If I understand the methodology correctly, he built a spider that recursively followed all outgoing recommendations. He was able to find fringe content when starting with mainstream content. But is that so strange? I remember playing the "Hitler wikipedia game" where you start on a random page and try to find the Hitler page with the minimum number of clicks and it's typically possible to get there in a handful of cl…
What I don't totally understand is how it determines the directed nature of the recommendation graph, and how the destinations are labeled. What does it mean to "end up" at "alt right" videos? Do the recommendations have to exclusively lean right or just mostly? What about higher vs lower recommendations? If there are recommendation patterns that go left but don't end up in a tight loop are they considered somehow? There could also be some dependence from the chosen starting video / topic. I'd be curious to see the results from a variety of starting points.
Basically, it's an interesting exploration but I'd like to see it with the rigor of a peer reviewed paper, more than just the NYT.