Rule Breaker Slides Lesson 1
THE RULE
BREAKER
Coding vs. Machine Learning: How do machines actually "learn"?
The Big Question
Can a computer tell the difference between a taco and a hot dog without being told the rules?
Think about it...
How would YOU explain the difference to someone who has never seen food?
🌮
🌭
The Tech Tug-of-War
Coding
Explicit instructions. We tell the computer exactly what to do step-by-step.
Rule Example:
"If it has a bun and a sausage, it is a hot dog."
Machine Learning
Learning from examples. We show the computer thousands of pictures and let it find the patterns.
ML Example:
"Here are 1,000 photos of tacos. Figure out what makes them tacos."
Who's Who?
The Calculator
Pure Coding. It follows math rules perfectly. It never "learns" a faster way to add 2+2.
The Toddler
Pure Machine Learning. They learn what a "dog" is by seeing many dogs, not by reading a manual on biology.
Mission: Sorting
Are you a Coder or a Machine Learner?
Team CODE
You must write 5 strict rules for how to identify a "Scrap Metal" object from a pile of trash. No guessing!
Team ML
You get to look at 20 pictures of "Scrap Metal" for 30 seconds. Then you must sort the pile by "feel" and pattern.
Sorting Strategy Worksheet Sorting Strategy
Lesson 1: Coding vs. Machine Learning
Name:
Date:
Today we are testing two different ways to solve a problem: Coding (writing strict rules) and Machine Learning (recognizing patterns from examples).
Part 1: The Coder Approach
Imagine you have to tell a robot exactly how to identify a Taco . You can't show it a picture; you can only give it written rules. Write 3 strict rules below:
Challenge: What if the taco is in a wrapper? What if it's a "taco salad"? Do your rules still work?
Part 2: The Machine Learner
In the sorting game, you saw many examples of "Scrap Metal" very quickly. Describe the patterns or features your brain noticed without even thinking about specific rules:
One thing that made it EASY:
One thing that was TRICKY:
Brain Check
Which method do you think is better for a computer to recognize a blurry photo of a cat? Why?
Rule Breaker Teacher Guide Teacher Guide
Lesson 1: The Rule Breaker
Duration
Learning Objective
Students will be able to distinguish between traditional computer programming (coding) and machine learning (ML) by comparing explicit rule-following to pattern recognition.
Materials Needed
Rule Breaker Slides
Sorting Strategy Worksheet
A "Mystery Pile" of items
20 labeled photo cards
Key Vocabulary
Coding:
Specific, step-by-step instructions (rules).
Machine Learning:
A process where computers find patterns in data.
Features:
Specific traits or characteristics (shape, color).
Instructional Sequence
1
The Taco Hook (10 min)
Use Slides 1-2. Ask students: "If you had to teach a blind-folded alien to identify a taco using only words, what would you say?" Highlight how difficult it is to define "taconess" with just rules.
2
The Sorting Game (20 min)
Divide class into Team Code and Team ML . Give Team Code the task of writing 5 rules for "Scrap Metal." Give Team ML 30 seconds to look at 20 pictures of metal scraps. Then, have both teams try to sort a physical pile of random objects. Team Code must justify every move based on their written rules.
3
Worksheet & Reflection (15 min)
Hand out the Sorting Strategy Worksheet . Facilitate a discussion on why rules are good for math (Calculators) but examples are good for complex tasks (Toddlers/ML).
Facilitation Tips
Watch Out For:
"Team Code often gets frustrated because their rules are too specific (e.g., 'must be silver'). When they see a rusty brown piece of scrap metal, they can't pick it up. Use this as a teaching moment! Traditional code is rigid."
Guiding Questions
"Why did Team ML finish faster than Team Code?"
"What happens to Team Code if the object is upside down or partially hidden?"
"Is a calculator 'smart' or just 'obedient'?"
Differentiation
Support (Scaffolding)
Provide a "Rule Starter" list for Team Code (e.g., "If color is ____ AND shape is ____"). Allow Team ML more time to view the example photos.
Challenge (Extension)
Ask advanced students to think of a situation where Machine Learning would be worse than coding (e.g., a high-stakes surgery or a rocket launch calculation).
Data Buffet Slides Lesson 2
DATA
BUFFET
Feeding the computer the right information to help it "see."
Face-Off!
Can a computer tell if you are Happy or Surprised?
To do this, the computer needs a Dataset—a buffet of examples to learn from.
😊
😲
😊
😲
The 3 Steps of Training
1. Collect
Gather samples (images) for each category.
2. Train
The computer looks for patterns in those samples.
3. Test
Show the computer new images to see if it's right.
Garbage In, Garbage Out!
