弱道德人工智能可行吗--从精神医学用途到道德增强

Is Weak Moral Artificial Intelligence Feasible?:From Psychiatric Use to Moral Enhancement in General

弱道德人工智能试图通过采取信息增强的方式达到道德增强的目的。试图探索一种苏格拉底助手式的弱道德人工智能的可行性。从精神医学中使用的聊天机器人具体例子引入,在预设这种精神医学应用前景的基础上追问一般性的道德信息增强是否可行。进一步探究弱意义上的道德增强绕过动机这一做法的合理性,借助于梳理哲学中历来关于道德动机的论争,并经由动机伦理和康德式义务论视角来论证,指出弱道德人工智能设想中隐含的悖论,在结论部分尝试性地给出解决这一悖论的可能方案。

The model of weak moral artificial intelligence(AI)is proposed for moral enhancement through info-enhancement rather than biomedical means.This paper aims to discuss the feasibility of a weak moral AI as Socratic assistant.Beginning with chatbots used lately in psychiatry,this paper makes a detailed inquiry on the feasibility of moral enhancement in a general sense,then further explores the reasonability of bypassing the motivations in a weak sense.It tries to show hidden paradox by appealing to traditional debates on moral motivation in philosophy and arguing through the perspective of motivation ethics and Kantian deontology.The conclusion part tries to provide some suggestions as alternatives to such paradox.