Model-Agnostic Fairness Regularization for GNNs with Incomplete Sensitive Information

2025-11-27

AI Summary

Existing fairness-aware graph neural networks (GNNs) commonly assume full observability of sensitive attributes (e.g., race, gender) for all nodes during training—a strong assumption often violated in practice due to privacy constraints and data incompleteness. Method: We propose a model-agnostic, differentiable fairness regularization framework that, for the first time, unifies equal opportunity and statistical parity as jointly optimizable objectives under partial sensitive attribute availability. Leveraging message passing, our method jointly exploits graph topology and node features to dynamically impute missing sensitive attributes and mitigate bias during training. Contribution/Results: Evaluated on five real-world datasets, our approach consistently improves fairness metrics—including ΔEO and ΔSP—while preserving or slightly enhancing classification accuracy. It thus achieves a superior fairness–accuracy trade-off compared to state-of-the-art baselines, demonstrating robustness and practical applicability in realistic, privacy-sensitive settings.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Graph-based Machine Learning
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Graph Neural Networks (GNNs) have demonstrated exceptional efficacy in relational learning tasks, including node classification and link prediction. However, their application raises significant fairness concerns, as GNNs can perpetuate and even amplify societal biases against protected groups defined by sensitive attributes such as race or gender. These biases are often inherent in the node features, structural topology, and message-passing mechanisms of the graph itself. A critical limitation of existing fairness-aware GNN methods is their reliance on the strong assumption that sensitive attributes are fully available for all nodes during training--a condition that poses a practical impediment due to privacy concerns and data collection constraints. To address this gap, we propose a novel, model-agnostic fairness regularization framework designed for the realistic scenario where sensitive attributes are only partially available. Our approach formalizes a fairness-aware objective function that integrates both equal opportunity and statistical parity as differentiable regularization terms. Through a comprehensive empirical evaluation across five real-world benchmark datasets, we demonstrate that the proposed method significantly mitigates bias across key fairness metrics while maintaining competitive node classification performance. Results show that our framework consistently outperforms baseline models in achieving a favorable fairness-accuracy trade-off, with minimal degradation in predictive accuracy. The datasets and source code will be publicly released at https://github.com/mtavassoli/GNN-FC.

Problem

• Addresses fairness in GNNs with incomplete sensitive attributes
• Proposes model-agnostic regularization for bias mitigation
• Ensures fairness-accuracy trade-off in node classification tasks

Innovation

• Model-agnostic fairness regularization for GNNs
• Handles incomplete sensitive attributes in graphs
• Integrates equal opportunity and statistical parity terms