How Does Embodied Artificial Intelligence Act As A Moral Agent

Artificial intelligence(AI)ethics is an important aspect of the study of AI philosophy,and its problems are often more difficult and complex than those of human ethics.Whether AI has the status of moral agent is the primary problem that researchers have to solve.This paper focuses on the core issue of “how can AI participate in moral practice as a moral agent”,and carries out many aspects of analysis and demonstration.How does AI acts as a moral agent corresponds to three questions:(1)Is AI a moral agent?(2)how does AI understand ethical rules?(3)how does AI practice ethical rules(implement ethical rules)? The first concerns whether AI is respected by people,and the last two concerns how AI respects people.These three questions correspond to three challenges of AI metaethics: moral metaphysics problem,the symbol grounding problem in morality and the moral frame problem.The underlying puzzle behind the three problems is the problem of machine mind.As long as the problem of machine mind is solved,all three problems can be solved.The problem of machine mind is a variety of other mind problem,so we need to find a solution of other mind problem that is suitable for the problem of machine mind.This paper analyzes and selects four popular theories of mindreading(theory theory,simulation theory,embodied simulation theory,and direct access theory),and gives a Wittgenstein style solution to other minds problem——embodied direct access theory——in combination with embodied cognitive theory.In addition,starting from the other minds problem,the paper refutes the argument of Searle's Chinese room,and defends the strong(embodied)AI with the help of the embodied direct access theory and embodied Turing test,thus giving the conditions for AI as a moral agent.After proving that AI has the status of moral agent,this paper uses four versions of moral Turing test,relevant evidence of moral psychology,moral conceptual metaphor studies and the embodied analysis of moral rules to demonstrate that AI can not onlyunderstand ethical rules,but also make corresponding moral practice,so gives a response to core issue of “how does embodied AI act as a moral agent”.