Blockmodeling for analysis of social structures: theoretical and methodological foundations

Abstract

The article discusses the features of blockmodeling as a class of methods for clustering network data in the analysis of social structures. Blockmodeling is considered as an approach to the analysis of social structure, which combines network components into groups (clusters) based on their equivalent structural positions. The basic concepts of blockmodeling are described – matrix, matrix image, cluster, clustering, position, block, blockmodel; an illustrating example is given. The concept of equivalence is presented, and two types of equivalence, structural and regular, are described. The main approaches of blockmodeling – indirect and direct – and related methods and algorithms are presented. For each approach, examples of the practical application in social sciences are provided. Other methods of blockmodeling (stochastic blockmodeling) and similar methods of subgroups detection in networks are mentioned. It is shown that the methodology of blockmodeling has heuristic potential for analyzing social structures and is promising for identifying cohesive groups and determining the role and structural positions of individuals within them. In conclusion, the open questions and limitations of this research methodology are discussed.