Analysis of opinion evolution based on non-Bayesian social learning

Abstract

In a reposting network on a social platform, users form their own opinions relying on personal experience as well as the influence of neighboring users. We thus use a non-Bayesian social learning model to optimally integrate the user's social observations into the Bayesian inference process, where the users receive information from their neighbors and update opinions to be a weighted linear combination of the Bayesian posterior opinion and the neighbors opinions. By crawling the microblog topic, we conduct simulations and empirical analysis of the repost network under the topic. The simulation results show that users' opinions tend to reach group consensus and are similar to the empirical results.