AGENT-BASED MODELING AS A TOOL FOR VERIFYING MIDDLE-RANGE THEORIES IN ANALYTICAL SOCIOLOGY

Abstract

This paper addresses the problem of methodological uncertainty that arises when attempting to verify middle-range theories in analytical sociology using traditional statistical methods. It substantiates the idea that the toolkit of agent-based modeling (ABM) makes it possible to bridge the gap between macro-theoretical constructs and micro-level empirical data through the formalization of behavioral rules of heterogeneous actors. The evolution of ABM applications in analytical sociology, which is focused on the explication and reconstruction of specific social processes and mechanisms, is traced. The article compares classical problem formulations (T. Schelling’s segregation model, M. Granovetter’s threshold models) and modern modifications of these approaches (M. Macy’s network models of political polarization), which demonstrate a transition from abstract thought experiments to formalized computational procedures. Special attention is paid to the analysis of foreign experience in using ABM in epidemiological research as a reference case that reveals the potential of the method for testing causal dependencies. The article highlights the problems associated with the adaptation of the agent-based approach in the Russian academic discourse and identifies the specific trajectory of its interdisciplinary development. Based on a review of publications, the dominance of economic schools (in particular, the scientific school of V. L. Makarov and A. R. Bakhtizin) and a relatively weak representation of the method in fundamental sociological research are observed. A separate section of the study is devoted to an overview of the key methodological challenges that a researcher faces when implementing agent-based models for verifying middle-range theories. Procedures for empirical validation are examined, and the phenomenon of YAAWN (Yet Another Agent-Based Model... Whatever... Nevermind...) is analyzed, which denotes the risks of arbitrarily assigning agent behavior algorithms. On the basis of the comparative analysis conducted, a conclusion is drawn about the prospects for the development of the method as a supplementary tool for testing middle-range theories and analyzing social processes, provided that quantitative empirical data are used as a statistical and algorithmic foundation for model construction.