Skip to main content
Back to module
Lesson 2 of 3

How TikTok Knows You

The Recommendation Engine Explained

The Secret Formula

STEP 1: COLLECT

  • What you watch (and for how long)
  • What you skip after 2 seconds
  • What you like, share, comment
  • What you search for
  • Your location, time of day, device
  • What "similar people" watch

STEP 2: ANALYZE

  • "You watch 80% of cooking videos"
  • "You skip political content in 3 seconds"
  • "People like you love tech tutorials"
  • "You watch more at 9 PM"

STEP 3: PREDICT

  • "There's an 85% chance you'll watch this"
  • Rank all possible videos by probability
  • Show the highest-ranked ones first

STEP 4: LEARN

  • Did you click? → Update prediction model
  • Did you watch full video? → Strong positive signal
  • Did you close app? → Maybe showed wrong content

⚠️ The Filter Bubble Problem

What it is: AI shows you more of what you already like, creating an echo chamber.

Example:

  • You watch one video about "AI replacing jobs"
  • Algorithm shows you 10 more scary AI videos
  • You never see "AI creating new jobs" content
  • Result: Your view becomes one-sided

How to Break Your Bubble:

1. Actively search for opposing views

2. Use "Not Interested" button on biased content

3. Follow diverse creators

4. Clear your watch history periodically

Exercise10 points0 attempts

What is the 'Filter Bubble' problem in recommendation algorithms?

Exercise10 points0 attempts

Which of the following is NOT data that recommendation algorithms typically collect?