.png){kind=icon}
.png){kind=icon}
Return
Large language models exhibit greater diagnostic anchoring than physicians in a forced-choice vignette study
.png){kind=icon}
PRE
AI
A
C
E
C
S
DOI：10.1016/j.ijmedinf.2026.106550.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• LLMs ranked anchor diagnoses first 55.6% vs 21.2% (residents) and 10.0% (attendings). • LLMs showed 3–4 fold higher odds of diagnostic anchoring than physicians. • Anchoring bias varied substantially across 8 evaluated LLM models. • Patient-suggested diagnoses disproportionately influence LLM diagnostic outputs.
Journal
IF:
4.1
Papers:
4.5K
Citations:
1.1W
Organization
Cited Papers