Group Fairness
Overview
Group fairness asks whether people in different demographic or other defined groups receive comparable predictions, errors, opportunities, or resources. Common approaches set targets such as demographic parity or equalized odds, then intervene through data preprocessing, training-time constraints, or post-processing; distributionally robust learning and privacy-aware methods address shifting or distributed data. The ideas also apply to clustering, graphs, rankings, recommendations, and resource allocation, where representation and exposure matter alongside prediction accuracy. A central tension is that fairness metrics can disagree, and improving one measure may reduce utility or shift harms, especially for small or intersecting subgroups.
Latest papers
157 papers16 papers in the last four weeks, down 11% on the four weeks before. 0.1% of all new papers.
When fairness metrics fail: A utility-based perspective on -fairness
Measuring Model-Induced Discrimination via Efficient Fairness Approximation
harmonic fairness via manifolds (HFM)'' based on distances between sets. Yet the direct calculation of distances might be too expensive to afford, reducing its practical applicability. Therefore, we devise an approximation algorithm named Approximation of distance between sets (ApproxDist)'' to facilitate accurate estimation of distances, and we further demonstrate its algorithmic effectiveness under certain reasonable assumptions. Empirical results indicate that the proposed fairness measure HFM reflects bias from both individual and group fairness aspects and that the proposed ApproxDist is effective and efficient.