The Assignment Design System
Good teaching, doubled down for the AI era.
None of the ideas in this system are new. Great teachers have been designing lessons this way for decades: start with a real question, put students in the driver’s seat, assess thinking instead of output, share work with a real audience.
AI didn’t change what good teaching looks like. It raised the cost of pretending that worksheets and five-paragraph essays measure understanding. This system is a practical guide for doubling down on what we already know works, across four phases that build on each other.
Phase 1
Wonder
Start with a question worth asking
Phase 2
Tools
AI as thinking partner, not answer machine
Phase 3
Evidence
Assess thinking, not output
Phase 4
Share
Social accountability and peer review
Wonder
“Start with a question worth asking”
This phase is about hooking students before any content delivery. Make lessons relevant to their lives, tie the work to something real, and show them why the topic actually matters. Students who see the point are far less likely to outsource the thinking to AI.
Core Principles
- ▸Every lesson begins with genuine curiosity, not a textbook chapter.
- ▸Students form hypotheses before receiving any information.
- ▸The question should be interesting enough that AI can’t simply "answer" it.
- ▸Cognitive dissonance is the goal — make students uncomfortable in productive ways.
In Practice
Show 3 mystery organisms from extreme environments. Ask: "What adaptations let these survive?" Students discuss in pairs before any lecture.
From: Biology — The Hook
Start with genuine curiosity — students form hypotheses before receiving any information.
Common Mistake
Starting with "Today we’re going to learn about…" and jumping straight into content delivery.
Try these moves · Wonder
Design the Task
Low lift · 5 min to add
Local Anchors
Require students to reference something AI can’t access: classroom discussions, local data they collect, or recent events specific to your context.
Rewrite for
Connect this text to something specific happening in our school, town, or recent news from our region within the last 30 days.
AI doesn’t know what happened in your classroom last Tuesday. When students must integrate local or personal context, the thinking can’t be outsourced.
Low lift · 5 min to add
Personal Connection
Ask students to connect the topic to their own lived experience, family history, or individual observation — material AI can’t fabricate.
Write an essay analyzing the theme of sacrifice in our class texts. Include one paragraph connecting this theme to a sacrifice you’ve witnessed or experienced. Explain how that personal experience shaped your understanding of the literary theme.
AI doesn’t have personal experiences, family stories, or individual observations. The personal element forces original thinking and authentic voice.
Low lift · 15 min to adapt
Creative Constraint
Add a specific format or angle limitation that makes the conventional response impossible. The constraint itself becomes a thinking tool.
Write your character analysis as a series of letters between you and the character at three different points in the text. In each letter, ask a genuine question. In their "response," analyze what the text reveals about their thinking at that moment.
AI is excellent at producing conventional responses. When you constrain the format in unusual ways, students must think creatively about structure and content.
3 of 42 cards in The Assignment Design Guide.
Tools
“AI as thinking partner, not answer machine”
This phase is about teaching students to use AI actively instead of passively. They drive the investigation, cross-check what the model gives them, and document how they used it. AI accelerates the work, but judgment still lives with the student.
Core Principles
- ▸Students use AI actively, not passively — they drive the investigation.
- ▸Every AI response must be evaluated, cross-referenced, and critiqued.
- ▸The task is designed so AI accelerates research but can’t replace judgment.
- ▸Students document their process: prompts used, outputs evaluated, decisions made.
In Practice
Students use AI to research their assigned environment, but must build an "evidence map" showing what AI got right, wrong, and oversimplified.
From: U.S. History — Source Investigation with AI
AI accelerates research but the critical evaluation — deciding what's trustworthy — is entirely human.
Common Mistake
Banning AI entirely, or letting students copy-paste AI outputs without evaluation.
Try these moves · Tools
Support the Process
Medium lift · Multi-day
Checkpoint System
Break the assignment into staged, graded deliverables spread across the timeline. Process becomes visible and impossible to compress into one sitting.
Pick the AI policy you’d attach to each stage of a research essay. (The book recommends mixing levels across stages.)
🟢 AI Use: EXTENSIVE — Explore angles with AI. Document prompts and outputs in your notebook.
You can’t outsource a month-long process to AI in one sitting. When checkpoints are graded and spaced out, shortcuts become obvious or impossible.
