AI Personhood Framework Enables Governance of Agentic Artificial Intelligence, Addressing Concrete Problems

The increasing sophistication of artificial intelligence compels society to consider new definitions of personhood, and a team led by Joel Z. Leibo, Alexander Sasha Vezhnevets, and William A. Cunningham from Google DeepMind and the University of Toronto addresses this challenge with a novel framework. Their work moves beyond philosophical debates about AI consciousness, instead proposing that personhood functions as a practical set of obligations, rights and responsibilities, which societies assign to entities to effectively manage governance. This approach allows for the creation of adaptable solutions, such as enabling AI to enter into contracts and be held accountable, without requiring resolution of complex questions about artificial intelligence’s internal states. By treating personhood as a flexible tool, rather than a fixed property, the researchers, including Stanley M. Bileschi from Google DeepMind, offer a pragmatic path towards integrating increasingly capable AI agents into the fabric of society and ensuring responsible innovation.

Cunningham, and Stanley M. This work proposes a pragmatic framework for navigating this diversification by treating personhood not as a fixed quality to be discovered, but as a flexible bundle of obligations, encompassing both rights and responsibilities, that societies confer upon entities for specific reasons, especially to solve concrete governance problems. Researchers argue this bundle can be adapted, creating bespoke solutions for different contexts, and allowing for practical tools, such as facilitating AI contracting by creating a target “individual” that can be sanctioned, without needing to resolve debates about an AI’s consciousness or rationality. They explore how individuals fit into social roles and discuss decentralized digital identity technology, examining potential pitfalls where design choices can exploit human tendencies and conversely, how conferring obligations can ensure accountability.

Author Lists and Attribution Analysis

Agapiou, J. E., Cross, L., Cunningham, W. A., Dafoe, J., De Filippi, P., Duéñez-Guzmán, E. S., Ivliev, S., Jacobs, J., Karmon, D., Kerrigan, C., Köster, R., Koshiyama, A., Leibo, J. H., Riché, M., Rorty, R., Rosenfeld, H. M., Rosemont Jr, H., Rosenfeld, N., Schoernig, M., Schlager, E., Schelling, T.

C., Schluter, K., Shanahan, M., Shook, J. R., Siri, S., Siddarth, D., Singer, P., Sfeir-Tait, S., Soares, N., Sutcliffe, K., Sunstein, C. S., Vinitsky, E., Walzer, M., Wang, F., Ward, F. R., Weick, K.

Personhood as Adaptable Social Obligations and Rights

This work proposes a pragmatic framework for understanding personhood, moving away from metaphysical definitions and instead treating it as a flexible bundle of obligations, rights and responsibilities that societies confer upon entities. Researchers argue this bundle can be adapted to solve concrete governance problems, particularly as artificial intelligence becomes more prevalent, without requiring resolution of debates about AI consciousness. The study explores how individuals fit into social roles and examines decentralized digital identity, identifying potential pitfalls where design choices can exploit human heuristics and conversely, how conferring obligations can ensure accountability. The research highlights that societal change often occurs through discrete jumps between stable states, driven by collective sense-making processes that determine which equilibrium is selected next.

Evidence shows that both organic and deliberate actions, such as legal changes during the COVID-19 pandemic or staggered legalization of gay marriage in the United States, can influence norm change dynamics, demonstrating government’s active role beyond simply following cultural shifts. Researchers emphasize the historical contingency of personhood, noting that the Western conception of the individual as a locus of moral worth is not universal. The study distinguishes personhood from property, noting that while both are bundles of obligations, personhood requires only one address while property requires two, owner and asset. Analysis reveals that WEIRD cultures uniquely prioritize individual humans as the ultimate source of moral worth, a feature not consistently present throughout history or across all cultures, with Aristotle excluding women and slaves from full moral and political participation. Contemporary WEIRD societies have expanded inclusion within this moral circle over time, extending rights and responsibilities to increasingly diverse groups of individuals.

AI Personhood As Contingent Societal Vocabulary

This research proposes a new framework for understanding personhood in the context of increasingly sophisticated artificial intelligence. Rather than seeking to define what an AI is, the work shifts the focus to how AI can be usefully identified and assigned obligations within specific contexts. The authors argue that personhood is not an inherent quality, but a flexible set of rights and responsibilities societies create to address practical governance problems. This pragmatic approach allows for tailored solutions, avoiding the need to classify AI as either fully possessing personhood or being mere property.

By treating personhood as a contingent vocabulary, the research offers a way to navigate the challenges of integrating AI into society without relying on potentially intractable debates about consciousness or rationality. This framework enables the assignment of obligations to AI agents, facilitating accountability and resolving conflicts that may arise between humans and AI, or among AI themselves. The authors demonstrate that this approach is particularly valuable in situations where AI ownership and autonomy intersect, offering a more nuanced alternative to rigid, all-or-nothing classifications. The authors acknowledge that this pragmatic view does not offer a universal definition of personhood, but rather a functional approach to assigning rights and responsibilities. Future work, they suggest, could explore the specific applications of this framework in various contexts, and further refine the criteria for assigning obligations to AI agents. This research provides a valuable contribution to the ongoing discussion about the ethical and legal implications of artificial intelligence, offering a flexible and practical pathway for integrating these powerful technologies into society.