CIExplainer++: Generating Causal and Interpretable Explanations for Graph Neural Networks

Short summary

CIExplainer introduces a causal inference method to identify which subgraph components most influence Graph Neural Network predictions. G2TeXplainer extends this by translating causal explanations into human-readable natural language. Evaluated across GNN architectures (GCN, GraphSAGE, GAT, GIN).

• •Perturbation-based causal method for explaining Graph Neural Networks using Potential Outcome Framework
• •G2TeXplainer converts causal subgraphs into natural language explanations with feature and relational information
• •Tested across multiple GNN architectures but no results or practical examples provided

Generated with AI, which can make mistakes.