Interpreting Graph Neural Networks via Unrevealed Causal Learning

Wanyu Lin, Hao Lan, Hao Wang, Baochun Li

Keywords: Model Explainability, Graph Neural Networks, Structural Casual Model

Abstract: This paper proposes a new explanation framework, called OrphicX, for generating causal explanations for any graph neural networks (GNNs) based on learned latent causal factors. Specifically, we construct a distinct generative model and design an objective function that encourages the generative model to produce causal, compact, and faithful explanations. This is achieved by isolating the causal factors in the latent space of graphs by maximizing the information flow measurements. We theoretically analyze the cause-effect relationships in the proposed causal graph, identify node attributes as confounders between graphs and GNN predictions, and circumvent such confounder effect by leveraging the backdoor adjustment formula. Our framework is compatible with any GNNs, and it does not require access to the process by which the target GNN produces its predictions. In addition, it does not rely on the linear-independence assumption of the explained features, nor require prior knowledge on the graph learning tasks. Empirically, we show that OrphicX can effectively identify the causal semantics for generating causal explanations, significantly outperforming its alternatives.

One-sentence Summary: This paper proposes a new explanation framework for generating causal explanations for any graph neural networks based on learned latent causal factors.