Information diffusion in networks with the Bayesian Peer Influence heuristic

Repeated communication in networks is often considered to impose large information requirements on individuals, and for that reason, the literature has resorted to use heuristics, such as DeGroot's, to compute how individuals update beliefs. In this paper we propose a new heuristic which we term the Bayesian Peer Influence (BPI) heuristic. The BPI accords with Bayesian updating for all (conditionally) independent information structures. More generally, the BPI can be used to analyze the effects of correlation neglect on communication in networks. We analyze the evolution of beliefs and show that the limit is a simple extension of the BPI and parameters of the network structure. We also show that consensus in society might change dynamically, and that beliefs might become polarized. These results contrast with those obtained in papers that have used the DeGroot heuristic