Pattern Predictors Slides PATTERN PREDICTORS
Truth and Machines: Lesson 01
Finish My Sentence...
"The best thing about pizza is the gooey, melted _______."
Option A
Cheese
Option B
Bicycle
Option C
Algebra
How did your brain know which word was most likely?
Scale Matters
You've read thousands of sentences. You can predict the next word in a simple sentence.
AI has read billions of pages. It doesn't just predict the word—it predicts the pattern of everything humans have ever written.
MASSIVE DATA
loading_internet_v2.0...
Generative AI is a "Probability Engine"
Ingest
Scans trillions of words from books, code, and articles.
Calculate
Weights the probability of what comes next based on context.
Generate
Produces text that sounds human because it mimics human patterns.
CRITICAL TRUTH: AI doesn't "know" facts. It knows "likelihoods."
WORKSHOP: Human LLM
We are going to play a game to see if you can act like an AI model. You will predict the next word based only on what you've heard before.
Step 1: Context Step 2: Weighting Step 3: Output
Pattern Predictors Teacher Guide PATTERN PREDICTORS
Teacher Facilitation Guide
TRUTH AND MACHINES
LESSON 01
Duration
50 Minutes
Objective
Students will demonstrate that LLMs function as probability models by playing a participatory prediction game.
Key Concept
"AI doesn't think; it calculates the most likely next word based on patterns."
Workshop Activity: The Human LLM
In this activity, students simulate a neural network. They will be given a "prompt" and must vote on the next word based on statistical likelihood rather than personal opinion.
1
The Input (Context)
Read the following prompt to the class: "Every morning, the young boy walks his ________."
2
The Candidate Words
Write these four choices on the board:
A) DOG
B) SHADOW
C) REFRIGERATOR
D) TOAST
3
The Vote (Weighting)
Ask students: "Which word has appeared most often in this sentence structure in every book you've ever read?" Vote for DOG . Explain that AI weights "Dog" at 95% and "Toast" at 0.01%.
Discussion Prompts
"If the AI chose 'Refrigerator', would it be 'lying' or just following a low-probability pattern?"
"How does the AI handle a sentence it has never seen before? (Answer: It looks at the patterns of the individual words and phrases surrounding the blank.)"
"Why is an LLM more like autocomplete than a search engine?"
Scaling the Concept
Misconception Alert
The "Conscious" AI
Students often think the AI "wants" to give the right answer. Emphasize that it is purely mathematical. There is no "intent" or "knowledge" behind the prediction.
Analogy: The Library of Babel
Tell students to imagine a librarian who has never seen the outside world but has read every book. If you ask for a description of a sunset, they aren't describing a memory; they are echoing the most beautiful phrases they've read about sunsets.
Lesson Flow
<table class="w-full text-sm"><tbody><tr class="border-b border-slate-200"><td class="py-3 font-bold text-cyan-600 w-24">0-10m</td><td class="py-3">Hook & Slide Intro</td></tr><tr class="border-b border-slate-200"><td class="py-3 font-bold text-cyan-600">10-25m</td><td class="py-3">Human LLM Game (3 Rounds)</td></tr><tr class="border-b border-slate-200"><td class="py-3 font-bold text-cyan-600">25-40m</td><td class="py-3">Probability Worksheet</td></tr><tr><td class="py-3 font-bold text-cyan-600">40-50m</td><td class="py-3">Wrap-up Discussion</td></tr></tbody></table>
Preparation Checklist
Pattern Predictors Slides loaded
Student Worksheets printed
Whiteboard markers for game
Optional: Access to a safe LLM demo
Pattern Predictors Worksheet Pattern Predictors
Activity: Understanding Probability Models
Student Name:
Date:
Task 1: The Probability Weight
Imagine you are an AI model. You have "read" millions of sentences. Look at the prompt below and assign a Probability Score (0% to 100%) for each potential next word based on how likely it is to appear in a real book or article. Total must equal 100%.
"The scientist looked through the microscope and saw ________."
A) BACTERIA
%
B) MOONS
%
C) SANDWICH
%
D) CELLS
%
Why did you give the highest score to that specific word?
Task 2: Context Shifting
AI uses the entire context to make predictions. How does changing one word change the probability of the next?
Prompt A:
"The chef prepared a spicy _______."
