Faithful, Sufficient and Understandable: Rethinking Graph Counterfactual Explanations via Discrete Diffusion Inversion

2026-08-12

AI Summary

This work addresses the lack of interpretability in graph neural networks within high-stakes scenarios, where existing methods struggle to generate counterfactual explanations that are both manifold-consistent and span the full edit space. We propose the first approach leveraging a discrete denoising diffusion model combined with a novel discrete inversion mechanism to produce distribution-aware, structurally minimal, and valid counterfactuals—strictly respecting node/edge types and domain-specific constraints such as chemical valency. Our method ensures edits remain on the data manifold while comprehensively covering the feasible edit space, and we introduce a unified evaluation framework to assess explanation quality. Experiments across four benchmarks demonstrate substantial improvements over state-of-the-art methods, particularly in molecular graph tasks, where our approach generates high-quality, interpretable, and chemically plausible counterfactual examples.

Abstract

Graph Neural Networks (GNNs) achieve strong predictive performance on graph-structured data across domains such as chemistry, biology, and network analysis, yet they provide no intrinsic explanation of their predictions. This limits their adoption in high-stakes and safety-critical settings. Counterfactual explanations address this by revealing the minimal structural modifications that would change a model's prediction. On graphs, however, such a modification is hard to produce. The search space is discrete and combinatorial, and a valid answer must respect categorical node and edge types together with domain rules such as chemical valency in the case of molecular graphs. Existing explainers give up one of two things. Either edits are not held on the data manifold, or the search does not span the full edit space. We propose Graph Diffusion Counterfactual Explanation via Inversion (GDCE-I), which gives up neither. A discrete denoising diffusion model with a novel discrete inversion scheme enables distribution-aware edits leveraging the whole domain edit space. We further address the incomplete and inconsistent evaluation of graph counterfactuals by deriving a framework of explanation desiderata and applying it to every method under one shared protocol. Across four benchmarks, GDCE-I outperforms related work by a large margin on the defined framework. For the molecular domain, we further qualitatively show that GDCE-I attains interpretable in-distribution solutions.

Problem

• Graph Neural Networks
• Counterfactual Explanations
• Discrete Search Space
• Data Manifold
• Domain Constraints

Innovation

• discrete diffusion
• counterfactual explanation
• graph neural networks
• manifold-preserving editing
• inversion-based generation