A data scientist proposes using 'equal accuracy' as the sole fairness metric for a healthcare AI triage system. Why is this approach INSUFFICIENT from a fairness perspective?
Select an answer to reveal the explanation.
Short Explanation
You can’t rely on equal overall accuracy alone because it can hide very different false negative rates across groups. In healthcare, a false negative can mean a missed diagnosis, so one group can suffer disproportionate harm even when accuracy looks equal.
Full Explanation
Equal overall accuracy can mask highly unequal false negative rates across demographic groups. In healthcare, false negatives mean missed diagnoses, and if one group has a significantly higher false negative rate, that group experiences disproportionate harm even when overall accuracy is equal. The multi-class computation objection is incorrect because equal accuracy can be computed for any classification task. The large-dataset impracticality objection is not a fundamental limitation. The legal-standard objection is factually incorrect.