Uncovering the Structural Fairness in Graph Contrastive Learning

TL;DR

This paper investigates the structural fairness of graph contrastive learning (GCL), revealing its inherent fairness advantages over GCN and proposing a novel augmentation method GRADE to further improve fairness for low-degree nodes.

Contribution

The paper provides a theoretical analysis of GCL's fairness properties and introduces GRADE, a new augmentation technique to enhance fairness for low-degree nodes.

Findings

GCL representations are inherently fairer to degree bias than GCN.

GCL's fairness is due to intra-community concentration and inter-community scatter.

GRAde improves fairness for low-degree nodes in experiments.

Abstract

Recent studies show that graph convolutional network (GCN) often performs worse for low-degree nodes, exhibiting the so-called structural unfairness for graphs with long-tailed degree distributions prevalent in the real world. Graph contrastive learning (GCL), which marries the power of GCN and contrastive learning, has emerged as a promising self-supervised approach for learning node representations. How does GCL behave in terms of structural fairness? Surprisingly, we find that representations obtained by GCL methods are already fairer to degree bias than those learned by GCN. We theoretically show that this fairness stems from intra-community concentration and inter-community scatter properties of GCL, resulting in a much clear community structure to drive low-degree nodes away from the community boundary. Based on our theoretical analysis, we further devise a novel graph augmentation method, called GRAph contrastive learning for DEgree bias (GRADE), which applies different strategies to low- and high-degree nodes. Extensive experiments on various benchmarks and evaluation protocols validate the effectiveness of the proposed method.