FUGNN: Harmonizing Fairness and Utility in Graph Neural Networks

TL;DR

FUGNN introduces a spectral graph learning method that balances fairness and utility in GNNs by spectrum truncation and eigenvector optimization, backed by theoretical analysis and experiments on real datasets.

Contribution

This work presents FUGNN, a novel spectral GNN approach that harmonizes fairness and utility through spectrum truncation and eigenvector optimization, guided by spectral graph theory.

Findings

FUGNN outperforms baseline methods on six real-world datasets.

Spectral truncation reduces sensitive feature impact while maintaining utility.

Eigenvector optimization via transformer enhances node representation quality.

Abstract

Fairness-aware Graph Neural Networks (GNNs) often face a challenging trade-off, where prioritizing fairness may require compromising utility. In this work, we re-examine fairness through the lens of spectral graph theory, aiming to reconcile fairness and utility within the framework of spectral graph learning. We explore the correlation between sensitive features and spectrum in GNNs, using theoretical analysis to delineate the similarity between original sensitive features and those after convolution under different spectra. Our analysis reveals a reduction in the impact of similarity when the eigenvectors associated with the largest magnitude eigenvalue exhibit directional similarity. Based on these theoretical insights, we propose FUGNN, a novel spectral graph learning approach that harmonizes the conflict between fairness and utility. FUGNN ensures algorithmic fairness and utility by truncating the spectrum and optimizing eigenvector distribution during the encoding process. The fairness-aware eigenvector selection reduces the impact of convolution on sensitive features while concurrently minimizing the sacrifice of utility. FUGNN further optimizes the distribution of eigenvectors through a transformer architecture. By incorporating the optimized spectrum into the graph convolution network, FUGNN effectively learns node representations. Experiments on six real-world datasets demonstrate the superiority of FUGNN over baseline methods. The codes are available at https://github.com/yushuowiki/FUGNN.