Learning Media Credibility from Facebook Data

• In Event: Crowdsourcing

Abstract

The production, consumption, and dissemination of online news is of growing interest among scholars studying democracy, but much difficulty lies in the study of news quality in a comparative perspective. Many problems plague cross-country studies, but studies of media credibility are particularly susceptible to issues such as different political environments, language barriers, differing cultural contexts, and differing media regulation. This study leverages a new dataset published by Facebook through the Social Science One initiative to study the prevalence of fake news on Facebook in over twenty countries in Europe. We use a supervised model (trained on US data) to predict the credibility of a given news domain based on users’ feedback and behavior. We then analyze the prevalence of unreliable news and the news diet of Facebook users in states of varying democratic quality. We find an East-West divide between countries in Europe with higher proportion of unreliable news in former socialist countries, as well as in the UK. Furthermore, we find that more recently registered news domains and those registered outside of the country are more likely to be perceived as the sources of unreliable information.

Authors

• Tom Paskhalis, New York University

• Kevin Aslett, New York University

• Zhanna Terechshenko, New York University

• Joshua A. Tucker, New York University

• Jonathan Nagler, New York University

• Richard Bonneau, NYU