Auditing Radicalization Pathways on YouTube

EXECUTIVE SUMMARY

  • The study investigates the existence of radicalization pathways on YouTube, specifically how users may progress toward more extreme content.

  • This is the first large-scale audit providing quantitative evidence of user radicalization on the platform.

RESEARCH OBJECTIVES

  • To assess user radicalization on YouTube and the impact of its recommendation algorithm.

  • The authors aim to answer three central questions:

    • RQ1: How have these channels grown on YouTube in the last decade?

    • RQ2: To what extent do users gravitate towards more extreme content?

    • RQ3: Do algorithmic recommendations influence users towards more extreme content?

METHODOLOGY

  • The study encompasses an analysis of 330,925 videos from 349 channels, categorized into:

    • Media

    • Alt-lite

    • Intellectual Dark Web (I.D.W.)

    • Alt-right

  • Data collection involved:

    1. Identifying channels from these communities via seed channels, keywords, and YouTube’s search functionality.

    2. Collecting metadata, comments, video/channel recommendations (non-personalized).

    3. Total of 72M+ comments analyzed.

    4. Recommendation graphs were constructed to study the relationships between channel types.

FINDINGS

GROWTH OF COMMUNITIES

  • Alt-lite, I.D.W., and Alt-right communities have experienced significant growth in terms of views, likes, and comments since 2015.

  • The Alt-right community, while historically dormant pre-2016, saw a steep rise in activity coinciding with the 2016 U.S. presidential elections.

USER INTERSECTION AND MIGRATION

  • Users tend to migrate from I.D.W. & Alt-lite to Alt-right content over time.

  • 50% of users commenting on Alt-right channels also engaged with Alt-lite and I.D.W. channels, suggesting a shared user base.

  • Users initially engaging with milder content (I.D.W. or Alt-lite) began consuming Alt-right content particularly post-2016.

  • Weekend exposure to Alt-right content increased significantly among users commenting exclusively on I.D.W. or Alt-lite channels, with rising percentages traced back to these channels.

RECOMMENDATION ALGORITHM

  • The analysis of recommendation algorithms indicates that Alt-lite and I.D.W. content serves as a gateway to Alt-right channels.

  • Users can navigate from milder to extreme content primarily through channel recommendations, rather than video recommendations.

  • Reachability from I.D.W. to Alt-lite channels is approximately 25% after five steps; from Alt-lite to Alt-right, the reachability is around 4%.

DISCOURSE

  • Discourse regarding radicalization pathways is tied to broader societal mistrust in mainstream media, with rising consumer engagement on platforms like YouTube as alternatives.

  • The terms ‘Alt-lite’ and ‘Alt-right’ are defined; the former as deniers of explicit white supremacist stereotypes while the latter as more extreme proponents of racial nationalism.

DATA OVERVIEW

  • Tabulated datasets were provided, including:

    • Growth metrics on likes, views, and comments across communities (Tables 4 and 5).

    • Average engagement metrics showing that Alt-right channels have nearly 1 comment per 5 views.

CONCLUSIONS

  • There is a systematic pathway from milder content consumption to Alt-right radicalization on YouTube.

  • Evidence suggests that user radicalization is not only present but has significantly increased, influenced by YouTube's algorithmic recommendations that facilitate access to extreme content.

  • The study calls for further exploration into the causes of radicalization mechanisms beyond mere statistical analysis, emphasizing the need to analyze the content and narratives perpetuated within these communities over time.