Skip to Main Content
Aims of social simulation include predicting a future, finding reasons of current situations, or noticing problems of a target society. Recently this research field attracts a lot of researchers from computer science, artificial intelligence, economics, political science, and son on. However, most simulations are not based on actual data. That is, models of target societies are not tuned or identified by the actual data collected from their target societies. This is because the cost of collecting actual data from their target societies is expensive. Therefore, most researches tend to show their simulation results by varying values of parameters in their model and explain several scenarios according to the corresponding parameter values. From these simulation results, we are able to learn several lessons about the nature of target societies, though, it is difficult to see a quantitative results or consequences for them. In this talk, we show several trials to design and develop models of social simulations based on actually collected data from target societies. We employ multi-agent simulation models for our social simulations, and show simulation results for polling place assignment, for hospital scrap-and-build, and for national pension problems. We also show an approach for collecting actual data by using a web-based approach.