Towards a Theory of AI Personhood

Abstract

I am a person and so are you. Philosophically and legally, we sometimes grant personhood to non-human animals, and even to entities such as rivers and corporations. But when, if ever, should we ascribe personhood to AI systems? In this paper, we outline necessary conditions for AI personhood, focusing on agency, theory-of-mind, and self-awareness. We discuss evidence from the machine learning literature regarding the extent to which contemporary AI systems, such as language models, satisfy these conditions. We argue that no current AI system could plausibly be considered a person.