Low lift · 1 class period
Studio Time
Dedicate class time for students to work on the assignment with you present — observing, coaching, unsticking — while thinking actually happens.
Rewrite for
Tuesday: bring your outline and sources. You’ll draft your introduction and first body paragraph in class with support available. 🔴 No AI during studio time.
When core thinking happens in class, you see it happen. The blank-page paralysis that drives AI use at 11pm gets resolved with teacher support.
Low lift · 5 min to add
Revision Memos
Require a brief written explanation of what changed between drafts and why — making the student’s revision thinking visible on its own.
Submit your final draft with a 1/2- to 1-page revision memo. Address: What two or three significant changes did you make from your checkpoint draft? Which piece of feedback had the biggest impact? What’s one thing you’re still not satisfied with, and why? Memo is 10% of the grade.
AI can revise text, but it can’t authentically explain a student’s genuine revision decisions. The memo reveals whether they engaged with feedback.
3 of 42 cards in The Assignment Design Guide.
Evidence
“Assess thinking, not output”
This phase is about what you measure. Ask for decisions, reasoning, and defense rather than for information any model can produce. When the grade rewards the thinking behind the answer, students have to do the thinking.
Core Principles
- ▸Assessment asks for decisions and reasoning, not information AI can produce.
- ▸There is no single "right answer" — the quality is in the argument.
- ▸Students must defend their thinking orally, in writing, or through creation.
- ▸The assessment captures thinking transformation — something AI literally cannot fake.
In Practice
Students design a hypothetical organism for a novel environment, justifying every adaptation with evolutionary principles and citing which sources informed each decision.
From: English — Audience Testing
Persuasion is measured by actual audience response, not a rubric. This mirrors how persuasion works in the real world.
Common Mistake
Grading essays or problem sets that AI can complete perfectly in seconds.
Try these moves · Evidence
Gather Evidence
Low lift · 10–30 min
In-Class Writing
Include a brief supervised in-class writing component tied to the main assignment. Direct evidence of what students can articulate in their own words.
Pick the level you’d attach to the in-class write.
🔴 AI Use: NONE — This is your thinking, in the moment. Notes and the text are allowed. No AI, no internet.
The writing doesn’t need to be perfect. The goal is verification, not performance. Grade for thinking, not polish.
Medium lift · 5–7 min per student
Oral Defense (Mini-Viva)
Have students explain their work, defend their choices, and answer questions in a short 3–7 minute conversation. AI can’t show up to class.
Rewrite for
Walk me through your main argument in your own words. Which piece of evidence was most important? If someone disagreed with your thesis, what would you say?
Students who genuinely did the work can explain their thinking. Students who heavily relied on AI struggle to articulate reasoning. Conversation, not interrogation.
Low lift · 15 min
Teach It Back
Have students explain a concept or process to someone who doesn’t already understand. You can’t teach what you don’t know.
Write your lab report AND create a 3-minute explanation of your findings for 8th graders who haven’t taken this class. Your explanation should help them understand both what you discovered and why it matters. Film it or present to a peer. 15% of the lab grade.
Explaining requires deeper processing than summarizing. Students who relied heavily on AI struggle to adapt explanations to different audiences or answer follow-ups.
3 of 42 cards in The Assignment Design Guide.
Share
“Social accountability and peer review”
This phase is about closing the loop with an audience. Students present, defend, and revise their work in front of peers. A live conversation with a room of real people is a kind of accountability AI simply cannot provide for them.
Core Principles
- ▸Students present, defend, and revise their thinking with peers.
- ▸Oral argumentation and live Q&A can’t be outsourced to AI.
- ▸Peer feedback creates social accountability for original thinking.
- ▸Reflection on what changed in their thinking is the real assessed artifact.
In Practice
Students pitch proposals to peers who rate persuasiveness and identify weak points. Designs are revised based on real audience feedback.
From: Algebra II — Error Analysis
Error analysis builds metacognition — understanding math, not just doing math.
Common Mistake
Submitting a final product without ever explaining or defending the thinking behind it.
Ready to redesign your lessons?
Test your own lesson plan with the AI Analyzer, or go deeper with excerpts from the Assignment Design Guide.