Modelling and analysis of social contagion in dynamic networks

In this paper an agent-based social contagion model with an underlying dynamic network is proposed and analysed. In contrast to the existing social contagion models, the strength of links between agents changes gradually rather than abruptly based on a threshold mechanism. An essential feature of the model – the ability to form clusters – is extensively investigated in the paper analytically and by simulation. Specifically, the distribution of clusters in random and scale-free networks is investigated, the dynamics of links within and between clusters are determined, the minimal distance between two clusters is identified. Moreover, model abstraction methods are proposed by using which aggregated opinion states of clusters of agents can be approximated with a high accuracy. These techniques also improve the computational efficiency of social contagion models (up to 6 times).