Which principle BEST guides an auditor assessing whether an AI system's outputs are sufficiently explainable for the use case?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you should match explainability to the stakes and to what affected parties need to understand and contest. Low-risk content recommendations don't need the same depth as a housing-eligibility model, and demanding closed-form derivations for every prediction isn't realistic.
Full Explanation
The correct principle is that explainability requirements should be proportional to the decision's stakes and the need of affected parties to understand and contest outcomes. Low-risk content recommendations need less explainability than a model determining housing eligibility. Requiring closed-form mathematical derivations for every prediction is unrealistic for modern ML models. Limiting explainability to personal-data models under GDPR narrows it to one legal basis. Calling neural networks inherently unexplainable is an overstatement, since post-hoc techniques exist.