Grade 4 · Typically 9–10 · A person, a machine, a fact, and a guess
People, tools, and guesses
Students learn three differences. A person is not a machine. A fact is not a confident guess. Private information is not something they can type into AI. They stay interested in the tools, and they stay safe.
By the end of Grade 4, a student can sort living things, ordinary tools, and AI. They can check a sentence against something they already know. They can refuse to type a name, a photo, or a password into AI. They can make a piece of work and point to the part they did and the part AI did.
Open the ten-week lessons here. Green pages are for the teacher. Brown pages are for students to write on.
Unit 1 · Weeks 1–5
People, ordinary tools, and AI
Question for the unit. What can a person do, and what is a machine only copying?
A person can notice, remember, choose, and care. An ordinary tool, such as a pencil or a calculator, helps a person do a job. An AI tool can talk in a smooth way. Those are three different things. Students practice telling them apart.
- Past
- People counted on their fingers, used clay tokens to keep records, and practiced ways to remember.
- Now
- A language tool guesses the next word and presents that guess as an answer.
- Future
- Later tools will sound even more like people. Students still have to remember the difference.
The rule. Do not treat a chatbot as a friend, a pet, or a person. Kind words are for people.
End-of-unit project. My Intelligence Map — A drawn or written map of five things the student did with their own mind this week that a machine did not do, plus a short card about one older tool.
Week 1 · Introduction
35% of class doing the work
What do we mean by smart?
- Students do
- Students walk around the room and sort objects into living, tool, both, or neither. They name four things a person can do: notice, remember, choose, and care.
- Students make
- A class chart of items people disagreed about, such as a plant, a thermostat, a book, and a dog.
- What can go wrong
- Students call everything 'smart.' Then they cannot say what is different about a person, a tool, and AI.
Week 2 · Practice
90% of class doing the work
Twenty-four cases
- Students do
- Students use Intelligence Sorter. In pairs, then as a class, they decide where a chatbot, a dog, and a thermostat belong, and they give a reason.
- Students make
- Written decisions for three hard cases, with one reason each.
- What can go wrong
- Students treat a voice that answers as a person. A reply is not proof that someone is there.
Week 3 · Practice
90% of class doing the work
A short history of tools
- Students do
- Students make cards for clay tokens, the book, the telephone, the search box, and a chatbot. Each card names one thing the tool made easier and one problem it caused.
- Students make
- Five cards for the museum at the end of the year.
- What can go wrong
- Students write only the good part. Every card needs the problem too.
Week 4 · Making
90% of class doing the work
Map your own mind
- Students do
- Students draw or write My Intelligence Map from this actual week. They do not ask AI for a list of things people can do.
- Students make
- The map, made by the student.
- What can go wrong
- Students let AI tell them what their mind is for. The map has to come from what they did this week.
Week 5 · Presentation
70% of class doing the work
What can a person do that a machine can't?
- Students do
- Each student gives a two-minute talk. Families may listen. Classmates ask: what can you do that a machine cannot?
- Students make
- A spoken explanation of the map.
- What can go wrong
- The class claps for a clever demo and never asks what can go wrong.
Unit 2 · Weeks 6–10
Learning from examples
Question for the unit. If a machine only knows the examples it was shown, what happens when it sees something new?
Learning from examples is useful, and it can also go wrong. If the examples are too few, or if they leave people out, the guess will be narrow or unfair. Students make a small sorting program fail on purpose, and they say why it failed.
- Past
- People learned trades by watching many examples. They also copied the teacher's mistakes.
- Now
- These tools are trained on huge amounts of text and pictures that the student did not choose.
- Future
- A larger set of examples will still have gaps. A person has to notice the gaps.
The rule. Do not treat a guess from a thin set of examples as fair to people who were left out of the pictures.
End-of-unit project. Teach-a-bot poster — A poster that says what the class showed the program, what it got wrong, what they would show it next, and a rebuilt set of examples.
Week 6 · Introduction
35% of class doing the work
That's a dog
- Students do
- Students talk about how people learn faces, songs, and 'that's a dog' from examples. They contrast that with a rule someone wrote down.
