Teach the AI
The AI plays a student who explains a concept the group has just learned and applies it in two examples. The group's job is to judge the answer: what is right, what is missing, what is subtly wrong. They explain their corrections back to the AI. Finding the AI's mistakes tests how well people really understand the concept.
Use at the end of a learning block to check and deepen understanding, when a quiz would be too shallow.
Solves: People nod along to new concepts but cannot tell a correct application from a plausible wrong one.
- 1
Step 1: Give each group the AI-as-student prompt and the concept they will test (3 min).
- 2
Step 2: Ask the AI to explain the concept and give two applications, for example as a short scene or story (3 min).
- 3
Step 3: The group marks what the AI got right, wrong and left out (8-10 min).
- 4
Step 4: The group writes back its correction and explanation to the AI (5 min).
- 5
Step 5: Plenary: each group shares the most instructive mistake it found (5-10 min).
- Use it only after people have learned the topic; without knowledge they cannot spot the errors.
- If the AI gets it all right, ask groups to explain precisely why each part is right.
- Warn that the AI may argue when corrected, and that groups can end the exchange once they have made their point.
- Groups each test a different concept and swap annotated outputs for a peer check.