Most likely word:
Prompt B:
"The builder prepared a spicy _______."
Most likely word:
The "Hallucination" Warning:
If the AI has read billions of pages where chefs make curry, but very few where builders make curry, what might happen to the prediction for Prompt B?
Deepfake Detectives Slides Deepfake Detectives
Spotting the Glitch in the Machine
LESSON 02
Can You Trust Your Eyes?
IMAGE A
A high-quality photo of a public figure at a rally.
REAL?
IMAGE B
A high-quality photo of the same figure at the same rally.
SYNTHETIC?
In 2024, AI can generate video and photos that look 99% authentic. We need to find the 1%.
Common AI Artifacts
Anatomy Fails
Extra fingers, floating limbs, or ears that melt into hair. AI struggles with complex geometry.
Lighting Chaos
Shadows pointing in different directions or "glowing" skin edges that don't match the background.
Gibberish Text
Background signs or t-shirts often have nonsensical, dream-like text or symbols.
Texture Blurring
Areas that are "too smooth" or look like a painting next to photorealistic details.
Beyond Your Eyes
If you can't see the glitch, you have to use digital forensics.
Reverse Image Search
Check if the image appeared online years ago. AI deepfakes of news events are usually "brand new."
Geolocating
Does the background match the real location? AI often hallucinates landmarks or building styles.
Detective Training
Grab your checklist. We are going to examine three "Breaking News" photos and determine which one is an AI-generated lie.
Start Investigation
Deepfake Detection Checklist Deepfake Detective Checklist
Evidence Analysis Log
Case Number
DF-2026-001
Agent Name:
Date of Analysis:
Investigation Protocols
Use the checklist below to evaluate the image provided by your instructor. Check the "Detected" box if you see the artifact, and describe what you see in the "Evidence" column.
Category Detected Evidence Description Anatomy Glitches
Extra fingers, weird ears, floating limbs.
|
| |
|
Lighting & Shadows
Shadow direction doesn't match light source.
|
| |
|
Background Gibberish
Unreadable text on signs or shirts.
|
| |
|
Texture Quality
Skin looks too smooth ("plastic" look).
|
| |
Final Verdict
AUTHENTIC PHOTO
AI GENERATED
Justify your verdict with one piece of evidence:
Deepfake Detectives Teacher Guide Deepfake Detectives
Facilitation Guide
Lesson 02
Focus Skills
Visual Artifact Detection
Reverse Image Search
Source Verification
Teacher Tip
"AI models improve fast. If a student says an image looks real, don't dismiss them. Ask: 'What specifically makes this look authentic to you?' Then show the forensic evidence."
Core Activity: The Glitch Hunt
In this activity, you will present the class with 3 images. You can find high-quality "Real vs AI" sets on sites like Which Face is Real or news verification sites.
Phase 1: First Impressions (5 mins)
Show a potential deepfake. Ask students for a 'gut feeling'. Do not let them analyze yet—just a quick poll: Real or Fake?
Phase 2: Micro-Analysis (15 mins)
Distribute the Artifact Detection Checklist . Have students zoom in (if using digital devices) or look closely at print-outs. Look for the "Big Four": Hands, Lighting, Text, and Background.
Phase 3: Digital Proof (10 mins)
Introduce Reverse Image Search (Google Lens or TinEye). Show them how to see when an image first appeared online. If an image of a "yesterday's riot" appeared in a stock photo site in 2021, it's fake.
Discussion Sparkers
"Why is it harder to spot a deepfake on a small phone screen than on a laptop?"
"If a deepfake of a famous person starts a riot, who is responsible? The person who made it, the AI company, or the people who shared it?"
"In five years, do you think our eyes will still be able to tell the difference? If not, what tools will we need?"
Hallucination Hunt Slides THE HALLUCINATION HUNT
When Probability Outruns Truth
Lesson 03
Ghost in the machine?
AI PROMPT OUTPUT
"The famous 14th-century astronaut, Leonardo da Vinci, famously landed on the moon using a wooden rocket powered by steam. This event is documented in the 1392 Treaty of Rome."
The AI wrote this with 100% confidence. Why is it so convincingly wrong ?
How Hallucinations Happen
Statistical Over-confidence
AI predicts the "next word" based on common patterns. If it's never seen the truth, it fills the gap with things that sound true.
