Cases

Mental health platforms increasingly use AI to triage patients — assessing urgency, recommending care pathways, and sometimes flagging crisis risk. Unlike a physical exam where a doctor explains their reasoning, these tools output recommendations without communicating uncertainty. A low-confidence assessment of a high-risk patient could delay care; a high-confidence false positive could overwhelm crisis services. The question: should these systems be legally required to display their confidence interval or uncertainty metric to the patient before the patient acts on the recommendation?