Towards a Unified Framework for Fair and Stable Graph Representation Learning

As the representations output by Graph Neural Networks (GNNs) are employed in real-world applications, it becomes important to that these representations are fair and stable. In this work, we a key connection between counterfactual fairness and stability and it to propose a novel framework, NIFTY (uNIfying Fairness and), which can be used with any GNN to learn fair and stable. We introduce a novel objective function that simultaneously for fairness and stability and develop a layer-wise weight using the Lipschitz constant to enhance neural message passing in. In doing so, we enforce fairness and stability both in the objective as well as in the GNN architecture. Further, we show theoretically our layer-wise weight normalization promotes counterfactual fairness and in the resulting representations. We introduce three new graph comprising of high-stakes decisions in criminal justice and financial domains. Extensive experimentation with the above datasets demonstrates efficacy of our framework.