Procedural Fairness in Machine Learning
Ziming Wang, Changwu Huang, Xin Yao
TL;DR
This paper addresses the gap in ML fairness research by defining procedural fairness and distinguishing it from distributive fairness, proposing formal definitions for individual and group procedural fairness. It introduces GPFFAE, an FAE-based metric that measures differences in model decision logic across groups by comparing SHAP/FAE explanations with maximum mean discrepancy and permutation testing; the method requires sampling paired data points from two groups. The authors validate GPFFAE across nine datasets (synthetic plus eight real-world), analyze its relationship with distributive fairness, and show how it can identify unfair features to support targeted mitigation. They propose two mitigation strategies—removing unfair features via retraining and reducing their influence through an explanation-loss penalty—demonstrating substantial improvements in procedural fairness with only modest impacts on accuracy and occasional distributive fairness gains, highlighting practical implications for fair, responsible AI.
Abstract
Fairness in machine learning (ML) has received much attention. However, existing studies have mainly focused on the distributive fairness of ML models. The other dimension of fairness, i.e., procedural fairness, has been neglected. In this paper, we first define the procedural fairness of ML models, and then give formal definitions of individual and group procedural fairness. We propose a novel metric to evaluate the group procedural fairness of ML models, called $GPF_{FAE}$, which utilizes a widely used explainable artificial intelligence technique, namely feature attribution explanation (FAE), to capture the decision process of the ML models. We validate the effectiveness of $GPF_{FAE}$ on a synthetic dataset and eight real-world datasets. Our experiments reveal the relationship between procedural and distributive fairness of the ML model. Based on our analysis, we propose a method for identifying the features that lead to the procedural unfairness of the model and propose two methods to improve procedural fairness after identifying unfair features. Our experimental results demonstrate that we can accurately identify the features that lead to procedural unfairness in the ML model, and both of our proposed methods can significantly improve procedural fairness with a slight impact on model performance, while also improving distributive fairness.