A financial regulator asks a bank to explain why its AI credit model denied a specific customer's application. The model is a deep neural network with 50 layers. Which approach BEST satisfies the regulator's request while remaining technically honest?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you need to show the regulator which inputs pushed this specific applicant toward denial. SHAP values do that by breaking the prediction into feature contributions, and you can translate those contributions into plain language without pretending the deep network is simple.
Full Explanation
The SHAP values approach is correct because SHAP provides individualized feature-level explanations for specific predictions, translating each feature's contribution to the denial into plain language. That satisfies regulatory explanation requirements while staying technically grounded. The raw weight matrices approach is wrong because those weights are meaningless to regulators. The unexplainable-neural-networks claim is wrong because post-hoc explainability methods exist. The retrospective linear model approach is wrong because it is post-hoc rationalization, not an honest explanation of the actual model's decision.