How Faithful are Self-Explainable GNNs?
This work addresses the reliability of interpretability methods for graph data, which is incremental as it critiques existing self-explainable GNNs without introducing new models.
The paper analyzes the faithfulness of self-explainable graph neural networks (GNNs) by evaluating multiple models and metrics, identifying limitations in both the models and evaluation approaches.
Self-explainable deep neural networks are a recent class of models that can output ante-hoc local explanations that are faithful to the model's reasoning, and as such represent a step forward toward filling the gap between expressiveness and interpretability. Self-explainable graph neural networks (GNNs) aim at achieving the same in the context of graph data. This begs the question: do these models fulfill their implicit guarantees in terms of faithfulness? In this extended abstract, we analyze the faithfulness of several self-explainable GNNs using different measures of faithfulness, identify several limitations -- both in the models themselves and in the evaluation metrics -- and outline possible ways forward.