Asymptotic learning on Bayesian social networks (Q2447283)

This paper studies the asymptotic learning on Bayesian social networks represented by directed graphs. The aim of the agents is to estimate a binary state of the world \(S\in\{0,1\}\), where each of the two possibilities occurs with probability \(1/2\). Initially, the agents are provided with two private signals which are informative with respect to \(S\) and i.i.d. conditioned on \(S\). At time \(t\), each agent \(v\) chooses an action 0 or 1, conditioned on the information available to \(v\) at time \(t\). One of the main results shows that when initial private beliefs are non-atomic, then the optimal action sets of the players are identical at the limit \(t\rightarrow\infty\). When such agreement is guaranteed then the agents learn the state of the world with high probability, if the number of agents is large. This phenomenon is referred to as asymptotic learning in this paper.