Improve explainability step

app.py

@@ -93,8 +93,11 @@ with demo:
         - a user's personal information in order to evaluate his/her credit card eligibility;
         - the user’s bank account history, which provides any type of information on the user's 
         banking information relevant to the decision (here, we consider duration of account);
-        - and credit scoring agency information, which represents any other information (here, employment 
-        history) that could provide additional insight relevant to the decision.
+        - and credit scoring agency information, which represents any other information (here, 
+        employment history) that could provide additional insight relevant to the decision.
+        
+        Please always encrypt and send the values (through the buttons on the right) once updated
+        before running the FHE inference.
         """
     )
 
@@ -250,7 +253,7 @@ with demo:
     # Button to send the encodings to the server using post method
     execute_fhe_button.click(run_fhe, inputs=[client_id], outputs=[fhe_execution_time])
 
-    gr.Markdown("# Client, Bank and Credit Scoring Agency setup")  
+    gr.Markdown("# Client, Bank and Credit Scoring Agency decryption")  
     gr.Markdown(
         """
         Once the server completed the inference, the encrypted output is returned to the user.
@@ -288,20 +291,23 @@ with demo:
     gr.Markdown("## Step 6 (optional): Explain the prediction.")
     gr.Markdown(
         """
-        In case the credit card is likely to be denied, the user can run a second model in order to 
-        Explain the prediction better. More specifically, this new model indicates the number of
-        additional years of employment that could be required in order to increase the chance of 
+        In case the credit card is likely to be denied, the user can ask for how many years of 
+        employment would most likely be required in order to increase the chance of getting a 
         credit card approval.
+        
         All of the above steps are combined into a single button for simplicity. The following 
-        button therefore encrypts the same inputs (except the years of employment) from all three 
-        parties, runs the new prediction in FHE and decrypts the output. 
+        button therefore encrypts the same inputs (except the years of employment, which varies) 
+        from all three parties, runs the new prediction in FHE and decrypts the output. 
+        
+        In case the following states to try a new "Years of employment" input, one can simply 
+        update the value in Step 2 and directly run Step 6 once more.  
         """
     )
     explain_button = gr.Button(
         "Encrypt the inputs, compute in FHE and decrypt the output."
     )
     explain_prediction = gr.Textbox(
-        label="Additional years of employed required.", max_lines=1, interactive=False
+        label="Additional years of employed required.", interactive=False
     )
 
     # Button to explain the prediction

backend.py

@@ -31,6 +31,10 @@ from settings import (
 
 from utils.client_server_interface import MultiInputsFHEModelClient
 
+# Define the messages associated to the predictions
+APPROVED_MESSAGE = "Credit card is likely to be approved ✅"
+DENIED_MESSAGE = "Credit card is likely to be denied ❌"
+
 # Load pre-processor instances
 with (
     PRE_PROCESSOR_USER_PATH.open('rb') as file_user, 
@@ -395,8 +399,7 @@ def get_output_and_decrypt(client_id):
             output = numpy.argmax(output_proba, axis=1).squeeze()
             
             return (
-                "Credit card is likely to be approved ✅" if output == 1 
-                else "Credit card is likely to be denied ❌",
+                APPROVED_MESSAGE if output == 1 else DENIED_MESSAGE,
                 encrypted_output_short,
             )
 
@@ -431,14 +434,12 @@ def explain_encrypt_run_decrypt(client_id, prediction_output, *inputs):
     bin_index = YEARS_EMPLOYED_BIN_NAME_TO_INDEX[years_employed]
 
     # If the bin is not the last (representing the most years of employment), we run the model in 
-    # FHE for each bins "older" than the given bin, in order. Then, we retrieve the first bin that
-    # changes the model's prediction to "approval" and display it to the user.  
+    # FHE for each bins "older" or equal to the given bin, in order. Then, we retrieve the first 
+    # bin that changes the model's prediction to "approval" and display it to the user.  
     if bin_index != len(YEARS_EMPLOYED_BINS) - 1:
-
-        output_predictions = []
         
-        # Loop over the bins "older" than the input one
-        for years_employed_bin in YEARS_EMPLOYED_BINS[bin_index+1:]:
+        # Loop over the bins starting with "older" or equal to the given bin
+        for years_employed_bin in YEARS_EMPLOYED_BINS[bin_index:]:
             
             # Send the new encrypted input
             pre_process_encrypt_send_cs_agency(client_id, years_employed_bin, employed)
@@ -449,34 +450,34 @@ def explain_encrypt_run_decrypt(client_id, prediction_output, *inputs):
             # Retrieve the new prediction
             output_prediction = get_output_and_decrypt(client_id)
 
-            is_approved = "approved" in output_prediction[0]
-            output_predictions.append(is_approved)
-        
-        # Re-send the initial credit scoring agency inputs in order to avoid unwanted conflict (as sending
-        # some inputs basically re-writes the associated file on the server side)
-        pre_process_encrypt_send_cs_agency(client_id, years_employed, employed)
-        
-        # In case the model predicted at least one approval
-        if any(output_predictions):
+            # If the bin made the model predict an approval, share it to the user 
+            if "approved" in output_prediction[0]:
 
-            # Retrieve the first bin that made the model predict an approval 
-            first_approved_prediction_index = numpy.argmax(output_predictions)
-            years_employed_bin_needed = YEARS_EMPLOYED_BINS[first_approved_prediction_index + bin_index + 1]
+                # If the approval was made using the given input, that means the user most likely 
+                # tried the bin suggested in a previous explainability run. In that case, we 
+                # confirm that the credit card is likely to be approved
+                if years_employed_bin == years_employed:
+                    return APPROVED_MESSAGE
 
-            return (
-                f"Having at least {years_employed_bin_needed} more years of employment would "
-                "increase your chance of having your credit card approved."
-            )
+                # Else, that means the users is looking for some explainability. We therefore 
+                # suggest to try the obtained bin
+                return (
+                    DENIED_MESSAGE + f" However, having at least {years_employed_bin} years of "
+                    "employment would increase your chance of having your credit card approved."
+                )
 
+        # In case no bins made the model predict an approval, explain why
         return (
-            f"Increasing the number of years of employment up to {YEARS_EMPLOYED_BINS[-1]} years "
-            "does not seem to be enough to get an approval based on the given inputs. Other inputs "
-            "like the income or the account's age might have bigger impact in this particular case."
-        )
+            DENIED_MESSAGE + " Unfortunately, increasing the number of years of employment up to "
+            f"{YEARS_EMPLOYED_BINS[-1]} years does not seem to be enough to get an approval based "
+            "on the given inputs. Other inputs like the income or the account's age might have "
+            "bigger impact in this particular case."
+        )  
 
+    # In case the user tried the "oldest" bin (but still got denied), explain why
     return (
-        f"You already have the maximum amount of years of employment ({years_employed} years). "
-        "Other inputs like the income or the account's age might have bigger impact in this "
-        "particular case."
+        DENIED_MESSAGE + " Unfortunately, you already have the maximum amount of years of "
+        f"employment ({years_employed} years). Other inputs like the income or the account's age "
+        "might have a bigger impact in this particular case."
     )