← BackGrades 5-6 | Chapter 2: AI in Action

AI Recommendations 🎬

How YouTube, Netflix, TikTok Know What You Like

Session 4: Recommendation Algorithms

The Recommendation Problem 🤔

Netflix has 10,000 hours of content. YouTube has 500 hours uploaded EVERY SECOND. How do these platforms know what YOU want to watch?

The Solution: AI Recommendations

AI builds a profile of your interests based on:

  • What videos you watch
  • How long you watch each video
  • What you like/dislike (thumbs up/down)
  • What you skip
  • What you search for
  • When you watch (time of day, season, holidays)

Collaborative Filtering 👥

Netflix uses a technique called collaborative filtering - finding people similar to you and recommending what THEY liked.

How It Works:

  1. You watch "Breaking Bad" and like it
  2. AI finds 10,000 other users with similar taste
  3. AI checks what those users also liked
  4. Recommends those shows to you

Example: If 95% of people who liked "The Office" also liked "Parks and Recreation", the algorithm will recommend it to you!

Your Digital Profile 📊

Every platform creates a detailed user profile based on your behavior:

What's In Your Profile?

  • Your watch history (every video you've clicked)
  • Time spent on each video
  • Videos you paused, rewound, or skipped
  • Likes and comments
  • Shares and saves
  • Search queries
  • Device type and internet speed
  • Location and time zone

The Filter Bubble Problem ⚠️

Here's the dark side: recommendation algorithms can trap you in a filter bubble.

How Filter Bubbles Work:

  • Algorithm shows you more of what you like
  • You only see content matching your interests
  • You never encounter opposing viewpoints
  • Your beliefs get reinforced, not challenged
  • Over time, your views become more extreme

Example: If you watch one conspiracy theory video, YouTube's algorithm might recommend increasingly extreme videos, trapping you in a rabbit hole.

Algorithm Addiction 🎮

Tech companies use AI to keep you engaged as long as possible. This is called engagement maximization.

Tactics Used:

  • Auto-play next video (without asking)
  • Ending videos with cliffhangers
  • Showing notifications about new videos
  • A/B testing (trying different thumbnails to see what gets more clicks)
  • Infinite scroll (never reaching the end)

Why? More time watching = more ads shown = more money made!

Critical Thinking Questions 🤔

Consider This:

  • Is personalization good or bad? It's convenient, but limits your world view
  • Who benefits? You get content you like. Companies make money from ads
  • The cost: Your data. Algorithms track everything you do
  • Solution: Be intentional. Seek diverse content. Don't just follow recommendations

What We Learned 🎓

  • Recommendation algorithms use your behavior to predict preferences
  • Collaborative filtering finds similar users and recommends their favorites
  • Platforms create detailed user profiles to predict preferences
  • Filter bubbles trap you in one viewpoint
  • Algorithms are optimized for engagement, which means keeping you watching
  • Be aware and seek diverse content intentionally

Understand Recommendations! 🎉

AI predictions are powerful but come with hidden costs

Next Session: How AI Sees Images