Assignment Redesign Lab

Can AI Pass Your Class?

Time to think like a student. You'll try to use AI to cheat on your own assignments — then redesign them to be AI-resilient.

Why This Matters

If you can use AI to pass your assignment in 30 seconds, your students already know how. And they're not just pasting text — they're snapping photos of worksheets, voice-chatting with AI on their phone, and running ChatGPT in a tab next to their Google Doc. This isn't about catching cheaters — it's about designing assignments that make AI a tool, not a shortcut.

The Reality

The AI Vulnerability Spectrum

Not all assignments are equally vulnerable. Here's how common assignment types stack up:

Most Vulnerable

AI can complete these perfectly with zero effort — even from a photo

Recall questions, definitions, basic summaries, fill-in-the-blank, standard essays on common topics, any worksheet a phone camera can see

Partially Vulnerable

AI produces plausible output but lacks personal depth

Analysis essays, code without explanation, research without synthesis, generic projects

Most Resilient

AI struggles because these require personal context

Debugging live, peer code reviews, personal reflections on process, in-class presentations, iterative portfolios

The goal isn't to ban AI. It's to design assignments where using AI thoughtfully is part of the learning — and copy-pasting isn't enough.

Know the Playbook

How Students Actually Use AI

It's not just copying from ChatGPT anymore. Here's what's really happening:

Photo → Answer

Snap a photo of the worksheet, upload it to ChatGPT or Gemini, and get a completed version back in seconds. Works for handwritten and printed assignments.

Voice Chat

Students read the prompt aloud to AI on their phone and get a spoken answer back. No typing, no screen to spot. Looks like they're on a phone call.

Split Screen

Google Doc on one side, AI on the other. Students generate a paragraph, rephrase it slightly, paste it in. Repeat section by section.

Rewrite & Humanize

Generate the answer, then ask AI to "rewrite this like a 10th grader would write it" or "add some mistakes so it looks real." Defeats most detectors.

The Challenge

Try to Beat Your Own Assignment

Pick one of your real assignments. Then use any AI tool to try to complete it. Be honest — how much could a student fake?

Step 1: Pick an Assignment

Choose something you actually use — a homework, a project prompt, a test, a lab. The more real it is, the more useful this exercise will be.
Step 2: Feed It to AI — Every Way You Can

Don't just paste text. Try what students actually do: take a photo of the worksheet with your phone and upload it to ChatGPT. Open a Google Doc and AI side by side. Try voice chat — just read the assignment aloud to Gemini. Use Gemini, Claude, or ChatGPT — whichever your students would reach for.
Step 3: Grade It Honestly

Would this pass? Would you catch it? What grade would it get? Be brutally honest. Note what it got right and what was clearly missing.

Solo Work 15:00

Table Talk

What Did You Find?

Discuss with your table, then share your reflections:

Assignment You Tested *

What Did AI Get Right? *

Would You Catch It? *

What Was Missing? *

The Playbook

Strategies for AI-Resilient Assignments

Require Personal Context

"Use a specific example from our class discussion" or "reference your own code from Lab 3." AI doesn't know what happened in your room.

Assess the Process

Ask for drafts, revision history, or a reflection on what they changed and why. AI can produce a final product but can't fake a real journey.

Add a Live Component

Pair written work with an in-class presentation, a live demo, or a peer code review. If they can't explain it, they didn't learn it.

Make AI Part of It

"Use AI to generate a first draft, then critique and improve it." When AI use is the assignment, there's nothing to hide.

Examples

Before & After

Here's what a simple redesign looks like in practice:

Before

"Write a Python function that sorts a list."

AI completes this in 2 seconds. Any student can paste the prompt and submit the output. No learning required.

After

"Write a sort function, then explain line-by-line to a partner. Record a 60-second video walking through your logic and one mistake you made along the way."

AI can write the code, but it can't fake the video, the mistakes, or the partner conversation.

Before

"Write an essay about the causes of the Civil War."

AI produces a polished, generic essay instantly. Hard to distinguish from student work.

After

"Use AI to generate an essay on the causes of the Civil War. Then write a critique: what did it get right? What's oversimplified? What perspective is missing? Use sources from our class readings."

Now the student has to actually think — and reference class-specific material AI doesn't have.

Your Turn

Redesign Your Assignment

Take the same assignment you just tested. Now make it AI-resilient using what you've learned.

Step 1: Identify the Weakness

Where exactly did AI succeed? Was it the format (essay, multiple choice)? The lack of personal context? The absence of a live component?
Step 2: Pick a Strategy

Add personal context, assess the process, add a live component, or make AI use explicit. You don't need all four — one or two changes is usually enough.
Step 3: Rewrite the Prompt

Write the new version of the assignment. Keep it the same topic and learning objective — just change how students demonstrate understanding.
Step 4: Test It Again

Feed the redesigned assignment back to AI. Is the output still passable? If not, you've won. If yes, iterate — you know the drill.

Redesign Time 20:00

Share Your Work

Submit Your Redesign

Share your before & after so we can learn from each other. All submissions go to a shared collection.

Gallery Walk

Redesign Showcase

See what everyone came up with. Click any card to read the full before & after.

0 redesigns submitted

Loading submissions…

Wrap Up

Key Takeaways

AI Isn't Going Away

Students will use these tools whether we like it or not. The question isn't "how do we prevent it" — it's "how do we teach alongside it."
Design > Detection

AI detectors are unreliable and create adversarial classrooms. Better assignment design is more effective and builds trust.
Small Changes, Big Impact

You don't need to reinvent every assignment. Adding one personal context requirement or one live component is often enough to shift from "AI-completable" to "AI-assisted."
You Now Have a Revised Assignment

You walked in with an assignment. You're walking out with a better one. That's the whole point.

The best assignments in an AI world aren't the ones AI can't do. They're the ones where using AI well is part of the learning.