AI as Student
AI as Student is a learning exercise in which the usual roles are reversed: an AI chat assistant plays a student who explains a topic, and the learner acts as the teacher who judges what the explanation got right, got wrong or left out. It was described by Ethan and Lilach Mollick in their 2023 paper 'Assigning AI: Seven Approaches for Students, with Prompts'. The exercise relies on the fact that explaining something to someone else exposes how well you understand it, and on the assistant's habit of producing answers that sound right but are subtly incomplete.
Facilitation script
- 1
Explain the role reversal and the task: the assistant explains, you assess. State clearly that the assistant can be wrong and can argue, and that participants have the final word.
4 min - 2
Hand out the prompt and let everyone start a chat. Participants name the topic and choose how the assistant should illustrate it.
4 min - 3
Participants read the explanation and the two applications and check them against notes and reference material, marking what is right, wrong and missing.
8 min - 4
Participants write their assessment to the assistant in their own words, respond once if it argues back, then close the chat.
8 min - 5
Group debrief: collect the errors that were caught, the points that were hard to explain and any place where the reference material had to settle a dispute.
10 min - 6
Close by naming the concepts that need another round of teaching and agree how they will be covered.
4 min
Tips
Test the prompt on your topic with the assistant your group will use before the session, because output quality and the tendency to argue differ between tools and between runs.
Tell participants up front that a correct answer from the assistant is not a failed exercise: explaining exactly why it is correct is just as demanding.
Have reference material within reach so that disputes are settled by the source and not by whoever sounds more confident.
Common pitfalls
Running it on a topic people have not practised, so errors go unnoticed and participants leave with a wrong mental model
Letting participants accept the assistant's counter-argument because it sounds confident, which rewards fluency over accuracy and undermines the point of the exercise
Skipping the reference material, so that nobody can verify who is right and the debrief turns into opinion
Treating an accurate answer as nothing to do, when the learning lies in naming precisely why the explanation and the examples are correct
Not testing the prompt beforehand, so some assistants refuse the role or skip the waiting steps and the first ten minutes go to troubleshooting
Variations
The original is an individual student assignment; the paired set-up and the group debrief are additions that suit a training room. In pairs, one person types and the other checks against the notes, then they swap for a second concept. For a whole-group version, project one chat, let the group agree aloud on what in the answer is right, wrong and missing, and only then have one person type that feedback into the chat. Remote groups can work in breakout rooms and paste their corrections into a shared document for the debrief.
Where it fits
When to use it
Participants have been taught a concept and practised it, and you want to see whether they can explain it and spot a flawed explanation
A training module is ending and a quiz would only test recall, not fluency
The group uses AI assistants at work and tends to accept their answers without checking
You have a mixed room where everyone can work at their own pace on the same concept
People claim to know a topic because they have heard about it, and you want them to test that claim themselves
When not to use it
The topic is new to the group: people who do not know the material cannot spot the errors and may remember the wrong version, so teach it first
The content is confidential or contains personal data that should not be typed into an external AI tool
You need a graded, comparable assessment: every chat produces different output, so use a fixed test or a practical task instead
The aim is discussion between people about differing views: use Think-Pair-Share or Socratic Questioning, where the group itself does the explaining
Related methods
Frequently asked questions
What is AI as Student?▾
AI as Student is an exercise in which an AI chat assistant plays a student who explains a topic and the human learner plays the teacher. The learner assesses the explanation and tells the assistant what was right, wrong or missing. It was described by Ethan and Lilach Mollick in a 2023 paper on classroom uses of AI.
Why let an AI explain instead of asking participants to explain to each other?▾
The assistant produces an explanation and examples in seconds, so every participant gets something concrete to assess at the same time. Its answers are often plausible but incomplete, which makes them good material for checking. Peer explanation is still valuable; the two can be combined by working in pairs.
What are the risks of AI as Student?▾
The main risk is that learners who do not know the topic well miss the errors or remember the assistant's flawed examples. The assistant may also refuse the role, misunderstand the request or argue with a correct critique. Use it only after instruction and practice, and keep reference material at hand.
How long does AI as Student take?▾
The chat itself takes about ten to fifteen minutes per concept. With an introduction and a group debrief, plan 20 to 40 minutes. The source paper describes an individual assignment and gives no fixed duration.
Does the assistant actually learn from the correction?▾
No. The 'teaching' is a device for the learner: organising knowledge well enough to correct someone else is what builds understanding. The assistant's reply to the feedback matters much less than the quality of the learner's explanation.
Plan your next workshop with AI
Workshop Weaver helps you combine methods like AI as Student into a complete, timed agenda in minutes.
Try it freeMethod descriptions on Workshop Weaver are original content written by our team, based on established facilitation practices. This method was inspired by work from Ethan Mollick and Lilach Mollick, 'Assigning AI: Seven Approaches for Students, with Prompts' (2023).