir / "compatibility.yaml"), + } + + +# ============================================================ +# Text / rendering utilities +# ============================================================ + + +def article(word: str) -> str: + """Return 'a' or 'an' appropriate for the following word.""" + if not word: + return "a" + return "an" if word[0].lower() in "aeiou" else "a" + + +def strip_trailing_period(text: str) -> str: + """Remove a single trailing period — for safe mid-sentence embedding.""" + return text.rstrip().rstrip(".") + + +def render_prompt_sequence(hierarchy_sequence: list, prompt_types: list) -> str: + """Render a prompt-hierarchy sequence as a human-readable arrow chain. + + Looks up each item in the canonical prompt_types list; falls back to + underscore-to-space conversion for composite IDs like '0s_delay'. + """ + id_to_name = {pt["id"]: pt["name"] for pt in prompt_types} + rendered = [] + for seq_id in hierarchy_sequence: + if seq_id in id_to_name: + rendered.append(id_to_name[seq_id].lower()) + else: + rendered.append(seq_id.replace("_", " ")) + return " -> ".join(rendered) + + +def parse_skill_numeric_range(skill_text: str): + """Extract (min, max) numeric range from skill text. + + Handles patterns like '1-5', '1–5' (en-dash), 'to 10', '1-20'. + Returns None if no range is present. + """ + s = skill_text.lower() + m = re.search(r"(\d+)\s*[-–]\s*(\d+)", s) + if m: + lo, hi = int(m.group(1)), int(m.group(2)) + return (min(lo, hi), max(lo, hi)) + m = re.search(r"to\s+(\d+)", s) + if m: + return (1, int(m.group(1))) + return None + + +# ============================================================ +# Shared stimulus pools (used by DTT-like array-based generators) +# ============================================================ + +STIMULUS_POOLS = { + "objects": [ + "ball", "cup", "shoe", "spoon", "book", "car", "block", "brush", "key", + "phone", "hat", "sock", "apple", "crayon", "scissors", "glue", "pencil", + "napkin", "plate", "towel", "jacket", "backpack", "bottle", "blanket", + "toothbrush", + ], + "colors": ["red", "blue", "green", "yellow", "orange", "purple", "pink", "brown", "black", "white"], + "shapes": ["circle", "square", "triangle", "rectangle", "star", "diamond", "oval", "heart"], + "animals": [ + "dog", "cat", "bird", "fish", "horse", "cow", "pig", "sheep", "rabbit", + "frog", "bear", "lion", "elephant", "monkey", "duck", "chicken", + ], + "actions": [ + "running", "jumping", "eating", "sleeping", "reading", "writing", + "drawing", "singing", "clapping", "washing", "brushing", "pouring", + ], + "letters": list("ABCDEFGHIJKLMNOPQRSTUVWXYZ"), + "numbers": [str(n) for n in range(1, 21)], +} + +SKILL_TO_POOL = [ + ("color", "colors"), + ("shape", "shapes"), + ("animal", "animals"), + ("letter", "letters"), + ("number", "numbers"), + ("numeral", "numbers"), + ("count", "numbers"), + ("action", "actions"), +] + + +def choose_stimulus_pool(skill_target: str) -> str: + """Pick the stimulus pool whose keyword appears in the skill name; default 'objects'.""" + s = skill_target.lower() + for keyword, pool in SKILL_TO_POOL: + if keyword in s: + return pool + return "objects" + + +def sample_stimuli( + skill_target: str, array_size_n: int, rng: random.Random +) -> dict: + """Pick target + distractor stimuli, respecting any numeric range in the skill text.""" + pool_name = choose_stimulus_pool(skill_target) + pool = list(STIMULUS_POOLS[pool_name]) + + rng_tuple = parse_skill_numeric_range(skill_target) + if rng_tuple is not None: + lo, hi = rng_tuple + if pool_name == "numbers": + pool = [str(n) for n in range(lo, hi + 1)] + elif pool_name == "letters": + pool = pool[: max(hi, 3)] + + n_targets = min(3, len(pool)) + targets = rng.sample(pool, n_targets) + remaining = [x for x in pool if x not in targets] + n_distractors = min(max(array_size_n - 1, 