Echo chambers are particularly salient in polarised debates regarding politics, societal issues or conspiracy theories. They tend to foster animosity between opposite sides, and fuel reinforcement of pre-existing beliefs. In this talk, I present a framework to quantify and control the echo chamber effect, based on the Voter Model. I quantify the echo chamber effect via the intensity of exposure to congruent opinions. This relies on the computation of discord probabilities between agents, and I demonstrate how to compute them exactly in a highly general setting: in any given directed, weighted network with multiple opinions, zealots, and individual update rates. This calls for a generalised definition of the active links density, to take into account long-range, weighted interactions. The impact of the network topology on active links density is investigated on real-life and synthetic networks. Then, I show how adding links between antagonistic communities impacts the echo chamber effect and the opinion diversity. Depending on the level of bias of the groups, I uncover widely different behaviours. Finally, I introduce a macroscopical method to increase the diversity of content users are exposed to, while accounting for potential backfire effects. Even in the absence of user-level features, the method is able to impact the echo chamber effect and opinion diversity. This research contributes insights to the benefit of the debate on regulation of online social platforms.
Juan Fernández Gracia Contact form