Fairness-Aware Graph Filter Design

TL;DR

This paper introduces a fair graph filter to reduce bias in graph-based machine learning tasks, demonstrating improved fairness and stability without sacrificing utility on real-world networks.

Contribution

It proposes a novel fair graph filter design based on bias analysis, optimizing bias mitigation while maintaining utility in graph learning tasks.

Findings

Effective bias reduction in node classification

Comparable utility to baseline algorithms

Enhanced stability in predictions

Abstract

Graphs are mathematical tools that can be used to represent complex real-world systems, such as financial markets and social networks. Hence, machine learning (ML) over graphs has attracted significant attention recently. However, it has been demonstrated that ML over graphs amplifies the already existing bias towards certain under-represented groups in various decision-making problems due to the information aggregation over biased graph structures. Faced with this challenge, in this paper, we design a fair graph filter that can be employed in a versatile manner for graph-based learning tasks. The design of the proposed filter is based on a bias analysis and its optimality in mitigating bias compared to its fairness-agnostic counterpart is established. Experiments on real-world networks for node classification demonstrate the efficacy of the proposed filter design in mitigating bias, while attaining similar utility and better stability compared to baseline algorithms.