Fairness in Graph Mining: A Survey

Abstract

Graph mining algorithms have been playing a significant role in myriad fields over the years. However, despite their promising performance on various graph analytical tasks, most of these algorithms lack fairness considerations. As a consequence, they could lead to discrimination towards certain populations when exploited in human-centered applications. Recently, algorithmic fairness has been extensively studied in graph-based applications. In contrast to algorithmic fairness on independent and identically distributed (i.i.d.) data, fairness in graph mining has exclusive backgrounds, taxonomies, and fulfilling techniques. In this survey, we provide a comprehensive and up-to-date introduction of existing literature under the context of fair graph mining. Specifically, we propose a novel taxonomy of fairness notions on graphs, which sheds light on their connections and differences. We further present an organized summary of existing techniques that promote fairness in graph mining. Finally, we summarize the widely used datasets in this emerging research field and provide insights on current research challenges and open questions, aiming at encouraging cross-breeding ideas and further advances.

p2pGNN: A Decentralized Graph Neural Network for Node Classification in Peer-to-Peer Networks

“…We consider the first two of these topics as the communities whose members we aim to rediscover, for instance as if they were new keywords whose entries we needed to populate. In line with many fairness benchmarks (Dong et al, 2022), we protect recommendation results with respect one of the topics-in our case, the one with the most members-so that they obtain on average equal posteriors to the rest of recommended nodes.…”

“…Fairness has also been recently explored for the outcomes of graph neural networks (Dai and Wang, 2020;Ma et al, 2021;Dong et al, 2021;Dai and Wang, 2021;Zhang et al, 2022;Dong et al, 2022). Approaches are typically trained to produce fair predictions (e.g., recommendations, link predictions) by a) incorporating traditional notions of fairness as regularizers to differentiable relaxations of accurate node classification (e.g., cross-entropy minimization), b) imposing similar fairness constraints on the training process, or c) rebalancing algorithmic components to debias the inference process.…”

Learning causality with graphs

“…Generally, a fair machine learning-based predictor aims to mitigate the discrimination of model prediction against certain demographic subpopulations regarding sensitive attributes such as race, gender, and age. Recently, fairness on graph mining is also an emerging field (Dong, Ma, Chen, and Li 2022). Among existing notions of fairness, counterfactual fairness (Kusner et al 2017) measures the fairness of predictors from a causal perspective by comparing the predictions of each individual from the original data and the counterfactuals in which the sensitive attributes of this individual had been modified to a different value.…”

Graph-Free Knowledge Distillation for Graph Neural Networks

“…Group fairness requires that each subgroup should receive their fair share of interest according to the output GNN predictions [23]. Various explorations have been made towards achieving a higher level of group fairness for GNNs [7]. Decoupling the output predictions from sensitive attributes via adversarial learning is one of the most popular approaches among existing works [31,3].…”