From Consensus to Social Learning in Complex Networks
Abstract:
Over the past few years there has been a rapidly growing interest in analysis, design and optimization of various types of collective behaviors in networked dynamic systems. Collective phenomena (such as flocking, schooling, rendezvous, synchronization, and agreement) have been studied in a diverse set of disciplines, ranging from computer graphics and statistical physics to distributed computation, and from robotics and control theory to social sciences and economics. A common underlying goal in such studies is to understand the emergence of some global phenomena from local rules and interactions.
In this talk, I will expand on such developments and present and analyze models of consensus and agreement in random networks as well as new models for information aggregation tailored to social networks that go beyond existing "consensus-based" models. Specifically I will present some surprising analytical results on the first two moments of consensus value when the network changes randomly. Furthermore, I will present a model of social learning in which each agent acts like a rational Bayesian agent with respect to her own observations, but exhibits a bias towards the average belief of its neighbors. When the underlying social network is strongly connected all agents reach consensus in there beliefs. Moreover, I will show that when each agents observed signal is independent from others, each agent will "learn" like a Bayesian who has access to all information.