Scale and Topology Effects on Agent-Based Simulation: A Trust-Based Coalition Formation Case Study
Luis G. Nardin (Universidade de São Paulo, Brazil), Luciano M. Rosset (Universidade de São Paulo, Brazil) and Jaime S. Sichman (Universidade de São Paulo, Brazil)
Copyright: © 2014
The exploitation of Agent-Based Social Simulation (ABSS) full capabilities often requires massive computing power and tools in order to support the achievement of breakthrough results in social sciences. Lately, this issue has being addressed by the release of several high-performance computing agent-based simulation tools; however, they have not been used for exploring critical issues, such as ABSS results invariance and universality. Hence, in order to advance this topic, this chapter provides an invariance analysis, considering scale and topology, of a model that incorporates the concepts of trust and coalition formation, in which agents are placed on a square lattice interacting locally with their neighbors and forming coalitions. By varying the environment size, its topology, as well as the neighborhood topology, it is identified in the experimental scenario that apparently the only parameter that affects the simulation dynamics is the neighborhood topology.
Among the few related work that approaches robustness, Flache and Hegselmann (2001) relaxed the assumption of rectangular grid structures and analyzed how the use of irregular (non-rectangular) structures, such as hexagonal or triangular cells, affects the migration and influence dynamics in cellular automaton models. They concluded that both dynamics are robust to structure variation.
Urbig et al. (2008) proposed an opinion dynamics model that allows the variation of the number of peers that meet at once in order to analyze changes on the speed of the opinions adaptation. They investigated the effect of such changes in different population sizes concluding that the number of meeting peers affects the tendency towards consensus.