人工智能必须考虑风土性因素——以“电车难题”为母题

Artificial Intelligence,Trolley Problem and the Involvement of Cultural-geographical Factors

对于与自动驾驶有关的伦理学规范的讨论,往往以“电车难题”为母题。不过,西方学界处理电车难题的思路,主要就是在义务论思想与功利论思想之间打转,而这种视野上的局限性也影响了西方人工智能专家开发用以处理道德两难的自动推理系统的思路。而在上述处理进路中被普遍忽视的,便是“风土性因素”在紧急避险的过程中所发挥的作用——尽管这些因素的存在,已经得到了大量经验证据的支持。在这里,“风土性因素”可以被理解为在特定文化与地理场域中具有特定特征的司机关于如何进行紧急避险的缄默习惯。从立法角度看,未来可能出现的针对全自动驾驶系统的法律法规,应当在合适的尺度上容纳上述“风土性因素”,以便降低可能的执法成本。而这一诉求自然就会倒逼今天我国的应用伦理学研究者对我国各地区司机所默认的紧急避险原则进行系统的调研与统计。而从技术角度看,要在为自动驾驶技术匹配的人工智能系统中以某种足够灵活的方式实现“风土性”原则,我们就需要对现有的基于符号人工智能与深度学习的主流技术路径进行大胆挑战。

The tolley problem has long played a salient role in the studies of the ethical codes applied to automated driving vehicles.The framework underlying the mainstream narrative for formulating the problem is either deontological or consequentialist.However,neither framework can do justice to the culural-geographical factors involved in the mental procedure for handling moral dilemmas,although solid empirical evidence supports the existence of these factors.This essay argues that the laws and regulations applied to automated driving vehicles must include these cultural-geographical factors to fit the local cultures.Due to this consideration,a comprehensive survey of local human drivers'atitudes towards varieties of trolley-problem-alike-moral dilemmas should be carried out before enacting relevant laws.Furthermore,AI scientists also need to develop new approaches to include cultural-geographical factors in automated driving softwares.