On Completeness-aware Concept-Based Explanations in Deep Neural Networks

TL;DR

This paper introduces a completeness measure for concept-based explanations in DNNs, proposes a method to discover complete and interpretable concepts, and validates its effectiveness through metrics and user studies on synthetic and real datasets.

Contribution

It defines the notion of completeness for concept explanations, and develops a method to discover a complete set of interpretable concepts for DNNs.

Findings

The proposed method effectively finds complete and interpretable concepts.

Validation shows improved explanation quality on synthetic and real datasets.

User studies confirm the interpretability and completeness of the explanations.

Abstract

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)