医疗领域大模型伦理风险识别、治理及前瞻研究
刘浏 , 张馨 , 张琪琪 , 刘洁 , 刘子裕 , 王惠玲 , 易汉希 , 王维圆 , Diabate Ousmane , 王俊普
中国工程科学 ›› 2025, Vol. 27 ›› Issue (6) : 54 -67.
医疗领域大模型伦理风险识别、治理及前瞻研究
Ethical Risk Identification, Governance, and Foresight of Medical Foundation Models
医疗领域正经历以多模态大模型为特征的智能化转型,然而技术赋能与伦理风险共生并存,医疗领域大模型的深度应用使伦理风险日益凸显,亟需聚焦医疗领域大模型,全面识别其伦理风险并探索有效的治理路径。本文深入探讨了医疗领域大模型应用过程中存在的数据隐私风险、算法决策风险、主体关系风险、社会公平风险等四大核心伦理风险,并结合典型案例加以解析:提出了基于“数据 ‒ 算法 ‒ 应用 ‒ 法律”四位一体的医疗领域大模型治理框架,涵盖构建数据治理体系、创新算法治理机制、建设临床应用规范、完善法律监管框架4个方面;分析了医疗领域大模型发展面临的关键技术挑战和政策挑战。最后,展望了医疗领域大模型的未来方向,包括探索基于区块链医疗数据确权、开发轻量化模型普惠基层医疗、构建“政产学研医”协同生态系统,以期为推动医疗领域大模型技术规范健康发展、保障患者权益及完善医疗伦理治理体系提供理论与实践支撑。
The medical field is undergoing an intelligent transformation characterized by the emergence of multimodal foundation models. While technological empowerment brings unprecedented opportunities, it is accompanied by profound ethical risks. The deep application of medical foundation models has amplified such risks, necessitating comprehensive identification and effective governance strategies. This study systematically examines four core categories of ethical risks in medical foundation model applications: data privacy risks, algorithmic decision-making risks, subject-relation risks, and social equity risks, illustrated through representative cases. To address these challenges, a data‒algorithm‒application‒law integrated governance framework is proposed, encompassing the establishment of robust data governance systems, innovation of algorithmic governance mechanisms, formulation of clinical application standards, and improvement in legal and regulatory frameworks. Furthermore, the study analyzes key technological and policy challenges constraining the development of medical foundation models. Looking ahead, the study outlines potential future directions, including the exploration of blockchain-based medical data ownership confirmation, development of lightweight models to promote equitable healthcare at the grassroots level, and construction of collaborative ecosystems integrating government, industry, academia, research, and healthcare. These efforts are intended to provide theoretical foundations and practical pathways for fostering the normative and sound development of medical foundation model technologies, ensuring patient rights, and enhancing the ethical governance system in the healthcare domain.
大模型 / 医疗大模型 / 伦理风险 / 风险识别 / 治理路径
foundation model / medical foundation model / ethical risk / risk identification / governance approaches
| [1] |
Rajpurkar P, Chen E, Banerjee O, et al. AI in health and medicine [J]. Nature Medicine, 2022, 28(1): 31‒38. |
| [2] |
中国信息通信研究院知识产权与创新发展中心, 中国信息通信研究院科技伦理研究中心. 人工智能伦理治理研究报告 [R]. 北京: 中国信息通信研究院, 2023. |
| [3] |
Intellectual Property and Innovation Development Center of China Academy of Information and Communications Technology, Science and Technology Ethics Research Center of China Academy of Information and Communications Technology. Research report on AI ethics governance [R]. Beijing: China Academy of Information and Communications Technology, 2023. |
| [4] |
余玉刚, 王耀刚, 江志斌, 智慧健康医疗管理研究热点分析 [J]. 管理科学学报, 2021, 24(8): 58‒66. |
| [5] |
Yu Y G, Wang Y G, Jiang Z B, et al. Analysis of research hotspots of intelligent health care management [J]. Journal of Management Sciences in China, 2021, 24(8): 58‒66. |
| [6] |
Maddox T M, Embí P, Gerhart J, et al. Generative AI in medicine—Evaluating progress and challenges [J]. New England Journal of Medicine, 2025, 392(24): 2479‒2483. |
| [7] |
郭华源, 刘盼, 卢若谷, 人工智能大模型医学应用研究 [J]. 中国科学(生命科学), 2024, 54(3): 482‒506. |
| [8] |
Guo H Y, Liu P, Lu R G, et al. Research on medical application of artificial intelligence large model [J]. Science in China (Vitae), 2024, 54(3): 482‒506. |
| [9] |
