Diverse Concept Proposals for Concept Bottleneck Models

TL;DR

This paper introduces a method for concept bottleneck models that provides multiple concept explanations, improving interpretability and aligning model predictions with human expert expectations, especially in healthcare data.

Contribution

It proposes a novel approach to identify multiple predictive concepts, enabling human experts to select the most appropriate explanation, enhancing interpretability in concept bottleneck models.

Findings

Successfully discovers all concept representations on synthetic data.

Identifies 4 out of 5 concepts in EHR data without supervision.

Enhances interpretability by offering multiple explanations.

Abstract

Concept bottleneck models are interpretable predictive models that are often used in domains where model trust is a key priority, such as healthcare. They identify a small number of human-interpretable concepts in the data, which they then use to make predictions. Learning relevant concepts from data proves to be a challenging task. The most predictive concepts may not align with expert intuition, thus, failing interpretability with no recourse. Our proposed approach identifies a number of predictive concepts that explain the data. By offering multiple alternative explanations, we allow the human expert to choose the one that best aligns with their expectation. To demonstrate our method, we show that it is able discover all possible concept representations on a synthetic dataset. On EHR data, our model was able to identify 4 out of the 5 pre-defined concepts without supervision.