The online information environment provides citizens with an abundance of news and other content relevant for political opinion formation that varies considerably in terms of journalistic quality and political extremity. Politicians, journalists and academics have speculated that by facilitating individualized processes of opinion formation, the current high-choice environment of online news websites, social media and other information intermediaries is related to audience fragmentation and political polarization. Yet, research on the interplay of media exposure and dynamic polarization processes is sparse and at times contradictory. One of the reasons is that methods traditionally employed by social scientists to measure exposure such as survey-based self-reports inadequately capture online behavior.

Our project brings together techniques from computational social science with theories and methods from communication and political science. At its core is a year-long web tracking of a representative sample of approximately 1,500 German citizens from an online access panel. The participants are surveyed regularly on their demographics, political opinions, and media use and consent to having their visited websites tracked. Through web crawling and automated text analysis of those website contents we are able to measure the source, issue, and actor diversity of the information participants are exposed to and relate that to the development of their political opinions over time. With this longitudinal data, we systematically study the interplay between online information use and political opinions by statistically disentangling within-person changes. The findings have implications for academic and political debates on echo chambers, societal cohesion, and the regulation of online environments.