Assessing polarization in relational terms: diversity, homophily and communities

Contemporary societies are becoming increasingly polarized along socio-political lines. Unsurprisingly, social scientists’ attention to it has grown exponentially over the past decade. However, as it often happens with buzzwords, the notion of “polarization” is becoming increasingly heterogeneous and fuzzy, being applied to an ever wider set of related yet distinct empirical phenomena. In fact, clarity in definitions and empirical operationalizations is often missing, thus hampering knowledge accumulation. This article proposes a network-analytic structural approach for the study of what is arguably the least studied polarization-related phenomena: interactional polarization. The aim is to theoretically and empirically distinguish a truly “polarized” social setting from other possible structural scenarios, as well as to identify different pathways leading towards it. This would add much needed precision to the study of the severity, causes and consequences of polarization an enable longitudinal and cross-section comparisons. We propose a new analytical framework that simultaneously assesses three crucial dimensions: composition, homophily and structural topography. Based upon this framework, we set out a novel classification of 10 ideal-type network structures, including an ideal “polarized network”. The latter requires the concurrence of (i) high diversity (i.e. distant positions/attitudes), (ii) strong homophily (i.e. preference to establish and/or maintain social interactions with those alike), and (iii) a marked community structure (i.e. presence of subgroups that are more densely connected internally than with the rest of the networks). Furthermore, the article reviews specific procedures and metrics for the assessment of each dimension and discusses potential empirical applications for both socio-centric and ego-centric approaches.