Opinion formation among mobile agents

The evolution of public opinion has been widely studied to understand how atomic interactions between individuals cause opinions to evolve. However, while many studies have paid attention to the influence and interaction mechanisms, the vast majority of the literature assumes a static representation of immobile agents, ignoring the effect that physical proximity and mobility has on interactions, as observed in real-life. Mobility provides humans with the opportunity to meet and locally interact with a diverse range of people, which can heavily influence opinion spread in human societies. Considering both opinion and location dynamics on widely used opinion models, such as the Bounded Confidence model, can therefore result in more realistic understanding of the drivers that cause agreement and diversity. This thesis investigates both directed and random mobility, inspired by two fundamental concepts from psychology: homophily and cognitive dissonance. These theories can drive the response behaviours to agreement and disagreement in humans. We translate these as attraction and repulsion forces in our mobility model. Through incorporating these phenomena, we quantify the different outcomes that arise and propose new evaluation metrics for analysis in this context that capture the formation of opinions and communities, reflecting the self-organisation among the populations. Extensive simulation results demonstrate the impact of the random and directed mobility. The main findings show that opinion formation is highly insensitive to random mobility, showing similarity in behaviour to static modelling. This is a very important result because the literature usually applies this approach. Furthermore, we find that alternative psychological theories, as incorporated into mobility, impact differently on both the opinion and spatial organisation of the agents. As these parameters are varied, we find a distinct transition in behaviour. Finally, by combining and analysing all the results, we propose a novel classification approach for different outcomes of self-organisation in opinion models