Social-Behavioral Modeling for Complex Systems

5 How Big and How Certain? A New Approach to Defining Levels of Analysis for Modeling Social Science Topics

Matthew E. Brashears

Department of Sociology, University of South Carolina, Columbia, SC, 29208, USA

Introduction

When modeling social processes, what level of analysis is most appropriate? Should we stand back to a macro‐level and treat nation‐states as unitary entities? Should we stand at the micro‐level and attempt to model the knowledge, motivations, and choices of millions of simulated actors in the hope that they aggregate into consistent results? Or should we try to locate ourselves somewhere in between, in the meso‐level, where predictions are easier and data requirements are lighter? The answer is critical to producing useful results, but it is often not obvious which level of analysis offers the most promise. This problem is further exacerbated by the fact that each descriptor, micro‐, macro‐, and meso‐level, applies to a range of scales rather than just one, depriving these terms of analytic precision. Here, I argue that the appropriate level of analysis is often difficult to identify because our concept of levels is inadequate. The appropriate level of analysis depends not solely on the size of the subject to be modeled but also on the constancy of the system itself, or the degree to which its behavior varies between fully determined and complex. By explicitly considering the interaction between the size and the constancy of the system, we improve ...