东北大学学报(社会科学版) ›› 2021, Vol. 23 ›› Issue (3): 8-16.DOI: 10.15936/j.cnki.1008-3758.2021.03.002

• 科技哲学研究 • 上一篇    下一篇

“道德机器”的道德偏差与无人驾驶技术的伦理设计

钱圆媛   

  1. (中共中央党校(国家行政学院)哲学教研部,北京100091)
  • 发布日期:2021-05-21
  • 通讯作者: 钱圆媛
  • 作者简介:钱圆媛(1985-),女,重庆人,中共中央党校(国家行政学院)讲师,哲学博士,主要从事西方哲学研究。
  • 基金资助:
    国家社会科学基金重大资助项目(20&ZD041)。

Moral Deviations of “Moral Machines” and Ethical Approaches to Autonomous Vehicles

QIAN Yuanyuan   

  1. (Department of Philosophy, Party School of the Central Committee of C.P.C., Beijing 100091, China)
  • Published:2021-05-21
  • Contact: -
  • About author:-
  • Supported by:
    -

摘要: 无人驾驶技术的飞速发展催生了“电车困境”的现实新讨论。其中,MIT的“道德机器”实验声称,通过收集公众的道德偏好,有望解决无人驾驶汽车的道德算法问题。然而,该实验并不有助于无人驾驶技术的伦理设计,这部分源于试图抽象普遍地构建“电车困境”问题的方式,同时单靠一套自上而下的义务论或功利论去匹配数据结构和算法模拟,难以处理非线性、多因素、多冲突且情境敏感的现实道德问题。通过结合自下而上的机器学习与自上而下的符号推理,运用更相适配的联结主义网络与德性伦理学来发展具有诸良好特质或“德性”的无人汽车计算机系统,并与外部系统连接以达成整体系统的多功能融合,被认为是更有可能且更观照道德本性的无人驾驶伦理设计进路。

关键词: 无人驾驶汽车; 道德机器; 电车困境; 神经网络; 德性伦理学

Abstract: The rapid development of autonomous vehicles (AV) has brought the “trolley dilemma” into new discussions. Among them, MIT's “moral machines” claim the potential to figure out the moral algorithm of AV by collecting the public's moral preferences. However, this experiment does not contribute much to the ethical designs of AV, partly because of the abstract and universal way to construct the “trolley dilemma”. Also, it is difficult to deal with the non-linear, multi-factor, multi-conflict and context-sensitive moral problems by using a set of top-down deontological or utilitarian algorithm. In fact, combining bottom-up machine learning with top-down symbolic reasoning and using connectionist networks and virtue ethics to develop virtuous AV systems is considered to be more promising and fruitful in taking account of the essence of morality.

Key words: autonomous vehicle(AV); moral machine; trolley dilemma; neural network; virtue ethics

中图分类号: