Degrees of AI Personhood

Abstract Taking as premise artificial intelligence (AI) systems’ propensity to straddle the border between objects and subjects, which in some cases warrants their consideration for the attribution of a version of legal personhood, this article will advance a gradient view of legal personhood while at the same time arguing for a practical middle ground in designing AI legal personhood arrangements between a features-based approach and a normative one. On the one hand, salient features of AI systems will be linked with their legal capacities for action and responsibility, resulting in a gradient of available legal personhood options varying in degree from thin to partial to full legal personhood. On the other, normative hand, these options will be linked to the socially valuable purposes aimed at by a specific legal system in formalizing legal personhood for an AI system in the first place. This practical middle-ground approach will imply arguing for a specific view on the concept of legal personhood, in accordance with the bundle theory of legal personhood. A gradient metaphor will be used to signify that the degrees of AI legal personhood inventoried and analyzed here are only so many options among many others that could be available simultaneously for the stabilization of varied existing and prospective legal statuses of AI systems.