Overview

Intersectional fairness examines how model outcomes and errors differ for people at the overlap of attributes such as race, gender, age, or health status, where single-attribute averages can hide harms. Common approaches include subgroup audits with metrics such as demographic parity and equalized odds, fairness-constrained optimization, representation learning, and preprocessing or postprocessing interventions. These methods are evaluated in settings including healthcare prediction, face and speech analysis, language models, recommendation, and clustering, with attention to both predictive utility and subgroup outcomes. A central challenge is that fine-grained groups may have sparse data, while fairness metrics and mitigation effects can conflict or vary across contexts, so interventions need careful, repeated evaluation.

Latest papers

31 papers6 papers in the last four weeks, up 100% on the four weeks before. 0.0% of all new papers.

Jun 28, 2025

How Reliable are Fairness Audits with Unreliable Data?

Fairness audits are a key component of responsible machine-learning deployment. Yet, audit-recommendation reliability under incomplete protected-label access is still poorly understood. In this work, we focused on protected-label missingness in fairness mitigation audits. We introduced a seed-calibrated stress test to separate missingness effects from seed-to-seed movement already present under complete labels. Across ACS/Folktables tasks, missingness settings that retain some protected labels usually do not move selected mitigation methods beyond a complete-label seed-to-seed baseline. At 00% protected-label access, candidates collapse to an empirical-risk-minimization baseline and deterministic tie-breaking rather than revealing a broad missingness effect. We also found that threshold optimization can turn fairness gains on a single protected axis into intersectional harm above a seed baseline, and this threshold-optimizer finding persists under random-forest validation. Overall, our results highlight that protected-label missingness should be reported with seed-null calibration, candidate-set context, and intersectional consequences before it is treated as evidence of audit fragility.