Network Theory
My interest in social network theory centers on questions of culture and inequalities. I elaborate on three projects here.
Tie-formation mechanisms like homophily, reciprocity, transitivity, and preferential attachment operate to varying degrees across contexts. Romantic ties tend to be homophilous by age, education, and ethnicity, and to remain exclusively dyadic. Scientific citations, in contrast, fall into highly unequal preferential attachment patterns. Triangulation is nearly absent in advice and patronage networks but features prominently in friendship ties. Building on relational sociology and field theory, my co-author, Jan Fuhse, and I argue in a paper published at Social Networks that such variation in patterns of connectivity derives from the meaning attached to relationships, which varies by context. Tie-formation logics are institutionalized rules-of-the-game and susceptible to change over time. Accordingly, the mechanisms of tie-formation behind empirical network patterns have to be connected to the cultural rules in particular fields.
In an ongoing project, I argue that preferential attachment processes in social networks should differ from those in other kinds of networks (such as website links). The typical assumption in preferential attachment is that actors infer worth from popularity – more popular actors are perceived as possessing attributes of greater worth and hence preferred as interaction partners. I elaborate on this baseline model by theorizing that actors undertake an ‘evaluation’ of popularity prior to extending ties. Two factors are especially salient: anticipated risk of proposed ties being rejected and degree of spillover between choices. Rejection risks are elevated when tie formation is expensive. This pushes preferences downstream toward available alters producing localized evaluation. Spillovers between choices are high when Matthew Effects are in play or tie-seekers look to form ties that are deemed legitimate by onlookers. Here, preferences become focalized toward popular actors producing preferential evaluation. Using simulations to examine effects of rejection initiated by popular actors alongside localized and preferential evaluation, I find that localization is more effective in containing inequalities than rejection from the top. Regardless of rejection, preferential evaluation is especially likely to consolidate the status of highly popular actors, akin to consecration. The results suggest that preferential attachment models in hierarchical social networks need to be sensitive to social conditions and their effects on evaluation of popularity.
Sociologists have persuasively argued that cultural meaning can be interpreted by analyzing the system of relations connecting cultural materials. Yet, the methodologies available for investigating the relational structure of cultural materials using survey data remain inadequate. Recent advances in the study of culture using such data have tended to focus on the classification of persons based on shared tastes. I introduce a new method to investigate the dual problem – the structure of relationships between cultural tastes and practices based on their sharedness by persons. I draw on a data mining technique called ‘association rules,’ which uses Boolean methods to generate antecedent and consequence relationships. I convert the generated relationships to directed graphs and analyze the structure of the graphs using network analysis. The proposed methodology has advantages over other techniques that use Boolean methods such as entailment analysis and Galois lattices. It can be used to investigate (1) datasets comprising many cultural preferences and practices; (2) preferences expressed using non-binary coding such as likert scales; (3) subsets of respondents based on demographic attributes such as age or socioeconomic class; and (4) reasonably large-sized datasets. I illustrate the technique on two well-known arts’ participation survey datasets for the United States.