Overview
Block II: High-Stakes Decisions (Weeks 5, 6, 8)
Central question: When an algorithm is right, why might a clinician still be right not to use it?
Big idea: Clinical performance is not clinical usefulness. A model can outperform experts on clean cases and still harm patients in practice if it enters the wrong workflow, sees the wrong population, gives advice at the wrong time, or quietly shifts responsibility without anyone admitting it.
Block focus: Understand how clinicians make decisions under uncertainty and how AI can safely support human expertise.
Being right on paper is not the same as improving care. A recommendation enters a workflow with incomplete records, interruptions, responsibility, and existing relationships of trust. We therefore ask not only whether an answer is correct, but whether using it improves care.