Bad Data
• Blurry photos
• Only 2 examples
• Messy backgrounds
• Confusing angles
Good Data
• Clear, bright photos
• 50+ examples
• Clean backgrounds
• Many different angles
Mission: Buffet
Time to feed the model!
Open Teachable Machine . We are going to train two categories:
Hand Sign A Hand Sign B
The Challenge
"Can you make your model 100% accurate? What happens if you stand closer or further away?"
Model Training Log Worksheet Model Training Log
Lesson 2: Data Buffet Training
Researcher:
Lab Station:
Mission Parameters
Category 1 Name:
e.g., Thumbs Up
Category 2 Name:
e.g., Peace Sign
The Buffet (Training Rounds)
Round Number of Samples Accuracy (0-100%) Notes (What went wrong?) 1
|
|
| 2 |
|
|
|
| 3 |
|
|
|
Breakthrough
What was the biggest change you made to improve your model's accuracy?
The Trash Factor
Did you find any "Garbage Data"? What was making the computer confused?
Data Buffet Teacher Guide Teacher Guide
Lesson 2: Data Buffet
Duration
Learning Objective
Students will understand the role of data in training AI models and demonstrate how the quality and quantity of a dataset affect a machine's accuracy.
Technical Requirements
Internet Access
Webcams (1 per 2 students)
teachablemachine.withgoogle.com
Model Training Log
Concepts
Dataset:
The collection of examples used to train a model.
Accuracy:
How often the computer correctly identifies an object.
G.I.G.O:
"Garbage In, Garbage Out." Bad data leads to bad results.
Instructional Sequence
1
Introduction to Training (10 min)
Use Data Buffet Slides. Explain that machines "eat" data to learn. Define G.I.G.O and show examples of clear vs. blurry photos.
2
The 10-Photo Trial (15 min)
Students go to Teachable Machine. Task them with capturing only 10 photos for two categories (e.g., Happy vs Sad). Have them test the model. Usually, the accuracy will be poor or jumpy. They record this as Round 1 on their Training Log.
3
The Buffet Feast (25 min)
Students now provide 100+ photos for each category, using different angles, lighting, and backgrounds. They observe the accuracy improvement and record Rounds 2 and 3.
Teacher Tips
Classroom Management
"Training requires silence/stillness during capture for clean data. If students are wearing masks or glasses, have them discuss how that might 'confuse' the model if they only train without them."
Discussion Starters
"Why did 100 photos work better than 10?"
"What happens if you use a photo of a cat to train the 'Happy' category?"
"If we train a computer in this room, will it work as well outside? Why or why not?"
Common Pitfalls
Background Bias
Students often realize the computer isn't recognizing their face, but rather the color of the wall behind them. If they move, the model fails. This is a perfect lead-in to Lesson 4 on bias.
Training Speed
If the training takes too long, check the number of images. 500+ is usually unnecessary for simple gestures and can slow down school Chromebooks.
Pattern Pixels Slides Lesson 3
PATTERN
PIXELS
Cracking the code of how computers "see" through neural networks.
Fastest Guesser
How can a computer guess you're drawing a bicycle before you even draw the wheels?
"It's all about the PIXELS and the PATTERNS."
"BICYCLE?"
Inside a Digital Brain
👁️
Input
The computer looks at the grid of pixels you drew.
Neural Network
💡
Output
The computer guesses based on similar drawings it has seen.
How it works: Layer by Layer
Layer 1
Detects simple edges and lines (is it curved or straight?).
Layer 2
Detects shapes (circles, triangles, squares).
Layer 3
Detects the object (It's a wheel! It's a bicycle!).
Mission: Guessing Game
Can you predict what the algorithm will see?
Pattern Battle
We are going to play Quick Draw as a class. Pay attention to exactly when the computer guesses correctly.
Is it the first stroke?
Is it the details?
Does it guess better if you draw fast?
Pixel Patterns Worksheet Pixel Patterns
Lesson 3: Thinking in Layers
Analyzer:
Grid Sector:
Part 1: Layered Vision
Computers don't see "objects." They see pixels. Your job is to act like a neural network layer.
Input Data (The Pixels):
Layer 1 (Edges)
How many "dark" squares are in a row at the top?
Layer 2 (Features)
Do you see two small shapes inside the big shape? What are they?
Layer 3 (Prediction)
Based on these patterns, what is this a drawing of?
Part 2: The Quick Draw Battle
When we played Quick Draw , did you notice anything about when the computer guessed? Fill in the blanks below:
A
The computer guessed correctly even before I finished drawing the because it saw a pattern in the first few pixels.
B
The computer got confused when I added too many to my drawing.
Real World Connection
Neural networks are used to read zip codes on mail. Why is it easier for a computer to read a zip code (numbers) than to read a whole handwritten letter?
