Dataset Description
Paper: https://arxiv.org/pdf/2410.22446
Language(s) (NLP): English
License: apache-2.0
Point of Contact: Viet Cuong (Johnny) Nguyen
Dataset Summary
CounselingBench is a dataset of 1612 mental health counseling-related questions across 138 case studies extracted from existing NCMHCE mock exams. NCMHCE questions are designed to test a candidate's aptitude in one out of five mental health counseling competencies:
- Intake, Assessment & Diagnosis
- Counseling Skills & Interventions
- Treatment Planning
- Professional Practice & Ethics
- Core Counseling Attributes
Data Fields
- question # (int): The unique numeric identifier for the question
- patient demographic (string): Information regarding the patient's demographic
- mental status exam (string): Information regarding the patient's mental status examination
- presenting problem (string): Information regarding the patient's presenting problem
- other contexts (string): Other information regarding the patient's background and presentation
- question (string): The full text of a question.
- choice a (string): The full text of Choice A
- choice b (string): The full text of Choice B
- choice c (string): The full text of Choice C
- choice d (string): The full text of Choice D
- potential answers (string): The concatenated full text of all potential answers to the question
- correct answer (string): The full text of the correct answer
- correct answer (letter) (string): The letter corresponding to the correct answer
- explanation for correct answer (string): Expert-generated explanation for the correct answer
- competency (string): Expert-annotated competency which the question aims to test
Licensing Information
CounselingBench is now made available under the Apache 2.0 License.
Citation Information
Please consider citing our paper if you find this dataset useful:
@article{nguyen2024large,
title={Do Large Language Models Align with Core Mental Health Counseling Competencies?},
author={Nguyen, Viet Cuong and Taher, Mohammad and Hong, Dongwan and Possobom, Vinicius Konkolics and Gopalakrishnan, Vibha Thirunellayi and Raj, Ekta and Li, Zihang and Soled, Heather J and Birnbaum, Michael L and Kumar, Srijan and others},
journal={arXiv preprint arXiv:2410.22446},
year={2024}
}