Will the Personalization of Online Social Networks Foster Opinion Polarization?

Pundits and scholars warned that the personalization of the web and in particular of online social networks fosters interaction between likeminded users and amplifies the polarization of political opinions. We criticize, however, that this warning is based on one particular polarization model and that an alternative and equally prominent theory implies the exact opposite effect, predicting that personalization fosters consensus rather than polarization. We develop a general modeling framework to compare the predictions of the competing models. Using agent-based modeling, we formally demonstrate that the two theories make contradicting predictions and study the conditions of polarization. In conclusion, we call for empirical research on the competing assumptions of the two models, discussing major roadblocks and novel methods to overcome them.