0), len(remaining)) + distractors = rng.sample(remaining, n_distractors) if n_distractors > 0 else [] + return {"targets": targets, "distractors": distractors} + + +# ============================================================ +# Example envelope helpers +# ============================================================ + + +import hashlib + + +def make_example_envelope( + *, + system_content: str, + user_content: str, + assistant_content: str, + task_type: str, + gold_labels: dict, + provenance: dict, +) -> dict: + """Assemble the final JSONL example dict with id + meta envelope. + + Envelope is a deterministic function of its inputs: identical inputs produce + an identical envelope, including the example_id. The example_id is + ``sha256(user_content + assistant_content)[:16]`` computed on the *published* + (stripped) message content, so any consumer can verify it directly from the + JSONL row. Per-run wall-clock timestamps are deliberately NOT included so + that running the generator twice with the same seed produces byte-identical + output. + """ + system_content = system_content.strip() + user_content = user_content.strip() + assistant_content = assistant_content.strip() + example_hash = hashlib.sha256( + (user_content + assistant_content).encode() + ).hexdigest()[:16] + return { + "messages": [ + {"role": "system", "content": system_content}, + {"role": "user", "content": user_content}, + {"role": "assistant", "content": assistant_content}, + ], + "meta": { + "task_type": task_type, + "example_id": example_hash, + "gold_labels": gold_labels, + "provenance": provenance, + }, + } diff --git a/src/prepare_curation.py b/src/prepare_curation.py new file mode 100644 index 0000000000000000000000000000000000000000..04cab98f4d0cb28d3bb58db3b6fe54548e054ad5 --- /dev/null +++ b/src/prepare_curation.py @@ -0,0 +1,151 @@ +"""Render curation_pool.jsonl as a readable review document. + +This used to drive a hand-curation workflow with decisions.csv; we now +use the whole curation pool as our test set (see compile_curation.py). +review.md is still produced as a human-readable browse of the test +corpus so clinical issues can be flagged ad hoc and fixed in the +generator configs. + +Usage: + uv run python src/prepare_curation.py +""" + +import json +import sys +from pathlib import Path + +REPO_ROOT = Path(__file__).resolve().parent.parent +POOL_PATH = REPO_ROOT / "data" / "splits" / "curation_pool.jsonl" +REVIEW_MD = REPO_ROOT / "docs" / "curation" / "review.md" + + +def category_of(example: dict) -> str: + """Short category label for a candidate.""" + gl = example["meta"]["gold_labels"] + if example["meta"]["task_type"] == "teaching_program": + return gl.get("method", "?") + return gl.get("pattern_class", "?") + + +def gold_labels_markdown(example: dict) -> str: + """Render gold labels inline for the review markdown.""" + gl = example["meta"]["gold_labels"] + bits = [] + for k, v in gl.items(): + if isinstance(v, dict): + inner = ", ".join(f"{ik}->{iv}" for ik, iv in v.items()) + bits.append(f"**{k}:** {{{inner}}}") + elif isinstance(v, bool): + bits.append(f"**{k}:** {str(v).lower()}") + else: + bits.append(f"**{k}:** `{v}`") + return " · ".join(bits) + + +def provenance_markdown(example: dict) -> str: + """Render a short provenance line.""" + prov = example["meta"].get("provenance", {}) + bits = [] + if "n_sessions" in prov: + bits.append(f"{prov['n_sessions']} sessions") + if "n_behaviors" in prov: + bits.append(f"{prov['n_behaviors']} behaviors") + if "has_ioa_session" in prov: + bits.append("IOA ✓" if prov["has_ioa_session"] else "no IOA") + if "has_abc_data" in prov: + bits.append("ABC ✓" if prov["has_abc_data"] else "no ABC") + return " · ".join(bits) if bits else "" + + +def render_candidate(idx: int, example: dict) -> str: + """Render one candidate as a markdown block.""" + task = example["meta"]["task_type"] + cat = category_of(example) + example_id = example["meta"]["example_id"] + user_content = example["messages"][1]["content"] + assistant_content = example["messages"][2]["content"] + prov_str = provenance_markdown(example) + prov_line = f"\n*{prov_str}*\n" if prov_str else "\n" + + return f""" +--- + +### Candidate {idx:03d} | `{task}` | `{cat}` | id `{example_id}` + +**Gold labels:** {gold_labels_markdown(example)} +{prov_line} +