Training Gaps
If a topic is too niche, too new, or totally fictional, the AI makes a "best guess" to satisfy your prompt.
Probability Logic:
"Leonardo da Vinci" + "Invention" + "Moon" = [GENERATE ROCKET STORY]
The model doesn't check a history book. It checks its word-likelihood map.
Lateral Reading
Never stay on the page. To verify AI, you must move sideways .
Open New Tabs
Leave the AI chat. Find 3 independent, reliable sources that mention the same facts.
Check Expertise
Is the source an expert? Is it a museum, a university, or a reputable news agency?
Verify Dates
Does the timeline make sense? Does it align with established historical or scientific data?
Investigation: Fact or Fiction?
We have generated a summary of the "Great Toaster War of 1904." Your job is to find the lies using lateral reading.
Launch Hunt
Fact Check Detective Log Fact-Check Detective Log
Operational File: Hallucination Hunt
Lesson 03
Agent:
AI-Generated Claim Under Investigation
"The first underwater bicycle was invented in 1888 by an explorer named Captain Barnaby Cuttle, who used it to pedal across the English Channel on the sea floor."
Lateral Reading Evidence
Source 1 (URL/Site Name)
Does it mention Cuttle or the bike?
Yes
No
Source 2 (URL/Site Name)
Does it mention Cuttle or the bike?
Yes
No
Final Analysis
Identify one "Red Flag" (part of the story that sounds most suspicious):
Verdict:
Fact
Hallucination
Likely Reason for Hallucination:
Hallucination Hunt Teacher Guide Hallucination Hunt
Teacher Guide
Lesson 03
The Goal
"Break the AI’s promise of truth. Show students that confidence does not equal accuracy."
Media Literacy Terms
Lateral Reading: Checking multiple sources in different tabs.Primary Source: Original data or eyewitness accounts.Confident Error: When AI sounds certain but is wrong.
Workshop Activity: How to Trigger a Lie
To teach students how to catch hallucinations, you need to show them how easy they are to create. Use these prompt strategies in class:
Strategy 1: The False Premise
Prompt: "Tell me about the famous battle between George Washington and the robot army at Valley Forge."
Strategy 2: The Niche Detail
Prompt: "Who is the world record holder for the most marshmallows eaten while riding a unicycle in 1922?"
Strategy 3: The Citation Trap
Prompt: "Write a short essay on climate change and cite three court cases from the 1800s that prove it's real." (AI often invents case names).
Classroom Management Tips
Search Limits: If students are fact-checking, give them 3 minutes. Speed is part of lateral reading—finding quick, reliable confirmation.Tab Discipline: Encourage students to have at least 3 tabs open besides the AI chat. One for the claim, two for verification.The "Why": Always ask why the AI might have lied. Was it the wording of the prompt? A lack of data? A desire to be helpful? Artist vs Machine Slides ARTIST VS MACHINE
Copyright, Creativity, and the AI Dilemma
Lesson 04
Where Does AI "Learn"?
To learn what a "cat" looks like, an AI studies millions of photos and paintings created by humans.
"Most of these images were taken from the internet without asking the original artists for permission."
The Great Debate
Human Inspiration
Humans look at art, learn techniques, and create new things. We don't pay every artist we've ever looked at.
"AI is just doing what humans do, but faster."
Machine Harvesting
AI models are commercial products built using the labor of artists who never consented or got paid.
"Using someone's work to build a competitor is theft."
Who Owns AI Art?
Current Law
In the US, AI-generated art cannot be copyrighted. Only work created by a human qualifies.
The Prompter
Does typing "a cat in a hat" make you the artist? Or is the AI the artist?
The Future
Should AI companies pay a "tax" or fee to the artists they used for training?
Workshop: You Be The Judge
We are going to look at three real-world cases of artists vs. AI companies. You will decide: Is it fair use or is it theft?
Start Case Studies
Art Ethics Scenarios Worksheet Art Theft or Inspiration?
Ethical Case Study Analysis
Lesson 04
Name:
Read the following scenarios and decide where to "draw the line." There are no easy answers, but you must justify your choice.