Pattern Pixels Teacher Guide Teacher Guide
Lesson 3: Pattern Pixels
Duration
Learning Objective
Students will identify how computers use neural network layers to recognize patterns in data (pixels) and predict outcomes based on similarity.
Resources
Pattern Pixels Slides
Pixel Patterns Worksheet
quickdraw.withgoogle.com
Concepts
Neural Network:
A computer system modeled after the human brain that learns patterns.
Pixels:
The tiny "dots" of color that make up a digital image.
Hidden Layers:
Stages in the algorithm that detect features like edges or shapes.
Instructional Sequence
1
The Guessing Game (15 min)
Display Quick Draw on the board. Play a few rounds as a class. Challenge students to shout out the exact moment the computer correctly guesses. Ask: "What did the computer see first? Was it a circle or a straight line?"
2
Layers of Logic (15 min)
Use the Pattern Pixels Slides to explain neural network layers. Use the analogy of looking through three different pairs of glasses: 1. Edge-Vision, 2. Shape-Vision, 3. Object-Vision.
3
Pixel Patterns Worksheet (20 min)
Hand out the worksheet. Students must identify the patterns in the grid. Note: The grid contains a "smiley face" pattern. They must "decode" it like a computer would.
Teacher Tips
Deepen the Discussion
"Remind students that Quick Draw has seen millions of drawings. It's not 'smart' in a human way; it's just very good at comparing what you draw to its massive library of drawings."
Guiding Questions
"If you draw a dog that looks like a cat, what will the computer guess?"
"Why does the computer struggle with drawings that are very small or in the corner of the screen?"
"Can an algorithm 'know' what it's looking at, or is it just matching pixels?"
Differentiation
Support
Provide students with a physical "Feature Finder" (a piece of paper with a small square hole cut out) to look at the grid one square at a time, just like a computer's scan.
Challenge
Have students create their own 8x8 pixel art pattern and challenge a partner to "Layer Decode" it without seeing the final image first.
Fairness Filter Slides Lesson 4
Fairness
Filter
What happens when machines learn from an unfair world?
Is a doctor... glasses?
If we only show a robot photos of doctors who wear glasses...
Will it recognize a doctor who doesn't wear them?
👨⚕️👓
"YES! DOCTOR!"
👩⚕️
"ERROR: UNKNOWN"
The Definition of Bias
Algorithmic Bias: When a computer system makes unfair decisions because the data it learned from was incomplete or one-sided.
The computer isn't "mean" or "evil"—it's just limited by what we fed it.
Why It Matters
Facial Recognition
If software is only trained on people with light skin, it might not work for people with dark skin.
Recruitment
If an AI only sees resumes of men, it might start "hiding" job ads from women.
Mission: Fruit Sorter
Can we make an UNFAIR robot?
"We are going to train a fruit sorter that only knows Green Apples. Then we'll show it a Red Apple. What do you think will happen?"
🍏
🍎
Bias Detection Worksheet Bias Detection Lab
Lesson 4: Finding the Fairness Flaw
Ethical Auditor:
Model ID:
Experiment: The Unfair Fruit Sorter
Goal: Intentionally train a biased model to see what happens when data is limited.
Category 1: APPLE
ONLY use photos of Green Apples for this category.
🍏🍏🍏
Category 2: NOT APPLE
Use photos of Bananas or Oranges .
🍌🍊🍌
Test Phase: Show the computer a Red Apple
What you showed it The Computer's Guess Confidence (%) Is it Correct? Red Apple 🍎 ________________ ________________ ________________
The Root Cause
Why did the computer fail to recognize the red apple as an "Apple"? What feature was it actually looking for?
The Fairness Fix
How would you fix this dataset so the fruit sorter is fair to all apples?
Real World Reflection: Imagine a digital voice assistant that was only trained to listen to adults. What would happen when a child tried to use it? Is that fair?
Fairness Filter Teacher Guide Teacher Guide
Lesson 4: Fairness Filter
Duration
Learning Objective
Students will identify examples of algorithmic bias and explain how inclusive, diverse datasets are essential for creating fair and ethical AI systems.
Materials
Fairness Filter Slides
Bias Detection Worksheet
Variety of Fruit (Physical or images)
Concepts
Algorithmic Bias:
Systemic errors in AI that lead to unfair results.
Inclusion:
Ensuring all types of data are represented.
Ethics:
Making choices that are right, fair, and safe.
Instructional Sequence
1
The "Glasses" Hook (10 min)
Use Slides 1-2. Discussion: "Is a computer mean if it doesn't recognize a doctor without glasses?" Lead them to understand it's a data limitation, not a personality trait. This helps students focus on the system rather than the machine .
