Relational Concept Bottleneck Models

TL;DR

Relational Concept Bottleneck Models (R-CBMs) are a new class of interpretable deep learning models that effectively handle relational tasks, combining the strengths of Concept Bottleneck Models and Graph Neural Networks.

Contribution

This paper introduces R-CBMs, a novel framework that unifies interpretability with relational reasoning, capable of representing CBMs and GNNs, and demonstrates their effectiveness across diverse relational tasks.

Findings

R-CBMs match the performance of black-box relational models.

They enable concept-based explanations and test-time interventions.

They perform well in out-of-distribution and limited data scenarios.

Abstract

The design of interpretable deep learning models working in relational domains poses an open challenge: interpretable deep learning methods, such as Concept Bottleneck Models (CBMs), are not designed to solve relational problems, while relational deep learning models, such as Graph Neural Networks (GNNs), are not as interpretable as CBMs. To overcome these limitations, we propose Relational Concept Bottleneck Models (R-CBMs), a family of relational deep learning methods providing interpretable task predictions. As special cases, we show that R-CBMs are capable of both representing standard CBMs and message-passing GNNs. To evaluate the effectiveness and versatility of these models, we designed a class of experimental problems, ranging from image classification to link prediction in knowledge graphs. In particular we show that R-CBMs (i) match generalization performance of existing relational black-boxes, (ii) support the generation of quantified concept-based explanations, (iii) effectively respond to test-time interventions, and (iv) withstand demanding settings including out-of-distribution scenarios, limited training data regimes, and scarce concept supervisions.