Manufacturing Misogyny: How The YouTube Recommendation Algorithm Radicalizes Young Men

Abstract

In the last year, there has been an explosion of misogynistic, homophobic, and transphobic content creators across YouTube and other social media platforms led by Andrew Tate. This content has been viewed billions of times, including by many impressionable adolescents who have been spewing Tate’s hateful talking points to their classmates and teachers. With 26% of U.S. adults getting political content from YouTube and three-quarters of teens visiting the site daily, it is important to investigate how users are exposed to this extreme content to understand the risk of algorithmic radicalization on the platform.

In this thesis, we take a network approach to investigate the presence of radicalization pipelines on YouTube toward misogynistic content. Drawing on data from the YouTube Data API v3 and direct webpage scraping of YouTube, we construct our network through a multi-step crawl of YouTube recommendations and visit over 30,000 unique YouTube videos. We also propose a novel method of personalized analysis of social media websites through the application of fabricated browser cookies.

Through our network analysis, we identify evidence for a radicalization pipeline on YouTube that flows through anti-Social Justice Warrior (Anti-SJW) content to the “Manosphere,” or channels with Tate-like content. Our novel approach to creating a personalized network for an “Anti-SJW user identity” results in stronger radicalization pathways, providing supporting evidence for algorithmic reinforcement of radicalization. Our study, like work that has investigated similar pipelines, demonstrates the need for a reorientation of algorithmic incentives around a focus on social externalities over engagement maximization.