Abstract: The structure of information plays a pivotal role in game theory, influencing decision-making and
outcomes in dynamic games and multi-agent systems. Intriguingly, more information does not always
lead to improved social utility, giving rise to counterintuitive paradoxes. While classical multi-agent
learning frameworks rely on well-defined and structured information settings, real-world applications
often present unstructured or incomplete information landscapes. This presentation investigates both
the theoretical underpinnings and practical approaches to addressing these complexities, with a
particular focus on conjectural learning, an adaptive, subjective framework that allows agents to form
and refine conjectures about their environment and unknown elements. We will explore applications
spanning misinformation dynamics in social networks, strategies for cyber deception, and robotics.
Additionally, the talk will examine the emerging potential of large language models (LLMs) as tools
to enhance and revolutionize multi-agent learning in these contexts.

Biography: Quanyan Zhu is an assistant professor in the Department of Electrical and Computer Engineering
at New York University. He received the B.Eng. in honors electrical engineering with distinction
from McGill University in 2006, the M.A.Sc. from the University of Toronto in 2008, and the Ph.D.
from the University of Illinois at Urbana-Champaign in 2013. From 2013 to 2014, he was a
postdoctoral research associate in the Department of Electrical Engineering, Princeton University. He
is a recipient of many awards including the NSERC Canada Graduate Scholarship, the Mavis Future
Faculty Fellowships, and the NSERC Postdoctoral Fellowship. He spearheaded the INFOCOM
workshop on Communications and Control on Smart Energy Systems and the Midwest Workshop on
Control and Game Theory. His current research interests include optimal control, game theory,
reinforcement learning, network security and privacy, resilient control systems, and cyberphysical
systems.