A student says the answer is 42. The teacher does not say yes or no. She asks what would change his mind.
That single question is the whole lesson. Below is a short dialog, then a walk through the belief shifts happening underneath. No formulas required. The shape of Bayes' rule (Class E, Parr, Pezzulo, and Friston, 2022) stays visible in the way each person moves.
The dialog
Setting: a middle-school science room. The class is estimating how many paperclips a small magnet will lift. A student, Jo, has just written 42 on the board.
Teacher: What made you land on 42?
Jo: I picked it up last time and it grabbed a bunch. I think it was around forty something. Forty two felt right.
Teacher: Okay. Before we test it, what number would surprise you? What would make you say, that cannot be right?
Jo: If it was five. Or if it was two hundred.
Teacher: Good. So your belief has a range. Now, one thing before we test. Is the magnet the same one you used last time?
Jo: I think so. It looks the same.
Teacher: What if I told you this one has been used in the drawer for three weeks and might be weaker?
Jo: Then maybe less. Thirty?
Teacher: You just updated. Not because you did the experiment yet, but because a piece of information came in and you moved your guess. Let us test.
They test. The magnet lifts 26 paperclips.
Jo: I was way off.
Teacher: You were closer than you think. You started at 42, moved to 30 on the drawer news, and the answer is 26. Which of those two moves taught you more?
Jo: The drawer thing.
Teacher: Why?
Jo: Because I did not have to be wrong first. I just heard something and thought about it.
What actually happened
Three belief moves happened in that exchange, and each one has the same shape:
- Jo held a prior belief (about 42).
- Evidence arrived (the drawer story, then the test).
- Jo revised toward the evidence, weighted by how much he trusted it.
That is the shape of Bayesian updating (Class E, Parr, Pezzulo, and Friston, 2022). Belief times evidence, normalized, becomes new belief. In a classroom you do not need the equation. You need the discipline of asking, before the test: what would change your mind, and how much.
The teacher did two things that are worth naming:
- She asked for a falsifier before running the experiment. This is Class F thinking in the classroom: a prediction is not real unless it can be broken.
- She rewarded the smaller update (the drawer news) as much as the bigger one (the test). This is important. Students learn that updating on a good piece of information is not weakness. It is skill.
Why this belongs in the educator-readiness cluster
Teachers already do a version of this every day. The active-inference frame (Class E) gives it a shared vocabulary so a school can talk about it across grades and subjects. Prior, evidence, update. Prediction, check, revise. When a whole building uses the same words, students move between classrooms without losing the thread.
The Workbench (Class C, integration) supports this by asking students to log two things per lesson: what they predicted, and what changed their mind. Over a term, the log is a record of how a student's mind actually moves under evidence. That is more useful to a teacher than a stack of right-or-wrong answers.
This work sits on the attainable path toward General Natural Intelligence, natural not artificial. It is a working hypothesis with growing, evidence-classed evidence, tested in the open.
Next reads
- What Active Inference Actually Says for the ground under the dialog above.
- Prediction Then Check in Daily Lessons for a routine you can run tomorrow morning.
- Evidence Classes as Classroom Vocabulary for how A, B, C, E, F, and U get spoken in a lesson.
- If you want the full teacher pathway, see the /workshop page.