I am a computer scientist working on cooperative AI, multi-agent safety, and institutions for collective intelligence. My research asks how we can build, evaluate, and govern AIs that help humans, AIs, and institutions cooperate in complex multi-actor worlds.

I am interested in developing robust and generalizable AI whose capabilities are beneficial to humans, aligned with human values and principles, and safely integrated into society. My research is grounded in a simple premise: our world is inherently a multi-actor ecosystem shaped by incentives, norms, institutions, relationships, and collective action. This perspective motivates my broader goal of building AI systems and surrounding institutions that support cooperation-—the hallmark of intelligence that underpins human societies’ ability to coordinate effectively in addressing the challenges of shared existence.

I develop methods blending innovations in AI for multi-agent systems (e.g. reinforcement learning, generative agents, graph machine learning) and interdisciplinary topics spanning social science, anthropology, game theory and economics among others. My recent work increasingly explores how to evaluate AI systems in richer social and institutional environments: how they reason about other agents, respond to incentives and norms, participate in cooperative settings, and affect coordination among humans and AI systems.

I am incredibly fortunate to collaborate closely with Dylan Hadfield-Menell at MIT, David Parkes at Harvard University, Gillian Hadfield at Johns Hopkins University and Joel Leibo and other members of the the multi-agent team at Google Deepmind on these topics.