|
AI may be accurate, but does it actually change clinician behavior or improve patient decisions?
In this JAMA+ AI Conversations, Yun Liu, PhD, a research scientist at Google Research, spoke with JAMA+ AI Associate Editor Yulin Hswen, ScD, MPH, about what is behind the curtain in clinical AI systems.
Using a "Wizard of Oz" lens—a research method in human-computer interaction where participants interact with a computer model but are unaware that researchers are controlling the output—Dr. Liu and his colleagues question whether it matters if the intelligence comes from a model or a clinician, or whether accuracy alone is enough.
Listen now on Spotify | Apple Podcasts | YouTube | JAMA.com.
Editor’s Picks in this week’s JAMA+ AI Update:
- AI may facilitate mental health care by expanding access and improving diagnoses, but not everyone is convinced. Because AI tool performance isn’t always predictable, problems may only appear once real patients are affected. (JAMA Psychiatry)
- In a study of more than 20,000 primary care visits, AI ambient scribes captured more neuropsychiatric symptoms than human scribes, or visits without scribes. But clinicians were less likely to diagnose or treat depression during those visits, suggesting the importance of understanding how ambient scribe use may impact clinical practice patterns more generally. (JAMA Psychiatry)
- A recent study found a U-shaped link between social media use and teen well-being: heavy use in younger adolescents and no use in older teens, especially boys, were both tied to lower well-being. The findings suggest that balanced use, not extremes, may be healthiest, underscoring the need for more nuanced policy and research. (JAMA Pediatrics)
|
