[2603.26476] Shapley meets Rawls: an integrated framework for measuring and explaining unfairness

Computer Science > Machine Learning arXiv:2603.26476 (cs) [Submitted on 27 Mar 2026] Title:Shapley meets Rawls: an integrated framework for measuring and explaining unfairness Authors:Fadoua Amri-Jouidel, Emmanuel Kemel, Stéphane Mussard View a PDF of the paper titled Shapley meets Rawls: an integrated framework for measuring and explaining unfairness, by Fadoua Amri-Jouidel and Emmanuel Kemel and St\'ephane Mussard View PDF HTML (experimental) Abstract:Explainability and fairness have mainly been considered separately, with recent exceptions trying the explain the sources of unfairness. This paper shows that the Shapley value can be used to both define and explain unfairness, under standard group fairness criteria. This offers an integrated framework to estimate and derive inference on unfairness as-well-as the features that contribute to it. Our framework can also be extended from Shapley values to the family of Efficient-Symmetric-Linear (ESL) values, some of which offer more robust definitions of fairness, and shorter computation times. An illustration is run on the Census Income dataset from the UCI Machine Learning Repository. Our approach shows that ``Age", ``Number of hours" and ``Marital status" generate gender unfairness, using shorter computation time than traditional Bootstrap tests. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2603.26476 [cs.LG]   (or arXiv:2603.26476v1 [cs.LG] for this version)   https://doi.org/10.48550/arXiv.2603.26476 Focus to learn m...