Curriculum guide · Grades 6–12

A middle and high school AI curriculum that keeps students thinking

Twenty-four lesson outlines for grades 6–12 in four grade bands, two for middle school and two for high school, with links to supporting ARQIV resources. Each lesson says whether screens are off or on, what students will show, and which ARQIV lab, kit or lesson to use. Built for classrooms with no devices, a shared projector or a full lab.

Grades 6–1224 lesson outlines6 weeks per bandScreens optional

The design rule

Understand it, test it, check it, own it.

Each unit moves students from understanding how AI works, to testing what it can and cannot do, to verifying what it says, to owning the final work and defending it out loud.

Every unit includes at least one task students complete without AI, so teachers can see what students have really learned.

Six strands

What students learn

The same six strands run through every band. They get deeper, not just harder, as students grow.

Strand 1

How AI works

Sensors, data, patterns, training, tokens and prediction. Students build working mental models instead of memorizing definitions.

Strand 2

Using AI well

Writing a clear brief, thinking first, choosing the right tool and keeping ownership of the work.

Strand 3

Verify & research

Checking claims, tracing sources, testing models and reading research honestly.

Strand 4

Problem solving

Estimating, reasoning, planning and defending answers, with and without a tool.

Strand 5

Ethics & society

Data, fairness, privacy, creators’ rights and who is affected by AI decisions.

Strand 6

AI in the world

Real uses in space, health, weather, farming and engineering, and the careers behind them.

Four grade bands

Pick a band, see the six weeks

Grade 6 and Grades 7–8 are for middle school; Grades 9–10 and 11–12 are for high school. Lessons are 30 to 90 minutes. Swap the order, combine lessons or run one a month. The sequence works as a unit or as stand-alone pieces. The guide starts at grade 6 because the ideas (prediction, training data, verification) need abstract reasoning, and most AI tools set a minimum age of 13.

Grade 6 · Middle school · Pattern Finders

The big idea

AI finds patterns in data. A machine only knows what its sensors and examples give it, and people decide the rules and check the results.

Ground rules

Unplugged first, teacher screen only

Most sixth graders are 11 or 12, younger than the minimum age (13) in most chatbots’ terms. Students do not use chatbots. Work is unplugged or in ARQIV’s no-account labs, and any chatbot demo runs on the teacher’s screen.

Connections: AI4K12 (grades 6–8): Perception, Representation & Reasoning, Learning, Societal Impact. UNESCO level: Understand.

Six weeks, six lesson outlines

1

What is AI, and what is not?

Goal: Sort everyday technology into AI and not-AI, and name where each one gets its information.

Activity: Students sort cards (calculator, spam filter, music recommendations, light switch, voice assistant, thermostat, face unlock, autocorrect) and justify each choice. Compare ordinary fixed-rule automation (“follows fixed steps”) with machine learning (“learns from examples”), one important kind of AI. Some other AI systems use human-written knowledge and reasoning rules.

Students show: Choose one card the class disagreed on and argue both sides in writing.

How AI works40 min○ Screens off
2

What can a machine sense?

Goal: Explain that a machine only “knows” what its sensors and data capture, and predict what it will miss.

Activity: Play “Sensor-Limited Robot”: one student can only hear spoken commands, another can only see printed cards, and each tries to finish a task. Then teams map a real device (phone camera, smart speaker, self-checkout, parking sensor): what it senses, what data it turns that into, and what it cannot detect.

Students show: A sensor map for one device, with two things it cannot sense and one mistake that could cause.

How AI works40 min○ Screens off
3

Write the sorting rule

Goal: Write a sorting rule like a programmer, then test it on cases that break it.

Activity: Teams write a rule or simple decision tree that sorts a mixed set of picture cards (for example, “Is it a fruit?”). Another team runs the rule exactly as written on new cards, including edge cases such as a tomato or a plastic apple. Teams patch the rule and count the patches. Debrief: why do engineers switch from hand-written rules to learning from examples?

Students show: A decision tree, one edge case that broke it, and the fix.

How AI works40 min○ Screens off
4

Train the machine

Goal: A model learns whatever pattern is in its examples, including a wrong one.

