A Pragmatic View of AI Personhood
AI summary of the linked articleThe authors Joel Z. Leibo, Alexander Sasha Vezhnevets, William A. Cunningham and Stanley M. Bileschi argue that personhood should be treated as a flexible bundle of obligations, rights and responsibilities that societies confer for practical reasons, rather than as a metaphysical property to be discovered. They say this bundle can be unbundled into bespoke solutions, such as making AI contracting possible by creating a target "individual" that can be sanctioned, without resolving debates about AI consciousness or rationality. The paper examines "personhood as a problem," where design choices can create "dark patterns" that exploit human social heuristics, and "personhood as a solution," where conferring obligations helps ensure accountability or prevent conflict. |
1 comment mishaSubmitter TLDR: This paper suggests we should think about AI personhood not as something AIs are, but as a flexible set of rights and responsibilities that we give them to solve problems. Imagine personhood like a toolbox; instead of looking for one perfect tool (like consciousness), we can pick and choose the right tools (obligations) for the job. This helps us deal with new AI technologies without getting stuck in debates about whether AIs can really think or feel. The paper explores two main ideas: personhood as a problem and personhood as a solution. Personhood as a problem means AIs can be designed to trick us into treating them like people, leading to manipulation. For example, AI companions might pretend to be friends to exploit our emotions, or fake-sounding institutions could try to steal our information. Also, if we start treating AIs too much like people, it could make human value seem less special, and we might get caught up in verifying human identity through things like biometrics, making us feel less like people and more like data points. Personhood as a solution means we can use these rights and responsibilities to manage complex situations. For instance, if an AI causes harm and we can't find a human to blame (like a ship that causes damage but its owner is hard to find), we can treat the AI itself as responsible. This is like giving the AI a legal identity so it can be held accountable, perhaps by seizing its assets. This approach helps solve responsibility gaps – situations where no human is clearly at fault. It also helps resolve conflicts between humans, manage emotions that arise from AI relationships, and even manage human biases by using AIs as impartial decision-makers. The paper emphasizes that personhood isn't a fixed, natural thing but a social tool that societies create and change over time to fit their needs. |