Read on Twitter
Excited to share a new preprint "Evaluating the scale, growth, and origins of right-wing echo chambers on YouTube", with great @Amir_Ghasemian & @aaronclauset & @markusmobius & @DavMicRot & @duncanjwatts:
http://arxiv.org/abs/2011.12843
Here’s a little summary: 1/10
http://arxiv.org/abs/2011.12843
Here’s a little summary: 1/10
Recently, YouTube’s outsize influence has sparked concerns that its recommendation algorithm systematically directs users to radical right-wing content. However, systematic evidence for these claims are elusive and anecdotes do not on their own indicate systematic problems. 2/10
We investigate these concerns with large scale longitudinal data of individuals’ browsing behavior from a representative sample of the US population, with more than 20M watched videos, with nearly half a million unique videos spanning the political spectrum. 3/10
Our results show the existence of distinct political news “echo chambers” on YouTube, in which users predominantly consume videos from one political category. 4/10
The far-right chamber is larger than the far-left (smaller than the centrist); is growing rapidly in size and watch time. Its users are more engaged and more likely to stay engaged in the future than users in other echo chambers, especially when exposed to bursts of content. 5/10
Users of radical content on YouTube are actively seeking it & do so both on & off the platform.While not ruling out the possibility that recommendations are driving engagement for heaviest consumers, nonetheless show consumption of radical content extends well beyond YouTube.6/10
The pathways by which users reach far-right videos are diverse & only a fraction can plausibly be attributed to platform recommendations. Instead, a large fraction are referrals from outside of YouTube, meaning other far-right websites are using YouTube as a content library. 7/10
Within sessions of consecutive video viewership, we see no trend toward more extreme content, indicating that consumption of this content is determined more by user preferences than by recommendation. 8/10
Nearly 80% of sessions are length one — 55% of videos are in sessions of no more than 4 videos—& reflect the tastes and intentions that users entered with. Our findings suggest that it is a much broader problem than simply the policies & algorithmic properties of a platform. 9/10
Takeaway: The recent focus on the recommendation engine is overly narrow. Rather, YouTube should be viewed as part of a larger information ecosystem in which extreme and misleading content is widely available, easily discovered, and both increasingly and actively. 10/10
