Lecture Series April 2017: Michael Mäs › view all

"Will the personalization of the web foster opinion polarization? Theoretical, empirical, and engineering challenges"

April 26, 2017 - 16:15h/4:15pm
University of Bremen, UNICOM, Conference Room (7.3280)
Mary-Somerville-Str. 9
28359 Bremen
Contact: Stephan Dochow
Series: BIGSSS Lecture Series
Event type: public

Michael Mäs gives a talk on "Will the personalization of the web foster opinion polarization? Theoretical, empirical, and engineering challenges" in the BIGSSS Lecture Series on April 26, 2017.

Michael Mäs is Assistant Professor at the Department of Sociology and the ICS at the University of Groningen.

The talk takes place at the University of Bremen, Unicom-Building 9, Conference Room (7.3280).


The Internet and in particular online social networks have revolutionized the ways in which we interact and exchange information with friends, colleagues, and business partners. However, observers of the Internet point to potentially problematic dynamics that these technologies might trigger off or amplify. In particular, it has been warned that the personalization of social-network sites and online search-engines might critically affect the distribution of political opinions in a societies and might lead to opinion polarization in the worst case. To date, however, there is no empirical support for the hypothesis that personalization increases opinion polar- ization. The debate is based on anecdotal evidence and theoretical models of opinion dynamics. What is more, this modeling work focused on one particular set of models. In this presentation, I demonstrate that there is an alternative set of models that implies the exact opposite effect, predicting that personalization fosters consensus rather than opinion polarization. To this end, we developed an agent-based model that allowed us to compare the competing models of opinion polarization. With a simulation experiment we demonstrated that the models make contradicting predictions and studied the conditions under which the models predict that personalization affects opinion distributions. From the results of the study we draw conclusions about the design of empirical studies on the relationship between personalization and polarization. In addition, we discuss alternative ways of redesigning personalization algorithms in order to prevent opinion polarization.

About the BIGSSS Lecture Series:

Each semester the Graduate School invites a mix of established and young scholars to present their work to the students and the faculty of the School as well as to the wider interested public. Taking place every other week, the Lecture Series is the central meeting point for the entire Graduate School and provides an excellent opportunity for engaging in intensive, interdisciplinary, scholarly debate.