Who to share information with? A model of strategic diffusion in social networks

I study information diffusion in a social network where a third party can control the precision of information as well as who initially receives information. Applications include spreading of news by digital media outlets, lobbyists persuading senators to contribute to a project and others. A designer engages in a bayesian persuasion game with multiple agents, but is constrained to send information privately to a subset of agents. The agents can communicate this information to each other through links in a social network. The designer relies on this word-of-mouth communication channel to diffuse information. The optimal precision of information sent by the designer reflects a fundamental tradeoff: precise information increases diffusion, but reduces the designer’s ability to manipulate the agents’ beliefs to- wards his objective. The optimal seeding strategy involves choosing an agent with the highest “influence”, a novel centrality measure that is determined in endogenously by information accuracy and the network structure. As preferences become more diverse, I show that there exists equilibrium with sub- optimal spreading where the designer caters only to agents relatively more aligned towards his objective.