On Completeness-aware Concept-Based Explanations in Deep Neural Networks
This work addresses the need for more interpretable and complete explanations in AI for users like researchers and practitioners, though it is incremental as it builds on existing concept-based methods.
The paper tackles the problem of concept-based explainability in deep neural networks by introducing a notion of completeness to quantify how well a set of concepts explains model predictions, and proposes a method to discover interpretable concepts with ConceptSHAP for importance scoring, validated on synthetic and real-world datasets with metrics and user studies showing effectiveness.
Human explanations of high-level decisions are often expressed in terms of key concepts the decisions are based on. In this paper, we study such concept-based explainability for Deep Neural Networks (DNNs). First, we define the notion of completeness, which quantifies how sufficient a particular set of concepts is in explaining a model's prediction behavior based on the assumption that complete concept scores are sufficient statistics of the model prediction. Next, we propose a concept discovery method that aims to infer a complete set of concepts that are additionally encouraged to be interpretable, which addresses the limitations of existing methods on concept explanations. To define an importance score for each discovered concept, we adapt game-theoretic notions to aggregate over sets and propose ConceptSHAP. Via proposed metrics and user studies, on a synthetic dataset with apriori-known concept explanations, as well as on real-world image and language datasets, we validate the effectiveness of our method in finding concepts that are both complete in explaining the decisions and interpretable. (The code is released at https://github.com/chihkuanyeh/concept_exp)