The American Medical Association (AMA) has adopted policies aimed at shaping the development of artificial intelligence systems and their integration into healthcare, according to several releases from their annual meeting.
AMA’s recommendations are intended to ensure practicing physicians are involved in the development of AI applications and receive the training needed to use them effectively.
The policies seek to make developers of autonomous AI applications, not physicians, liable for medical errors involving the technologies.
With more than two-thirds of healthcare executives predicting artificial intelligence will have a big effect on their organizations over the next three years, the relationship between physicians and technologies looks set to change in the near- to mid-term.
To prepare for (and influence) those changes, AMA adopted a raft of policies and recommendations at its annual meeting. Some of the policies center on physician training while others deal more broadly with the development, adoption and oversight of healthcare AI systems.
Both sets of policies feature points relevant to developers of AI systems. Having adopted the policy recommendations, AMA will advocate for AI developers, not physicians, to be held accountable when problems arise from the use of the technology, given AI developers "are in the best position to manage issues of liability arising directly from system failure or misdiagnosis."
While AMA is clear on who should be responsible, other authorities are still debating the topic. At a European Commission on Civil Law Rules and Robotics meeting earlier this year, a European politician asked, "who will be responsible if the robots make a mistake in surgery or if an autonomous car crashes," adding that the issues pose new challenges for regulators.
AMA wants to see policies enacted that reduce the risk of AI applications that cause harm coming to market. The medical association is advocating for regulations "based on risk of harm and benefit" and reimbursement decisions that factor in safety, efficacy, privacy and security. AMA thinks payer coverage should help physicians transition to new systems and offer them chances to give feedback, adding that nobody should be penalized for not using AI while the situation is "in flux."
The training policies seek to ensure physicians who want to use AI are equipped to do so. Specific AMA policies include a push for the provision of educational materials to "guard against inadvertent dissemination of bias that may be inherent in AI systems."