Implicit Polarization on YouTube

• In Event: The Perils and Promise of Social Media Discourse and Information

Abstract

How information spreads on social media and what individual characteristics or behaviors drive migration to radical echo chambers is a topic of much debate in political science. As the most widely used platform in the US with the most representative user base (Pew Research 2021), as well as a host for embedded content across the web, the wide ranging impact of YouTube makes it a prime platform for understanding this phenomenon. Though researchers have addressed questions of polarization and extreme content through a combination of surveys and interventions, these as well as large scale explorations of behavior over time focus on desktop data and access to a set of pre-labelled channels. This paper uses novel longitudinal data on individuals' YouTube consumption from mobile data provider mFour to consider how patterns of user viewing behavior, the networks of YouTube channels they watch, and their demographics influence viewing over time. In particular, we consider whether an individual’s history of watching YouTube videos with implicit messages regarding polarizing topics, as found using content analysis of video transcripts, increases the likelihood that they will watch more videos of this type in the future. We further evaluate data such as the ads they are served on YouTube and their activity on other social networking sites to obtain a holistic understanding of their internet presence. Finally, we discuss the distinction between results related to individual choices and behaviors and those related to YouTube’s recommendation algorithm as well as implications for policy and future research.

Author

• Erin Elizabeth Walk, University of Pennsylvania