Causal Concept Explanations for Deep Neural Models

Josh Rountree, Pulkit Verma, Oswin So +2

Concept-based explanations in neural models are important for the development, certification, and adoption of autonomous systems. Yet, interpretability remains correlational. Witho…