Towards a Unified Framework for Fair and Stable Graph Representation Learning

TL;DR

This paper introduces NIFTY, a unified framework for GNNs that enhances fairness and stability in graph representations by combining a novel objective function with layer-wise weight normalization, supported by theoretical guarantees and new datasets.

Contribution

The paper proposes a novel framework, NIFTY, that unifies fairness and stability in GNNs through a new objective and weight normalization, with theoretical and empirical validation.

Findings

NIFTY improves fairness and stability in GNN representations.

Layer-wise weight normalization promotes counterfactual fairness.

Experimental results on new datasets demonstrate effectiveness.

Abstract

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