NETWORK SIMULATION MODELS Major

Accurate simulation models of social networks enable over time prediction, statistical testing of hypothesis about social groups, and enables researchers to develop more informed hypotheses for human experimentation, by evaluating them and reasoning about them using simulation. Several methods for simulating social network behavior are compared. The multi-agent simulation Construct is shown to be an excellent approach for simulating social behavior. The stochastic engine closely resembles an independently developed statistical framework for dynamic, temporal networks called the Link Probability Model (LPM). This paper illustrates that Construct, a multi-agent network model for the coevolution of agents and socio-cultural environments, is a viable choice for most network simulation needs because it is based on the LPM concepts which perform well on empirical data. Further, with the ability to add additional network dependence, Construct is able to leverage flexibility to produce statistically greater conjectures on network structure and knowledge diffusion than alternatives.