Webster P. Six ways large language models are changing healthcare [J]. Nature Medicine, 2023, 29(12): 2969‒2971. |
| [10] |
Moor M, Banerjee O, Abad Z S H, et al. Foundation models for generalist medical artificial intelligence [J]. Nature, 2023, 616(7956): 259‒265. |
| [11] |
Singhal K, Azizi S, Tu T, et al. Large language models encode clinical knowledge [J]. Nature, 2023, 620(7972): 172‒180. |
| [12] |
Xiang J X, Wang X Y, Zhang X M, et al. A vision-language foundation model for precision oncology [J]. Nature, 2025, 638(8051): 769‒778. |
| [13] |
陈晓红, 刘浏, 牛雅娟, 数智病理平台构建及服务模式研究 [J]. 中国工程科学, 2025, 27(2): 304‒314. |
| [14] |
Chen X H, Liu L, Niu Y J, et al. Digital intelligence pathology platform and its service pattern [J]. Strategic Study of CAE, 2025, 27(2): 304‒314. |
| [15] |
国家卫生健康委员会办公厅, 国家中医药管理局综合司, 国家疾病预防控制局综合司. 卫生健康行业人工智能应用场景参考指引 [EB/OL]. (2024-11-06)[2025-07-17]. https://www.nhc.gov.cn/wjw/c100175/202411/5bcb3c4edd064e31ac5d279caf5830f4.shtml. |
| [16] |
General Office of the National Health Commission, Comprehensive Department of the National Administration of Traditional Chinese Medicine, Comprehensive Department of the National Disease Control and Prevention Administration. Reference guidelines for artificial intelligence application scenarios in the health industry [EB/OL]. (2024-11-06)[2025-07-17]. https://www.nhc.gov.cn/wjw/c100175/202411/5bcb3c4edd064e31ac5d279caf5830f4.shtml. |
| [17] |
陈晓红, 刘浏, 袁依格, 医疗大模型技术及应用发展研究 [J]. 中国工程科学, 2024, 26(6): 77‒88. |
| [18] |
Chen X H, Liu L, Yuan Y G, et al. Technology and application development of medical foundation model [J]. Strategic Study of CAE, 2024, 26(6): 77‒88. |
| [19] |
Thirunavukarasu A J, Ting D S J, Elangovan K, et al. Large language models in medicine [J]. Nature Medicine, 2023, 29(8): 1930‒1940. |
| [20] |
Obermeyer Z, Powers B, Vogeli C, et al. Dissecting racial bias in an algorithm used to manage the health of populations [J]. Science, 2019, 366(6464): 447‒453. |
| [21] |
Rocher L, Hendrickx J M, de Montjoye Y A. Estimating the success of re-identifications in incomplete datasets using generative models [J]. Nature Communications, 2019, 10: 3069. |
| [22] |
施锦诚, 王国豫, 王迎春. ESG视角下人工智能大模型风险识别与治理模型 [J]. 中国科学院院刊, 2024, 39(11): 1845‒1859. |
| [23] |
Shi J C, Wang G Y, Wang Y C. Artificial intelligence foundation model risk identification and governance model from the ESG perspective [J]. Bulletin of Chinese Academy of Sciences, 2024, 39(11): 1845‒1859. |
| [24] |
中华人民共和国国务院. 新一代人工智能发展规划 [EB/OL]. (2017-07-08)[2025-07-17]. https://www.gov.cn/gongbao/content/2017/content_5216427.htm. |
| [25] |
State Council of the People's Republic of China. Next generation artificial intelligence development plan [EB/OL]. (2017-07-08)[2025-07-17]. https://www.gov.cn/gongbao/content/2017/content_5216427.htm. |
| [26] |
中华人民共和国国家互联网信息办公室, 中华人民共和国国家发展和改革委员会, 中华人民共和国教育部, 生成式人工智能服务管理暂行办法 [EB/OL]. (2023-07-10)[2025-07-18]. https://www.gov.cn/zhengce/zhengceku/202307/content_6891752.htm. |
| [27] |
Cyberspace Administration of China, National Development and Reform Commission, Ministry of Education of the People's Republic of China, et al. Interim measures for the management of generative artificial intelligence services [EB/OL]. (2023-07-10)[2025-07-18]. https://www.gov.cn/zhengce/zhengceku/202307/content_6891752.htm. |
| [28] |
李强. 政府工作报告 [EB/OL]. (2025-03-12)[2025-07-18]. https://www.gov.cn/yaowen/liebiao/202503/content_7013163.htm. |
| [29] |
Li Q. Government work report [EB/OL]. (2025-03-12)[2025-07-18]. https://www.gov.cn/yaowen/liebiao/202503/content_7013163.htm. |
| [30] |
国家药品监督管理局. 人工智能医用软件产品分类界定指导原则 [EB/OL]. (2021-07-01)[2025-07-18]. https://www.nmpa.gov.cn/xxgk/ggtg/ylqxggtg/ylqxqtggtg/20210708111147171.html. |
| [31] |