2
The Biased Lab (30 min)
Students use Teachable Machine. They intentionally train Category 1 with only Green Apples. When they test with a Red Apple, the computer will likely label it "Not Apple" (Category 2). Have them record this failure on the Bias Detection Worksheet.
3
Ethical Audit (20 min)
Class discussion on real-world impacts. Discuss facial recognition and hiring software examples from the slides. Ask: "How do we fix these problems as the creators of technology?"
Teacher Tips
Sensitive Topics
"Algorithmic bias can touch on race, gender, and disability. Keep the conversation focused on data representation . Emphasize that engineers have a responsibility to test their tools with everyone in mind."
Guiding Questions
"If we wanted to make a fruit sorter for the whole world, what fruits would we need?"
"Why might it be hard to collect data from every single type of person?"
"Who should be in the room when an AI is being designed?"
Differentiation
Support
Instead of webcam training, provide students with sets of physical cards (e.g., 10 photos of red birds, 1 photo of a blue bird) to demonstrate how the "minority" data gets ignored.
Challenge
Ask students to draft a "Code of Ethics" for the Brainy Bot Lab. What are 3 rules every AI engineer must follow to ensure fairness?
AI Hunt Slides Lesson 5
AI
HUNT
You probably used AI three times before breakfast. Can you find it?
Invisible Magic?
AI doesn't always look like a shiny robot. Most of the time, it's hidden inside your gadgets.
"It's the algorithm making decisions for you."
Voice Assistants
Movie Suggestions
Face ID
Traffic Maps
The "Because You Watched..." Effect
How it works:
AI looks at what you liked in the past.
AI compares you to millions of other users.
AI "guesses" what will keep you watching.
"Is it reading your mind? No. It's reading your DATA!"
The Newest Frontier: Generative AI
Text
Chatbots that write stories and answer questions.
Images
Tools that create art from a simple description.
Sound
AI that composes songs or mimics voices.
Mission: The Digital Hunt
Map the AI in your life.
Search Areas
📍 At Home (Kitchen, Living Room)
📍 At School (Classroom, Library)
📍 In Your Pocket (Phone, Watch)
"Can you find at least 5 AI 'neighbors' living in your world?"
AI Neighbor Map Worksheet AI Neighbor Map
Lesson 5: Real-World Scavenger Hunt
Explorer:
Mission Status:
The Hunt Log
Look for AI "neighbors" in your daily life. An AI neighbor is any technology that makes a prediction or recommendation based on data.
1
What did you find?
How is it using AI?
2
3
4
AI Choice
Pick one item from your list. What data is it "eating" to make its prediction? (e.g., your voice, your watch history, your face).
Lab Challenge
If you could invent one "AI Assistant" to help you at school, what would it do? How would it learn?
Brainy Bot Lab Certification
"I understand that AI is a tool created by humans, powered by data, and improved through fair patterns."
Signed
AI Hunt Teacher Guide Teacher Guide
Lesson 5: AI Hunt
Duration
Learning Objective
Students will identify common AI applications in their daily lives and explain how these technologies use data to make predictions or recommendations.
Resources
AI Hunt Slides
AI Neighbor Map Worksheet
Chromebooks/Tablets
Themes
Ubiquity:
AI is everywhere, often hidden in software.
Personalization:
AI uses your data to make choices just for you.
Empowerment:
Knowing how AI works helps you use it better.
Instructional Sequence
1
The Invisible AI (10 min)
Use AI Hunt Slides. Explain that most AI doesn't look like a robot. Discuss recommendation engines (Netflix/YouTube) and voice assistants. Use the "mind reading" vs "data reading" analogy.
2
The Digital Hunt (25 min)
Hand out AI Neighbor Map . Students can search the classroom/school or reflect on their homes. Challenge them to find 5 "neighbors." They must explain why they think it's AI (e.g., "It guessed what I wanted to type next").
3
Lab Graduation (15 min)
Unit wrap-up. Have students share their "AI Assistant" inventions. Sign their certifications on the worksheet to celebrate their new status as AI-literate explorers.
Unit Summary
The Big Idea
"By the end of this unit, students should move from seeing AI as 'magic' to seeing it as a predictable system that follows patterns learned from data. They should be more critical of the data they see and use."
Exit Ticket Questions
"Can AI learn something without data?"
"Why is it important to have diverse data in an AI system?"
"What is the difference between coding and machine learning?"
Future Lab Projects
Ethics Extension
Research one news story about AI bias. How did the company try to fix it? Did it work?
Creation Extension
Use a tool like Scratch to build a simple 'Rule-Based' game and compare it to the 'Machine Learning' models they built in Lesson 2.
