人工智能技术目前已经在全世界的范围内引起广泛的伦理讨论，如何规范人工智能，尤其是人工智能的物质形态——智能机器，成了一个迫切的问题。但是伦理学对于智能机器以何种角色参与到伦理讨论中，没有很好的界定。而道德能动性正是判断智能机器是否具有道德可考虑性的关键。因此，本研究关注的核心问题是：智能机器的道德能动性何以可能？ 第一，道德主体问题。在现代性技术社会中，人的主观自我已经不是社会道德交往的基础，反而社会性、共识性的自我，即道德身份，成了交往的基础。本文认为，人与机器存在道德自我趋同的情况，是因为道德他者的技术性缺席。人们认为智能机器不是道德能动者，是由于对于他者性的理解偏差，从而对智能机器产生智能偏见、期望偏见与地位偏见。第二，道德能力问题。一般认为智能机器具有道德推理能力，但是缺失人类所具有的、最关键的道德情感能力。本文认为，智能机器道德能力的基础应该是直觉，这种机器直觉基于情感转向之后的情感内涵以及直觉的知觉经验解释，即基于社会普遍期望与道德沟通的机器直觉。道德期望是连接自我与他者的方式，可以表征双方的道德意向性，是建立他者型道德能动者的基础。第三，道德地位问题。道德共同体是经典伦理学的基本语境，是对于每个道德准则的主体条件与适用范围的社会性描述。在智能机器是否能进入道德社会、并被社会成员接受的问题上，道德共同体主张一种普遍意义上的人类中心主义，即中心主义。本文认为，我们应该主张智能机器作为一种差异性的边缘成员，从而具有作为他者而应该具有的权利。智能机器的外观设计目前遵循拟人化的进路，拟人化设计带来问题的根源在，认为类人的形象才是道德能动者的形象。本文认为，人工物作为他者的图像应该是社会性且差异化的，其互动方式应该是他者型的。社会机器人的设计应该遵循以他者性为核心的差异化设计路径。总之，智能机器应该具有道德相关性意义上的道德能动性，即智能机器作为他者而成为道德主体，被他人期望而具有道德能力，享有反中心化共同体中的道德成员地位，这是一种新的讨论道德能动性的维度。

Artificial intelligence （AI） has been raised wide concern around the world. How to regulate AI, especially the intelligent machines as the physical form of AI, turns to be a crucial problem. Ethics of technology, however, has not considered it sufficiently that which character the intelligent machine should be granted for participating in ethical situations. From the perspective of ethics, moral agency is the key to judging whether intelligent machines are morally considerable.Therefore, in this thesis, our main question is that, could intelligent machine granted moral agency? We will discuss that in three aspects.For moral subjectivity, we should start from the moral self of human beings. In modern technical society, the subjective self is no longer the main foundation of social communication. Oppositely, the social and consensus self, that is, moral identity, has become the basis of communication. We argue that convergence between the moral self of man and machine is due to the technical absence of the moral other. Redefining ourselves and others from the perspective of others, we will find that the autonomy and intentionality of moral subjects are defective. Even from responsibility, on account of misunderstanding the otherness, people just think that intelligent machines are not moral agents, which creating intelligent prejudice, expectation bias and status bias on intelligent machines.Second, for the intelligent ability of intelligent machines, primarily we should discuss the difference between moral affections and moral rationality. It is generally believed that intelligent machines have the ability to reason morally, but lack the most critical ability of moral emotion that only belong to human beings. We propose that the basis of moral ability of intelligent machine should be the intuition. This machine intuition is based on the definition of affection after the affective turn and the perceptual experience explanation of intuition, that is, based on the social common expectation and moral communication. Moral expectations are the core of the intrinsic representation between the self and the other, and the basis for the establishment of moral agents as others.Third, for the moral status of intelligent machines, the prior to consider about is moral community. The moral community constructs the basic context of classical ethics, which is a social description of the subjective conditions and practical scope of each moral standard. This centralism of moral community decides whether intelligent machines can enter the moral society and be accepted by members of society. The radical views of machine ethics are opposed to this centralized moral community. We believe it should be advocated that intelligent machines as a marginal member of difference and thus are empowered as others.The appearance design of intelligent machines currently follows an anthropomorphic approach. The key problem of anthropomorphic design is that the image of the human being is equivalent to the image of the moral agent. We argue that, the image of artifacts as others should be more public and distinctive; the way of interaction should be others-type instead of emotional projection. The anthropomorphic design brings about the fallacy of self, and the dissimilar design should go against the uncanny valley, and construct a dissimilar design approach centered on others.Consequently, intelligent machines should have the moral agency in the sense of moral relevance, in other words, intelligent machines should turn into moral subjects as others, have moral ability due to being expected by others, and be granted the status of moral membership in an anti-centralized community, which could be a new dimension that discusses moral agency.