National Medical Products Administration. Guidelines for the classification and definition of artificial intelligence medical software products [EB/OL]. (2021-07-01)[2025-07-18]. https://www.nmpa.gov.cn/xxgk/ggtg/ylqxggtg/ylqxqtggtg/20210708111147171.html. |
| [32] |
Zhang Z Q, Yan C, Malin B A. Membership inference attacks against synthetic health data [J]. Journal of Biomedical Informatics, 2022, 125: 103977. |
| [33] |
杨善林, 丁帅, 顾东晓, 医疗健康大数据驱动的知识发现与知识服务方法 [J]. 管理世界, 2022, 38(1): 219‒229. |
| [34] |
Yang S L, Ding S, Gu D X, et al. Healthcare big data driven knowledge discovery and knowledge service approach [J]. Journal of Management World, 2022, 38(1): 219‒229. |
| [35] |
罗映宇, 朱国玮, 钱无忌, 人工智能时代的算法厌恶: 研究框架与未来展望 [J]. 管理世界, 2023, 39(10): 205‒233. |
| [36] |
Luo Y Y, Zhu G W, Qian W J, et al. Algorithm aversion in the era of artificial intelligence: Research framework and future agenda [J]. Journal of Management World, 2023, 39(10): 205‒233. |
| [37] |
McCradden M D, Joshi S, Mazwi M, et al. Ethical limitations of algorithmic fairness solutions in health care machine learning [J]. The Lancet Digital Health, 2020, 2(5): 221‒223. |
| [38] |
Benjamin R. Assessing risk, automating racism [J]. Science, 2019, 366(6464): 421‒422. |
| [39] |
赵玲玲, 郭遥. 智能医疗机器人伦理风险: 类型、成因与防控策略 [J]. 医学与哲学, 2023, 44(12): 35‒39. |
| [40] |
Zhao L L, Guo Y. Ethical risks of intelligent medical robots: Types, causes, prevention and control strategies [J]. Medicine & Philosophy, 2023, 44(12): 35‒39. |
| [41] |
Maliha G, Gerke S, Cohen I G, et al. Artificial intelligence and liability in medicine: Balancing safety and innovation [J]. The Milbank Quarterly, 2021, 99(3): 629‒647. |
| [42] |
Brin D, Sorin V, Vaid A, et al. Comparing ChatGPT and GPT-4 performance in USMLE soft skill assessments [J]. Scientific Reports, 2023, 13: 16492. |
| [43] |
Cordeiro J V. Digital technologies and data science as health enablers: An outline of appealing promises and compelling ethical, legal, and social challenges [J]. Frontiers in Medicine, 2021, 8: 647897. |
| [44] |
唐露源, 谢士尧, 徐源. 大模型产业创新生态系统构建及风险识别研究 [J]. 科学学研究, 2025, 43(9): 1972‒1981. |
| [45] |
Tang L Y, Xie S Y, Xu Y. Research on the construction and risk identification of large language model industrial innovation ecosystem [J]. Studies in Science of Science, 2025, 43(9): 1972‒1981. |
| [46] |
Cao K, Xia Y D, Yao J W, et al. Large-scale pancreatic cancer detection via non-contrast CT and deep learning [J]. Nature Medicine, 2023, 29(12): 3033‒3043. |
| [47] |
汪琛. 医疗人工智能伦理治理的问题、困境与求解 [J]. 科学学研究, 2025, 43(2): 414‒422. |
| [48] |
Wang C. Research on governance over ethics of AI4health/medicine: Problems, dilemmas and solutions [J]. Studies in Science of Science, 2025, 43(2): 414‒422. |
| [49] |
李润生. 论医疗人工智能的法律规制——从近期方案到远期设想 [J]. 行政法学研究, 2020 (4): 46‒57. |
| [50] |
Li R S. On the legal regulation of medical artificial intelligence—From short-term plan to long-term plan [J]. Administrative Law Review, 2020 (4): 46‒57. |
| [51] |
Imrie F, Davis R, van der Schaar M. Multiple stakeholders drive diverse interpretability requirements for machine learning in healthcare [J]. Nature Machine Intelligence, 2023, 5(8): 824‒829. |
| [52] |
World Health Organization. Ethics and governance of artificial intelligence for health [EB/OL]. (2021-06-28)[2025-07-18]. https://www.who.int/publications/i/item/9789240029200. |
| [53] |
陈晓红, 刘晓亮, 王俊普, 从AI驱动到元宇宙赋能: 病理诊断管理模式创新与扩展 [J/OL]. 中国管理科学, 1‒14[2025-09-17]. https://doi.org/10.16381/j.cnki.issn1003-207x.2023.1211. |
| [54] |
Chen X H, Liu X L, Wang J P, et al. From AI-driven to metaverse-empowered: Innovation and expansion of pathological diagnosis management models [J/OL]. Chinese Journal of Management Science, 1‒14[2025-09-17]. https://doi.org/10.16381/j.cnki.issn1003-207x.2023.1211. |
芙蓉实验室项目(2024PT5111)
/
| 〈 |
|
〉 |