Toward the Realization of Real-Scale Social Simulations

Real-Scale Social Simulation (RSSS) aims to reproduce societal phenomena at near-real granularity by modeling individual decision-making and scaling up to groups, organizations, and entire societies for policy and institutional analysis. Since detailed administrative microdata are often inaccessible due to privacy constraints, constructing a Synthetic Population from publicly available tabulations becomes a key enabling technology. This paper reviews Synthetic Reconstruction (SR) approaches for generating microdata from statistical tables, covering both sample-based methods such as IPFP and sample-free alternatives including MCMC-based simulation methods, fitness-based synthesis, and simulated annealing (SA). It further outlines advances in SA-based household synthesis toward real-scale (nationwide) generation through improved initialization and search strategies. In addition, the paper proposes a practical workflow to enrich the synthetic population with extra attributes—such as fine-grained location, employment/industry, and income—by combining statistically grounded assignment with iterative adjustment to minimize discrepancies against multiple tabulations. By positioning the resulting synthetic population as “virtual real-world data,” the study provides a methodological foundation for high-fidelity social simulations in domains such as transportation, disaster management, and infectious disease modeling.