mlsanityData to 14 Oct 2026

From Fragile to Certified: Wasserstein Audits of Group Fairness Under Distribution Shift

arXiv 2509.26241 ↗First posted 30 Sep 2025cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Group-fairness metrics (e.g., equalized odds) can vary sharply across resamples and are especially brittle under distribution shift, undermining reliable audits. We propose a Wasserstein distributionally robust framework that certifies worst-case group fairness over a ball of plausible test distributions centered at the empirical law. Our formulation unifies common group fairness notions via a generic conditional-probability functional and defines varepsilon-Wasserstein Distributional Fairness (varepsilon-WDF) as the audit target. Leveraging strong duality, we derive tractable reformulations and an efficient estimator (DRUNE) for varepsilon-WDF. We prove feasibility and consistency and establish finite-sample certification guarantees for auditing fairness, along with quantitative bounds under smoothness and margin conditions. Across standard benchmarks and classifiers, varepsilon-WDF delivers stable fairness assessments under distribution shift, providing a principled basis for auditing and certifying group fairness beyond observational data.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 33 of its 55 references that are in the map. First posted on arXiv 30 Sep 2025.

Builds on

33 of 55 references.

References · 33

  1. 2014Deep Learning Face Attributes in the Wildcited by 3,693
  2. 2018Fairness Without Demographics in Repeated Loss Minimizationcited by 364
  3. 2015Data-driven Distributionally Robust Optimization Using the Wasserstein Metric: Performance Guarantees and Tractable Reformulationscited by 291
  4. 2018A comparative study of fairness-enhancing interventions in machine learningcited by 208
  5. 2019Wasserstein Fair Classificationcited by 131
  6. 2020Fair Regression with Wasserstein Barycenterscited by 100
  7. 2019Distributionally Robust Optimization: A Reviewcited by 76
  8. 2023Recent Advances in Optimal Transport for Machine Learningcited by 66
  9. 2020Robust Fairness under Covariate Shiftcited by 58
  10. 2021Sample Selection for Fair and Robust Trainingcited by 48
  11. 2020A Distributionally Robust Approach to Fair Classificationcited by 43
  12. 2020Active Sampling for Min-Max Fairnesscited by 33
21 more
  1. 2023Arbitrariness and Social Prediction: The Confounding Role of Variance in Fair Classificationcited by 32
  2. 2020Can I Trust My Fairness Metric? Assessing Fairness with Unlabeled Data and Bayesian Inferencecited by 31
  3. 2020A Statistical Test for Probabilistic Fairnesscited by 25
  4. 2021Testing Group Fairness via Optimal Transport Projectionscited by 23
  5. 2023Re-weighting Based Group Fairness Regularization via Classwise Robust Optimizationcited by 20
  6. 2020Auditing ML Models for Individual Bias and Unfairnesscited by 18
  7. 2020Wasserstein-based fairness interpretability framework for machine learning modelscited by 14
  8. 2024Fairness in Survival Analysis with Distributionally Robust Optimizationcited by 14
  9. 2024Towards Fairness-Aware Adversarial Learningcited by 11
  10. 2021Wasserstein Robust Classification with Fairness Constraintscited by 10
  11. 2018Confidence Intervals for Testing Disparate Impact in Fair Learningcited by 9
  12. 2021Robust Fairness-aware Learning Under Sample Selection Biascited by 9
  13. 2023Fair Wasserstein Coresetscited by 8
  14. 2023FairWASP: Fast and Optimal Fair Wasserstein Pre-processingcited by 5
  15. 2024Universal generalization guarantees for Wasserstein distributionally robust modelscited by 5
  16. 2022SLIDE: a surrogate fairness constraint to ensure fairness consistencycited by 3
  17. 2024Wasserstein Projection Tests: Higher-Order Asymptotics and Connections to Empirical Likelihoodcited by 2
  18. 2025Towards Fair Class-wise Robustness: Class Optimal Distribution Adversarial Trainingcited by 2
  19. 2023Uncertainty in Fairness Assessment: Maintaining Stable Conclusions Despite Fluctuationscited by 1
  20. 2023Dr. FERMI: A Stochastic Distributionally Robust Fair Empirical Risk Minimization Frameworkcited by 1
  21. 2024A Distributionally Robust Optimisation Approach to Fair Credit Scoringcited by 1

Built on by

Not cited yet.