Quantifying Explanation Quality in Graph Neural Networks using Out-of-Distribution Generalization

2026-02-07

AI Summary

This work addresses the limitation of existing evaluation metrics for post-hoc explanations of graph neural networks (GNNs), which struggle to assess whether explanations capture true causal mechanisms. To this end, the authors propose the Explanation-Generalization Score (EGS), a novel framework that leverages out-of-distribution (OOD) generalization as a proxy for causal validity. Specifically, EGS trains a GNN solely on the identified explanatory subgraph and evaluates its predictive stability under OOD shifts, thereby establishing a new paradigm for explanation quality grounded in feature invariance. Empirical results demonstrate that EGS effectively discriminates among explainers based on their ability to recover causal substructures, significantly outperforming conventional fidelity-based metrics on both synthetic and real-world datasets.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Causality
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Evaluating the quality of post-hoc explanations for Graph Neural Networks (GNNs) remains a significant challenge. While recent years have seen an increasing development of explainability methods, current evaluation metrics (e.g., fidelity, sparsity) often fail to assess whether an explanation identifies the true underlying causal variables. To address this, we propose the Explanation-Generalization Score (EGS), a metric that quantifies the causal relevance of GNN explanations. EGS is founded on the principle of feature invariance and posits that if an explanation captures true causal drivers, it should lead to stable predictions across distribution shifts. To quantify this, we introduce a framework that trains GNNs using explanatory subgraphs and evaluates their performance in Out-of-Distribution (OOD) settings (here, OOD generalization serves as a rigorous proxy for the explanation's causal validity). Through large-scale validation involving 11,200 model combinations across synthetic and real-world datasets, our results demonstrate that EGS provides a principled benchmark for ranking explainers based on their ability to capture causal substructures, offering a robust alternative to traditional fidelity-based metrics.

Problem

• Explanation Quality
• Graph Neural Networks
• Causal Variables
• Out-of-Distribution Generalization
• Post-hoc Explanations

Innovation

• Explanation-Generalization Score
• Out-of-Distribution Generalization
• Causal Explanation
• Graph Neural Networks
• Feature Invariance