| 400 | 0 | 275 |
| 下载次数 | 被引频次 | 阅读次数 |
伦理治理是保障教育人工智能可持续健康发展的基石。当前教育人工智能伦理治理以“风险应对”模式为主导,该模式历经风险辨识、风险归类与风险预判三个阶段,推动了人工智能与教育协同发展,但在教育价值实现、伦理风险识别、治理体系构建上面临效度瓶颈而亟须转型。治理目标决定治理模式的本质、方向与路径,面对治理模式的转型诉求,教育人工智能伦理治理目标应由“风险消解”转向“技术向善”,带动治理模式转型。在转型过程中,教育人工智能的概念范畴是衔接治理目标与实践的关键。鉴于技术与社会的复杂互动,教育人工智能应超越“技术制品集合”的单维范畴,从社会-技术系统视角把握其概念本质、构成维度及互动关系。基于教育人工智能社会-技术系统范畴下技术工件、行动者与制度三大核心维度及六种互动机制,应将“联合优化”确立为伦理治理的实践路向,通过贯穿生命周期的价值对齐、协同适配的素养提升、健全灵活的制度设计,实现教育人工智能社会-技术系统的整体性向善。
Abstract：The ethical governance of artificial intelligence in education bears directly on its sustainable and healthy development. At present, such governance is predominantly guided by a “risk response” model. Having evolved through three stages-risk identification, risk classification, and risk prediction, this model has promoted the coordinated development of artificial intelligence and education. However, it now faces pressing validity bottlenecks in realizing educational values, identifying ethical risks, and constructing governance systems, and thus urgently requires transformation. Governance objectives determine the essence, direction, and pathways of governance models. In response to the demand for transformation, the objective of ethical governance for educational artificial intelligence should shift from “risk elimination” to “technology for good,” so that a reorientation of goals may drive the transformation of governance models. In this process, defining the concept of educational artificial intelligence is key to linking governance objectives with practice. Given the complex interactions between technological and social factors, educational artificial intelligence should move beyond the one-dimensional Categories of a “collection of technological artifacts” and instead reconstruct its conceptual essence, constituent dimensions, and interactive relationships from the perspective of a socio-technical system. Based on this socio-technical system framework, which encompasses three core dimensions-technological artifacts, actors, and institutions-as well as six interaction mechanisms among them, joint optimization should serve as the practical approach to ethical governance. Through lifecycle-wide value alignment, literacy enhancement for coordinated adaptation, and sound yet flexible institutional design, the educational AI socio-technical system can be advanced toward holistic ethical betterment.
[1]白钧溢.国际教育人工智能伦理框架研究的回顾与镜鉴[J].清华大学教育研究,2024,(2):50-59.
[2]The Ethics of Artificial Intelligence in Education:Practices, Challenges, and Debates[M].Oxford:Taylor&Francis,2022.80-84.
[3]逯行,黄荣怀.教育数字化转型期的现代化风险观照及其治理研究[J].清华大学教育研究,2023,(3):54-64.
[4]王佑镁,王旦等.敏捷治理:教育人工智能伦理治理新模式[J].电化教育研究,2023,(7):21-28.
[5]杨俊锋,褚娟.人工智能教育应用的伦理风险和规范原则[J].中国教育学刊,2024,(11):21-27.
[6]王佑镁,王欣颖等.教育领域生成式人工智能应用的伦理风险管理框架研究[J].电化教育研究,2024,(10):28-34+42.
[7]Ihde D.Technology and Prognostic Predicaments[J].AI&Society,1999,13(1):44-51.
[8]教科信[2026]1号文件,教育部等五部门关于印发《“人工智能+教育”行动计划》的通知[Z].
[9][14][新西兰]史蒂夫·马修曼.王志弘,高郁婷译.技术与社会理论[M].新北:群学出版有限公司,2023.10+50.
[10][17] Johnson D G,Verdicchio M.The Sociotechnical Entanglement of AI and Values[J].AI&Society,2025,40(1):67-76.
[11][美]凯特·克劳福德.丁宁等译.技术之外:社会联结中的人工智能[M].北京:中国科学技术出版社,2024.7.
