My Video Went Viral. Here's Why

Veritasium very recently spoke about algorithm optimizing for click through rate in the context of the bad. this video It seems okay in this case though.

Let me Godwin it up >We're optimizing for click through rate. and so was Hitler in ~1923. Optimizing something that affects how other people act (same happens on YT Veritasium:My Video Went Viral this video ) leads to bad outcomes (radicalization, race to the bottom, catering to lowest denominator etc).

This seems like it takes some notes from Veritasium's theory on YouTube's recommendation algorithm which he posted after his initial reservoir shade balls video went viral. (edited for clarity) this video for his theory.

Immediate counter-arguments: Most professional YouTubers complain about having to tailor their videos to suit the recommendations algorithms. Historically, many short videos were padded out to just over 10 minutes in order to reach the minimum threshold to insert a midroll advertisement; today, many videos are as long as possible to take advantage of the algorithmic prioritisation of view time over view count. There is widespread dissatisfaction with the recent necessity to use clickbait thumbnails and titles, due to YouTube's increasing weighting of click-through-rate. this video Most analysts believe that the high weight on view time is causing the YouTube algorithms to promote the kind of videos that their heaviest users watch . The problem is that these heavy users aren't representative of the general population - they're far more likely to be interested in conspiracy theories or extremist politics. link The vast majority of YouTube video views are as a result of algorithmic decisions by YouTube - what to show in the recommendations bar, what to prioritise on the home page, what to put at the top of search results. Those algorithms are not and cannot be neutral. These algorithms aren't just shaping which videos people watch; YouTube creators freely admit to tailoring the content of their videos to fit the recommendations algorithm.

What if YouTube just does the same, but the goal of the algorithm is not to maximize like/dislike ratios, but actual click-through rate and watch time? [1] It's an ad company after all. Also, the domain of TikTok videos seems very limited in comparison to YouTube videos, so your like-based algorithm might result in bad user retention with this broad strategy of showing random videos. So the question really is: If YouTube's algorithm is more sophisticated than TikTok's and TikTok is trying to do the same, would they still be as transparent about it? [1] this video

Regardless how to define what's garbage and what's not, YouTube seems to remove demonetized videos from recommendations. Since I don't think that good content is always necessarily advertiser-friendly, a lot of good videos are just not getting any attention because a few big companies decided that they're not helpful to market their products. Example: Historical or educational videos which are about controversial topics or just have the word "war" in their title. Basically, their recommendations algorithm favors profit from advertising over personalized recommendations because it seems to be designed to maximize click-through rate and watch time for videos with many ads. [1] [1] this video

Can't agree more except for the long-videos thing. The way YouTube works now encourages clickbait crying-for-a-click titles and silly-shocked-faces thumbnails because click-through-rate plays a huge role in video success now. In his video [0], that youtuber explains that and why he was going to shift toward using clickbait titles in his channel. Because otherwise, you would not thrive on YT. For long videos, I think sometimes even 10 minutes are not enough to convey a substantial message in a video. But I think you mean those youtubers that prolong their videos longer—without added value—just for more ads. [0]: this video

>but I saw a video of his talking about this. He doesn't actually like these titles, but claims to have proof they improve viewership counts Veritasium also had video on this topic: this video

> The way most entertainers get known is just through sheer luck or connections. Derek Muller has a couple of videos talking about the success of his Veritasium YouTube channel (and work, luck, and success generally): this video link

The Veritasium channel does this with all its videos; it lets Derek test which thumbnail/titles are effective and which aren't. He talks about it in this video this video at about the 15:50-16:00 mark

You Sir, are a Gentleman and a Scholar :-) >Now everything on the internet is "content," a hyper optimized form of human intellectual output that is devoid of art, minimal on information, excreted instead of created, measured to the lowest common denominator, monetized yet worthless, just enough bait on the hook for the next lizard brain click. I could not have said it better myself ! The Internet has become a place to publish/share every trivial thought that has ever popped into every mediocre mind. There is no self-regulation/self-censoring anymore. The Noise is just overwhelming and it is extremely mentally fatiguing just trying to get at any kind of Signal. You just give up and start numbing your mind with all the junk available. PS: Here is Veritasium explaining the insidious effects of "Youtube Platform" Algorithms - this video Now extrapolate the same idea to other dominant platforms like Facebook, Google etc. and you realize that "The Internet" has become a platform for Expression and not necessarily Information. PPS: The book Networked Life: 20 Questions and Answers by Mung Chiang gives an inside algorithmic view of how the above platforms work.

This article is only half-right. Some of the other things you don't see because the algorithm doesn't promote them are: - scientists and educators that make 1 hour long videos to explain how something is incorrect. Example: link (less than 0.1% reach of the original) - long threads of detailed analysis with cross-references - long educational material that explaines e.g. confounders in order to explain why some statistical results observed are not possible to interpret in a certain way The reason you don't see this kind of content is simple: the level of its engagement is low. Its not clickbaity, its not exciting, it doesn't have a rebel perspective, it doesn't claim to uncover something nobody has ever seen before. Its often kind of a downer too, pointing out other people's mistakes. Social media sites make their money through engagement. The longer you stay on their site, the more ads they can show you. As such, their default algorithms promote engagement at any cost . It turns out that this trains creators to spread clickbaity, potentially misleading, false but exciting and novel "information" (see paper [1] as well as Veritasium's youtube specific explanation [2]) You know what other thing spreads a lot? Outrage. Extreme reactions to spread of misinformation are also engaging. People will fight over them due to strong feelings: one side frustrated by seeing the effort put into debunking having absolutely no effect (because it doens't get promoted by the algorithms, nobo

Just like we recognized tabloids for what they were, we need to recognize social media for being the same phenomenon disguised in a "grassroots" cloak (i.e. crowdsourced tabloids) There is nothing organic about the speech that happens on YouTube. Veritasium explains the shift that happened when youtube switched to an algorithmic model from the subscriber model: this video - you need to carefully manipulate people to click your thumbnail based on the picture and the video title. Making outrageous but believable shit up to sensationalize your content is a well known tactic of tabloids (which does continue today to some extent). YouTube and other social media incentivise exactly that.

Committed to, actually. He made a video about how they were going to start doing it more: this video . Also, more recently: link .

Veritasium has a video about this and I’ve seen other creators like GothamChess and maybe even Tom Scott talk about how they have to create clickbait thumbnails otherwise they get no views. The YouTube algorithm punishes creators who don’t use clickbait. this video