Towards learning to explain with concept bottleneck models: mitigating information leakage

TL;DR

This paper proposes using Monte-Carlo Dropout in concept bottleneck models to produce soft concept predictions that avoid information leakage, thereby improving model interpretability and trustworthiness.

Contribution

It introduces a novel application of Monte-Carlo Dropout to mitigate information leakage in soft concept predictions within concept bottleneck models.

Findings

Monte-Carlo Dropout effectively reduces information leakage

Soft concept predictions become more reliable and interpretable

Improved trust in concept-based explanations

Abstract

Concept bottleneck models perform classification by first predicting which of a list of human provided concepts are true about a datapoint. Then a downstream model uses these predicted concept labels to predict the target label. The predicted concepts act as a rationale for the target prediction. Model trust issues emerge in this paradigm when soft concept labels are used: it has previously been observed that extra information about the data distribution leaks into the concept predictions. In this work we show how Monte-Carlo Dropout can be used to attain soft concept predictions that do not contain leaked information.