How does YouTube learn what I want to watch? YouTube basically knows what types of videos you watch and how long you watch them. It has a lot of other information as well, but it does its best to match new videos with the type of video you are always watching.
All social network sites work on the same principle. We think that YouTube is a video site, Google is a search engine, Facebook is a way to keep in touch with our friends, and Instagram is a way to see photos of beautiful places and people, but they are not. They are all first and foremost advertising sites. They are all worth billions of dollars because of advertising. The longer people stay on their sites, the more adverts they can show and the more money they can make. In order to get people to stay for longer, they have to give us what we want and make us stay. If I only like cats and YouTube shows me nothing but dog videos, I won’t stay on the site for very long and Google won’t make any money from showing me adverts. Therefore, YouTube needs to know that I like cats, and what kind of cats, and then show me as many cat videos as possible. So, how does YouTube know that I like cats?
YouTube needs to be able to do two things. Firstly, it has to be able to tell what kind of videos I like. Secondly, it has to be able to tell what a video is about. Every action you make on YouTube is recorded and added to your profile. YouTube knows what kind of videos you search for. It knows what videos you watch, and it knows what videos it offers you that you don’t watch. It knows which videos you like, and it knows which parts of videos you skip over. It knows how long you watch a video for. All of this information can be plugged into an algorithm to make it more accurate and each day that you watch more videos, the algorithm will become even more accurate. If you like cats, but you only ever watch videos of orange cats, the YouTube algorithm will start to only offer you orange cats. Then, if you only watch the videos where the orange cat is wearing a hat, your homepage will only have videos of orange cats wearing hats. Occasionally the algorithm will throw in other cats to see if you are really happy with the cat wearing orange hats and through this process, it will narrow down its image of what you want to see. This is excellent for the viewer because we only get videos we want to see. It is also excellent for YouTube and Google because we will watch far more videos if they are ones we like. It is not quite so good if you want to be exposed to new things. There is something called “confirmation bias”, which is where you only get information that supports the views you already have. For example, if you think the Earth is flat and you only get videos supporting your idea that the Earth is flat, you will double down on what you believe.
The second part is YouTube needs to know what a video is about. There are several ways that they can do this. The first is simply by looking at the information the uploader included. They use tags, a title, a description, categories, and other things. All of this helps YouTube narrow down what the video is about. Then YouTube can analyze the audio of the video for key words, phrases, and ideas. Then YouTube can analyze the video visually to find images and content. All of this will give YouTube a pretty good idea of what the video is about and allow it to recommend videos about hat wearing orange cats to you.
However, YouTube has a few other tricks up its sleeve. Firstly, it can look at the comments on a video and analyze those. The comments will give it more of an idea what the video is about but, more than that, the comments will tell YouTube what kind of person is watching and engaging with that video and whether they are engaging positively or negatively. Secondly, YouTube can look at its vast data resources and know what other people watched before they watched this video and what they watched afterwards. That will also help inform it of the contents. The third trick is that YouTube’s algorithm, or AI, is fluid. It is constantly looking for data to improve itself. It compares you to what you used to look at, and to what other people look at. If you start looking at orange cats wearing shoes instead of hats, the algorithm will quickly adapt and change the videos you get on your homepage. It is a very clever system that is designed to keep you on the site as long as possible. It is highly addictive. It could be compared to heroin that can constantly adapt itself to please you. And this is what I learned today.
Try these
Sources
https://buffer.com/resources/youtube-algorithm
https://www.repustate.com/blog/youtube-video-analysis
https://www.makeuseof.com/tag/how-do-social-networks-make-money-case-wondering
Photo by Viralyft : https://www.pexels.com/photo/man-using-internet-in-smartphone-16852624/