The Impact of AI on Professional Decision-Making: Evidence from Medical Scenarios

How to cite

Wang, Xingjian; Miao, Meng. 2025. "The Impact of AI on Professional Decision-Making: Evidence from Medical Scenarios." AEA RCT Registry. AEARCTR-0017455. https://doi.org/10.1257/rct.17455-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Xingjian Wang* , Meng Miao

* Primary investigator

Registry ID
AEARCTR-0017455
DOI
https://doi.org/10.1257/rct.17455-1.0
First registered
Last updated
Status
In development
Trial dates
2025-12-15 – 2026-12-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/17455

Abstract

AI is rapidly penetrating various professional fields, but it remains unclear how human experts respond to algorithmic suggestions. This study focuses on the core theoretical questions of human-robot collaboration: Are AI suggestions adopted as decision-making aids, or do they trigger defensive resistance among professionals?
We use medical scenarios as an empirical environment to test doctors' responses to AI diagnostic suggestions provided by patients through standardized patient (SP) field experiments.

Experimental design

This project plans to systematically investigate the influence mechanism of artificial intelligence diagnostic suggestions on doctors' professional decision-making through standardized patients (SP) field experimental methods. The study will be conducted in real medical scenarios to obtain behavioral data of experts collaborating with AI in their natural state.
The project will recruit a certain number of standardized patients through professional actor recruitment channels and community volunteer networks in a voluntary manner. All standardized patients will receive full project instructions and professional training before participation, and can participate in formal experiments only after passing the training.
The experiment will be conducted in the routine outpatient clinic of the hospital, and standardized patients will be treated according to a unified disease script (common diseases such as osteoporosis, prediabetes, hypertension, etc.). The experiment adopts a randomized group design: the control group standardizes patients without mentioning AI; Treatment group 1 presented AI diagnosis suggestions before the doctor's initial consultation; Treatment group 2 presented AI recommendations after the doctor's initial consultation. AI suggestions are divided into two categories: high-quality and controversial, with cross-random assignment.
Standardized patients fill in a structured record form immediately after the end of the visit, recording the doctor's diagnosis, prescription, examination recommendations, consultation duration, and key information of doctor-patient interaction, but standardized patients do not perform any real medical operations.

Design details

Randomization method
randomization done by computer algorithm
Randomization unit
individual
Clusters
500 individuals
Observations
500 individuals
Arms
200 individuals Treatment group 1, 200 individuals Treatment group 2, 100 individuals Control
Primary outcomes
Consistency of doctor's plans with AI recommendations

Cited by

No citation data available for this registration yet.

Keywords