Exploring Concept Subspace for Self-explainable Text-Attributed Graph Learning

TL;DR

This paper proposes Graph Concept Bottleneck (GCB), a self-explainable graph learning method that uses meaningful concepts for interpretation, achieving interpretability and robustness comparable to black-box models.

Contribution

Introduces GCB, a novel concept bottleneck approach for interpretable text-attributed graph learning, enhancing explanation quality and robustness.

Findings

GCB achieves accuracy comparable to black-box GNNs.

GCB provides more faithful and concise explanations.

GCB outperforms under distribution shifts and data perturbations.

Abstract

We introduce Graph Concept Bottleneck (GCB) as a new paradigm for self-explainable text-attributed graph learning. GCB maps graphs into a subspace, concept bottleneck, where each concept is a meaningful phrase, and predictions are made based on the activation of these concepts. Unlike existing interpretable graph learning methods that primarily rely on subgraphs as explanations, the concept bottleneck provides a new form of interpretation. To refine the concept space, we apply the information bottleneck principle to focus on the most relevant concepts. This not only yields more concise and faithful explanations but also explicitly guides the model to "think" toward the correct decision. We empirically show that GCB achieves intrinsic interpretability with accuracy on par with black-box Graph Neural Networks. Moreover, it delivers better performance under distribution shifts and data perturbations, showing improved robustness and generalizability, benefitting from concept-guided prediction.