This week’s Coffee Break AI Challenge asks you to do something that feels counter-intuitive at first:
use AI to deliberately design wrong answers.
Not silly ones.
Not obviously incorrect ones.
But the kind of answers that look reasonable, sound confident, and reveal exactly where a learner’s thinking has gone slightly off track.
Because if you can predict misconceptions, you’re in a much better position to teach around them. And it turns out AI is quite good at producing convincing mistakes — which is precisely why this works as a teaching tool.
The Challenge
Use AI to generate plausible misconceptions for a topic you teach, then use them to support discussion, diagnosis and deeper thinking.
The aim isn’t to catch students out.
It’s to surface misunderstandings early and make thinking visible — something most teachers already do instinctively, just at speed.
How To Do It
Step 1: Pick a Familiar Problem
Choose a concept that regularly causes confusion.
This could be:
- a maths method students half-remember
- a scientific process that gets muddled
- a grammatical rule that’s often overgeneralised
- an exam question with a common trap
Paste the concept, question or objective into your AI tool.
Step 2: Ask for Misconceptions (Not Answers)
Use a prompt along these lines:
Generate 3–5 common misconceptions or plausible incorrect responses that students might give for this concept.
Explain why each one is tempting and what misunderstanding it reveals.
Do not provide the correct answer.
That final instruction matters more than it looks.
Step 3: Apply Professional Judgement
Some of what comes back will be useful.
Some of it won’t.
Occasionally it will invent a misconception you’ve never actually seen in your lessons..
That’s fine.
Your very important role here is to:
- select what genuinely reflects your learners
- reword anything that needs tightening
- ignore anything that doesn’t fit your context
As ever, the AI generates possibilities.
You decide what has value.
Step 4: Use Them Deliberately
Those misconceptions can now be used as:
- diagnostic starter questions
- multiple-choice distractors
- discussion prompts
- hinge questions
- “Which answer would you pick — and why?” tasks
Students often engage more readily with almost-right answers than with perfect ones.
Why This Works
This challenge quietly reinforces a few important ideas:
- mistakes are not random — they’re logical steps taken too far
- understanding why something is wrong matters as much as knowing the right answer
- AI can support pedagogy when it’s carefully constrained
- teachers still shape the learning, the narrative and the judgement
It also sidesteps a lot of AI anxiety. There are no shortcuts here — just better questions.
Time Taken
A couple of minutes to prompt the AI.
A few more to refine what comes back.
A noticeable improvement in the quality of classroom conversation.
Bonus Idea
Try this as a short CPD activity.
Give staff a shared concept.
Ask them to generate misconceptions using AI.
Compare results.
It’s a surprisingly effective way to surface subject-specific thinking and how different teachers anticipate misunderstanding.
Challenge Complete
Designing misconceptions on purpose feels odd until you remember: teachers have always done this.
AI just gives you a faster first draft.
Used carefully, it becomes a diagnostic tool rather than a shortcut — and once again, the human stays firmly in the loop.
Want Support Using AI Responsibly in Your School?
If your school or Trust wants practical, evidence-based CPD on using AI to support learning without undermining thinking or integrity, we run short sessions focused on exactly this balance.
At the very least, you’ll leave confident that AI hasn’t stolen the thinking — because you never handed it over in the first place.
It’s ideal for learning content, planning and workload reduction — while keeping teachers firmly in charge of the thinking.
Enjoyed this challenge? – Check out the others in the series