P. Nallari

Explainable Recommendation in Graph Neural Networks Using Propagation Path Analysis and Counterfactual Graph Editing

A post-hoc explainability framework for LightGCN is proposed combining two complementary techniques: Propagation Path Analysis, which decomposes recommendation scores by propagation layer to attribute influence to specific training interactions, and Counterfactual Graph Editing, which identifies the most influential user-item edges through structural sensitivity analysis and targeted edge removal.