YouTube could be one of the most powerful radicalizing tools of this century, given its billion users and algorithms that recommend ever more extreme videos
Was a Key Factor Missed in the Study of Viral Lies? Jillian D'Onfro / CNBC : How YouTube search pushes people toward conspiracy theories and extreme content Nick Douglas / Lifehacker : How to Hide YouTube's Related Videos Natasha Lomas / TechCrunch : Report calls for algorithmic transparency and education to fight fake news Facebook: Zeynep Tufekci : My latest for the New York Times: Recommender algorithms … Tweets: @niemanlab : “Videos about vegetarianism led to videos about veganism. Videos about jogging led to videos about running ultramarathons. It seems as if you are never ‘hard core’ enough for YouTube's recommendation algorithm.” http://www.nytimes.com/... Zeynep Tufekci / @zeynep : YouTube may well be operating as a giant radicalizing engine through its recommendation algorithm—leading people down a rabbit hole of misinformation, hoaxes and incendiary content. My latest for the New York Times on one of the most overlooked issues. http://www.nytimes.com/... http://twitter.com/... Dan CoopEUr / @danielwcooper : True fact: I watched a couple of Norm MacDonald clips. The right rail filled with more Norm, then videos about libertarianism crept in, then ones titled “Why SJWs are disgusting.” You can go from standup to racism in an hour. http://twitter.com/... Maya Wiley / @mayawiley : #Google “algorithm seems to have concluded that people are drawn to content that is more extreme than what they started with — or to incendiary content in general.” #BigData http://twitter.com/... Sanho Tree / @sanhotree : I know a pastor who became a gold bug after getting sucked into a @YouTube echo chamber. He converted all his assists into gold, started buying guns (to protect his gold), and has turned into a crazy prepper who stockpiles food while awaiting global collapse—like Jeebus intended. http://twitter.com/... @nathanjurgenson : “As we click and click, we are carried along by the exciting sensation of uncovering more secrets and deeper truths. YouTube leads viewers down a rabbit hole of extremism” http://www.nytimes.com/... Chris Albon / @chrisalbon : “Let's watch a video on t-tests!” [two weeks later] “The fundamental problem with deep adversarial networks is...” http://twitter.com/... Jason Willick / @jawillick : Nature of human appetite—desire to perpetually heighten sensation, whether in lifestyle or politics or pornography—creates dark incentives for internet content creators. http://twitter.com/... Ethan Strauss / @sherwoodstrauss : You've probably noticed this. Watch something on YouTube and YouTube's algorithm suggests you watch more extreme versions. This impersonal algorithm might be doing more to rile up society than any lone impassioned human http://twitter.com/... NYT Opinion / @nytopinion : For all its lofty rhetoric, Google is an advertising broker, selling our attention to companies that will pay for it. The longer people stay on YouTube, the more money Google makes. http://www.nytimes.com/... http://twitter.com/... Paul Krugman / @paulkrugman : Interesting, important, and scary. I personally use YouTube almost entirely to watch live musical performances; now I have to wonder whether the recommendation engine is somehow exaggerating my tastes. But the political side is important. https://nyti.ms/2Df1IoJ Zeynep Tufekci / @zeynep : So to people taking issue with the word radicalizing. What I am talking about here is radicalization through misinformation, lies and hoaxes—not radicalization in the sense of challenging power or outside of mainstream in a grounded, radical manner. That's not the dynamic. John Sipher / @john_sipher : YouTube's “algorithm seems to have concluded that people are drawn to content that is more extreme than what they started with — or to incendiary content in general.” Sugar and fat for the brain. https://nyti.ms/2GeTMa6 Thomas G. Dietterich / @tdietterich : If this is due to machine learning, it is the most clear-cut case yet of optimizing the wrong objective. I'll bet it becomes a case study in future AI Ethics classes. Great article by @zeynep https://www.nytimes.com/... Dan Gillmor / @dangillmor : Sometimes I think the tech platforms actually want government interference, if not control. But if you think things are bad now, just wait until that happens. Meanwhile, read @zeynep's latest, on YouTube: http://www.nytimes.com/... Dare Obasanjo / @carnage4life : YouTube has mostly escaped criticism as a social media app whose engagement focused algorithm encourages terrible stuff to spread but it should. Their algorithm optimizes for watch time independent of content outside of policing porn & copyright violations. http://twitter.com/... Claire Lehmann / @clairlemon : And it's not just politics: “Videos about vegetarianism led to videos about veganism. Videos about jogging led to videos about running ultramarathons.” https://www.nytimes.com/... Zeynep Tufekci / @zeynep : Folks, please go read this whole thread below. All of it. And then remember that Chromebooks, which come with access to YouTube (and that I use because great sturdy secure machines!), are half the K-12 market in the US. http://twitter.com/... Radley Balko / @radleybalko : This definitely seems like a problem. But it's hard to conceive of a regulatory solution that wouldn't be worse. Government officials determining what is and isn't “radical” is a dangerous road. http://www.nytimes.com/... Malachy Browne / @malachybrowne : YouTube often “fed far-right or far-left videos to users who watched relatively mainstream news sources.” https://nyti.ms/2Df1IoJ Jason Abbruzzese / @jasonabbruzzese : I think this is the thing that could really change the narrative from “this is bad” to “this is an emergency.” Kids watch a ton of YouTube and I think parents are starting to realize that it's a gateway to some pretty gnarly stuff. https://twitter.com/... See also Mediagazer
Context & Ripple Effects
This NYT argument lands mid-arc in a running debate over what YouTube's recommendation engine actually does. James Bridle had already documented the platform automatically serving traumatizing content to young children, and months later former far-right extremists described being radicalized as teenagers by the same algorithm that keeps surfacing extremist videos.
What makes this moment consequential is that the causal story was still contested: a later report argued right-wing content spreads through supply and demand rather than the recommendation system itself, while a former alt-right creator explained how he deliberately tailored confrontational videos for the echo chamber the algorithm rewards.
First-order effects
- YouTube faces an immediate enforcement bind: curbing misinformation means punishing its own stars, a step the platform has not shown it can take broadly despite its recent policy shift.
- Advertisers and mainstream creators sit downstream of the same watch-time optimization, since incendiary content earns the engagement the algorithm pays out for.
Second-order effects
- Creators rationally adapt to the incentive structure: gaming commentators use anger to gain views, and researchers say those videos are amplified by the same recommendations.
- If YouTube polices supply without touching the ranking system, demand migrates to whatever formats still ride the algorithm — shifting the battleground from individual videos to recommendation design itself.
Third-order effects
- The dispute over whether the algorithm causes extremism or merely reflects demand is becoming the central question regulators and transparency advocates must answer before they can mandate fixes.
- If engagement-optimized recommenders are treated as radicalization infrastructure, platforms like Google-owned YouTube face structural pressure toward algorithmic accountability rather than case-by-case takedowns.
The trend: Engagement-optimized recommendation systems are moving from growth engines to objects of public-health-style scrutiny, with the cause-versus-mirror debate over radicalization setting the terms of future regulation.