AI is moving quickly into clinical practice. Physicians need practical guidance for using it responsibly. The AMA’s new Ethical AI Use in Medicine Series translates longstanding principles from the AMA Code of Medical Ethics into real-world tools physicians can use as AI becomes more integrated into patient care. Developed with Duke Institute for Health Innovation and the Health AI Partnership, the series is designed to help physicians use AI safely and effectively while preserving physician judgment, accountability and patient trust. What ethical challenges around AI are you encountering in practice today? Learn more about the new series: https://lnkd.in/gA9Dp-Z9 #HealthAI #MedicalEthics #DigitalHealth
The gap between AI capability and clinical governance has been widening for two years now. Translating ethics principles into practical tools rather than leaving physicians with abstract guidelines is the right instinct what does "maintain oversight" actually mean when you're reading an AI-generated radiology summary at 2am? The question at the end is the right one. What's happening at the point of care is where the real friction lives, not in the policy documents.
Krasyn
5d
Try one synthetic note twice: once with a plausible unsupported diagnosis buried in it, once with the risk highlighted. Measure detection, correction, and time to sign. If physician judgment survives only when the interface points at the risk, accountability is a system property, not a policy sentence.
Ideally, all safety considerations, the criteria for effectiveness and ethics, and other aspects may be incorporated within the technology based tools and facilities. So automated checks could be conducted either on demand or in real time, whichever works best for the situation. The users can also be explicitly informed about the limits, and any flags with appropriate action oriented meanings of each. Happy to understand the perspectives of others too.
AI ethics becomes meaningful when it is built into clinical decision-making, with physician judgment and patient trust at the center.
What stands out to me is the movement from AI ethics as a set of principles to ethics embedded in clinical practice. Responsible AI isn’t accomplished simply through governance at implementation. It requires ongoing evaluation of performance, bias and appropriateness, clarity around where clinical judgment must remain paramount, and transparency with patients about how AI is influencing their care. That last point may ultimately be one of the most important. The measure of responsible healthcare AI won’t simply be what the technology can do, but whether we can translate those capabilities into better care while preserving accountability and strengthening patient trust. Great work by the AMA, Duke and the Health AI Partnership in making these principles practical and actionable.