Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FairGT: A Fairness-aware Graph Transformer

Renqiang Luo, Huafei Huang, Shuo Yu, Xiuzhen Zhang, Feng Xia

TL;DR

FairGT addresses fairness issues in Graph Transformers by introducing two fairness-aware encodings: a structural topology encoding built from the adjacency matrix’s eigenvectors and a node feature encoding based on a sensitive-feature complete graph to preserve independence of sensitive attributes. The authors provide theoretical support showing that using the largest-magnitude eigenvectors and k-hop sensitive information enhances fairness while maintaining predictive power. Empirically, FairGT achieves superior fairness (lower \Delta_{\\text{SP}}) and competitive accuracy across five real-world datasets (NBA, Bail, German, Credit, Income) compared with Graph Transformers, GNNs, and fairness-aware baselines. The method offers a practical fairness extension for GTs with favorable training cost and clear ablation validation of its components.

Abstract

The design of Graph Transformers (GTs) generally neglects considerations for fairness, resulting in biased outcomes against certain sensitive subgroups. Since GTs encode graph information without relying on message-passing mechanisms, conventional fairness-aware graph learning methods cannot be directly applicable to address these issues. To tackle this challenge, we propose FairGT, a Fairness-aware Graph Transformer explicitly crafted to mitigate fairness concerns inherent in GTs. FairGT incorporates a meticulous structural feature selection strategy and a multi-hop node feature integration method, ensuring independence of sensitive features and bolstering fairness considerations. These fairness-aware graph information encodings seamlessly integrate into the Transformer framework for downstream tasks. We also prove that the proposed fair structural topology encoding with adjacency matrix eigenvector selection and multi-hop integration are theoretically effective. Empirical evaluations conducted across five real-world datasets demonstrate FairGT's superiority in fairness metrics over existing graph transformers, graph neural networks, and state-of-the-art fairness-aware graph learning approaches.

FairGT: A Fairness-aware Graph Transformer

TL;DR

FairGT addresses fairness issues in Graph Transformers by introducing two fairness-aware encodings: a structural topology encoding built from the adjacency matrix’s eigenvectors and a node feature encoding based on a sensitive-feature complete graph to preserve independence of sensitive attributes. The authors provide theoretical support showing that using the largest-magnitude eigenvectors and k-hop sensitive information enhances fairness while maintaining predictive power. Empirically, FairGT achieves superior fairness (lower \Delta_{\\text{SP}}) and competitive accuracy across five real-world datasets (NBA, Bail, German, Credit, Income) compared with Graph Transformers, GNNs, and fairness-aware baselines. The method offers a practical fairness extension for GTs with favorable training cost and clear ablation validation of its components.

Abstract

The design of Graph Transformers (GTs) generally neglects considerations for fairness, resulting in biased outcomes against certain sensitive subgroups. Since GTs encode graph information without relying on message-passing mechanisms, conventional fairness-aware graph learning methods cannot be directly applicable to address these issues. To tackle this challenge, we propose FairGT, a Fairness-aware Graph Transformer explicitly crafted to mitigate fairness concerns inherent in GTs. FairGT incorporates a meticulous structural feature selection strategy and a multi-hop node feature integration method, ensuring independence of sensitive features and bolstering fairness considerations. These fairness-aware graph information encodings seamlessly integrate into the Transformer framework for downstream tasks. We also prove that the proposed fair structural topology encoding with adjacency matrix eigenvector selection and multi-hop integration are theoretically effective. Empirical evaluations conducted across five real-world datasets demonstrate FairGT's superiority in fairness metrics over existing graph transformers, graph neural networks, and state-of-the-art fairness-aware graph learning approaches.
Paper Structure (22 sections, 25 equations, 3 figures, 7 tables)

This paper contains 22 sections, 25 equations, 3 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Graph Transformers
  4. Fairness-aware GNNs
  5. Preliminaries
  6. Notations
  7. Fairness Evaluation Metrics
  8. Transformer
  9. The Design of FairGT
  10. Theoretical Findings Underpinning FairGT
  11. Structural Topology Encoding
  12. Node Feature Encoding
  13. FairGT for Node Classification
  14. Complexity of FairGT
  15. Experiments
  16. ...and 7 more sections

Figures (3)

  • Figure 1: The illustration of FairGT.
  • Figure 2: The accuracy and $\Delta_{\text{SP}}$ of FairGT w.r.t. different parameter $l$ values.
  • Figure 3: The accuracy and $\Delta_{\text{SP}}$ of FairGT w.r.t. different parameter $t$ values.