🤖 AI Summary
Existing research on fairness in graph link prediction has largely focused on homophily bias, overlooking broader topological biases and thereby limiting the generalizability of fairness interventions. This work proposes the first benchmark framework for fair link prediction that systematically incorporates non-homophilous topological biases. The framework formalizes a taxonomy of topological biases, introduces a controllable graph generation mechanism to instantiate diverse structural settings, and comprehensively evaluates both classical and fairness-aware models across varied topologies. Experimental results reveal that current methods are highly sensitive to non-homophilous structural biases, highlighting a critical gap in existing approaches. By establishing a new paradigm for structure-aware fairness evaluation in graph learning, this study provides both empirical insights and methodological foundations for future research in equitable graph representation learning.
📝 Abstract
Graph link prediction (LP) plays a critical role in socially impactful applications, such as job recommendation and friendship formation. Ensuring fairness in this task is thus essential. While many fairness-aware methods manipulate graph structures to mitigate prediction disparities, the topological biases inherent to social graph structures remain poorly understood and are often reduced to homophily alone. This undermines the generalization potential of fairness interventions and limits their applicability across diverse network topologies. In this work, we propose a novel benchmarking framework for fair LP, centered on the structural biases of the underlying graphs. We begin by reviewing and formalizing a broad taxonomy of topological bias measures relevant to fairness in graphs. In parallel, we introduce a flexible graph generation method that simultaneously ensures fidelity to real-world graph patterns and enables controlled variation across a wide spectrum of structural biases. We apply this framework to evaluate both classical and fairness-aware LP models across multiple use cases. Our results provide a fine-grained empirical analysis of the interactions between predictive fairness and structural biases. This new perspective reveals the sensitivity of fairness interventions to beyond-homophily biases and underscores the need for structurally grounded fairness evaluations in graph learning.