Abstract: Regardless of the perspective of internal mind or external performance,the judgment on the moral role and function of artificial intelligence agents is confronted with an insurmountable theoretical dilemma.Therefore,it is necessary to look at the moral practice of artificial intelligence agents from the perspective of interaction construction.The moral practice of artificial intelligence agents is gradually realized through the interaction among social regulations,human beings,artificial non-intelligence agents,other artificial intelligence agents and technical specifications.This process is not only an ongoing maturation of the “intelligence”,but also the process of artificial intelligence agents showing their moral potentials and realizing their goodness.From the perspective of engineering technology,artificial intelligence agents can carry out the symbolic transformation of moral principles,norms and demands in the practical context through three paths:top-down,bottom-up,and hybrid.Besides,they can realize the information exchange through interaction and the output of “goodness”.From the social perspective,the intention value will be written and evaluated in the design stage of artificial intelligence agents,forming a script for the moral output of artificial intelligence agents.In the stage of their promotion and application,the realization of artificial intelligence agents' intention value must be adjusted according to the social mainstream value,and it is necessary to create a social atmosphere that is conducive to the value adjustment function of artificial intelligence agents.In the stage of usage,the moral practice of artificial intelligence agents needs to be constructed together with their users,and the intention value and realization value are integrated through interaction.

Key words: artificial intelligence agents, moral practice, interaction construction, the practice context