Rulenet: Mapping the structure of cultural preferences using association-rules and network graphs

Abstract

Sociologists have persuasively argued that cultural meaning can be interpreted by analyzing the systems of relations that measure the so-called ‘going together’ of cultural materials. Research investigating cultural tastes and preferences has used this approach to interpret consumption patterns as relational systems using a variety of techniques including multidimensional scaling, two-mode network analysis, and variable correlation networks. I contribute to this growing set of tools by describing and demonstrating the use of a datamining technique with scant history of use within sociology, called ‘association-rules.’ The key contribution of this technique is that it generates directed relationships between variables (e.g., preference for opera → preference for ballet), which has several advantages over existing techniques that conceptualize relationality in terms of mutual presence. I show how such ‘one-sided’ clustering (A goes with B, but B may not go together with A) can be represented and analyzed as network graphs, an approach I call ‘Rulenet.’ I discuss how the proposed technique can provide relatively novel insights into the organizations of tastes, less feasible via other techniques, and illustrate Rulenet on two well-known cultural participation survey datasets for the United States: (1) The Survey of Public Participation in the Arts (SPPA) from 2017 and (2) The General Social Survey Culture Module from 1993 (GSS).