结构互动程序：人工智能时代“电车难题”伦理新解

doi:10.13438/j.cnki.jdxb.2026.04.017

摘要/Abstract

摘要：面对人工智能时代的伦理挑战，传统的义务论与功利主义均存在局限。义务论因恪守抽象律令而难以应对具体情境，功利主义则因无法量化幸福与价值而陷入实践困境。二者因忽视个体性动机和具体情境等因素交织，对多元价值进行单一标准排序，导致面对复杂伦理生活时可能“束手无策”。若直接将其嵌入算法中，则都可能导致道德上不可接受的后果。近年提出的古德尔三阶段模型与风险管理方案虽试图超越传统，但仍偏重客观可量化的风险因素，未能妥善处理道德主体性与公正性问题，容易陷入伦理悖论。可见，伦理困境的解决不能仅依赖既定原则或计算，而应立足于具体情境中多种因素的动态互动与结构化过程。由此提出“结构互动程序”视角，主张伦理规范是在特定情境因素反复互动、竞争与演化中逐渐生成的稳定解法。对于“电车难题”这类极端低频情境，因其缺乏足够的互动频次与社会结构化过程，难以形成普适的道德算法；人工智能伦理应注重情境因素的分解、多元反馈的引入以及互动框架下的演化学习，才能在复杂现实中形成具有实践生命力的伦理方案。该框架并非提供终极道德答案，而是强调在人工智能时代应对伦理困境时须重视情境互动与演化过程的实践必然性。

关键词： 人工智能, 自动驾驶, 应用伦理, “电车难题”, 结构互动程序, 情境主义

Abstract: Traditional deontology and utilitarianism both prove inadequate in addressing ethical challenges brought by artificial intelligence.Bound by abstract moral imperatives,deontology fails to adapt to real-world circumstances,whereas utilitarianism runs into practical hurdles since happiness and value defy quantitative measurement.Overlooking the interplay between individual motivations and contextual specifics,the two doctrines apply a uniform criterion to rank plural values,leaving them ineffective amid intricate ethical realities.Either theory,if coded into algorithms directly,may produce morally objectionable outcomes.Recent efforts including the three-stage Goodall model and relevant risk management frameworks seek to move beyond conventional theories,yet they still overemphasize objectively quantifiable risks.Neglecting moral subjectivity and impartiality,these approaches are prone to ethical paradoxes.Clearly,resolving ethical dilemmas requires more than rigid principles or numerical calculations;it hinges on the dynamic interaction and structuring of diverse factors within given contexts.This paper therefore puts forward the concept of "structural interaction procedure",arguing that ethical norms evolve into stable resolutions through continuous interaction,competition and iteration of contextual elements.For extreme,low-frequency scenarios like the"trolley problem",the lack of sufficient interaction and social structuring makes it impossible to develop a universal moral algorithm.To formulate ethically sound and workable solutions for complex real-world scenarios,AI ethics needs to focus on dissecting situational variables,incorporating multi-dimensional feedback,and advancing evolutionary learning within interactive frameworks.This framework is not intended to deliver definitive moral answers.Instead,it underscores the practical imperative of valuing contextual interaction and evolutionary dynamics when tackling ethical dilemmas in the artificial intelligence era.

Key words: artificial intelligence, autonomous driving, applied ethics, trolley problem, structural interaction procedure, contextualism

王倩茹. 结构互动程序：人工智能时代“电车难题”伦理新解[J]. 吉首大学学报（社会科学版）, 2026, 47(4): 168-176.

Wang Qianru. Structural Interaction Procedure:A New Ethical Interpretation of the"Trolley Problem" in the Age of Artificial Intelligence[J]. Journal of Jishou University(Social Sciences Edition), 2026, 47(4): 168-176.

