Graphs Counterfactual Explainability: A Comprehensive Landscape (AAAI 2024 Tutorial)
Tutorial at the Thirty-Eighth AAAI Conference on Artificial Intelligence (AAAI 2024), February 20–21, 2024, Vancouver, Canada.
Graph Neural Networks (GNNs) have proven highly effective in graph-related tasks, including Traffic Modeling, Learning Physical Simulations, Protein Modeling, and Large-scale Recommender Systems. However, their ``black-box’’ nature limits their interpretability in high-stakes applications. This tutorial explored Graph Counterfactual Explainability (GCE) as a means to provide transparent explanations for GNN predictions.
Speakers
- Mario Alfonso Prado-Romero
- Bardh Prenkaj
- Giovanni Stilo
Topics Covered
- Foundations and formal definitions of Graph Counterfactual Explanations
- Taxonomy and categorization of GCE methods
- Evaluation metrics and protocols for GCE
- Practical demonstration with the GRETEL framework
- Current research challenges and future directions
Highlights
Based on the ACM Computing Survey: A Survey on Graph Counterfactual Explanations: Definitions, Methods, Evaluation, and Research Challenges (ACM CSUR, 2024).