📄 中文摘要
该研究认为理性的人并不拥有目标，理性的人工智能也不应当拥有目标。人类的行为之所以被视为理性，并不是因为我们将其指向某个最终的目标，而是因为我们将行为与实践相一致。这些实践包括一系列的行为、行为倾向以及行为评估标准。通过这种方式，德性伦理为人工智能的对齐提供了一种新的视角，强调了行为与社会规范之间的关系，而非单纯的目标导向。
德性伦理在人工智能对齐中的合理有效性
出处: The Reasonable Effectiveness of Virtue Ethics in AI Alignment
发布: 2026年2月19日
📄 English Summary
The Reasonable Effectiveness of Virtue Ethics in AI Alignment
The essay argues that rational individuals do not possess goals, and rational AIs should not have goals either. Human actions are considered rational not because they are directed towards some final 'goals', but because they align with practices, which encompass networks of actions, action-dispositions, and action-evaluation criteria. This perspective suggests that virtue ethics offers a new lens for AI alignment, emphasizing the relationship between actions and social norms rather than a purely goal-oriented approach.