2016/07/22 by Deepanshu Vasal, Vasal, Deepanshu, Achilleas Anastasopoulos +1
Decision Sciences · Economics, Econometrics and Finance · Physics and Astronomy · #Computer Science and Game Theory (cs.GT) #Economic Policies and Impacts #Economic theories and models #FOS: Computer and information sciences #FOS: Economics and business #FOS: Electrical engineering #Game Theory and Applications #Opinion Dynamics and Social Influence #Systems and Control (eess.SY) #Theoretical Economics (econ.TH) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1607.06847
openalex publication_date 2016/07/22 · openalex created_date 2021/02/01 · openalex updated_date 2026/08/01
We study the problem of Bayesian learning in a dynamical system involving\nstrategic agents with asymmetric information. In a series of seminal papers in\nthe literature, this problem has been investigated under a simplifying model\nwhere myopically selfish players appear sequentially and act once in the game,\nbased on private noisy observations of the system state and public observation\nof past players' actions. It has been shown that there exist information\ncascades where users discard their private information and mimic the action of\ntheir predecessor. In this paper, we provide a framework for studying Bayesian\nlearning dynamics in a more general setting than the one described above. In\nparticular, our model incorporates cases where players are non-myopic and\nstrategically participate for the whole duration of the game, and cases where\nan endogenous process selects which subset of players will act at each time\ninstance. The proposed framework hinges on a sequential decomposition\nmethodology for finding structured perfect Bayesian equilibria (PBE) of a\ngeneral class of dynamic games with asymmetric information, where user-specific\nstates evolve as conditionally independent Markov processes and users make\nindependent noisy observations of their states. Using this methodology, we\nstudy a specific dynamic learning model where players make decisions about\npublic investment based on their estimates of everyone's types. We characterize\na set of informational cascades for this problem where learning stops for the\nteam as a whole. We show that in such cascades, all players' estimates of other\nplayers' types freeze even though each individual player asymptotically learns\nits own true type.\n