Integrating statistical data into social simulation

Objectives

Within the FOSSR project for an Italian scientific open cloud, we aim to provide an automated service to initialize agent-based models representative of the Italian population. The method simulates human values and cognitions in virtual agents interacting in artificial populations. 

Manipulating the conditions where agents interact, researchers can map the emergence of dynamic collective phenomena (e.g. polarization), not possible with only empirical data.

Description

We implement the Iterative Proportional Fitting algorithm, equal to raking, to synthetically reconstruct a new joint distribution of agents’ socio-demographics from independent data sources.

Results

We validated the algorithm with marginal distributions of ISTAT 2022 gender and age data in Italy, managing to reproduce their empirical joint distribution in artificial populations. We will present current and next steps to extend the algorithm to nested and multilevel data.

References

Paolillo, R., & Paolucci, M. (2024). FOSSR Deliverable 5.8 - Complex modelling and artificial populations for agent-based simulations. Zenodo. https://doi.org/10.5281/zenodo.1063880