🤖 AI Summary
This paper addresses representational bias against minority groups in network centrality ranking caused by missing edges. We first formally define “minority group representation” and propose a group-dependent edge-missing error model. Based on this model, we develop a statistical testing framework to detect ranking bias and design an asymptotically consistent correction method that jointly optimizes fairness and ranking accuracy. Our approach integrates centrality measures, stochastic graph modeling, and hypothesis testing theory—requiring only partial network observations. Experiments on synthetic and real-world contact networks demonstrate that the proposed method significantly improves the representational accuracy of minority groups in top-ranked positions (average gain of 32%), while preserving overall ranking quality. The core contribution is a testable and correctable theoretical and algorithmic foundation for structural fairness in graph learning, bridging fairness-aware inference with incomplete network data.
📝 Abstract
Networks, composed of nodes and their connections, are widely used to model complex relationships across various fields. Centrality metrics often inform decisions such as identifying key nodes or prioritizing resources. However, networks frequently suffer from missing or incorrect edges, which can systematically centrality-based decisions and distort the representation of certain protected groups. To address this issue, we introduce a formal definition of minority representation, measured as the proportion of minority nodes among the top-ranked nodes. We model systematic bias against minority groups by using group-dependent missing edge errors. We propose methods to estimate and detect systematic bias. Asymptotic limits of minority representation statistics are derived under canonical network models and used to correct representation of minority groups in node rankings. Simulation results demonstrate the effectiveness of our estimation, testing, and ranking correction procedures, and we apply our methods to a contact network, showcasing their practical applicability.