Read the web edition of Stat’s AI Prognosis Newsletter, a subscriber exclusive guide to artificial intelligence in healthcare and medicine. Sign up and stream in your inbox every Wednesday.
Patients prefer messages written by AI, but when they are told that messages are being drafted AI, they don’t like them much, a new study published yesterday at JAMA Network Open said.
Duke University researchers looked at more than 1,400 patients on the Health System’s Patient Advisory Committee to see how much they liked the AI draft or the various human drafted patient portal messages. The message came with disclosure that the message was written by AI, written by humans, or that there was no disclosure at all. The patient was more pleased with the answer that was said to have been written by a human. Or, despite preferring the content of AI-generated messages that are longer, more detailed, and potentially empathetic, it was not disclosed.
This article is for stat+ subscribers only
In addition to unlocking this article, we also provide in-depth analysis, newsletters, premium events and news alerts.
Do you already have an account? Login
View all plans
To read the rest of this story, subscribe to stat+.
Subscribe
