An organization's model risk management team is performing validati…

Question

An organization's model risk management team is performing validation on a neural network that classifies transaction fraud. The team cannot explain why specific transactions are flagged. Which validation technique should the team employ to address explainability requirements under SR 11-7?

Accepted Answer

Apply post-hoc explainability methods such as SHAP values to attribute feature importance to individual predictions — Here's the deal — b is correct because post-hoc explainability methods (SHAP, LIME, or similar) can provide feature attribution for individual predictions from complex models, satisfying SR 11-7 requirements for understanding model outputs without requiring the model to be inherently interpretable. A is unnecessary and may sacrifice predictive power.

Answer

Rewrite the neural network as a logistic regression model to achieve full transparency

Answer

Validate only on aggregate population metrics and waive individual-level explainability requirements

Answer

Document that the model is a black box and proceed with deployment after management sign-off

An organization's model risk management team is performing validation on a neural network that classifies transaction fraud. The team cannot explain why specific transactions are flagged. Which validation technique should the team employ to address explainability requirements under SR 11-7?

Short Explanation and Infographic

Full Explanation