Model Explainability Methods (SHAP/LIME)
FRM Part 2 glossary · Current Issues
Definition
Post-hoc explainability: Shapley attributions / local linear surrogates.
Model Explainability Methods (SHAP/LIME) in plain English
Modern risk models (like ones deciding who gets a loan) can be 'black boxes' — they spit out an answer but won't tell you why, which is scary when real money and fairness are on the line. SHAP and LIME are two tricks for cracking that box open AFTER the model is built (that's what 'post-hoc' means). SHAP borrows a fair-division idea from game theory: it treats each input (income, age, credit history) like a player on a team and calculates how much each one 'contributed' to the final score. LIME does something simpler — it zooms in on one specific prediction and fits a plain, easy-to-read straight-line model just around that case, so you can see which factors mattered for THAT decision. Think of it as asking, 'Okay, but why did YOU personally get rejected?' and getting an honest, itemized receipt.
Also called: SHAP / LIME
Related FRM Part 2 terms
FRM Part 2 practice questions · FRM Part 2 flashcards · FRM Part 2 study guide