Subgraph Filtering for Fair Graph Neural Networks

arXiv cs.LG Papers

Summary

This paper proposes Subgraph Filtering for Fair Graph Neural Networks (SF-GNN), a framework that mitigates structural bias in GNNs by filtering edges to improve fairness-accuracy trade-offs.

arXiv:2608.26437v1 Announce Type: new Abstract: Graph neural networks (GNNs) can exhibit unfair behavior even when sensitive attributes are excluded from node features, because graph topology and message passing propagate group-correlated signals under sensitive homophily. Existing fairness-aware GNN methods mainly constrain representations or prediction distributions at a global level, without explicitly controlling the local structural pathways through which biased information propagates during aggregation. We propose Subgraph Filtering for Fair Graph Neural Networks (SF-GNN), a lightweight and architecture-agnostic framework that mitigates structural bias at its source. SF-GNN identifies bias-prone edges by combining sensitive homophily with structural propagation amplifiers, including hub participation and triadic closure. It then incorporates stochastic edge filtering into each message-passing step to selectively downweight or remove these edges while preserving the remaining graph structure. Training further incorporates a statistical-parity regularizer with a warm-up schedule to stabilize optimization. Experiments on five benchmark datasets show that SF-GNN achieves consistent fairness improvements while maintaining competitive predictive performance, leading to a better fairness--accuracy trade-off than recent fairness-aware GNN baselines.
Original Article
View Cached Full Text

Cached at: 08/28/26, 09:39 AM

# Subgraph Filtering for Fair Graph Neural Networks
Source: [https://arxiv.org/abs/2608.26437](https://arxiv.org/abs/2608.26437)
[View PDF](https://arxiv.org/pdf/2608.26437)[HTML \(experimental\)](https://arxiv.org/html/2608.26437v1)

> Abstract:Graph neural networks \(GNNs\) can exhibit unfair behavior even when sensitive attributes are excluded from node features, because graph topology and message passing propagate group\-correlated signals under sensitive homophily\. Existing fairness\-aware GNN methods mainly constrain representations or prediction distributions at a global level, without explicitly controlling the local structural pathways through which biased information propagates during aggregation\. We propose Subgraph Filtering for Fair Graph Neural Networks \(SF\-GNN\), a lightweight and architecture\-agnostic framework that mitigates structural bias at its source\. SF\-GNN identifies bias\-prone edges by combining sensitive homophily with structural propagation amplifiers, including hub participation and triadic closure\. It then incorporates stochastic edge filtering into each message\-passing step to selectively downweight or remove these edges while preserving the remaining graph structure\. Training further incorporates a statistical\-parity regularizer with a warm\-up schedule to stabilize optimization\. Experiments on five benchmark datasets show that SF\-GNN achieves consistent fairness improvements while maintaining competitive predictive performance, leading to a better fairness\-\-accuracy trade\-off than recent fairness\-aware GNN baselines\.

## Submission history

From: Haohui Lu \[[view email](https://arxiv.org/show-email/87f59061/2608.26437)\] **\[v1\]**Wed, 26 Aug 2026 22:40:31 UTC \(148 KB\)

Similar Articles

Schreier-Coset Graph Rewiring

arXiv cs.LG

Introduces Schreier-Coset Graph Rewiring, a group-theoretic method to rewire graphs for GNNs, mitigating over-squashing by improving spectral gap and effective resistance. Empirical results show significant reduction in effective resistance across learning tasks.