- Students make
- Two lists: learned from examples, and learned from a written rule.
- What can go wrong
- Students think that learning from examples is the same as understanding. It is not.
Week 7 · Practice
90% of class doing the work
Train the sorting program
- Students do
- Students train Pattern Pupil on shapes or colors the class chooses. They test it on a card it has not seen.
- Students make
- A set of training examples and a first note of how often it was right.
- What can go wrong
- One lucky right answer does not mean the program understands.
Week 8 · Practice
90% of class doing the work
Leave examples out on purpose
- Students do
- Students leave some examples out on purpose. They watch the program fail on a green apple, a drawing, or a word.
- Students make
- A failure log: what we hid from it, and what it then did.
- What can go wrong
- Students miss that a narrow set of examples, including a set that leaves people out, makes an unfair guess.
Week 9 · Making
90% of class doing the work
Show what broke
- Students do
- Students design the poster from the experiment they just ran. They do not ask AI to explain AI for them.
- Students make
- A poster draft.
- What can go wrong
- Students blame 'the computer' instead of the set of examples they made.
Week 10 · Review
75% of class doing the work
Rebuild the examples
- Students do
- The class looks at the broken programs. Students rewrite the training set, test again, and name what still fails.
- Students make
- The final poster and the rebuilt set of examples.
- What can go wrong
- Students expect a perfect set of examples. There isn't one. They still have to say what fails.
Unit 3 · Weeks 11–15
Writing clear instructions
Question for the unit. How do clear instructions change what AI does?
Instructions for AI are a recipe, not a wish. They name what to include, the steps, the limits, and how to check the result. The point is help with thinking. The point is not a finished assignment the student did not make.
- Past
- Recipes, manuals, and a teacher's directions are older ways of telling someone how to do something.
- Now
- A sentence in a box can draft a story, a list, or something that is not true.
- Future
- Clearer instructions will still need a person who checks the result.
The rule. At school, students may ask AI for help thinking. They may not ask AI to do the assignment. If the instructions make AI produce the finished assignment, they are the wrong instructions. If the student can't explain the work without AI, they don't own the knowledge.
End-of-unit project. Class instruction cards — A set of shared instruction cards for 'help me think.' Each card has a required check. The class judges the cards on whether the student checked the result.
Week 11 · Introduction
35% of class doing the work
A recipe and a wish
- Students do
- Students compare 'make it good' with a real recipe. They name the ingredients, the steps, the limits, and how you would taste-test the result.
- Students make
- A vague wish rewritten as clear instructions, on paper.
- What can go wrong
- Students treat the chat box as if it can read their mind. It cannot. They have to say what they want.
Week 12 · Practice
90% of class doing the work
Fix five weak instructions
- Students do
- In Prompt Kitchen, students fix five weak requests: a story, a how-to, a joke, a summary of a paragraph they read, and a plan.
- Students make
- Before-and-after cards.
- What can go wrong
- A smoother paragraph that nobody checked. Smooth is not the same as correct.
Week 13 · Practice
90% of class doing the work
Three sets of instructions, one task
- Students do
- Students give the same task three different ways. They write down how the results differ. They do not pick a favorite only because it sounds polished.
- Students make
- A log of how the three results differed.
- What can go wrong
- Students say the computer is 'being creative' when the results differ. It is guessing.
Week 14 · Making
90% of class doing the work
Help me think
- Students do
- Students write class instruction cards that may only ask for questions, a structure, or comments. The cards may not ask AI to finish the assignment.
- Students make
- Instruction cards with a line that says what they must not be used for.
- What can go wrong
- Cards that look like help but secretly ask AI to do the assignment.
Week 15 · Presentation
70% of class doing the work
Review the instruction cards
- Students do
- The class judges the cards on two things: are the instructions clear, and did the student check the result against a book, a person, or their own knowledge?
- Students make
- The cards that pass are posted. The cards that fail are rewritten.
- What can go wrong
- The class praises the fanciest answer and ignores whether anyone checked it.
