Sociodemographic studies on human migration phenomena are mostly based on surveys and censuses, which significantly increases the research costs. This scenario becomes even worse when the study involves migration and social networks, which often lacks on representative data and consensually accepted concepts by demographers and sociologists. In this paper we propose a new multi-evolutionary agent model dedicated to social simulations, mainly for those problems where higher order dynamic behaviors (e.g. secondary emergent phenomena) are important to the investigated phenomenon. Its usefulness lies on its multilevel evolutionary adaptability which enables it to capture multiple parallel phenomena. To verify our hypothesis we applied the model to Brazilian internal human migration phenomenon and to the influence of social networks on migration flows. This is followed by a comparative analysis against simulations carried out based on a non-evolutionary cognitive agent model. Results show that the proposed model was able to rise secondary migration-related phenomena such as countermigration flows. Experiments with the cognitive agent also produced the emergence of migration flows but no secondary phenomenon was observed which was the case with our approach. Furthermore, results also pointed out a significant influence exerted by information exchange inside social networks on migration flows.