[12]Leveson N G.Engineering a Safer World:Systems Thinking Applied to Safety[M].Cambridge:The MIT Press,2016.13.
[13]Dobbe R,Wolters A.Toward Sociotechnical AI:Mapping Vulnerabilities for Machine Learning in Context[J].Minds and Machines,2024,34(2):12-64.
[15]白钧溢,于伟.超越“共识”:教育人工智能伦理原则构建的发展方向[J].中国电化教育,2023,(6):9-17+24.
[16]Geels F.Co-evolution of Technology and Society:The Transition in Water Supply and Personal Hygiene in the Netherlands(1850-1930)—A Case Study in Multi-level Perspective[J].Technology in Society,2005,27(3):363-397.
[18]Kudina O,Van de Poel I.A Sociotechnical System Perspective on AI[J].Minds and Machines,2024,34(3):21-31.
[19]Baxter G, Sommerville I. Socio-technical Systems:From Design Methods to Systems Engineering[J].Interacting with Computers,2011,23(1):4-17.
[20]Adomavicius G, Bockstedt J, et al. Understanding Evolution in Technology Ecosystems[J].Communications of the ACM,2008,51(10):117-122.
[21]Holmes W,Tuomi I.State of the Art and Practice in AI in Education[J].European Journal of Education,2022,57(4):542-570.
[22][32][美]W·理查德·斯科特.姚伟等译.制度与组织:思想观念、利益偏好与身份认同(第4版)[M].北京:中国人民大学出版社,2020.
[23][27] Van de Poel I.Embedding Values in Artificial Intelligence(AI)Systems[J].Minds and Machines,2020,30(3):385-409.
[24]Vermaas P E, Houkes W. Technical Functions:A Drawbridge between the Intentional and Structural Natures of Technical Artefacts[J]. Studies in History and Philosophy of Science Part A,2006,37(1):5-18.
[25]Di Maio P.Towards a metamodel to support the joint optimization of socio technical systems[J].Systems,2014,2(3):273-296.
[26]闫坤如.人工智能价值对齐的价值表征及伦理路径[J].伦理学研究,2024,(4):94-100.
[28]Van den Hoven J,Vermaas P E,et al.Early Engagement and New Technologies:Opening up the Laboratory[M].Dordrecht:Springer Netherlands,2013.55-95.
[29]工业和信息化部等.工业和信息化部等部门公开征求对《人工智能科技伦理管理服务办法(试行)》的意见[EB/OL].https://www.miit.gov.cn/jgsj/kjs/jscx/gjsfz/art/2025/art＿092a447008f340d3abd55819b8c8e5cf.html,2025-08-22.
[30]白钧溢,于伟.人工智能伦理教育的三重境界[J].中国电化教育,2025,(3):11-19.
[31]Herrmann T,Pfeiffer S.Keeping the Organization in the Loop:A Socio-technical Extension of Human-centered Artificial Intelligence[J].AI&Society,2023,38(4):1523-1542.
[33]Dekker S. Drift into Failure:From Hunting Broken Components to Understanding Complex Systems[M].Boca Raton:CRC Press,2016.6.
(1)弗兰克·吉尔斯(Frank Geels)指出,技术、行动者与制度三个维度在实践中始终相互关联,但出于分析的实用考虑,将其两两组合探讨,有助于更清晰地揭示社会-技术系统的运作逻辑。
(1)必须说明的是,联合优化并非意味着技术工件、行动者与制度在任何条件下都能实现同步最优,而是在承认价值张力、利益冲突与认知局限的前提下,通过对关键维度及其相互作用的协同调适,尽可能避免单一维度优化引发的系统失衡。
基本信息:
中图分类号:G434
引用信息:
[1]白钧溢,于伟.从“风险应对”到“技术向善”:教育人工智能伦理治理的社会-技术重构[J].中国电化教育,2026,No.475(08):101-109.
基金信息:
中国博士后科学基金第78批面上资助项目“基于价值嵌入的教育人工智能伦理治理研究”（课题编号：2025M781598）研究成果
2026-08-06
2026-08-06