科学技术哲学研究

随着人工智能自主性的不断提高，出现了人工智能道德责任的归属问题。责任主体一元论认为，人类应该承担全部的道德责任；责任主体二元论认为，道德责任应该由人类和人工智能共同承担；人工智能自我责任论则基于人工智能可发展的未来，认为人工智能作为具有智能和自主决策能力的实体，其所产生的道德责任问题应该归责于人工智能本身。究其三种争论的本质是对道德责任概念本身的不同理解，故产生了对道德责任归责对象的争议。斯特劳森从实践的角度出发，认为道德责任的归属问题应该是在“反应型态度”的表达上展开的。基于斯特劳森反应型态度理论分析人工智能责任归属问题，使道德责任归属问题更加明确。人工智能只能作为“客观态度”的处理对象，不被视为道德责任的承担者。而人类作为人工智能的研发、管理和使用者被视为“反应型态度”的对象，将承担全部的道德责任。

With the increasing autonomy of artificial intelligence, the problem of attribution of moral responsibility in artificial intelligence appears. The monism of responsible subject holds that humans should bear all moral responsibility; the dualism of responsible subject holds that moral responsibility should be shared by humans and artificial intelligence; the self-responsibility theory of artificial intelligence, based on the future that artificial intelligence can develop, holds that as an entity with intelligence and autonomous decision-making ability, the problem of moral responsibility generated by artificial intelligence should be attributed to artificial intelligence itself. The essence of the three kinds of arguments is the different understanding of the concept of moral responsibility itself, so there is a dispute about the objects of moral responsibility imputation. From the perspective of practice, Strawson believes that the problem in the attribution of moral responsibility should be developed on the expression of “reactive attitudes”. Based on Strawson's reactive attitudes theory, this paper analyzes the problem of responsibility attribution in artificial intelligence and makes the problem in the attribution of moral responsibility more clearly. Artificial intelligence can only be treated as an object of “objective attitudes” and is not seen as a bearer of moral responsibility. Humans, as developers, managers and users of artificial intelligence, are regarded as the object of “reactive attitudes” and will bear the full moral responsibility.