| 引用本文：杨娇,郭蕊.信息生态系统视角下生成式人工智能辅助临床决策的人机协同研究进展及展望[J].中国卫生政策研究,2025,18(12):40-48 |
|
| 信息生态系统视角下生成式人工智能辅助临床决策的人机协同研究进展及展望 |
| 投稿时间：2025-11-27 修订日期：2025-12-12 PDF全文浏览 HTML全文浏览 |
| 杨娇，郭蕊 |
| 首都医科大学公共卫生学院 北京 100069 |
| 摘要:生成式人工智能（Generative Artificial Intelligence，GAI）正加速融入医疗服务体系，并正在重塑医疗信息生态系统。本文基于信息生态理论，从信息环境、信息流与信息人三个层面系统梳理GAI辅助医生临床决策的人机协同研究进展。分析发现，在信息环境层面，GAI技术能力快速演化，但医疗准入政策相对滞后；在信息流层面，现有评估范式难以反映真实临床场景中诊断推理的动态性与逻辑性；在信息人层面，人机协作中医生的信息权衡等关键机制难以明晰，协同效果存在较高不确定性。基于此，本文提出三方面研究展望：一是构建国家级评估与监管框架，并持续开展场景化测评；二是建立诊断全流程动态评估与产品全生命周期的质量监测体系；三是深入开展医生与GAI的人机协同互动与认知过程研究。 |
| 关键词:生成式人工智能 大语言模型 临床决策支持 |
| 基金项目:国家自然科学基金面上项目（72574152）；首都卫生管理与政策研究基地开放性课题（2025JD01） |
|
| Generative Artificial Intelligence for clinical decision-making： Progress and prospects of Human-AI collaboration from an information ecology perspective |
| YANG Jiao, GUO Rui |
| School of Public Health， Capital Medical University， Beijing 100069， China |
| Abstract:Generative Artificial Intelligence （GAI） is rapidly integrating into healthcare service systems and is reshaping the medical information ecosystem. Based on information ecology theory， this study systematically reviews the research on human-AI collaboration in GAI-assisted clinical decision-making across three dimensions： the information environment， information flows， and information actors. Our analysis shows that， at the level of the information environment， GAI's capabilities are advancing rapidly， while regulatory and admission policies remain lagging. At the level of information flows， existing evaluation paradigms struggle to capture the dynamic and logical nature of diagnostic reasoning in real clinical settings. At the level of information actors， key mechanisms such as physicians’ information weighing in human-AI collaboration remain unclear， leading to considerable uncertainty in collaborative performance. Based on these findings， the study proposes three research directions： （1） establishing a national evaluation and regulatory framework with ongoing vignette-based assessments； （2） developing a dynamic evaluation system covering the entire diagnostic process and quality monitoring across the GAI product lifecycle； and （3） conducting in-depth studies on the human-AI interaction and cognitive processes between physicians and GAI. |
| Key words:Generative Artificial Intelligence Large language model Clinical decision support |
| 摘要点击次数: 82 全文下载次数: 27 |
|
|
|
