Abstract: “Value alignment” is a significant approach for the ethical governance of artificial intelligence. This approach holds a belief that when artificial intelligence aligns with human values, the behavioral choices it makes will conform to the standards of “right”. The meta-ethical stance that upholds “rightness” judgments as dependent on “value” judgments has been controversial in ethical studies since the beginning of the 20th century. The theoretical predicaments it faces, such as the inability to provide an analyzable explanation of values, the difficulty in accommodating duties within moral judgments, and the incapacity to justify the value of actions, are precisely the roots of the main practical challenges that artificial intelligence value alignment encounters. The opposite stance in meta-ethics, explains “value” through “rightness” and then “rightness” by “reasons”. This approach offers an analyzable explanation of values, balancing consequences and duties in moral considerations in a straightforward manner. It reveals the way moral principles works in practical reasoning and effectively resolves the difficulties faced by theories that indicate “rightness” by “value”. There have been various attempts to “align” the “reasons” of human being in the ethical design of artificial intelligence. Such attempts help to explore the fundamental factors that prompt artificial intelligence systems to act correctly, while clarifying the way “values” plays in moral reasoning.