具身性、认知语言学与人工智能伦理学

Embodiment,Cognitive Linguistics and Ethics of Artificial Intelligence

在主流的人工智能伦理学研究中,很少有人意识到:将伦理规范的内容转化为机器编码的作业,在逻辑上必须预设一个好的语义学理论框架,而目前主流人工智能研究所采用的语义学基础理论却恰恰是成问题的。文章主张在人工智能与人工智能伦理学的研究中引入认知语言学的理论成果,并在此基础上特别强调"身体图式"在伦理编码进程中所起到的基础性作用。依据此番立论,并主张:机器伦理学的核心关涉将包括对人工智能的"身体"——而不仅仅是"心智"——的设计规范,即必须严肃考虑"怎样的外围设备才被允许与中央语义系统进行恒久的接驳"这一问题。

The issue on how to make ethical codes ＂computable＂ in order to build artificial moral agents （AMA） is fundamentally a semantic issue on how to semantically represent the relevant norms by appealing to some proper algo- rithms. However, unfortunately, the mainstream semantic theories employed in artificial intelligence （AI） are problematic in this aspect or another; and more unfortunately, the mainstream studies in the ethics of artificial intelligence haven＇t realized the relevance of semantic considerations to ethics yet. What the author of this paper intends to propose include three points. Firstly, cognitive linguistics （CL）, especially its notion of ＂cognitive schema＂, could be a useful resource for building the needed semantic framework for AI; secondly, due to the fundamental status of the notion of ＂embodiment＂ in the whole CL-narrative, we can hardly imagine any semantic representations of ethical codes in the case of of designing AMAs if these representations are not based on certain physical features of the ＂bodies＂ of AMAs~ thirdly, due to the preceding considerations, issues like ＂what kind of peripheral equipment is allowed to be permanently connected to the information-processing center of a certain AMA＂ should be put on the table for all ethieists of AI.