GESD: Beyond Outcome-Oriented Fairness
The paper titled 'GESD: Beyond Outcome-Oriented Fairness' introduces a new fairness metric for machine learning called Group-level Explanation Stability Disparity (GESD). This metric aims to address the limitations of existing outcome-oriented fairness metrics by focusing on the stability and sensitivity of model explanations across different subgroups. The authors demonstrate that GESD can effectively capture discrepancies in explanation quality and improve fairness in predictive modeling through a multi-objective optimization framework.
- ▪GESD is a procedural-oriented fairness metric that measures disparities in model explanations across subgroups.
- ▪The metric is explainer-agnostic and model-agnostic, extending fairness analyses to explainability.
- ▪Empirical results show that GESD captures group-wise discrepancies in explanation quality.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.15295 (cs) [Submitted on 14 May 2026] Title:GESD: Beyond Outcome-Oriented Fairness Authors:Gideon Popoola, John Sheppard View a PDF of the paper titled GESD: Beyond Outcome-Oriented Fairness, by Gideon Popoola and 1 other authors View PDF HTML (experimental) Abstract:Machine learning (ML) algorithms are increasingly deployed in high-stakes decision-making domains such as loan approvals, hiring, and recidivism predictions. While existing fairness metrics (e.g., statistical parity, equal opportunity) effectively quantify outcome-oriented disparities, they offer limited insight into the procedure or explanation behind biased decisions.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
