TikTok Knows You Better Than You Know Yourself

Summary by Krater

An explainer on how TikTok's algorithm uses watch time and behavioral data to track user interests and curate content recommendations through collaborative filtering.

From the video

Answers: How does TikTok's algorithm know what you like so well?

• TikTok algorithm
• collaborative filtering
• Skinner box and variable rewards
• watch time tracking
• user behavior analysis
• psychology of scrolling

What it concludes

• TikTok does not rely on active signals like likes, follows, or comments, but tracks watch time down to fractions of a second.
• Collaborative filtering matches your behavioral patterns with thousands of other users to predict what you will watch next.
• Uncertain and unpredictable rewards, similar to Skinner box experiments, make scrolling compulsive.
• The algorithm spaces out high-interest videos with mediocre ones to keep users engaged.
• The lack of cohesive narrative in a fast-paced feed prevents the brain from consolidating memories, making time spent on the app seem to vanish.