The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
The paper 'The Clinician's Veto' examines trust and risk in autonomous AI prescribing. It finds that clinicians distinguish between real clinical ambiguity (aleatoric) and model ignorance (epistemic. To ensure safety, the authors propose three requirements: calibrated action-gating, differentiated uncertainty communication, and inferential transparency. This shifts AI from a self-governing agent to a supervised decision-support tool.