A02 - How Do You Actually Review a Protocol Involving Artificial Intelligence (AI)?

CLICK HERE to view captioning for this session! 

Much attention has been paid to the ethical issues posed by the use of AI in humans subjects research. Yet, most sessions raise more questions than answers and, at the end of the day, HRPP/IRB staff and reviewers are left to roll up their sleeves and review these protocols desks fast and furiously. How should HRPPs adapt their approaches and policies and train their staff to best review these protocols? How does an IRB reviewer actually review a protocol involving AI when the technology changes nearly every week? This session will provide practical guidance to IRBs and oversight bodies on how to review research involving AI, including how to determine if human subjects research definitions apply, how to assess FDA device regulations, and how to apply the federal regulatory criteria for approval to human research studies involving AI. Additionally, the session will examine special ethical considerations raised by research involving AI, including privacy and data ownership concerns related to the use of large, unconsented datasets, return of individual research results, and algorithmic bias. Finally, the session will discuss the scope and limitations of IRB review, and how oversight bodies must work together to effectively review research involving AI. For example, regulations restrict the IRB's consideration of future risks, yet AI algorithms raise exactly those concerns related to future risks posed by algorithmic bias and potential dual use.