The emergence of Graph Neural Networks (GNNs) has greatly advanced the development of recommendation systems. Recently, many researchers have leveraged GNN-based models to learn fair representations for users and items. However, current GNN-based models suffer from biased user–item interaction data, which negatively impacts recommendation fairness. Although there have been several studies employing adversarial learning to mitigate this issue in recommendation systems, they mostly focus on modifying the model training approach with fairness regularization and neglect direct intervention of biased interaction. In contrast to these models, this article introduces a novel perspective by directly intervening in observed interactions to generate a counterfactual graph (called FairGap) that is not influenced by sensitive node attributes, enabling us to learn fair representations for users and items easily. We design FairGap to answer the key counterfactual question: “Would interactions with an item remain unchanged if a user’s sensitive attributes were concealed?”. We also provide theoretical proofs to show that our learning strategy via the counterfactual graph is unbiased in expectation. Moreover, we propose a fairness-enhancing mechanism to continuously improve user fairness in the graph-based recommendation. Extensive experimental results against state-of-the-art competitors and base models on three real-world datasets validate the effectiveness of our proposed model.
16 June 2025 in IEEE Transactions on Knowledge and Data Engineering
As trustworthy AI continues to advance, the fairness issue in recommendations has received increasing attention. A recommender system is considered unfair when it produces unequal outcomes for different user groups based on user-sensitive attributes (e.g., age, gender). Some researchers have proposed data augmentation-based methods aiming at alleviating user-level unfairness by altering the skewed distribution of training data among various user groups. Despite yielding promising results, they often rely on fairness-related assumptions that may not align with reality, potentially reducing the data quality and negatively affecting model effectiveness. To tackle this issue, in this paper, we study how to implement high-quality data augmentation to improve recommendation fairness. Specifically, we propose FairDgcl, a dynamic graph adversarial contrastive learning framework aiming at improving fairness in recommender system. First, FairDgcl develops an adversarial contrastive network with a view generator and a view discriminator to learn generating fair augmentation strategies in an adversarial style. Then, we propose two dynamic, learnable models to generate contrastive views within contrastive learning framework, which automatically fine-tune the augmentation strategies. Meanwhile, we theoretically show that FairDgcl can simultaneously generate enhanced representations that possess both fairness and accuracy. Lastly, comprehensive experiments conducted on four datasets demonstrate the effectiveness of the proposed FairDgcl.
Collaborative Filtering (CF) methods commonly use negative sampling to improve preference learning by contrasting observed interactions with unobserved items. While effective, conventional practice implicitly treats all unclicked items as equally informative, regardless of their semantic group (e.g., genre or category). This overlooks a critical limitation: models may exploit coarse group distinctions rather than genuine fine-grained preferences, thereby inducing exposure imbalances across item groups. Such imbalances constitute a violation of item-side fairness, which seeks equitable exposure and evaluation for items from different semantic groups. When negative samples are drawn predominantly from groups semantically distant from a user’s positives, the learning signal becomes biased and comparisons unfair. We therefore revisit negative sampling through the lens of item-side fairness and argue that genuine fairness requires context-aware sampling that ensures like-for-like comparisons within each semantic group. To this end, we introduce FairNS, a diffusion-based sampling framework that generates negative samples within the same semantic group as the user’s positives, encouraging fair intra-group contrasts that respect group integrity. By centering training on these intra-group comparisons, FairNS mitigates cross-group bias and enables the recommender to learn more precise user preferences. FairNS is optimized via a bi-level objective that jointly refines the sampling mechanism and the recommendation model. Experiments on three benchmark datasets show that FairNS achieves a favorable fairness–accuracy tradeoff.
11 November 2026 in IEEE Transactions on Knowledge and Data Engineering