A theoretical framework for conducting multi-level studies of complex social systems with agent-based models and empirical data

TL;DR

This paper introduces a formal framework that connects agent-based models with empirical data across multiple levels of social systems, enabling more rigorous validation of complex simulations.

Contribution

It presents a novel formal structure for mapping agent-based models to empirical phenomena and validates models through multi-level statistical relationships.

Findings

Framework effectively links models to empirical data

Simulation sampling aligns with observed phenomena

Validation via multi-level statistical models

Abstract

A formal but intuitive framework is introduced to bridge the gap between data obtained from empirical studies and that generated by agent-based models. This is based on three key tenets. Firstly, a simulation can be given multiple formal descriptions corresponding to static and dynamic properties at different levels of observation. These can be easily mapped to empirically observed phenomena and data obtained from them. Secondly, an agent-based model generates a set of closed systems, and computational simulation is the means by which we sample from this set. Thirdly, properties at different levels and statistical relationships between them can be used to classify simulations as those that instantiate a more sophisticated set of constraints. These can be validated with models obtained from statistical models of empirical data (for example, structural equation or multi-level models) and hence provide more stringent criteria for validating the agent-based model itself.