Abusing the Recommended’s Algorithm

It’s no secret that social media algorithms have gone beyond easy understanding. Nobody really knows what makes something ‘blow up’ on social media websites outside of simple tricks like ‘popular take, made smarmy by the poster’, ‘re-posted content stolen from somewhere else’, ‘inflammatory or incorrect information being yelled into the camera’, and more. When something works, it works; when it doesn’t work, there’s often not really a clear reason.

Similarly, why Google will offer up some mysterious video with 90,000 views to someone is also not well known. It gets shown to a couple of people, those people watch for a fraction of a second longer, it gets shown to more people, the cycle repeats until the video is viral.  You see it all the time, comments asking why Google thought that the viewer would be into pickling or drywall repair when they normally watch LetsPlayers or Dramatube content. However, by commenting, they’re reinforcing whatever connection Google thought it made, and demonstrating to it that there is some connection. It doesn’t make sense and even people working specifically with the algorithm for Google can’t fully isolate what series of decisions the code made in order to do that. It’s essentially a black box, which is not great. As a bonus, it makes selling adspace also a gamble, because the videos some company wants to be shown on are not always the videos in their price range even when they should be. Nobody likes dumping money into a black box except when they’re in Vegas!

That said, some people treat this bizarre and esoteric recommendation system as a sort of astrology in itself. Well, why did Youtube show you this video over that one? Does the machine itself know something you don’t? Are you subliminally sending out signals you’re going to need a video about making bread soon? Does that have something to do with the weather your area is about to get? Is this the effect of the “Dark Patterns” you’ve doubtlessly heard about by now?

It really does feel like the machine is trying to steer you into content sometimes. Youtube will spit up a video for you, and put it front and center for your recommendeds for weeks! You may have to actively tell it you don’t want to look at the thumbnail anymore. The home page with no account signed in is a maelstrom of chaos, but Youtube knows you well enough to keep a relatively limited cycle of videos going for you.

Some video makers are responding to this tendency by making themed, timed, or otherwise “marked” videos for the supernaturally-inclined. It sounds pretty harmless. The title may be something like “if you’re seeing this October 3rd, this video is for you!” and if that were all it was, someone making some themed content and then releasing it, then whatever. They want views, someone is looking for some reassurance, it works out. It’s already a little sus that there’s money involved (a lot of these have ads enabled) but whatever.

It starts to get a little more annoying when the video itself is actually using a generic thumbnail so the title can get swapped around, so instead of the 3rd, the next day it’s the 4th. Other services are not made to handle changes like this: once you upload a video to TikTok, that’s how the video is. You can change the description and the title (I think) but you can’t swap out the actual video itself, and even the patently superstitious are not falling for a generic video when more specific – but limited – ones are already out there. It makes the ruse obvious. Similarly, videos and tweets posted to X or Bluesky are already not date-bound like Youtube and TikTok are because of the retweeting, and Instagram is already set up to show you the freshest stuff first and everything else after.

These videos get uploaded, and then changed basically every day for a week or more until the video cycles out of ‘fresh uploads’ and into ‘seasoned content’ where it’s no longer recommended willy-nilly for going viral.

Is this actually a problem? Of course not. Ultimately this all means nothing, and if someone finds meaning or doesn’t, then it’s that person’s problem. It’s annoying. It may mislead them if the message isn’t newspaper-horoscope levels of generic, but it’s basically a drop in the bucket of the other stuff (cough cough Cambridge Analytica) the hidden algorithms are doing.