+User message (click to expand) + +``` +{user_content} +``` + +
+ +
+Assistant response (click to expand) + +``` +{assistant_content} +``` + +
+""" + + +def main() -> int: + if not POOL_PATH.exists(): + print(f"Curation pool not found: {POOL_PATH}", file=sys.stderr) + print("Run `uv run python src/split_data.py` first.", file=sys.stderr) + return 1 + + candidates: list[dict] = [] + with open(POOL_PATH) as f: + for line in f: + candidates.append(json.loads(line)) + + # Group by category for diversity visibility. + grouped_raw: dict[str, list[dict]] = {} + for ex in candidates: + key = f"{ex['meta']['task_type']} / {category_of(ex)}" + grouped_raw.setdefault(key, []).append(ex) + + ordered: list[tuple[int, dict, str]] = [] + idx = 0 + for key in sorted(grouped_raw.keys()): + for ex in grouped_raw[key]: + idx += 1 + ordered.append((idx, ex, key)) + + REVIEW_MD.parent.mkdir(parents=True, exist_ok=True) + with open(REVIEW_MD, "w") as f: + f.write("# Corpus Review — TRACE test corpus\n\n") + f.write(f"**{len(candidates)} examples** held out as the evaluation corpus.\n\n") + f.write("## How to use this file\n") + f.write("- Browse candidates below to spot clinical-accuracy issues " + "(wrong topography, unrealistic scenario, inconsistent data, etc.).\n") + f.write("- When you find one, note the candidate id (shown next to the candidate title). " + "Fixes are applied by editing the generator configs in `configs/` and regenerating.\n") + f.write("- The test + sanity splits are compiled from this pool by " + "`uv run python src/compile_curation.py` (no per-example decisions required).\n\n") + f.write("## Distribution\n\n") + f.write("| Category | Count |\n|---|---|\n") + for key in sorted(grouped_raw.keys()): + f.write(f"| {key} | {len(grouped_raw[key])} |\n") + f.write("\n---\n\n") + f.write("## Candidates\n") + current_group = None + for idx, ex, key in ordered: + if key != current_group: + f.write(f"\n### Group: {key}\n") + current_group = key + f.write(render_candidate(idx, ex)) + + print(f"Wrote: {REVIEW_MD} ({len(candidates)} candidates)") + print("Next: run `uv run python src/compile_curation.py` to produce test.jsonl + sanity.jsonl.") + return 0 + + +if __name__ == "__main__": + raise SystemExit(main()) diff --git a/src/split_data.py b/src/split_data.py new file mode 100644 index 0000000000000000000000000000000000000000..a8f2e48aa2e1819efc5090b578612827bbba4949 --- /dev/null +++ b/src/split_data.py @@ -0,0 +1,148 @@ +"""Dataset splitter — combines per-area JSONL files into stratified train / valid / curation pool. + +Stratification ensures: +- task_type ratio preserved across splits (60/40 teaching vs interp by default) +- For task 1: each method represented proportionally +- For task 2: each pattern represented proportionally + +Usage: + uv run python src/split_data.py + uv run python src/split_data.py --seed 42 --out-dir data/splits +""" + +import argparse +import json +import random +import sys +from collections import defaultdict +from pathlib import Path + +REPO_ROOT = Path(__file__).resolve().parent.parent +sys.path.insert(0, str(REPO_ROOT)) + +from src.generators.base import load_yaml + +CONFIG_DIR = REPO_ROOT / "configs" + + +def load_all_area_files(input_dir: Path, areas: list[str]) -> list[dict]: + """Load every per-area JSONL file; deduplicate by example_id (a content hash).""" + seen: set[str] = set() + combined: list[dict] = [] + n_raw = 0 + for area in areas: + path = input_dir / f"{area}.jsonl" + if not path.exists(): + print(f" Skipping {area}: {path} not found", file=sys.stderr) + continue + with open(path) as f: + for line in f: + n_raw += 1 + ex = json.loads(line) + eid = ex["meta"]["example_id"] + if eid in seen: + continue + seen.add(eid) + combined.append(ex) + dropped = n_raw - len(combined) + if dropped: + print(f" Deduplication: dropped {dropped} exact-duplicate example(s) ({n_raw} -> {len(combined)}).") + return combined + + +def stratify_key(example: dict) -> str: + """Return a stratification bucket key for an example.""" + task = example["meta"]["task_type"] + gl = example["meta"]["gold_labels"] + if task == "teaching_program": + return f"t1:{gl['method']}" + return f"t2:{gl['pattern_class']}" + + +def stratified_split( + examples: list[dict], + train_ratio: float, + valid_ratio: float, + curation_ratio: float, + rng: random.Random, +) -> tuple[list[dict], list[dict], list[dict]]: + """Stratified three-way split preserving per-bucket proportions.""" + buckets: dict[str, list[dict]] = defaultdict(list) + for ex in examples: + buckets[stratify_key(ex)].append(ex) + + train, valid, curation = [], [], [] + for key in sorted(buckets.keys()): + items = buckets[key] + rng.shuffle(items) + n = len(items) + n_train = int(round(n * train_ratio)) + n_valid = int(round(n * valid_ratio)) + n_curation = n - n_train - n_valid + # Guarantee at least 1 per split if bucket is large enough + if n >= 10: + n_valid = max(1, n_valid) + n_curation = max(1, n_curation) + n_train = n - n_valid - n_curation + train.extend(items[:n_train]) + valid.extend(items[n_train : n_train + n_valid]) + curation.extend(items[n_train + n_valid :]) + rng.shuffle(train) + rng.shuffle(valid) + rng.shuffle(curation) + return train, valid, curation + + +def write_jsonl(examples: list[dict], path: Path) -> None: + with open(path, "w") as f: + for ex in examples: + f.write(json.dumps(ex) + "\n") + + +def main(): + parser = argparse.ArgumentParser() + parser.add_argument("--seed", type=int, default=None) + parser.add_argument("--out-dir", type=str, default=None) + args = parser.parse_args() + + gen_config = load_yaml(CONFIG_DIR / "generation.yaml") + seed = args.seed if args.seed is not None else gen_config["seed"] + out_dir = Path(args.out_dir) if args.out_dir else REPO_ROOT / gen_config["output_dir"] + + splits_cfg = gen_config["splits"] + tr = splits_cfg["train"] + va = splits_cfg["valid"] + cu = splits_cfg["curation_pool"] + if abs(tr + va + cu - 1.0) >= 1e-6: + raise ValueError(f"Splits must sum to 1.0 (got {tr + va + cu})") + + # Collect enabled areas that have output files + enabled_areas = [a for a, c in gen_config["areas"].items() if c.get("enabled")] + examples = load_all_area_files(out_dir, enabled_areas) + print(f"Loaded {len(examples)} total examples from {len(enabled_areas)} areas.\n") + + rng = random.Random(seed) + train, valid, curation = stratified_split(examples, tr, va, cu, rng) + + write_jsonl(train, out_dir / "train.jsonl") + write_jsonl(valid, out_dir / "valid.jsonl") + write_jsonl(curation, out_dir / "curation_pool.jsonl") + + print("Splits written:") + print(f" train.jsonl {len(train):5d} ({100*len(train)/len(examples):.1f}%)") + print(f" valid.jsonl {len(valid):5d} ({100*len(valid)/len(examples):.1f}%)") + print(f" curation_pool.jsonl {len(curation):5d} ({100*len(curation)/len(examples):.1f}%)") + print() + print("Curation pool is the source for the author-curated test (100) + sanity (20) splits.") + + # Sanity: report per-bucket balance in curation_pool + bucket_counts = defaultdict(int) + for ex in curation: + bucket_counts[stratify_key(ex)] += 1 + print("\nCuration-pool stratification:") + for key, c in sorted(bucket_counts.items()): + print(f" {key:45s} {c:3d}") + + +if __name__ == "__main__": + main()