An administrator notices that a pilot model performs well on average but produces systematically worse outcomes for one customer segment. Which pillar has been violated, and what should be checked first?
Select an answer to reveal the explanation.
Short Explanation
When a model performs well on average but systematically worse for one customer segment, you're seeing the textbook fairness symptom. Go straight to the training data, feature choices, and thresholds that produced it.
Full Explanation
The correct answer is the fairness option: systematically worse outcomes for a segment point to fairness, and the diagnostic starting point is the data, features, and decision thresholds that could be driving the disparity. The robustness option is incorrect because corrupted inputs would typically show up as instability across the board, not a consistent segment-level pattern. The transparency option is incorrect because outdated documentation would not itself produce worse outcomes for a segment. The accountability option is incorrect because who approved the rollout does not explain why the disparity exists in the first place. For the exam, pair each pillar with the concrete diagnostic step it implies, not just its definition.