Unbiased Graph Embedding with Biased Graph Observations

TL;DR

This paper introduces a novel approach to unbiased graph embedding by learning from a bias-free underlying graph, effectively reducing bias while maintaining embedding utility.

Contribution

It proposes a new principled framework and two methods to uncover bias-free graphs for embedding, improving fairness without sacrificing utility.

Findings

Theoretical justification supports the proposed methods.

Experimental results outperform state-of-the-art solutions.

Methods effectively reduce bias while preserving embedding quality.

Abstract

Graph embedding techniques are pivotal in real-world machine learning tasks that operate on graph-structured data, such as social recommendation and protein structure modeling. Embeddings are mostly performed on the node level for learning representations of each node. Since the formation of a graph is inevitably affected by certain sensitive node attributes, the node embeddings can inherit such sensitive information and introduce undesirable biases in downstream tasks. Most existing works impose ad-hoc constraints on the node embeddings to restrict their distributions for unbiasedness/fairness, which however compromise the utility of the resulting embeddings. In this paper, we propose a principled new way for unbiased graph embedding by learning node embeddings from an underlying bias-free graph, which is not influenced by sensitive node attributes. Motivated by this new perspective, we propose two complementary methods for uncovering such an underlying graph, with the goal of introducing minimum impact on the utility of the embeddings. Both our theoretical justification and extensive experimental comparisons against state-of-the-art solutions demonstrate the effectiveness of our proposed methods.