You do not need a curriculum overhaul to try this. You need one lesson, ten minutes of it, and a willingness to write on the board something you do not know the answer to yet.
Teachers write to us saying the workshop looks interesting but the timing is wrong: a new textbook is landing, the department is mid-rewrite, the semester is already scoped. That is fair. This post is for that teacher. Pick a lesson you already teach well. Run it exactly as planned. Add two small moves and one small gate. See what happens. If nothing changes, you lose nothing. If something changes, you have your own evidence to bring to the next conversation.
The one lesson
Choose a lesson where students look at something: a graph, a passage, a demo, a worked example, a photograph, a data table. Almost any lesson qualifies. Keep the content unchanged.
Before you show them the thing, add move one: ask for a written prediction. One sentence, committed on paper or in a shared doc. Not a guess at the right answer, a claim about what they expect to happen or notice. "The line will curve up and flatten." "The character will lie to protect the brother." "The number of cases will drop after the policy change."
Then show them the thing. Let them work with it the way you had planned.
At the end, add move two: ask for a written update. One sentence, beneath the prediction. "The line curved up and kept going, I was half right." "The character did lie, but not for the reason I said." "The number of cases went up before it went down, I did not expect the bump."
That is the whole change. Two sentences per student, bookending a lesson you were already going to teach. The pattern comes from active inference, where an agent reduces uncertainty by making a prediction and revising it against what it senses (Class E, Parr, Pezzulo, Friston, 2022). The classroom move is very old, guess and look and update, but the written commitment is the load-bearing piece. Without it, students quietly edit their prediction after the fact and the update never actually happens. There is more depth to this routine in prediction-then-check-in-daily-lessons, including how to hold the loop steady when a student refuses to commit on paper.
The one gate
Pick one place in the lesson where a student cannot go forward until they have written something down. This is the gate. It is not a grade. It is a threshold.
Good gate placements for a single lesson:
- Between predict and check. No prediction on paper, no access to the demo, the graph, or the text.
- Between check and update. No update written, no exit from the room, or no move to the next problem.
- Between individual work and group discussion. No written stance, no seat at the group.
Pick one. Not all three. The gate is small on purpose. It is a request, in physical form, that the student commit before the room lets them proceed. Design notes for tuning gates in a middle school setting are collected in gate-design-for-middle-school, including what to do with the student who freezes at the threshold.
What to watch for
You are not measuring learning gains today. You are watching your own room.
- Who wrote a prediction they clearly did not believe, to get past the gate (Class C, configuration signal)? Note them. That is useful information about the room's psychological safety, not a failure of the routine.
- Which predictions were confident and wrong? Those are the moments you want. A student who commits to a wrong prediction and updates in writing is doing the exact thing the routine is built to protect.
- Which predictions were vague, "something will happen"? Vagueness is a hedging move. The gate is what makes hedging visible.
Log three lines at the end of the day. What you saw. What surprised you. What you would change next time. That log is your data. It is more honest than any vendor's dashboard.
What this is not
This is not a claim that a single lesson changes a student, a classroom, or a semester. It is a low-cost, honest probe. You are testing whether the posture, predict on paper, check against evidence, update in writing, fits your room and your subject. If it does, there is a next step. If it does not, you have learned something real about your room without having reorganized anything.
There is also no AI in this lesson. No tool. No model. No app. This is a pedagogical posture from an active-inference framing (Class E), and it runs on paper.
The next step, if you want one
If the one lesson goes somewhere interesting, do it again next week with a different lesson. If after three tries you want the underlying model made explicit, and a small cohort of teachers thinking alongside you, that is what the workshop is for.
- /workshop: the small-cohort program where the routine and its gates are taught with the active-inference model behind them.
- prediction-then-check-in-daily-lessons: the fuller version of the routine, with subject-by-subject notes.
- gate-design-for-middle-school: how to place, tune, and soften gates for early adolescents.
One lesson. Two written sentences. One gate. That is a real week's work.