Update my_pages/multiverse.py

my_pages/multiverse.py

@@ -122,6 +122,7 @@ def render():
     add_instruction_text(
         """
         Visually explore the multiverse of AI models to judge loan applications.<br> 
+        We are using a publicly available loan approval dataset.<br>
         Make a choice, and scroll down to see the properties of the trained model.
         """
     )
@@ -310,10 +311,13 @@ def render():
     uniq_count_acc = np.sum(condition_acc)
     
     add_red_text(f"""
-        <b>Multiplicity of your predictions:</b> Your chosen model rejected {uniq_count_rej} loans that would have been accepted by the majority of the other models.
-        Similariy, your chosen model accepted {uniq_count_acc} loans that would have been rejected by the majority of the other models.<br><br>
-        You might want higher multiplicity, to avoid competiting for the same loan applicants with others and to give a chance to unique applicants.<br><br>
-        On the other hand, you might instead want lower multiplicity, to take lower risks and to avoid facing a justification crisis, i.e., 
-        not able to explain why you rejected certain applicants when they were accepted by everyone else.<br><br>
-        Whether you want higher or lower multiplicity depends on the context. However, auditing multiplicity in any AI problem is extremely important!
+        <b>Multiplicity of your predictions:</b><br>
+        Number of loans accepted by the majority, but rejected by you: {uniq_count_rej}<br>
+        Number of loans rejected by the majority, but accepted by you: {uniq_count_acc}<br><br>
+        <b>Reasons you might want lower multiplicity:</b><br>
+        To take lower risks and to avoid facing a justification crisis, i.e., 
+        not able to explain why you rejected an applicant who would have been accepted by most other models.<br><br>
+        <b>Reasons you might want higher multiplicity:</b><br>
+        To avoid competiting for the same loan applicants with others.<br> 
+        To give a chance to unique applicants and deal with the concerns of homogenization.<br><br>
     """)