The Role of Algorithms and Selective Exposure in Promoting Extremism on YouTube

• In Event: Non-Stop: Platform Affordances, Diffusion, and the News

Abstract

Political news and information are more accessible today than ever before, but many observers worry that new technologies are amplifying ideologically extreme, hateful, and conspiratorial ideas. In this project, we will examine the consumption of such content via YouTube, which has been identified as a potential radicalizing force promoting extremism to millions worldwide. The platform relies on its recommendation algorithm to drive traffic to videos and keep users engaged, but this approach could inadvertently boost extremist content. This paper will therefore seek to measure exposure to this content and the role that algorithms play in amplifying it.

Specifically, our study will measure real-world exposure to extreme, hateful, and conspiratorial content on YouTube among a nationally representative sample of Americans. We will also conduct the most comprehensive audit of the YouTube algorithm to date, allowing us to identify to what extent personalized recommendations increase exposure to objectionable content. By combining traditional social science methods (a survey) with computational methods (web traffic data and an algorithm audit), this study will provide the most systematic measurement to date of exposure to extreme, hateful, and conspiratorial content on YouTube. In addition, we will exploit the longitudinal nature of our data to assess the extent to which YouTube’s algorithm leads people into spirals of exposure (so-called “rabbit holes”) over time.

Authors

• Brendan Nyhan, Dartmouth College

• Jason A. Reifler, University of Exeter