Collapsibility of Performance Metrics in Clinical Predictive AI
Background: Population level assessments of predictive artificial intelligence (AI) can conceal performance disparities across subgroups. Fairness evaluations commonly rely on performance analyses across subgroups. However, some performance metrics are non-collapsible, meaning that the overall population performance value does not equal the weighted average of subgroup specific values. Objective:…
We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.