参考文献

[1] Lin P.Why Ethics Matters for Autonomous Cars?[M].Autonomous Driving,2016：69-85.
[2] Goodall N J.Can you program ethics into a self-driving car?[J].IEEE Spectrum,2016,53(6)：28-58.
[3] Virilio P.The Original Accident[M].Polity Press,2007:10.
[4] 伊曼努尔·康德.道德形而上学原理[M].苗力田,译.上海：上海人民出版社，2002：18-53.
[5] Shulman C,Rovane C.Moral uncertainty and the possibility of morality[J].Philosophical Studies,2009,143(1)：1-17.
[6] Ecoffet A,Lehman J.Reinforcement learning under moral uncertainty[C].Proceedings of the 38th International Conference on Machine Learning,2021：2926-2936.
[7] Robinson P.Moral disagreement and artificial intelligence[J].AI & Societ,2024,39(5)：2425-2438.
[8] Lockhart T.Moral uncertainty and its consequences[M].Oxford:Oxford University Press,2000：10-15.
[9] Birnbacher D.Analytische Einfuehrung in die Ethik[M].Berlin:Walter de Gruyter,2007：270-290.
[10] Noothigattu R,Gaikwad S,Awad E,et al.A Voting-BasedSystem for Ethical Decision Making[J].Proceedings of the AAAI Conference on Artificial Intelligence，2018：1587-1594.
[11] Harris J.The Immoral Machine[J].Cambridge Quarterly of Healthcare Ethics,2020,29(1)：71-79.
[12] Ethics Commission on Automated and Connected Driving.Report of the Ethics Commission:Automated and Connected Driving.Federal Ministry of Transport and Digital Infrastructure[R/OL].[2025-01-02].https://www.bmvi.de/SharedDocs/EN/publications/report-ethics-commission.pdf?__blob=publicationFile.
[13] Bayertz K.Praktische Philosophie als angewandte Ethik[M].in K,1994：19.
[14] Nyholm S,Smids J.The Ethics of Accident-Algorithmsfor Self-Driving Cars:An Applied Trolley Problem?[J].Ethical Theoryand Moral Practice,2016,19(5)：1275-1289.
[15] Goodall N J.Awayfrom Trolley Problems and toward Risk Management[J].Applied Artificial Intelligence,2016,30(8)：810-821.
[16] Geisslinger M,Poszler F,Betz J,et al.Autonomous Driving Ethics:from Trolley Problem to Ethics of Risk[J].Philosophy & Technology,2021(7729):1-23.
[17] Rawls J.A Theory of Justice[M].Cambridge:The Belknap Press of Harvard University Press,1971：136-142.
[18] Hardin R.Morality within the Limits of Reason[M].Chicago:University of Chicago Press,1988：26-28.
[19] Schelling T C.The Strategy of Conflict[M].Cambridge:Harvard University Press,1980：89-91.
[20] Smith J M.Evolution and the Theory of Games[J].Cambridge:Cambridge University Press,1982：96-101.
[21] Bernard W.Moral Luck：Philosophical Papers 1973—1980[M].Cambridge:Cambridge University Press,1981.
[22] Amartya S.Consequential Evaluation and Practical Reason[J].Journal of Philosophy,2000(9)：477-502.
[23] Axelrod R.The Evolution of Cooperation[M].New York:Basic Books,1984：33-39.
[24] Alexander J M.The Structural Evolution of Morality[M].Cambridge:Cambridge University Press，2007：5-8.
[25] Awad E,Souza S,Kim R,et al.The moral machine experiment[J].Nature,2018：563:59-64.
[26] Bostrom N,Yudkowsky E.The ethics of artificial intelligence[M]//Frankish K,Ramsey W M.The cambridge handbook of artificial intelligence.Cambridge:Cambridge University Press,2014：316-334.
[27] 闫瑞峰,邢畅.人工智能价值对齐的伦理限度与实践进路[J].西南民族大学学报(人文社会科学版),2025(10):154-162.
[28] 李伟,李侠.如何对人工智能道德运气进行道德评价?[J].科学技术哲学研究,2025(4):9-16.