道德计算是否可能——对机器伦理的反思

Is Possible to Compute Morality——Reflection on Machine Ethics

机器伦理是解决人工智能伦理困境的新方案,其倡导者主张,道德是可计算的,通过计算不仅可确保机器合乎道德地行动,而且可更好地探求道德的本质。但通过对机器伦理诸假设的分析发现,道德计算在方法论上面临着严峻挑战,不仅无法确保机器的道德行动,而且会动摇机器伦理的必要性基础。更重要的是,道德计算主义必然将人类导向一种旁观者立场,这或许是一个满足机器伦理要求的道德解释,但对于人类来说,却是一个潜藏着系统性风险的道德陷阱。因此,基于道德计算主义的机器伦理并不是一个解决人工智能伦理困境的有效方案,我们仍需要将伦理规制的重心放在人类设计者身上。

Machine ethics is a new solution to the ethical dilemma of artificial intelligence. Its advocators claim that morality is computable,and that morality computation can not only ensure the ethical actions of machines,but also better explore the essence of morality. However,through the analysis of the assumptions of machine ethics,it is found that moral computationism faces severe challenges in methodology,not only failing to ensure the moral actions of machines,but also shaking the necessary foundation of machine ethics. More importantly,moral computationism will inevitably lead human beings to be a bystander,which may be a moral explanation to meet the ethical requirements of machines,but for human beings will be a moral trap with hidden systemic risks. Therefore,machine ethics based on moral computationism is not an effective solution to the ethical dilemma of artificial intelligence,and we still need to focus on human designers.