Scalable Agent-Based Modeling of Forced Migration

Abstract

Migration studies have a long history in sociology and the social science though the discipline has matured in notable ways over the past several decades. One such way is the attention that has been paid to forced migration resulting from security and conflict events worldwide. The study of forced migration is distinct from the study of voluntary migration, the topic of most research in migration studies since the late 19th century. Another way the field has matured is in the application of computational modeling and simulation methods to the problem domain to augment or complement theoretical or statistical analysis of migration. Despite recent advancements in these two areas, there remains a dearth of research around computational modeling and simulation methods as applied to the study of forced migration. There are many gaps the scientific community of sociologists and computational social scientists must fill before empirical models can be generalized and used for predictive purposes by aid organizations to make decisions about the allocation of resources in response to forced migration events worldwide. For instance, many computational models have been applied to the modeling of voluntary migration but far fewer to forced migration. Of the models developed for forced migration, many are only theoretical, few are empirical, and only one is designed to run at a scale applicable to ongoing forced migration events worldwide with tens of millions of migrants but does not consider social networks. The research presented in this dissertation fills some of these gaps by addressing limitations in theoretical knowledge and computational methodology with the application of an agent-based model to a real-world forced migration case study – forced Syrian migration into neighboring Turkey. This research makes the case that agent-based modeling is the appropriate social simulation approach to take for forced migration modeling and discusses recent developments in forced migration theory that have yet to be applied in empirical computational modeling contexts, including social networks. This research also addresses the scale issue, demonstrating an agent-based simulation that runs with up to 25M agents and applying this simulation to a real-world event with 4-6M at-risk persons. The following chapters summarize a two-fold contribution to the sociology, social science, and computational modeling scientific communities: (1) the first empirically tested agent-based simulation to model forced migration that considers migrant social networks and (2) methodological advances to the state-of-the-art of computational forced migration modeling and publicly available computational tools and methods to facilitate future research for researchers in this domain.