AI Personhood, Criminal Law, and Punishment

摘要

Abstract This article considers what kind of personhood AI agents and robots might have under criminal law. The answer depends on whether the personhood is grounded in the ontological qualities of entities or whether its application has functional values. Ontologically, two stages are identified. The first stage considers AI agents or robots as moral patients (for example, if they can experience pain), which could warrant protection under criminal law—granting them rights but not duties. The second stage involves the capacity for moral reasoning, enabling them to understand right from wrong. This stage also assesses their ability to be criminally punished. This requires not only the capacity to feel pain but also the ethical permissibility of imposing such punishment and the ability to comprehend punishment and pain. From a functional perspective, personhood could be assigned to AI entities irrespective of their internal qualities if it serves practical purposes. This could involve granting rights to protect legal values that might be threatened by deploying AI technologies. The functional rationale could also be applied to make artificial agents responsible—and by imposing punishment on them, to fill a responsibility gap. Despite the wide range of positions on how to respond to the emergence of AI agents and robots in the context of criminal law, it appears that there is not yet a satisfactory response and so more work needs to be done.