A Bayesian Model for False Information Belief Impact, Optimal Design, and Fake News Containment

Abstract

This work is a technical approach to modeling false information nature, design, belief impact and containment in multi-agent networks. We present a Bayesian mathematical model for source information and viewer's belief, and how the former impacts the latter in a media (network) of broadcasters and viewers. Given the proposed model, we study how a particular information (true or false) can be optimally designed into a report, so that on average it conveys the most amount of the original intended information to the viewers of the network. Consequently, the model allows us to study susceptibility of a particular group of viewers to false information, as a function of statistical metrics of the their prior beliefs (e.g. bias, hesitation, open-mindedness, credibility assessment etc.). In addition, based on the same model we can study false information "containment" strategies imposed by network administrators. Specifically, we study a credibility assessment strategy, where every disseminated report must be within a certain distance of the truth. We study the trade-off between false and true information-belief convergence using this scheme which leads to ways for optimally deciding how truth sensitive an information dissemination network should operate.