Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
We may never fully understand how Tiktok’s “For You” algorithm works or why it sends users to sometimes very specific rabbit holes. But the company does it a little easier To control which subjects are likely (or not) in the recommendations.
The company deploys a parameter to “manage subjects” in the application to all users in order to add more personalization to user flows. The framework, which the company began to offer the United States last year, is a set of cursors that allow people to specify the frequency to which they want types of videos to appear in their recommendations. The cursors cover a dozen general subjects, such as nature; food and drinks; Fashion and beauty; and pets. For each interest, you can drag the cursor to “see more” or “see less”.
Tiktok notes that these integrated controls “will not eliminate any subject entirely”, so the sliders will not prevent any sporting content from appearing in your flow. But society affirms that the framework “can influence the frequency to which they are recommended as your interests evolve over time”.
For those who want more granular control on their flows – at least when it comes to blocking specific content types – Tiktok also extends its keyword filtering option, which allows people to choose specific keywords that they do not want to see. Now the application will use AI to make these filters smarter by adding additional keywords for similar content. For example, if you choose to filter the “reshaping”, the application can also choose to add “renovations” to your filtered words. The application can also automatically block alternative subjects of subjects, such as “CAT” or “C @ t”. The company says it expects these “smart filters” “to become more and more precise because more and more people use it over time” and that it will ultimately allow users to add up to 200 keywords.
The two features are available now in Tiktok’s parameters in the application preferences section of the application.