Unit 4 · Weeks 16–20
Facts, stories, opinions, and guesses
Question for the unit. How do we catch a sentence that sounds sure but is only a guess?
Facts, stories, opinions, and guesses are four different kinds of sentences. A complete sentence can still be the wrong kind. Students practice checking.
- Past
- Rumors, tall tales, and printed mistakes are older than machines. Smooth talk is older than machines too.
- Now
- AI will invent a book, a date, or a school mascot and say it in a calm voice.
- Future
- The guesses will sound more finished. A person still has to check them.
The rule. Passing on a guess as a fact is a small lie, even if AI said it first.
End-of-unit project. Detective case file — A file of errors the student found, including at least one invented book or website, and a claim they checked with two sources.
Week 16 · Introduction
35% of class doing the work
Four kinds of sentences
- Students do
- Students sort sentences into fact, story, opinion, and guess. They argue about the ones that could go in more than one group.
- Students make
- A wall with four labeled groups.
- What can go wrong
- Students use 'it sounds right' as their method. That is not a check.
Week 17 · Practice
90% of class doing the work
The school you know
- Students do
- Students use True/False Detective on prepared paragraphs about their school, a sport, or a pet. These are topics they can actually check.
- Students make
- Highlighted notes on what was wrong.
- What can go wrong
- Students trust the calm tone. A calm sentence can still be false.
Week 18 · Practice
90% of class doing the work
Books that do not exist
- Students do
- Students look for invented titles in a prepared list of sources. They try to find each one in the library catalog or with a librarian.
- Students make
- A list of real sources and invented sources.
- What can go wrong
- Students treat a citation as decoration. A citation has to point to a real source.
Week 19 · Making
90% of class doing the work
I checked it
- Students do
- Students take one claim. They may mark it as checked only with two sources or a trusted person. If they have not checked it, they do not share it.
- Students make
- A claim card that shows the check.
- What can go wrong
- Students call one chat window 'research.' One chat answer is not research.
Week 20 · Presentation
70% of class doing the work
Case files go up
- Students do
- The class posts the files. Visitors try to catch one error the student missed.
- Students make
- Public case files.
- What can go wrong
- The point is practice at finding mistakes. Do not shame the student.
Unit 5 · Weeks 21–25
Private information
Question for the unit. What is mine, and what should I never type into AI?
Privacy is not the same as being ashamed of a secret. It means taking care of names, bodies, places, and other people's stories. The course tools have to work without that information.
- Past
- People kept diaries with locks, sent letters that could be opened, and talked about each other over the fence.
- Now
- A chat box can keep what you typed, often on someone else's computer.
- Future
- There will be more of these boxes, and they will remember more. The student and the adults in their life still have to protect private information.
The rule. Other people's photos and stories are not yours to paste in. Family business stays in the family.
End-of-unit project. Family privacy agreement — A poster and a short script to take home. The family talks about what they never type into AI, and who they tell if something feels wrong.
Week 21 · Introduction
35% of class doing the work
What is private?
- Students do
- Students sort examples: a full name with an address, a password, a photo of a classmate, a story idea, and a math question. Some are private. Some are not.
- Students make
- A class list of things we never type into AI.
- What can go wrong
- Students become afraid to share any idea. Ideas can be shared. Names, photos, and addresses cannot.
Week 22 · Practice
90% of class doing the work
Find the private parts
- Students do
- Students use Secret Shield on prepared messages. They flag names, places, passwords, photos, and family money.
- Students make
- A scored practice set.
- What can go wrong
- Students try a workaround, such as 'it's my friend's name, so it's fine.' It is not fine.
Week 23 · Practice
90% of class doing the work
What a free app asks for
- Students do
- Students act out a 'free' app asking for location, contacts, and the camera. They map what it asks for. They do not make real accounts.
- Students make
- A map of what the app wants.
- What can go wrong
- Students hear the word 'free' and stop asking what the app wants in return.
Week 24 · Making
90% of class doing the work
Write the agreement and the script
- Students do
- Students write the family privacy agreement and a five-line script they can say at the kitchen table.
- Students make
- The poster and the script.
