GCI: A (G)raph (C)oncept (I)nterpretation Framework
This addresses the problem of interpretability in AI for researchers and practitioners working with complex tasks like molecular property prediction, though it is incremental as it builds on existing concept extraction methods.
The paper tackles the challenge of interpreting and evaluating concepts extracted from Graph Neural Networks (GNNs) by introducing GCI, a framework that quantitatively measures alignment between discovered concepts and human interpretations, achieving a 0.76 AUCROC completeness score in molecular property prediction.
Explainable AI (XAI) underwent a recent surge in research on concept extraction, focusing on extracting human-interpretable concepts from Deep Neural Networks. An important challenge facing concept extraction approaches is the difficulty of interpreting and evaluating discovered concepts, especially for complex tasks such as molecular property prediction. We address this challenge by presenting GCI: a (G)raph (C)oncept (I)nterpretation framework, used for quantitatively measuring alignment between concepts discovered from Graph Neural Networks (GNNs) and their corresponding human interpretations. GCI encodes concept interpretations as functions, which can be used to quantitatively measure the alignment between a given interpretation and concept definition. We demonstrate four applications of GCI: (i) quantitatively evaluating concept extractors, (ii) measuring alignment between concept extractors and human interpretations, (iii) measuring the completeness of interpretations with respect to an end task and (iv) a practical application of GCI to molecular property prediction, in which we demonstrate how to use chemical functional groups to explain GNNs trained on molecular property prediction tasks, and implement interpretations with a 0.76 AUCROC completeness score.