生成式人工智能对认知自主性的侵蚀风险与应对策略Erosion Risks of Cognitive Autonomy by Generative AI and Their Countermeasures
黄振伟,肖巍
摘要(Abstract):
生成式人工智能凭借其信息检索、内容输出等优势提升了人类活动效率,但同时也带来了对认知自主性的侵蚀风险。其核心风险在于,生成式人工智能有可能重构认知主体形成判断的条件。从认知定向、认知探究和理由评估三个维度理解认知自主性,可以揭示生成式人工智能对认知自主性的侵蚀机制,并据此分析注意力俘获、认知外包和认知锚定三重风险。这些风险在具体使用场景中交叉呈现,并且可能由个体层面的使用风险转化为制度性风险。据此,应对认知自主性侵蚀风险的重点应当从事后归责前移至判断形成过程的保护,通过前置风险识别、过程性权利保障和场景化义务,构成逐级递进的完整应对框架,从而在客观技术依赖不断加深的情况下,保护用户的自主思考、判断和选择能力。
关键词(KeyWords): 认知自主性;生成式人工智能;大语言模型;科技伦理
基金项目(Foundation): 国家社会科学基金重大项目“构建人类卫生健康共同体的伦理路径研究”(21&ZD057)
作者(Author): 黄振伟,肖巍
DOI: 10.16112/j.cnki.53-1160/c.2026.05.121
参考文献(References):
- [1]Li J W,Gao Y,Yang Y Z,et al.Fundamental capabilities and applications of large language models:a survey[J].ACM Computing Surveys,2026,58(2):38.
- [2]刘鑫怡,徐峰,司伟攀.生成式人工智能应用伦理风险的形成机理及治理策略研究[J].中国软科学,2025(10):194-204.
- [3]毕文轩.生成式人工智能的风险规制困境及其化解:以ChatGPT的规制为视角[J].比较法研究,2023(3):155-172.
- [4]郑煌杰.生成式人工智能的伦理风险与可信治理路径研究[J].科技进步与对策,2025,42(12):38-48.
- [5]Izquierdo-Condoy J S,Arias-Intriago M,Tello-De-la-Torre A,et al.Generative artificial intelligence in medical education:enhancing critical thinking or undermining cognitive autonomy?[J].Journal of Medical Internet Research,2025,27.
- [6]Christman J.Autonomy in moral and political philosophy[M/OL]//Zalta E N,Nodelman U.The Stanford encyclopedia of philosophy.Fall 2025 ed.Stanford:Stanford University,2025.https://plato.stanford.edu/archives/fall2025/entries/autonomy-moral/.
- [7]康德.道德形而上学的奠基:注释本[M].李秋零,译注.北京:中国人民大学出版社,2013:69-71.
- [8]约翰·穆勒.论自由[M].孟凡礼,译.上海:上海三联书店,2019:9-10.
- [9]Jose B,Joseph D,Mohan V,et al.Outsourcing cognition:the psychological costs of AI-era convenience[J].Frontiers in Psychology,2025,16:1645237.
- [10]Laitinen A,Sahlgren O.AI systems and respect for human autonomy[J].Frontiers in Artificial Intelligence,2021,4:705164.
- [11]Prunkl C.Human autonomy at risk?An analysis of the challenges from AI[J].Minds and Machines,2024,34(3):26.
- [12]Council of Europe Committee of Ministers.Declaration by the Committee of Ministers on the manipulative capabilities of algorithmic processes [R/OL].2019-02-13.https://rm.coe.int/090000168092dd4b.
- [13]von Arnauld A,von der Decken K,Susi M.The Cambridge handbook of new human rights:recognition,novelty,rhetoric[M].Cambridge:Cambridge University Press,2020:389.
- [14]国家新一代人工智能治理专业委员会.新一代人工智能伦理规范[EB/OL].(2021-09-25)[2026-04-01].https://www.safea.gov.cn/kjbgz/202109/t20210926_177848.html.
- [15]Council of Europe.Guidance note on the implications of generative artificial intelligence for freedom of expression:CDMSI(2025)15rev[R/OL].2025-12-03.https://rm.coe.int/cdmsi-2025-15rev-guidance-note-on-the-implications-of-generative-artif/488029df80.
- [16]Winner L.Autonomous technology:technics-out-of-control as a theme in political thought[M].Cambridge:MIT Press,1978:227.
- [17]Vallor S.The AI mirror:how to reclaim our humanity in an age of machine thinking[M].New York:Oxford University Press,2024:120.
- [18]贝尔纳·斯蒂格勒.技术与时间[M].裴程,赵和平,印螺,等,译.南京:译林出版社,2023:645.
- [19]Stiegler B.Taking care of youth and the generations[M].Stanford:Stanford University Press,2010:97.
- [20]Bakir V,McStay A.Move fast and break people?Ethics,companion apps,and the case of Character.ai[J].AI & Society,2025,40(8):6365-6377.
- [21]Chaudhary Y,Penn J.Beware the intention economy:collection and commodification of intent via large language models[J].Harvard Data Science Review,2023(Special Issue 5):1-25.
- [22]余胜泉,汪凡淙.人工智能教育应用的认知外包陷阱及其跨越[J].电化教育研究,2023,44(12):5-13.
- [23]汪凡淙,汤筱玙,余胜泉.基于生成式人工智能的认知外包:交互行为模式与认知结构特征分析[J].心理学报,2025,57(6):967-986.
- [24]Bu??inca Z,Malaya M B,Gajos K Z.To trust or to think:cognitive forcing functions can reduce overreliance on AI in AI-assisted decision-making[J].Proceedings of the ACM on Human-Computer Interaction,2021,5(CSCW1):188.
- [25]丹尼尔·卡尼曼.思考,快与慢[M].胡晓姣,李爱民,何梦莹,译.北京:中信出版社,2012.
- [26]Dahl M,Magesh V,Suzgun M,et al.Large legal fictions:profiling legal hallucinations in large language models[J].Journal of Legal Analysis,2024,16(1):64-91.
- [27]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,等.生成式人工智能服务管理暂行办法[A/OL].(2023-07-10)[2026-04-01].http://www.gov.cn/zhengce/zhengceku/202307/content_6891752.htm.
- [28]国家互联网信息办公室,工业和信息化部,公安部,等.互联网信息服务算法推荐管理规定[A/OL].(2021-12-31)[2026-04-21].https://www.cac.gov.cn/2022-01/04/c_1642894606364259.htm.
- [29]科技部,教育部,工业和信息化部,等.科技伦理审查办法(试行)[A/OL].(2023-09-07)[2026-04-21].https://www.gov.cn/gongbao/2023/issue_10826/202311/content_6915814.html.
- [30]国家互联网信息办公室,工业和信息化部,公安部,等.人工智能生成合成内容标识办法[A/OL].(2025-03-07)[2026-04-21].https://www.gov.cn/zhengce/zhengceku/202503/content_7014286.htm.
- [31]吴楷文.场景化治理:人工智能操纵技术的法律因应[J].探索与争鸣,2026(2):123-132.