Promoting Fairness in Link Prediction with Graph Enhancement

TL;DR

This work tackles fairness in link prediction by shifting from debiasing predictors to learning a fairness-enhanced graph, $\mathcal{G}_f$, that promotes dyadic fairness while preserving utility. FairLink jointly optimizes a dyadic fairness regularizer with a gradient-distance utility term, $\mathcal{L}_{util}$, using a scale-sensitive distance $D = D_{\cos}+\gamma D_{\text{euc}}$, and learns $\mathcal{G}_f$ via an MLP-based adjacency model to enable effective, debiasing-free training of downstream link predictors. Empirically, FairLink achieves superior fairness metrics $Δ_{\mathit{DP}}$ and $Δ_{\mathit{EO}}$ across four large graphs, while maintaining competitive F1 and AUC scores close to predictors trained on the original graph, and demonstrates strong generalizability across GNN architectures. The approach offers a practically scalable and architecture-agnostic path to fair link prediction with reusable, fairness-enhanced graphs suitable for deployment in real-world systems.

Abstract

Link prediction is a crucial task in network analysis, but it has been shown to be prone to biased predictions, particularly when links are unfairly predicted between nodes from different sensitive groups. In this paper, we study the fair link prediction problem, which aims to ensure that the predicted link probability is independent of the sensitive attributes of the connected nodes. Existing methods typically incorporate debiasing techniques within graph embeddings to mitigate this issue. However, training on large real-world graphs is already challenging, and adding fairness constraints can further complicate the process. To overcome this challenge, we propose FairLink, a method that learns a fairness-enhanced graph to bypass the need for debiasing during the link predictor's training. FairLink maintains link prediction accuracy by ensuring that the enhanced graph follows a training trajectory similar to that of the original input graph. Meanwhile, it enhances fairness by minimizing the absolute difference in link probabilities between node pairs within the same sensitive group and those between node pairs from different sensitive groups. Our extensive experiments on multiple large-scale graphs demonstrate that FairLink not only promotes fairness but also often achieves link prediction accuracy comparable to baseline methods. Most importantly, the enhanced graph exhibits strong generalizability across different GNN architectures.

Figures (3)

• Figure 1: The overall framework of FairLink aims to learn a fairness-enhanced graph in which both fairness is promoted and utility is preserved. Initially, a synthetic graph $\mathcal{G}_{f}$ is created with the same size as the input graph $\mathcal{G}$ and random link connections. Both the input graph and the synthetic graph are then fed into a trainable link predictor. The gradient of the cross-entropy loss with respect to the predictor's parameters is computed for both $\mathcal{G}$ and $\mathcal{G}_{f}$. The optimization of $\mathcal{G}_{f}$ involves minimizing a fairness loss in conjunction with the gradient distance between $\mathcal{G}$ and $\mathcal{G}_{f}$.
• Figure 2: Fair link prediction objective in FairLink: Ensure equal probability for links between nodes from different sensitive groups and those from the same group.
• Figure 3: Trade-off between fairness and link prediction accuracy across four datasets. Results in the upper left corner, which exhibit both lower bias and higher accuracy, represent the ideal balance.