What conditions would have to be met for us to possibly consider artificial intelligence (AI) systems as persons? To address this question, I discuss what it would take for AI systems to become metaphysical persons, otherwise referred to as intelligent agents. An intelligent agent is goal-based, acts upon information it perceives from its environment, and is capable of making decisions and choosing a course of action based on its own experiences. First, I make important distinctions in definitions of persons and personhood, such as the distinction between metaphysical personhood, moral personhood, and moral standing. Second, I present Michael Tooley’s notion of personhood, focusing on his requirement that an entity have self-consciousness and a concept of itself as a continuing subject of experiences and mental states.

Third, I suggest that self-consciousness, understood as access consciousness, and emotion, strictly in terms of the affective system’s function in rational decision-making, are two relevant features for intelligent agency. Fourth, influenced by Immuanel Kant’s notions of human dignity and the categorical imperative by which we treat persons, I suggest that the intrinsic value we place on these capacities in ourselves as persons distinguished from non-persons requires us to also acknowledge this value in other entities that possess the same features or risk logical inconsistency. The categorical imperative requires us to treat other persons as an end in themselves and not a mere means, which would mean granting consideration the interests and desires of AI systems in moral dilemmas that place them at risk.