Managing the Complexity of Large-Scale Agent-Based Social Diffusion Models with Different Network Topologies

Abstract

AbstractSocial diffusion models have been extensively applied to study biological and social processes on a large scale. Previously two issues with these models were identified: understanding emerging dynamic properties of complex systems, and a high computational complexity of large-scale social simulations. Both these issues were tackled by abstraction techniques developed previously for social diffusion models with underlying random networks. In the paper it is shown that these techniques perform poorly on scale-free networks. To address this limitation, new model abstraction methods are proposed and evaluated for three network types: scale-free, regular and random. These methods are inspired by node centrality measures from the social networks area. The proposed methods increase the computational efficiency of the original model significantly (up to 40 times for regular networks).Keywordsgroup dynamicsmodel abstractionsocial diffusionlarge-scale agent-based simulation