Grad2Fair: A Gradient-driven Approach for Graph Fairness without Demographics

2026-07-16

AI Summary

This work addresses the challenge of ensuring group fairness in graph neural networks when ground-truth demographic attributes are unavailable—a common scenario where existing methods rely on inaccurate predicted attributes and thus exhibit limited effectiveness. To overcome this limitation, the authors propose Grad2Fair, a novel fairness optimization framework that operates without any demographic information. Grad2Fair introduces GradDist, a measure based on the local modal distance of gradient distributions from misclassified nodes, to quantify bias. It then employs a gradient-driven debiasing mechanism that directly leverages these gradients to guide fairness-aware optimization. By eliminating dependence on predicted demographic attributes, Grad2Fair achieves consistently superior and more stable group fairness across multiple real-world graph datasets, significantly outperforming current baselines.

Abstract

Graph neural networks (GNNs) frequently encounter group fairness issues, often yielding biased predictions against specific demographic groups defined by sensitive attributes such as gender or race. While this challenge has motivated extensive research, most existing solutions rely on the strong assumption that demographics are fully available. To bypass this strict requirement, a few recent studies have attempted to use predicted demographics as proxies to enforce fairness constraints. However, predicted demographics may be inaccurate, resulting in the failure to improve fairness. In this work, we investigate the problem of graph fairness without demographic information and avoid the utilization of predicted demographics. Motivated by our observation that the gradient distributions of misclassified nodes implicitly encode demographic information, we first propose GradDist, a gradient-based metric that quantifies bias by measuring the distance between local modes within these distributions. To mitigate this bias, we propose Gradient-to-Fairness (Grad2Fair), a gradient-guided approach for group fairness without demographics. Due to the potential demographics in gradients, Grad2Fair directly leverages gradients to debias and eliminates demographic prediction, thereby enabling stable fairness performance. Experiments on several real-world datasets demonstrate the effectiveness of Grad2Fair, as evidenced by superior performance over baselines in most cases. Our code is available at https://github.com/ZzoomD/Grad2Fair.

Problem

• graph fairness
• demographic-free fairness
• group fairness
• graph neural networks
• sensitive attributes

Innovation

• gradient-driven fairness
• graph neural networks
• demographic-free fairness
• bias mitigation
• Grad2Fair