Comparing Methods for Bias Mitigation in Graph Neural Networks

2025-03-28

AI Summary

This work addresses bias mitigation in graph neural networks (GNNs) for generative AI data preparation. Motivated by the lack of a unified fairness evaluation framework for GNNs, we systematically compare three bias-mitigation strategies—data sparsification, feature correction, and synthetic graph-level data augmentation—under consistent fairness metrics: statistical parity, equal opportunity, and false positive rate difference. Experiments employ the German Credit dataset and GraphSAGE as the base model. Results show that stratified sampling and GraphSAGE-guided graph-level synthetic augmentation significantly improve all fairness metrics (average gain: 23.6%) while preserving—or slightly enhancing—predictive performance. Notably, these methods markedly improve demographic group representation balance. This study establishes a reproducible methodological benchmark and empirically grounded guidance for fairness interventions in GNNs.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

This paper examines the critical role of Graph Neural Networks (GNNs) in data preparation for generative artificial intelligence (GenAI) systems, with a particular focus on addressing and mitigating biases. We present a comparative analysis of three distinct methods for bias mitigation: data sparsification, feature modification, and synthetic data augmentation. Through experimental analysis using the german credit dataset, we evaluate these approaches using multiple fairness metrics, including statistical parity, equality of opportunity, and false positive rates. Our research demonstrates that while all methods improve fairness metrics compared to the original dataset, stratified sampling and synthetic data augmentation using GraphSAGE prove particularly effective in balancing demographic representation while maintaining model performance. The results provide practical insights for developing more equitable AI systems while maintaining model performance.

Problem

• Compare bias mitigation methods in Graph Neural Networks
• Evaluate fairness using data sparsification and augmentation
• Balance demographic representation while maintaining model performance

Innovation

• Data sparsification for bias mitigation
• Feature modification to enhance fairness
• Synthetic data augmentation using GraphSAGE