Bridging the Fairness Divide: Achieving Group and Individual Fairness in Graph Neural Networks

TL;DR

The paper tackles bias in graph neural networks by jointly addressing group fairness and individual fairness within groups. It introduces FairGI, a framework combining a similarity-based intra-group fairness objective with adversarial and covariance-based losses to promote Equal Opportunity and Statistical Parity, while keeping high predictive accuracy. A new metric for evaluating individual fairness within groups is proposed, and extensive experiments on Pokec_n, NBA, and Credit show that FairGI outperforms baselines on both group and intra-group fairness and achieves strong population-level individual fairness. The work provides a practical path toward fair graph learning in applications like social networks and recommendation systems.

Abstract

Graph neural networks (GNNs) have emerged as a powerful tool for analyzing and learning from complex data structured as graphs, demonstrating remarkable effectiveness in various applications, such as social network analysis, recommendation systems, and drug discovery. However, despite their impressive performance, the fairness problem has increasingly gained attention as a crucial aspect to consider. Existing research in graph learning focuses on either group fairness or individual fairness. However, since each concept provides unique insights into fairness from distinct perspectives, integrating them into a fair graph neural network system is crucial. To the best of our knowledge, no study has yet to comprehensively tackle both individual and group fairness simultaneously. In this paper, we propose a new concept of individual fairness within groups and a novel framework named Fairness for Group and Individual (FairGI), which considers both group fairness and individual fairness within groups in the context of graph learning. FairGI employs the similarity matrix of individuals to achieve individual fairness within groups, while leveraging adversarial learning to address group fairness in terms of both Equal Opportunity and Statistical Parity. The experimental results demonstrate that our approach not only outperforms other state-of-the-art models in terms of group fairness and individual fairness within groups, but also exhibits excellent performance in population-level individual fairness, while maintaining comparable prediction accuracy.

Figures (4)

• Figure 1: A toy example for student admission model with gender as the sensitive attribute. The red color represents female students, which is also the protected group. The blue color denotes male students. Fig. \ref{['fig1']} (a) shows the machine learning model can only guarantee the group fairness. Fig. \ref{['fig1']} (b) shows the machine learning model can guarantee both group and individual fairness.
• Figure 2: Overview of FairGI. Our method comprises three main parts, i.e., an individual fairness module, an group fairness module and a GNN classifier for node prediction.
• Figure 3: Comparison of our method, our method without loss function of individual fairness within groups, our method without the optimization for EO.
• Figure 4: Comparison of our method, our method without loss function of individual fairness within groups, our method without the optimization for EO.

Theorems & Definitions (10)

• Definition 1
• Definition 2
• Definition 3
• Definition 4
• Proposition 5
• proof
• Theorem 6
• proof
• Theorem 7
• proof