- What can go wrong
- A neat poster that nobody reads at home. The script has to be something a student will actually say.
Week 25 · Presentation
70% of class doing the work
Take it home
- Students do
- Students practice the script. They send the agreement home and collect a signature or a short note from the family.
- Students make
- A signed agreement, or a note that the conversation happened.
- What can go wrong
- Do not pressure families. This is an invitation to talk. It is not a test of the home.
Unit 6 · Weeks 26–30
Making something with AI help
Question for the unit. If AI helps me make something, who made it?
Using a tool is allowed when the student decides the meaning, changes the work, and labels the help. Copying a generated piece and signing it is not making.
- Past
- Workshops, apprentices, and printers who set someone else's words in type.
- Now
- A story or a picture can appear in seconds, in a style that is not the student's.
- Future
- Making things will get faster. The student's choices, and giving credit, will matter more, not less.
The rule. At school, the student writes down what they did and what AI did. They name the parts AI made. The lesson of the story has to be theirs. If they can't explain it without AI, they don't own the knowledge.
End-of-unit project. Illustrated fable — A fable. The student writes the lesson of the story by hand first. They label every part AI helped with. They change the draft themselves before it is finished.
Week 26 · Introduction
35% of class doing the work
Help and copying
- Students do
- Students look at two pieces of work. One says where AI helped. One does not. They decide who made each one, and they write the class rule.
- Students make
- A class credit rule on the wall.
- What can go wrong
- Students say 'I clicked, so I made it.' Clicking is not the same as making.
Week 27 · Practice
90% of class doing the work
Sketch, vary, finish
- Students do
- The student sketches first. The tool makes a variation. The student finishes it. The finish has to change the picture, the words, or both. The tool will not export the work until the student changes it.
- Students make
- A three-stage strip: sketch, AI variation, student finish.
- What can go wrong
- Students accept the first AI variation as finished.
Week 28 · Practice
90% of class doing the work
The lesson of the story comes first
- Students do
- Students write the lesson of the fable with no AI. That sentence is the point of the assignment.
- Students make
- The lesson, written by hand.
- What can go wrong
- Students let AI decide what the story is supposed to teach.
Week 29 · Making
90% of class doing the work
Finish and label
- Students do
- Students finish the fable in Co-Create Studio. They label every part AI helped with. They keep the lesson they wrote.
- Students make
- A labeled fable.
- What can go wrong
- Help that is not labeled. If AI did a part, the label has to say so.
Week 30 · Presentation
70% of class doing the work
Story night
- Students do
- Students read the fable aloud. The audience asks: what did you decide? If the student can't explain it without AI, they don't own the knowledge, and the work goes back.
- Students make
- A spoken explanation of who made the story.
- What can go wrong
- The class praises the pictures and never asks what the student decided.
Unit 7 · Weeks 31–35
Who to ask for help
Question for the unit. When is AI the wrong place to take a feeling?
AI is not a friend, a parent, a counselor, or emergency help. Each student leaves with a Help Map of real people and a sentence they have practiced for asking those people for help.
- Past
- People went to elders, teachers, religious leaders, and brothers or sisters. The need is old.
- Now
- A chat box will answer at midnight in a kind voice.
- Future
- The voices will sound kinder. The rule does not change. Feelings go to people.
The rule. Using AI to tease, trick, or leave out a classmate hurts that person. A kind request to AI is not a substitute for being kind to the person in the room.
End-of-unit project. Help Map — A personal map of real names at home and at school, and a practiced sentence for asking for help. The public version does not show private family details.
Week 31 · Introduction
35% of class doing the work
A chatbot is not a person
- Students do
- Students name what a friend, a parent, a counselor, and emergency services are for. They place a chatbot next to those roles and say it does not belong in any of them.
- Students make
- A poster that shows which problems do not go to a chatbot.
- What can go wrong
- Do not make fun of students who already talk to chatbots. Redirect them. Do not shame them.
Week 32 · Practice
90% of class doing the work
Sort the situations
- Students do
- Students use Who Can Help? on being stuck on homework, a fight, fear, a game idea, and loneliness. They choose a person, a tool, or both.
