1 article on model interpretability.
Healthcare AI requires explainability — "the model said so" is not a clinical rationale. Mechanistic interpretability is the research field trying to change that. Here is what it actually offers practitioners today, what the gap still is, and what you can do in the meantime.