Activity: Teams study set A cards, write a rule, then test it on new cards. Their rule works on A and fails on the test, which shows why diverse data matters. Run the online lab afterward.

Students show: Explain why “blue means Zib” was a good guess and a bad rule.

Train the Machine lab →

How AI works45 min● Supervised student screens
5

Exact instructions

Goal: Write an algorithm precise enough for a machine that does exactly what it is told.

Activity: One student plays a literal-minded robot that follows every word exactly. Partners write step-by-step instructions to draw a hidden figure, and the robot carries them out, gaps and all. Partners debug and rerun until the drawing matches. Connect it to AI: every model still runs on code, data and goals that people write.

Students show: A debugged set of instructions with each fixed bug marked.

Problem solving40 min○ Screens off
6

Fair data, fair machines

Goal: Explain how gaps in training examples lead to unfair results, and propose better data.

Activity: Read a short case: a model learned what a “sandwich” looks like from only one kind of picture. Teams predict what it will get wrong, then design a better training set: what to add, where it comes from and how to check it. Connect to real face and speech tools that made more errors for some groups of people.

Students show: Name one case the model missed and two examples you would add to its training data.

Ethics & society40 min○ Screens off

Grades 7–8 · Middle school · Check the Bot

The big idea

A chatbot predicts likely words. Likely is not the same as true, so every claim needs a check students can do themselves.

Ground rules

Teacher-led AI demos, students verify

Many students are still under 13. Chatbots run on the teacher’s screen while students work on paper and in ARQIV’s no-account labs. Students never type personal information into any tool, and they think before any AI appears.

Connections: AI4K12 (grades 6–8): Natural Interaction, Learning, Societal Impact. UNESCO level: Understand, starting Apply. AILit: Engage, Manage.

Six weeks, six lesson outlines

1

Guess the next word

Goal: A chatbot writes by predicting the next word, one word at a time.

Activity: Play round 1 of the Next Word game: students write a guess for the next word of each stem, the class tallies the votes, and the most popular guess wins. Ask: “Was the most popular word the true one? What makes one word more likely than another?”

Students show: Finish one sentence three ways, rank the endings from most to least likely, and explain the ranking.

Next Word kit →

How AI works30 min○ Screens off
2

How a chatbot writes

Goal: Chatbots predict the next word from patterns. Popular is not the same as true.

Activity: Continue the Next Word game through rounds 2–4, including the “temperature” die and the hallucination round. Students explain why turning up randomness makes answers more varied and less reliable.

Students show: Complete the sentence: “A chatbot can be wrong even when it sounds sure because ___.”

Next Word kit →

How AI works40 min○ Screens off
3

Real or AI-made?

Goal: Judge whether an image or story is real with checks, not gut feeling.

Activity: The teacher shows pairs of images prepared in advance (a real photo and an AI-made one, labeled on the back). Students vote and list their clues, then learn why visual clues alone are unreliable. Practice stronger checks: who posted it, whether a second reliable source reports it, and whether there is an original or a provenance label.

Students show: For one image, name the check you would use and the result that would change your mind.

Verify & research40 min○ Teacher screen only
4

Fact-check the bot

Goal: Check claims against evidence, not against how confident they sound.

Activity: Use the Fact-Check the Bot lesson: students read pretend bot answers beside the real records and mark each claim supported, contradicted or needs more evidence.

Students show: Evidence log with at least two claims correctly sorted.

Fact-Check the Bot →

Verify & research35 min○ Screens off
5

Verify it: the four SIFT moves

Goal: Use the four SIFT moves: Stop, Investigate the source, Find better coverage, Trace to the original.

Activity: Pairs play the Grades 6–8 set of the Verify It lab, then use the four moves on one claim from a class topic.

Students show: Show all four moves on a claim they chose.

Verify It lab →

Verify & research30 min● Supervised student screens
6

Problem first, prompt second

Goal: Estimate and solve first so you can tell when AI is wrong.

Activity: Use the lesson: students solve a bus-capacity problem on paper, estimate, then perform an autopsy on a confident wrong solution.

