Should providers tell patients if they use AI in care

You’re reading the web edition of STAT’s AI Prognosis newsletter, our subscriber-exclusive guide to artificial intelligence in health care and medicine. Sign up to get it delivered in your inbox every Wednesday. 

Patients prefer messages that are written by AI, but they like them less when they’re told the message was AI-drafted, says a new study that came out yesterday in JAMA Network Open. 

Researchers from Duke University surveyed over 1,400 patients on the health system’s patient advisory committee about how much they liked various patient portal messages, which were either AI-drafted or human-drafted. The messages came with either the disclosure that the message was written by AI, was written by a human, or no disclosure at all. Patients were more satisfied with answers they were told were written by humans, or that came with no disclosure, even though they preferred the content of the AI-generated messages, which were longer, more detailed, and potentially more empathetic.