Polarization in the social sciences: Assortative mixing in social science collaboration networks is resilient to interventions

Abstract

Academic collaboration in the social sciences is characterized by a polarization between hermeneutic and nomological researchers. This polarization is expressed in different publication strategies. The present article analyzes the complete co-authorship networks in a social science discipline in two separate countries over five years using an exponential random graph model. It examines whether and how assortative mixing in publication strategies is present and leads to a polarization in scientific collaboration. In the empirical analysis, assortative mixing is found to play a role in shaping the topology of the network and significantly explains collaboration. Co-authorship edges are more prevalent within each of the groups, but this mixing pattern does not fully account for the extent of polarization. Instead, a thought experiment reveals that other components of the complex system dampen or amplify polarization in the data-generating process and that microscopic interventions targeting behavior change with regard to assortativity would be hindered by the resilience of the system. The resilience to interventions is quantified in a series of simulations on the effect of microscopic behavior on macroscopic polarization. The empirical study controls for geographic proximity, supervision, and topical similarity (using a vector space model), and the interplay of these factors is likely responsible for this resilience. The paper also predicts the co-authorship network in one country based on the model of collaborations in the other country.

Network polarization, filter bubbles, and echo chambers: an annotated review of measures and reduction methods

“…Soós and Kampis (2011) analyzed the leading Hungarian research organizations, comparing the diversity of their publication performance and the polarity of each researcher's profile. Leifeld (2018) analyzed two research traditions (the hermeneutic and the nomological) in the coauthorship network of researchers in Germany and Switzerland. A higher similarity between researchers leads to a greater probability of coauthorship, showing a homophilic behavior between hermeneutic and nomological researchers observed by philosophers of science.…”

Network polarization, filter bubbles, and echo chambers: An annotated review of measures, models, and case studies

“…Leifeld [82] analyzed two research traditions: the hermeneutic (which investigates meaningful cultural entities handed down by tradition) and the nomological (which formulates and verifies hypotheses concerning the laws governing empirical regularities) in the co-authorship network of researchers in Germany and Switzerland. The research questions were: (1) did the two alleged camps exist empirically, and (2) how entrenched (polarized) were they?…”

UK trading patterns within and between regions in the automotive sector—A network analysis

“…The model is estimated with a maximum pseudo-likelihood procedure (for further discussion of this estimation approach see [55,57]). ERGMs have been applied in a wide variety of empirical settings [58][59][60][61], including the analysis of international trade and investment patterns [62,63].…”