TikTok has two native feeds that offer different contents to the user: one shows the videos of the people a user follows and the other shows clips that might be interesting to the user, as understood by the social network algorithms. However, many users do not understand how TikTok selects the videos that are displayed in this recommendation column, called “For You” or FYP (For You Page). Below, Techidence explains how the algorithm works TikTok and how it chooses the videos from the feed For You.
How does the TikTok algorithm work?
The TikTok algorithm uses machine learning to determine which content a user can most easily engage with. To do this, it primarily identifies videos that are similar to those the user has interacted with before. But this recognition starts well before the interactions. To identify what kind of video would be considered attractive to the viewer, the system starts to capture information as soon as the viewer creates a profile on the social network.
TikTok explains that during account creation, each person should classify as interesting or not various topics such as pets and travel. After that, when the user opens the app for the first time, he or she will see eight popular videos referring to the topics that are considered fun. From there, a combination of factors will rank the videos and determine what will be shown next.
One of the main evaluation metrics is interaction with the videos. The system identifies as the interaction the act of liking, sharing, or leaving comments on the clips, as well as the accounts that each person follows. The content that the individual creates is also used in this scan.
However, besides the interactions, some details of the production are also taken into account. The algorithm analyzes subtitles, hashtags, effects, and sounds contained in the production so that it can later show clips that contain similar items. Language preferences, country setting, and the user’s device are also used to suggest new clips, but with less decision weight.
With all this data identified, TikTok can map a user’s preferences and create a pattern of videos considered interesting to them. Besides, the system also matches the individual’s tastes to accounts with similar interests, creating a common video network that will display clips considered enjoyable to both.
How does this impact each person’s feed?
TikTok explains that the “For You” feed can show relevant recommendations over time, but does not specify exactly how long. Also, it should be clarified that the algorithm analyzes user behavior within the social network as a whole, not just within the “For You” tab. In other words, when you browse the app and follow a friend of an acquaintance, this is also taken into account by the system to create its recommendations.
With all this information, TikTok claims that there is no single, “canned” feed within the social network. This means that even if people initially watch the same videos, the way they behave on the platform will determine the experience for each one. In other words, each person’s feed is unique and personalized for each person, and this happens precisely because of the algorithms.
Doesn’t the content become repetitive?
Despite using a logic that seeks to avoid redundancy, such as showing several videos in sequence from the same creator or with the same music (which could bore the user), TikTok admits that the recommendations created by the algorithms can create a “filter bubble”, showing the same type of video to people. This would prevent users from seeing more varied videos and from broadening their horizons within the app. The company, however, says that they are studying the behavior of these bubbles, how long they last, and what to do to break them when necessary.
Also, TikTok says that these bubbles can reinforce the establishment of rumors, misinformation, and conspiracy theories if the user does not have access to other points of view on the web because of the repetition of the algorithms. With this possibility, the company says it studies accounts and videos that may be linked to fake news and sends these contents to human reviewers so they can be managed before being shared on the “For You” tab.