Recommended Polarization: How YouTube's Algorithm Pushes Real People Content

• In Event: Media and Polarization

Abstract

To what extent does the YouTube recommendation algorithm push users toward more ideologically congruent content? Existing research on this question has relied on anonymous browsers paired with user-agnostic APIs, and has found very little evidence that the recommendation algorithms are to blame. We hired real YouTube users to navigate from a provided starting video and clicked on recommended videos according to a simple rule (i.e., always click the first video, second video, third video, etc.). We demonstrate that the YouTube recommendation algorithm does, in fact, push real users toward more ideologically congruent content by showing that Republican users are exposed to more conservative content than are Democratic users. Furthermore, we show that this difference is entirely due to the algorithm recommending more conservative content to Republicans than what would be shown due to statistical mean reversion. Our analysis provides an ecologically valid estimate of the degree to which recommendation algorithms influence the ideological content that the public consumes from the supply side.

Authors

• Megan Brown, New York University

• James Hodgdon Bisbee, NYU

• Richard Bonneau, NYU

• Joshua A. Tucker, New York University

• Jonathan Nagler, New York University