Among the things that can be considered as the goal of AI value alignment, the consensus values of all humanity, which transcend both individual users and communities, should be chosen. Among the many factors of the consensus values of all humanity, natural justice—the only truly existing moral law—should be chosen. In short, the goal of AI value alignment should be natural justice. Natural justice is this: a player views any game in which any player of the current game is involved and which has a structure similar to that of the current game as a stage game of the same indefinitely repeated game, and thus, from this perspective, his strategy (action) is (1) in the first round, to cooperate, (2) from the second round onwards, to reciprocate, i.e., to reward or to punish, but if he defected in the previous round, to correct his own fault. Using three methods, namely global overlapping consensus, the "veil of ignorance" thought experiment, and the social choice thought experiment, to determine the goal of AI value alignment, all lead to the same conclusion: the goal of AI value alignment should be natural justice. Several issues requiring further research are proposed.

artificial intelligence (AI)  /  value alignment  /  natural justice  /  moral law  /  game theory  /  ethics  /  cooperation  /  reciprocation