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

2026-04-13

AI Summary

This work proposes the Graph Concept Bottleneck (GCB) paradigm to address the limitations of existing graph learning methods, which often rely on subgraph-based explanations lacking semantic clarity and fidelity. GCB uniquely maps attributed graphs into a concept subspace composed of human-interpretable semantic phrases, enabling predictions through concept activations. By integrating the information bottleneck principle, the framework identifies and retains only the most informative concepts to guide decision-making. The approach achieves intrinsic interpretability without sacrificing predictive accuracy—matching the performance of black-box graph neural networks—while overcoming the constraints of conventional subgraph explanations. Moreover, GCB demonstrates enhanced robustness and generalization under distribution shifts and data perturbations, highlighting its reliability in real-world scenarios.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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.

Problem

• self-explainable
• text-attributed graph
• concept bottleneck
• interpretability
• graph learning

Innovation

• Graph Concept Bottleneck
• concept subspace
• self-explainable graph learning
• information bottleneck
• text-attributed graphs