TikTok is already one of the most popular applications today, which naturally creates suspicion about how the content reaches the user. For this, the video social network detailed how the content of the # ParaYou feed is recommended.
In new accounts, the user selects interest categories that help to adapt the recommendations. Thus, an initial feed with videos is built, and is adapted according to the interests of each one. But then what?
How does TikTok recommend content?
If new users don’t select categories, the feed is created with popular videos. From the first likes, comments, replays and people who follow, the recommendations can change and become more relevant.
“The # ParaYou feed is one of the features that define the TikTok platform, but we know there are questions about how recommendations are delivered to your feed,” said the social network in a statement.
She says the system “recommends content by rating videos based on a combination of factors”. But what factors are analyzed to “teach” the recommendation system?
- User interactions: videos you watch, like and share, accounts you follow, comments and published videos
- Video information: subtitles, sounds, hashtags
- Settings and account: language, country, device type
According to TikTok, these factors are recorded to obtain “better performance”, however “they receive less weight in the recommendation system compared to other measured data points”. What are these points?
She says that watching a long video from start to finish gets more weight in the recommendation than metrics like the same country as the viewer and the video creator. The company also cites that having more or less followers – or previous videos with high performance – is not a determining factor for the recommendation system.
TikTok points out that the user can also use the “I’m not interested” option or report (and hide) videos or accounts. The actions contribute to the next recommendations.
At another point, it is mentioned that the system merges the content, even if it is of interest. As an example, duplicate videos are not recommended, nor are videos followed by a single creator.