CoDy: Counterfactual Explainers for Dynamic Graphs

TL;DR

CoDy is a novel counterfactual explanation method for dynamic graph neural networks that identifies influential subgraphs to interpret model predictions, improving explainability in temporal graph applications.

Contribution

We introduce CoDy, a model-agnostic, instance-level counterfactual explainer for TGNNs that uses a Monte Carlo Tree Search to efficiently find explanatory subgraphs.

Findings

CoDy outperforms state-of-the-art baselines by 16% in AUFSC+.

It effectively identifies influential subgraphs for model interpretability.

Experimental results validate CoDy’s efficiency and accuracy.

Abstract

Temporal Graph Neural Networks (TGNNs) are widely used to model dynamic systems where relationships and features evolve over time. Although TGNNs demonstrate strong predictive capabilities in these domains, their complex architectures pose significant challenges for explainability. Counterfactual explanation methods provide a promising solution by illustrating how modifications to input graphs can influence model predictions. To address this challenge, we present CoDy, Counterfactual Explainer for Dynamic Graphs, a model-agnostic, instance-level explanation approach that identifies counterfactual subgraphs to interpret TGNN predictions. CoDy employs a search algorithm that combines Monte Carlo Tree Search with heuristic selection policies, efficiently exploring a vast search space of potential explanatory subgraphs by leveraging spatial, temporal, and local event impact information. Extensive experiments against state-of-the-art factual and counterfactual baselines demonstrate CoDy's effectiveness, with improvements of 16% in AUFSC+ over the strongest baseline.