Ethical Principles for AI in Healthcare

The advocacy of the Coalition for Healthcare AI is guided by the following principles for AI creators, derived from the bioethical principles Non-maleficence, Agency, and Accountability:

• Improve patient outcome as shown either by direct evidence linked clinical literature, and aligned with evidence based clinical standards of care/practice patterns from quality of care organizations, professional medical societies and patient organizations, while accounting for safety, efficacy and equity

• Design so the AI’s operations are maximally reducible to characteristics aligned with scientific knowledge of human clinician cognition, rather than proxy characteristics.

• Maximize traceability of patient derived data, and commensurate data stewardship, accountability, and authorization; including by adherence to accepted standards.

• Validate rigorously for safety, efficacy and equity, using preregistered clinical studies, by comparing the AI against clinical outcome, or outcome surrogates in the case of chronic diseases, in the intended clinical workflow and usage, as shown by either direct or linked evidence

• Align liability or other protections commensurate with indications for use and autonomy, without unduly burdening with liabilities beyond other comparable entities.