摘要
人工智能的道德地位问题是一个真正新的道德难题。解决这个难题将涉及既有的"他心难题"和其变种"机心难题"。现有解决他心难题的诸方案中,类比推论、最佳说明理论和判据论都遇到了各自难以克服的困难。改造"态度方案",可以得到纯粹态度-推论消除的方案:以纯粹态度方式直接经验他心,但以推论排除无心灵实体确认他心。然而,这个方案对人工智能的机心难题是失效的,间接推论无法将那些通过了图灵测试但并不具有心灵能力的人工智能排除出去,由此也就无法确认任何有心灵的人工智能实体。"机心难题"作为一个元伦理学或道德形而上学问题,可以在规范伦理学的层次上回避或绕过。大致在康德伦理学的范围内,在确立关于手段或工具的诸位伦理学规则之后,我们可以根据人格伦理学和手段伦理学的规则确立如下对待人工智能的"选言命令式"。一个实体及其运行,要么冒犯了人,要么尊重了人。应该将一个冒犯人的实体看作既非目的也非手段的实体:当它是心灵实体时,限制其自由;当它是非心灵实体时中止其运行。应当尊重一个尊重你的实体的运行规则:当它是心灵实体时,尊重其本身;当它是一个非心灵实体时,尊重制造它的人。
What moral status should be allocated to AIs is a really new problem in ethics, which deals with the problem of other mind and its variant,the one of machinery mind.Several solutions,i.e.analogical inference,theory of best explanation and criteria theory,have encountered difficulties.Modify the attitudinal approach,we can get a new approach named 'pure attitudes/inferential eliminations:We directly experience other mind by pure attitudes and eliminate those entities without minds by inferences.But with such an approach we can't eliminate those AIs passed the Turing Test.The problem of machinery mind as a mataethical or moral metaphysical one,can be avoided on the level of normative ethics.Within Kantianism,we can establish an ethics of means,and then by combination with Kantian ethics of persons,we can deduce a disjunctive imperative to treat AIs as follows.We should regard any entity offending persons as neither end nor means,when it is a one with mind,we should limit its freedom;when it is a one without mind,we should stop its working.We should respect any entity respecting persons. when it is a one with mind,we should respect itself;when it is a one without mind,we should respect the person who makes it.
出处
《社会科学》
CSSCI
北大核心
2018年第12期110-120,共11页
Journal of Social Sciences
基金
国家社科基金重大项目"基于虚拟现实的实验研究对实验哲学的超越"(项目编号:15ZDB016)
教育部人文社科规划基金课题"实践知性与行动动机"(项目编号:15YJA720010)的阶段性成果
关键词
人工智能
他心难题
机心难题
手段伦理学
选言命令式
Artificial Intelligence
Problem of Other Mind
Problem of Machinery Mind
Ethics of Means
Disjunctive Imperative