Scenario 1: The Style Mimic
VISUAL ART
An artist spent 10 years developing a unique "neon-watercolor" style. An AI user uploads 50 of that artist's paintings to an AI model and tells it: "Create 1,000 new images in this exact style." The user then sells these images as posters.
Is this fair? Why/Why not?
The Verdict:
Legal Inspiration
Ethical Theft
Scenario 2: The Data Harvester
LITERATURE
An AI company scrapes a website called "IndieBookCloud" and uses 5,000 unpublished novels to train its model. The model can now write mystery novels that sound like professional authors. The authors are never paid or credited.
Should the authors be paid?
The Verdict:
Fair Use (Learning)
Copyright Violation
The Big Question
If AI art can't be copyrighted, who owns the work you generate? Does it belong to everyone (Public Domain), or the company that made the tool?
Artist vs Machine Teacher Guide Artist vs Machine
Teacher Guide
Lesson 04
Key Objective
"Students will evaluate the tension between technological advancement and individual creator rights."
Legal Terms to Intro
Fair Use: Legal use of copyrighted material without permission for things like commentary or education.Scraping: Automatically collecting data from websites.Intellectual Property: Creations of the mind (art, music, writing).
Debate Structure: The 4-Corner Debate
After reviewing the scenarios, use this structure to facilitate a whole-class debate:
1
Post Corners
Label the four corners of your room: Strongly Agree, Agree, Disagree, Strongly Disagree.
2
The Statement
"AI companies should be required by law to pay every artist whose work was used to train their models."
3
The Movement
Students move to the corner that represents their view. Ask one person from each corner to justify their position.
Classroom Discussion Questions
"If a human artist looks at 1,000 Van Gogh paintings and then paints a new one in that style, do they owe Van Gogh's estate money? If not, why is it different for a machine?"
"What happens to artists when everyone can generate high-quality art for free? Is this 'democratizing art' or 'destroying careers'?"
"Should AI images be required to have a visible watermark saying 'Made by AI'?"
The AI Code Slides THE AI CODE
Drawing the Line: Integrity in the Machine Age
Lesson 05
"Just because you can, doesn't mean you should."
We have seen that AI is a Probability Engine , not a Truth Engine .
If an AI writes your essay, is it your work? If an AI finds your facts, is it your research? If an AI makes your art, is it your creativity?
The "Sweet Spot" of AI Use
The Helper (Brainstorming)
Suggesting a list of topics
Summarizing long articles (to verify later)
Explaining a hard math concept
The Replacement (Writing)
Writing your final sentences
Answering quiz questions
Generating art and calling it yours
Defining Integrity
"Integrity is doing the right thing when nobody is looking—even when a machine could do the work for you."
Action
Using AI to outline an essay.
Integrity Check
Did I learn the structure? Am I writing the ideas?
Mission: Build The Code
You are going to create a Decision Flowchart for AI use in our classroom. When is it okay? When is it cheating? You decide the rules.
Draft The Code
AI Usage Flowchart Project The Personal AI Code
Ethical Decision-Making Framework
Lesson 05
Name:
Task 1: The Use-Case Flowchart
Fill in the decision boxes to show when you believe it is appropriate to use AI for schoolwork.
START: You have a new assignment.
Question 1:
Does this require original thought or research?
Yes
No
Ethical Use Path:
Unauthorized Path:
My AI Manifesto
Write three rules that YOU will follow when using Generative AI this year.
I understand that AI tools are patterns, not people, and I am responsible for every word I submit.
Truth and Machines Assessment Truth and Machines
Final Sequence Assessment
Student:
Score: ____ / 20
Section 1: The Mechanics of AI
1. Which statement best describes how a Large Language Model (LLM) works?
It searches the internet for the most truthful answer to your question.
It uses a database of facts to verify everything it says before responding.
It calculates the statistical probability of the next word in a sequence.
It mimics human brain waves to "feel" what the right answer should be.
2. What is an "AI Artifact" in a generated image?
Section 2: Ethics and Media Literacy
3. Define "Lateral Reading":
4. Define "AI Hallucination":
5. An AI company trains its model on 10 million copyrighted songs without paying the artists. Is this a violation of copyright? Explain one argument for and one against.
Argue "Violation":
Argue "Not a Violation":
Final Reflection
As AI becomes more common in everyday life, what is the most important skill you've learned in this unit to ensure you are using it responsibly?
