Interpretability is the degree to which people can understand how a model works or why it produces particular outputs.
Interpretability supports debugging, trust, auditability, compliance, and responsible use, and it is related to but not identical with explainability.