How to Train Your Algorithm
This video, titled “How to Train Your Algorithm“, uses graphics and narration to explain how recommendation algorithms work and how users can influence the content they see.
The video begins with the text “How to Train Your Algorithm” displayed on screen, accompanied by graphics notes indicating narration about recommendation algorithms choosing most of what we see online.
Next, the screen shows two interlocked gears turning together. The view zooms out to reveal two people standing underneath the gears, one presenting as a boy on the left and one presenting as a girl on the right. A hockey stick drops out from between the gears and is caught by the boy, while a pink cake drops out and is caught by the girl. The boy and girl then move closer together and trade the items.
The scene changes to show the boy and girl from the previous scene assembling a more complicated set of gears. These gears are all different colours. Once they finish assembling the gears, they start turning, and a cat drops out. The narration explains that algorithms make guesses based on what they think they know about you, and if those guesses are wrong, they might not show you what you want to see. It’s highlighted that knowing how to train your algorithm allows you to see more things that make you feel good or fewer things that make you feel bad, or even just more cute cat videos.
The video then transitions to a scene titled “Start from scratch“. Visually, a person erases a chalkboard that is covered with gears and many complicated-looking numbers and equations. The narration advises starting from scratch, mentioning that some apps have privacy settings to forget your information and stop collecting it. Following this, a phone screen is shown. A finger swipes the “gear” icon to display a menu with sliders. The finger then moves some of the sliders from “on” (indicated in green) to “off” (indicated in grey). The narration adds that you can also use privacy apps and phone settings to limit how much apps learn about you.
The next section is titled “Send the right signals“. The narration explains that now that you have a clean slate, you can send the right signals and choose what the algorithm knows about you. A phone screen appears, displaying a video app similar to YouTube or TikTok. A video of a cat doing something cute is playing. A finger taps the “thumbs up” and Share icons located under the video. The narration explains that actions like sharing, liking, and watching videos to the end send a signal that you want to see more similar content. Subsequently, a new video appears showing a stereotypical French mime. The finger taps the “thumbs down” icon and then swipes the video to the left off the screen, revealing a new cat video. The narration further explains that if you don’t want to see more of something, some apps offer options like downvoting or indicating you’re not interested, and skipping videos quickly also sends a signal.
The video moves to a section titled “Vote with your views“. The visuals show a YouTuber preparing to make a video. The YouTuber has both a cat and a dog, both wearing cute accessories like bow ties, and the YouTuber is shown deciding between them. The narration emphasizes that your choices not only affect your feed but also guide decisions about which new shows and videos get made. The view then zooms out to show the gears from the beginning (scene 1-2). Above the gears, there is a big flashing sign with a dog on it. The narration mentions that many creators report making videos they believe algorithms will recommend rather than what they truly want to create. Therefore, the narration advises supporting the videos you enjoy, even if the algorithm doesn’t promote them.
The final scene shows the boy and girl from earlier giving a thumbs up. The narration concludes by stating that sending all these signals will help you successfully train your algorithm so that you receive more of the content you actually want.