Causally Reliable Concept Bottleneck Models
This addresses the need for more reliable and interpretable AI models in domains requiring causal reasoning, though it is incremental as it builds on existing concept-based architectures.
The paper tackled the problem that concept-based models lack causal reliability, limiting their reasoning and generalization, by proposing Causally reliable Concept Bottleneck Models (C^2BMs) that enforce reasoning through causally structured concepts, resulting in improved interpretability, causal reliability, and intervention responsiveness while maintaining accuracy.
Concept-based models are an emerging paradigm in deep learning that constrains the inference process to operate through human-interpretable variables, facilitating explainability and human interaction. However, these architectures, on par with popular opaque neural models, fail to account for the true causal mechanisms underlying the target phenomena represented in the data. This hampers their ability to support causal reasoning tasks, limits out-of-distribution generalization, and hinders the implementation of fairness constraints. To overcome these issues, we propose Causally reliable Concept Bottleneck Models (C$^2$BMs), a class of concept-based architectures that enforce reasoning through a bottleneck of concepts structured according to a model of the real-world causal mechanisms. We also introduce a pipeline to automatically learn this structure from observational data and unstructured background knowledge (e.g., scientific literature). Experimental evidence suggests that C$^2$BMs are more interpretable, causally reliable, and improve responsiveness to interventions w.r.t. standard opaque and concept-based models, while maintaining their accuracy.