Applying PageRank to YouTube’s Alt-Right Pipeline
In this 2018 Data and Society report, Rebecca Lewis identified a network of YouTube creators who constituted the oft-discussed alt-right pipeline on YouTube: creators whose content led predominantly young, white men into alt-right and white nationalist radicalization. The creators in the network range from conservative/libertarian to overt white nationalist, and lend each other credibility by appearing on each other’s shows. The YouTube users who fall into the pipeline come to adopt more and more extreme views via a process of gradual desensitization to radical ideas: the more times they hear an idea, the more they come to accept it as a baseline truth. Thus, when more radical creators such as Stefan Molyneux, who advocates for scientific racism, appear repeatedly on many other more moderate channels (where their ideas are not challenged), the viewers of those channels can become more radicalized and fall even deeper down the rabbit hole of reactionary content.
We can imagine modeling this network of YouTube channels as a bipartite graph consisting of hub channels and authority channels. In the real world, the “alternative influence network” as described by Lewis is nowhere near bipartite, but we can imagine a simplified version of the creator network to be. Your hub channels might be talk show hosts, debate hosts, or podcasters who frequently have guests join them in their videos. The authority channels would be more independent creators who make solo videos about topics they find interesting, but also appear in other channels’ videos. We add an edge to the graph every time an authority creator appears in a hub creator’s content. In this context, let’s imagine that we run the pagerank algorithm to determine the relative relevance of each channel and determine the order that videos pop up in a YouTube search or in one’s recommended videos. The authority creators that make the most guest appearances (especially on channels that host a lot of other guests) would garner the highest authority scores. The hub creators that host a lot of guests, especially popular guests who appear on many shows, would gain the highest hub scores. This becomes problematic in the context of the alt-right-pipeline when we consider what types of guests are the most interesting to have on hub creators’ shows: more radical individuals with more shocking views and behavior make the most entertaining guests, and are then more likely to be invited on lots of shows. No political commentator or debater wants to host someone who is completely rational and open to changing their mind: we want to see drama! This gives those dangerously radical creators high authority scores, and lends extra hub score to the creators that are willing to host them. In turn, creators are rewarded for their radicalness or willingness to host radical creators with more pagerank and better treatment by the YouTube algorithm. The result? More vulnerable young people are exposed to ideas can endanger society.
Original report: https://datasociety.net/wp-content/uploads/2018/09/DS_Alternative_Influence.pdf