A Completion-Aware Framework for Impactful Counterfactual Explainability in Graph Neural Networks
About
In this study, we propose a novel pipeline for generic, model-agnostic, local-level counterfactual explainability in graph neural networks (GNNs). Although counterfactual explainers capable of both adding and removing edges have emerged in recent years, the need for generic and efficient solutions remains unmet, particularly concerning qualitative explanation generation. Our approach couples progress in factual explainability with missing edge prediction models rooted in link prediction research, in order to enhance the quality, robustness and intuitiveness of explanations. A multi-faceted experimental analysis conducted on real-world and synthetic graph classification benchmarks, both binary and multi-label, demonstrates the advancements in comparison to state-of-the-art baselines across diverse metrics.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Graph Counterfactual Explanations | BA-2MOTIFS | Avg Minimality Score100 | 16 | |
| Motif Proximity Evaluation | BA-2MOTIFS | Motif Proximity Score90 | 10 | |
| Motif Proximity Evaluation | BA 2Motifs 3Classes | Motif Proximity Score90 | 10 | |
| Motif Proximity Evaluation | BA-3Motifs | Motif Proximity Score0.98 | 10 | |
| Graph Counterfactual Explanation | Minimality Score100 | 10 | ||
| Graph Counterfactual Explanations | BA-3Motifs | Avg Minimality Score1 | 10 | |
| Graph Counterfactual Explanations | BA-4Motifs | Average Minimality Score1 | 10 | |
| Graph Counterfactual Explanations | GRAPH-SST5 | Average Minimality Score99 | 10 | |
| Motif Proximity Evaluation | BA-4Motifs | Motif Proximity Score0.94 | 10 | |
| Graph Counterfactual Explanation | BA-3Motifs | Validity94 | 10 |