Evaluating Intersectional Fairness across Clinical Machine Learning Use Cases using Fairlogue and the All of Us Research Program
A new research paper introduces FairLogue, a toolkit designed to audit intersectional fairness in clinical machine learning models. The study applied FairLogue to two existing models using the All of Us dataset, evaluating disparities across combined demographic groups (race, gender, and their intersections). While intersectional analysis revealed larger disparities than single-axis evaluations, counterfactual diagnostics suggested these were largely comparable to random group membership, highlighting the necessity of intersectional auditing for deeper bias insights. AI
IMPACT Highlights the need for more nuanced fairness evaluations in clinical AI, potentially influencing future model development and auditing practices.