Construction of synthetic populations with key attributes: Simulation set-up while accommodating multiple approaches within a flexible simulation platform

In this paper, we describe our concept for overcoming the data barriers of building credible synthetic populations to assist the transformation between social theories and mathematical models. We specifically developed a 31-million-agent model of Afghanistan's population to demonstrate the ability to computationally control and analytically manipulate a system with the large number of agents (i.e., 10⁸) necessary to model regions at the individual level using the LandScan Global population database. Afghanistan was selected for this case study because gathering data for Afghanistan was thought to be especially challenging. The LandScan Global population database is used by a majority of key U.S. and foreign agencies as their database system for worldwide geospatial distribution of populations. Assigning attributes to disaggregated population was achieved by fusing appropriate indicator databases using two forms of aggregation techniques - geographical and categorical. A new approach of matching attributes to theoretical constructs was illustrated. The other data sources used include data on military and peacekeeper forces' loyalties, readiness, and deployment collected through a combination of UN and classified force projections; economic data collected at the national level and disaggregated using data fusion techniques; data on social attitudes, beliefs, and social cleavages through anthropological studies, worldwide polling, and classified sources; and data on infrastructure and information systems and networks.