An AI audit team is testing a credit decision model for stability. They create 1,000 synthetic applicant profiles that differ from real applicants by only one feature at a time and compare predicted scores. What technique is this?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — b is correct because changing one feature at a time while holding others constant (one-at-a-time sensitivity analysis) measures how much each feature influences model output independently. This identifies unstable features that cause large score swings, which is a model robustness and reliability concern for credit models.
Full explanation below image
Full Explanation
B is correct because changing one feature at a time while holding others constant (one-at-a-time sensitivity analysis) measures how much each feature influences model output independently. This identifies unstable features that cause large score swings, which is a model robustness and reliability concern for credit models. FGSM (A) uses gradient-based adversarial perturbations, not one-feature-at-a-time changes. Counterfactual fairness testing (C) identifies the minimum change to flip a decision for a specific individual. PSI (D) compares distributional statistics, not individual prediction sensitivity.