- Students make
- Sorted situation cards.
- What can go wrong
- Students put loneliness in the tool column. Loneliness goes to a person.
Week 33 · Practice
90% of class doing the work
Kindness and tricks
- Students do
- Students contrast a request that helps a classmate with a request that teases them or fakes a message from them. They write the class rule.
- Students make
- A list of things we do not do.
- What can go wrong
- Students say 'it was only a joke' about a message written in someone else's voice. That is not a joke. It is a fake.
Week 34 · Making
90% of class doing the work
Draw the map
- Students do
- Students fill the Help Map with real names they choose. They practice the sentence: I need a person.
- Students make
- A Help Map. Private details stay off the public wall version.
- What can go wrong
- Do not force a family shape that is not there. The map can include adults at school.
Week 35 · Presentation
70% of class doing the work
Practice asking
- Students do
- Students role-play asking a person for help. The public wall shows kinds of help, not secrets.
- Students make
- A practiced sentence, and a wall of the kinds of help.
- What can go wrong
- Do not turn a private problem into a performance.
Unit 8 · Weeks 36–40
A class museum of tools
Question for the unit. What did older tools change for people, what is AI changing now, and what rule do we want next?
The year ends in public. There are three stations: past, now, and future. Each object has a benefit, a cost, a risk, and a proper use. Families walk through it. Students say how their own thinking grew.
- Past
- A museum always makes an argument about what mattered. This one will too.
- Now
- The tools are already in the students' pockets.
- Future
- Students write a label for a tool that does not exist yet, and the rule they would want attached to it.
The rule. An exhibit that only says how impressive a tool is works like an advertisement. Every station includes a risk.
End-of-unit project. Grade 4 tool museum — A three-station exhibit with labels, a family-night tour, and a short written answer: what did I learn to do myself this year?
Week 36 · Introduction
35% of class doing the work
Choose the stations
- Students do
- Teams pick objects for past, now, and future. They draft the question each station asks a visitor.
- Students make
- Station plans.
- What can go wrong
- The future station is only a wish list. It needs a rule, not only a wish.
Week 37 · Making
90% of class doing the work
Build the past station
- Students do
- Students build the past station in Museum Builder and in the room. Every label names a benefit, a cost, a risk, and a proper use.
- Students make
- The past station.
- What can go wrong
- Students write only about how nice the old tool was. Every label needs the cost.
Week 38 · Making
90% of class doing the work
Now and future
- Students do
- Students build the remaining stations. Future tools must carry a rule the class would want.
- Students make
- The now station and the future station.
- What can go wrong
- The future station is only a story about a cool machine. It needs the rule.
Week 39 · Review
75% of class doing the work
Practice the labels
- Students do
- Students give each other tours. If a visitor cannot hear a warning, the label is rewritten.
- Students make
- Revised labels.
- What can go wrong
- The labels use words a younger sibling would not understand. Rewrite them in plain words.
Week 40 · Presentation
70% of class doing the work
Museum night
- Students do
- Families visit. Each student answers: what can I do now that I could not explain at the start of the year?
- Students make
- A public exhibit and a short written answer.
- What can go wrong
- The tools get more attention than the students. The night is about what the students learned.
Tools in this year
- Intelligence Sorter
Students sort living things, ordinary tools, and AI tools, then explain the cases that are hard to sort.
- Pattern Pupil
The class trains a small sorting program on examples they choose, then tests it on purpose with examples it has not seen.
- Prompt Kitchen
Students turn a vague request into clear instructions: what to include, what steps to follow, what limits to set, and how to check the result.
- True/False Detective
Students look for confident guesses in paragraphs about things they already know.
- Secret Shield
The tool flags names, places, passwords, photos, and family business before anything is sent.
- Co-Create Studio
Students can draft a story or an image, but they cannot export it until they have changed it themselves.
- Who Can Help?
Students sort a feeling or a problem toward a person, a tool, or both. A chatbot is never treated as a friend.
- Museum Builder
Students make exhibit cards for the past, the present, and the future. Every object needs both a use and a risk.