The Anatomy of Rabbit Holes: Studying Information Segregation in YouTube’s Recommendation Graph

The impact of YouTube's recommendation algorithm has been subject to much debate. Case studies and anecdotal evidence have suggested that algorithmic recommendations may lure the platform's users into informational ‘rabbit holes' where they are exposed to biased content and misinformation. However, empirical research on the phenomenon has been challenging to conduct because the effects of the algorithm are inherently confounded by user and creator behavior. In this chapter, we study the net effect of algorithmic recommendations through a comparative network approach. That is, we collect the recommendation links between videos for various political and non-political issues. The resulting networks are then compared on various structural network characteristics. If the algorithm creates rabbit holes, one would expect the recommendation networks of videos on conspiratorial content to be more modular, less centralized and intra-topically dense. We analyzed 15,455 videos and 154,311 recommendation links on 40 different topics, each belonging to one of four main categories: news, science, conspiracy, and non-controversial. On the macro level, we find that conspiratorial video networks are small, sparse and centralized, with their density partly explained by lower view counts and network size. The sentiment of their videos is negative on average–like news networks, but unlike science and non-controversial networks. However, sentiment does not but view count does predict how often videos are recommended.