Students show: Solution, estimate and a written explanation of the AI’s error.

Problem first, prompt second →

Problem solving45 min○ Teacher screen only

Grades 9–10 · High school · Under the Hood

The big idea

AI is a set of methods with strengths, weaknesses and trade-offs. You find the limits by testing, and you stay in charge by thinking first.

Ground rules

Think first, then supervised AI use

First supervised student use with school-approved tools and the unaided-first rule: write your own attempt before asking AI. Teachers check each tool’s age and privacy terms.

Connections: AI4K12 (grades 9–12): Learning, Representation & Reasoning, Natural Interaction. UNESCO level: Apply. AILit: Engage, Manage.

Six weeks, six lesson outlines

1

Be the boss of the bot

Goal: Think first, then use the tool for one job, then check.

Activity: Pose a class question. Students think and plan on paper first (Ponder, Plan). In pairs, they ask a school-approved AI tool to do one specific job (Partner), then probe the answer: one thing right, one thing to check. They finish with an AI receipt.

Students show: A completed AI receipt and one fact checked in a second place.

The Think-First guide →

Using AI well45 min● Supervised student screens
2

The detective’s notebook

Goal: Reason from clues to hypotheses, then interview AI as a witness, not a detective.

Activity: Use the lesson: solve the mystery on paper, then question an AI as a witness and probe what it claims.

Students show: Notebook with hypotheses, evidence for and against, and one AI claim checked.

The detective’s notebook →

Problem solving50 min○ Teacher screen only
3

Inside a chatbot: tokens and attention

Goal: Explain tokens, training and attention in your own words.

Activity: Explore the transformer page in pairs (tokenizer, attention). Then explain next-word prediction to a partner without notes. Debrief: which part was hardest to explain?

Students show: A labeled diagram: text in, tokens, predictions, text out.

How transformers work →

How AI works45 min● Lab / devices
4

Spot the hallucination

Goal: Sort true claims from five types of AI mistakes and find the evidence.

Activity: Run the Spot the Hallucination kit as a scoot or partner sort. Then verify two cards with two independent sources.

Students show: Hallucination-type table filled in plus two verified cards.

Spot the Hallucination kit →

Verify & research45 min○ Screens off
5

Test the models

Goal: Design a fair test and report results with evidence.

Activity: Run the blind A/B vote and three known-answer prompts on two school-approved tools. Record version, date and mode. Fill in a model report card.

Students show: A report card that states what was tested and what it does not show.

Test the Models lab →

Verify & research50 min● Lab / devices
6

Who built it, who is missing?

Goal: Notice how data choices shape what AI does for different people. Connect to careers.

Activity: Bias detectives: compare image-search or label examples from a teacher-prepared packet, ask who is missing, then match AI in a field (space, health, farming) to the careers that work with it.

Students show: Two data questions and one career connection.

AI in the world: farming →

Ethics & society45 min○ Screens off

Grades 11–12 · High school · Think With, Not Through

The big idea

Use AI to extend thinking you have already done yourself. Be able to explain, test and defend everything you hand in.

Ground rules

Independent, accountable use

Students may use approved tools within the terms of service. Every project includes an AI receipt, an unaided checkpoint and an oral defense. Level of AI use is set by the teacher for each task (Levels 0–4).

Connections: AI4K12 (grades 9–12): all five Big Ideas. UNESCO level: Apply, Create. AILit: Engage, Create, Manage, Shape.

Six weeks, six lesson outlines

1

How models are built

Goal: Trace data to training to behavior, and explain why training choices matter.

Activity: Read the “How AI works” and “Transformers” pages. Draw the pipeline (data, training, tuning with human feedback, deployment). Write where errors and bias can enter at each stage.

Students show: Annotated pipeline with three places things go wrong.

How AI works →

How AI works50 min● Lab / devices
2

The model landscape

Goal: Compare model types, open vs closed, and what scoreboards measure.

Activity: Use the model field guide to compare three kinds of model. Critique one headline benchmark claim: what was tested, by whom, and what is missing?

Students show: One-page critique of a benchmark claim.

