人工智能体的道德设计及其面临的挑战

Moral Design of Artificial Agent and Its Challenge

作为人工智能技术的产物,人工智能体具有强烈的社会性特质,人们希望它能遵守人类社会的道德规范和伦理要求,以与人类和谐相处,并更好地服务于人类,而对人工智能体进行道德设计,是让它遵守人类道德的重要路径之一。人工智能体的道德设计有三条进路,其一是“自上而下”,其二是“自下而上”,其三是“自上而下”与“自下而上”的结合,它们都强调了道德推理和决策能力的重要性。目前人工智能体道德设计面临的主要挑战为:人工道德主体难以实现、道德认知和选择困难、计算机算力的限制、情感计算的限制,以及新技术的运用风险等。道德设计要实现的不是完全道德主体,而是有限道德主体。考虑到人类福利,这个有限道德主体应该是尊重人类道德价值观的,是“向善”而安全的。

As a product of artificial intelligence technology,artificial agent has strong characteristics of social behavior.People hope that it can abide by moral norms and ethical requirements of human society,so that it can get along with human beings harmoniously and serve human beings better.The moral design of artificial agent is one of the important ways to make it comply with human morality.There are three approaches to the moral design of AI:one is“top-down”,another is“bottom-up”and the third is the combination of“top-down”and“bottom-up”.All of them emphasize the importance of moral reasoning and decision-making ability.At present,the main challenges faced by the moral design of artificial agents are as follows:the difficulty of realizing artificial moral agents,moral cognition and selection difficulty,the limitation of computer computing power,the limitation of emotional computing,and the risk of using new technologies.What moral design wants to achieve is not a complete moral subject,but a limited moral subject.Considering human welfare,this limited moral subject respects human moral values and is"good"and safe.