Title:Mitigating Disinformation in Social Networks through Noise

Abstract:An abundance of literature has shown that the injection of noise into complex socio-economic systems can improve their resilience. This study aims to understand whether the same applies in the context of information diffusion in social networks. Specifically, we aim to understand whether the injection of noise in a social network of agents seeking to uncover a ground truth among a set of competing hypotheses can build resilience against disinformation. We implement two different stylized policies to inject noise in a social network, i.e., via random bots and via randomized recommendations, and find both to improve the population's overall belief in the ground truth. Notably, we find noise to be as effective as debunking when disinformation is particularly strong. On the other hand, such beneficial effects may lead to a misalignment between the agents' privately held and publicly stated beliefs, a phenomenon which is reminiscent of cognitive dissonance.