Simulating Complex Social Behaviour With the Genetic Action Tree Kernel

Abstract:     
The concept of genetic action trees combines action trees with genetic algorithms. In this paper, we create a multi-agent simulation on the base of this concept and provide the interested reader with a software package to apply genetic action trees in a multi-agent simulation to simulate complex social behaviour. An example model is introduced to conduct a feasibility study with the described method. We find that our library can be used to simulate the behaviour of agents in a complex setting and observe a convergence to a global optimum in spite of the absence of stable states.

Keywords: multi-agent system, genetic algorithm, action trees, social simulation

JEL Classifications: C61, C63, C65

Accepted Paper Series