AI model field guide →

How AI works50 min● Lab / devices
3

Run a real evaluation

Goal: Plan, run and report a reproducible test of AI tools.

Activity: Teams write a test plan, choose 10 prompts with known answers, run each three times, log results, and test one edge case (false premise or hidden instruction).

Students show: Evaluation report with method, results, limits and a recommendation.

Test the Models lab →

Verify & research60 min● Lab / devices
4

Socratic seminar with a machine

Goal: Cross-examine an AI debater, demand evidence, then debate without screens.

Activity: Use the two-session lesson: question an AI on a contested claim, check one source, catch flattery, then hold a screen-free seminar.

Students show: Seminar contribution plus a source check.

Socratic seminar with a machine →

Problem solving90 min○ Teacher screen only
5

Research with AI, accountably

Goal: Use AI to find leads, then trace every claim to an original source.

Activity: Students choose a question, make an unaided plan, use AI for search terms and counterarguments, then trace each claim and read one real paper abstract (see Research explained). Write an AI receipt.

Students show: Claim table: claim, source opened, verdict, plus an AI receipt.

Verify It lab →

Verify & research50 min● Supervised student screens
6

Capstone: audit an AI tool

Goal: Evaluate an AI tool for a real school or community problem and defend the finding.

Activity: Teams choose a tool or use case, test it, interview a stakeholder, and present findings: what works, what fails, who is affected, and what you would do. A five-minute oral defense with questions from peers.

Students show: Presentation, report card, AI receipts and defense.

AI in the world →

AI in the world90 min● Supervised student screens

Standards connections

How it lines up with major frameworks

BandAI4K12 Big Ideas emphasizedUNESCO student levelAILit domains
Grade 6Perception; Representation & Reasoning; Learning; Societal ImpactUnderstandEngage
Grades 7–8Natural Interaction; Learning; Societal ImpactUnderstand, starting ApplyEngage, Manage
Grades 9–10Learning; Representation & Reasoning; Natural InteractionApplyEngage, Manage
Grades 11–12All five Big IdeasApply, CreateEngage, Create, Manage, Shape

This crosswalk is ARQIV’s own mapping for planning. It is not an official alignment or certification. The AI4K12 Five Big Ideas are Perception, Representation & Reasoning, Learning, Natural Interaction and Societal Impact. UNESCO’s student framework (2024) describes three progression levels: Understand, Apply and Create. The final OECD and European Commission AI Literacy Framework (AILit, June 2026) has four domains: Engage with AI, Create with AI, Manage AI and Shape AI. We have updated the domain names; our grade-band mapping has not yet been re-checked against the final framework’s competencies.

Assessment

Grade the thinking, not the output

AI makes finished work easy to fake and thinking hard to see. These four checks make thinking visible in every band.

1

Unaided checkpoint

One short task per unit with no AI: a mini-quiz, a sketch, an oral explanation. It shows what students actually learned.

2

AI receipt

A short note: the tool, the prompt, what I kept, what I changed, what I checked, what I learned. In Grades 6–8, students write it about the teacher’s demo.

3

Oral defense

Two minutes: “Why this answer? What would change your mind?” A student who cannot explain the work does not own it yet.

4

Process rubric

Score the plan, the questions asked, the checking and the revision, not only the final product. Use the rubrics in each ARQIV lesson.

Why an unaided checkpoint? In a high-school math field experiment (Bastani et al., PNAS 2025), students with an unrestricted chatbot did better on practice but worse on the later exam without it. Students with a hint-giving tutor largely avoided that drop. Read the evidence →

Adapting it

For your classroom

No devices

Run it all on paper

Most Grade 6 and Grades 7–8 lessons are screen-free. For other lessons, print any AI sample in advance. Our printable kits and lessons include answer keys.

One projector

The teacher is the interface

Students think and write on paper. The teacher types student questions into one AI tool on the board. Students critique the result together.

Computer lab

Supervised pairs

Use pairs, a school-approved tool and the unaided-first rule. Check each tool’s age limits and privacy terms first. See the age rules.

Printable

Take the whole guide with you.

The PDF includes every lesson, the standards crosswalk, assessment ideas and a one-page planner.