Controlling Opinion Bias in Online Social Networks

Voter models are commonly used in modeling opinion dynamics in applications such as the spread of ideologies and politics. It is well known that the binary version of these models, where the state (or opinion) of each node is 0 or 1, always leads to consensus. We consider an extension, in which some nodes are "stubborn," i.e., do not change their states based on other nodes. In such a system, the asymptotic average opinion could be between 0 and 1. The goal of this paper is to study the ease with which bias (i.e., the tendency of the opinion to become close to 0) can be controlled (so that the average opinion exceeds 0.5). We formalize a new parameter, called the Minimum Opinion Control Factor (MOCF), to capture this, and study it through analysis and simulations on real online and synthetic networks. Finally, we experimentally demonstrate the usefulness of combining the voter model with an independent cascade model in controlling bias and we explain these findings in terms of network structure.