“Auditing Radicalization Pathways on YouTube,” presented at the 2020 ACM Conference on Fairness, Accountability, and Transparency (FAccT) by Manoel Horta Ribeiro, Raphael Ottoni, Robert West, Virgilio Almeida, and Wagner Meira Jr., was one of the first large-scale quantitative tests of the widely discussed claim that YouTube pushes viewers toward extremism.
The researchers analyzed 330,925 videos on 349 channels, which they grouped into four categories: mainstream Media, the Alt-lite, the Intellectual Dark Web, and the Alt-right. They also examined more than two million video and channel recommendations collected between May and July 2019. Their main finding, drawn from user comment activity, was that the three non-mainstream communities increasingly shared a user base and that users consistently migrated over time from milder to more extreme content: a substantial share of people commenting on Alt-right videos had previously commented on Alt-lite and Intellectual Dark Web content. On the recommendation graph itself, they found Alt-lite content was easily reachable from Intellectual Dark Web channels, while Alt-right videos were reachable mainly through channel recommendations rather than direct video suggestions.
The study became a touchstone in the debate over whether recommendation algorithms cause radicalization or merely reflect demand that already exists. Later work, including studies using fresh accounts after YouTube changed its recommender, reached more mixed conclusions, and the question remains contested.
Why business readers should care: this paper modeled how to study a recommender’s real-world effects empirically rather than by anecdote, and it framed a debate - algorithm-driven versus demand-driven behavior change - that recurs across every engagement-ranked platform.