An agent-based model of opinion formation in social networks

Abstract

In this thesis, opinion dynamics within social networks are examined by means of an agent-based simulation model. The Social Judgement Theory based model is run on random, scale-free and small-world networks, which are three widely considered topologies in the literature, and the results are comparatively analyzed to understand the effect of network topology on the opinion formation behaviors of individuals. The observations indicate that the structural changes in the social network only affect the convergence rates of opinion dynamics. An influential external actor is also introduced to the model, and its impact on opinion dynamics is studied. One critical observation is that a moderate external opinion, compared to an extreme one, has a better chance to push the majority towards extreme opinions in some populations. In addition, a population that normally results in consensus is seen to split into extremes under the effect of intense extreme propaganda. It is observed that polarization dominates the opinion dynamics in the presence of close-minded agents, when populations having heterogeneous thresholds are considered.