A housing authority must justify to a review board why a specific application for rental assistance was denied by its scoring model. The team applies model explainability techniques to produce feature-level reasoning for that individual prediction. What does this evaluation approach provide that overall accuracy metrics do not?
Select an answer to reveal the explanation.
Short Explanation
Think of the difference between a report card and a teacher's note explaining exactly why one specific assignment got the grade it did — overall accuracy is the report card, but the review board needs the note about this one case. Explainability techniques trace which input features drove a single prediction, giving a case-specific rationale that a dataset-wide accuracy number simply can't provide. That's what actually holds up when someone asks "why this applicant, specifically?"
Full Explanation
Model explainability techniques applied at the individual-prediction level identify which input features contributed most to a specific outcome, producing a case-specific account that can be presented as the rationale behind that one decision, which is exactly what a review board needs when questioning a single denial rather than the model's behavior in aggregate. Overall accuracy metrics summarize performance across an entire dataset and say nothing about why any one particular applicant received the outcome they did, so they cannot substitute for a per-case explanation in this context. The option claiming explainability guarantees freedom from bias overstates what the technique does; explainability reveals what influenced a prediction, it does not certify that the influence pattern is fair or free of bias, which requires separate bias evaluation. The option treating an explanation as an automatic justification misassigns the human judgment role; the explanation supplies evidence for the review board to evaluate, it does not itself make the denial decision defensible without human review of whether the reasoning was appropriate. The option describing a single dataset-wide performance score restates what overall accuracy already provides, which is precisely the aggregate-level information the scenario says is insufficient for justifying one specific case. Scope note: explainability outputs should be reviewed for whether the influential features are themselves appropriate and non-discriminatory factors to rely on. Operational check: confirm the explanation's top contributing features for the denied application align with policy-eligible criteria, not a proxy for a protected characteristic.