Homophily and spread of misinformation in random networks

We adopt the framework of multi-type random networks, with type-specific linking probabilities, to study the spread of misinformation. A novel feature is that we distinguish the reach and influence of misinformation. The eventual outcome of misinformation spread is closely related to the size and type composition of the giant component of the network. We show that homophily always increases the probability that misinformation goes viral and its reach. However, the influence of misinformation (affecting people's beliefs or voting behavior) is non-monotonic in homophily: it increases with homophily when there is little homophily, but the result is reversed when the degree of homophily is already high. We also study the impacts of increasing political polarization and the ideology bias of misinformation.