TikTok

TikTok is a mobile app in which people publish and watch short videos. The videos usually last between fifteen seconds and three minutes. The crucial difference from older social networks lies in what you see when you open it. On Facebook or Instagram, for a long time you mainly got posts from people you yourself followed. TikTok instead shows an endless stream of videos that a piece of software has picked out for you. This software watches what you view and infers from that what interests you. The app is run by the Chinese company ByteDance, which launched it in 2016.

Why TikTok reshaped the social media industry

TikTok showed that a social network no longer needs social connections. Someone starting fresh has zero followers and still immediately gets fitting videos. Conversely, even a completely unknown user can reach millions of viewers if a video strikes a chord. On the older networks, by contrast, reach depended heavily on how many people followed you.

This principle was so successful that competitors copied it. Instagram built Reels, YouTube built Shorts, both very similar short-video sections. There too, a recommendation algorithm now mostly determines what you get to see. Experts therefore speak of a shift from social network to entertainment channel.

Economically, the app is enormously significant because attention can be converted into advertising. Over a billion people use TikTok regularly. At the same time, the app is under political pressure. In the US and the EU there have been debates for years about whether a Chinese company should be allowed to collect so much data about Western users.

What the recommendation algorithm observes

At the core is a program that makes predictions. For every available video, it estimates how likely you are to watch it all the way through. The video with the highest estimated probability comes up next. Such systems are called recommendation systems and belong to machine learning: they learn rules from data instead of having them prescribed by programmers.

You supply the most important data yourself, without noticing. How long do you watch a video? Do you swipe on after two seconds? Do you watch it twice? These silent signals are more informative than likes, because almost nobody consistently taps the heart. Added to this is information about the video itself: audio track, length, visual content, hashtags.

A common misconception is that TikTok has a single secret trick. In fact, the underlying technique has been known for years and is also used by Netflix or Amazon. The advantage lies in the format: a short video is rated within seconds, whereas a film only after two hours. As a result, the system collects hundreds of responses per user per day and adapts very quickly.

TikTok in the news and in everyday life

In everyday life, you encounter the system as the “For You” feed. It often feels remarkably accurate, but it narrows easily. Anyone who watches three videos on one topic all the way through will soon see mostly that topic. Experts call this kind of narrowing a filter bubble. You can counteract it by actively swiping videos away or marking them as uninteresting.

In business news, TikTok mainly comes up in three contexts. First, in the dispute over a possible sale or ban of the US business. Second, in EU Commission proceedings concerning youth protection and addictive mechanisms. Third, as an advertising market in which TikTok takes revenue away from Google and Meta.

Technically, the app is also a textbook example of artificial intelligence at mass scale. It contains tools for automatic subtitles, translations, and filters that alter faces in real time. Anyone who wants to understand how recommendation systems influence the behavior of millions of people will find here the clearest case at present.