Synthesized Population for Real-Scale Social Simulations

Real-Scale Social Simulation (RSSS) aims to reproduce real societies at near-real granularity by embedding populations comparable in size to the target region and modeling individual decision-making using attributes such as age, sex, and occupation. However, detailed administrative microdata are difficult to use due to privacy constraints, and access to census microdata often requires approval and additional anonymization work. This article reviews approaches for synthesizing artificial societies (synthetic populations) from publicly available tabulations, with an emphasis on sample-free, simulated-annealing (SA)–based generation. Building on a framework that minimizes discrepancies between real and synthetic tabulations—while accounting for household/family types and multiple constraints such as sex–age distributions, parent–child age gaps, and spousal age gaps—the article outlines methodological advances toward large-scale synthesis (e.g., improved initialization and search procedures). It further describes extensions that enrich synthetic agents with finer spatial placement (allocation to small areas and buildings using fundamental geospatial data) and additional socio-economic attributes, including employment status, industry, firm size, and income assignment based on official surveys. Finally, it clarifies practical rules for distributing and using synthetic population data to avoid misunderstanding and privacy concerns, including non-inclusion of real personal/household information, limited guarantees on unseen statistics, potential updates, the requirement to use multiple synthetic datasets for analysis, and prohibitions against publishing results that can be linked back to real households or individuals.