Impact of LLM Assistance on Physician Decision-Making

Participants completed clinical vignettes either with access to chatbot support or without chatbot support. The dataset consists of two files.

The first contains post-experiment survey data at the participant level, including pseudonymized participant identifiers, demographic and professional characteristics, previous use of generative AI, attitudes toward AI, and evaluations of the experimental tasks and chatbot support. The second contains clinical vignette data at the vignette-step level, including the clinical information and questions presented to participants, participant responses, expert rubric answers, chatbot-use measures, and scores assigned by independent reviewers.