FairGap: Fairness-Aware Recommendation via Generating Counterfactual Graph

• DOI: 10.1145/3638352
• Publication Date: 2023-12-22T11:31:30Z
• Authors: Wei Chen; Yiqing Wu; Zhao Zhang; Fuzhen Zhuang; Zhongshi He; Ruobing Xie; Feng Xia
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

The emergence of Graph Neural Networks (GNNs) has greatly advanced the development recommendation systems. Recently, many researchers have leveraged GNN-based models to learn fair representations for users and items. However, current suffer from biased user–item interaction data, which negatively impacts fairness. Although there been several studies employing adversarial learning mitigate this issue in systems, they mostly focus on modifying model training approach with fairness regularization neglect direct intervention interaction. In contrast these models, article introduces a novel perspective by directly intervening observed interactions generate counterfactual graph (called FairGap) that is not influenced sensitive node attributes, enabling us items easily. We design FairGap answer key question: “Would an item remain unchanged if user’s attributes were concealed?”. also provide theoretical proofs show our strategy via unbiased expectation. Moreover, we propose fairness-enhancing mechanism continuously improve user graph-based recommendation. Extensive experimental results against state-of-the-art competitors base three real-world datasets validate effectiveness proposed model.