Abstract

Kueper and Pandit (2025) describe potential benefits and harms of technologies that incorporate artificial intelligence (AI), including bias and equity issues, effects on end-users and downstream impacts on quality of care and cost. They advocate for an iterative, life cycle approach in developing and monitoring “trustworthy” AI. Their model suggests that safe and effective deployment of AI requires “training” for end-users but leave ill-defined what such training might entail. The design of learning programs to facilitate safe incorporation of AI into healthcare must be proactive and deliberate and not an afterthought.