Ask a student what they think will happen, in writing, before you show them anything. That single move changes the shape of the next ten minutes.
The pattern is old. Guess, look, update. What active inference gives us is a reason it works, and a scaffold precise enough to run in a math class on Tuesday and a history class on Wednesday. In Parr, Pezzulo, and Friston (2022), agents reduce uncertainty by generating predictions and revising them against sensed evidence (Class E). A classroom that runs on prediction, then check, is putting that loop where students can see it, name it, and take it home.
The ten-minute pattern
The routine has three moves. Keep them short. Keep them visible on the board.
-
Predict. One sentence, written, committed. "I think the graph will curve up and then flatten." "I think the character lies to protect the brother." "I think the salt will dissolve faster in the warm cup." A prediction is not a guess about the right answer. It is a claim about what will happen next.
-
Check. Run the experiment, read the passage, watch the demo, work the problem. The teacher does not narrate what students should be noticing. Students look for what would confirm or contradict their own written line.
-
Update. One sentence, written under the first. "The graph curved up and kept going, I was half right." "The character did lie, but not for the reason I gave." "The salt dissolved faster, and the water got colder while it did, I did not expect that."
The whole loop fits inside ten minutes for most subjects. The written commitment is the load-bearing piece. Without it, students unconsciously edit their prediction after they see the evidence, and the update never happens.
Why this maps to active inference
An active-inference learner is not a blank slate that absorbs the lesson. They are a generative model that already has predictions about the world, running all the time, whether the teacher invites them out or not (Class E). The routine surfaces those predictions and gives them somewhere to go.
The update step is where the learning lives. In the framework, the size of the update is proportional to how surprising the evidence was against the prior (Class E). A prediction that turned out exactly right is fine. A prediction that turned out wrong in an interesting way is a gift, and students learn to say so out loud without shame. This is what evidence classes as classroom vocabulary is for: giving learners language to tag their own claims by how they know them.
What it changes in the room
Two things shift, in our experience running this in a small pilot cohort (Class C):
The first is that students stop performing certainty. When the routine treats a wrong prediction as data, not as a grade, the pressure to sound sure evaporates. The second is that the teacher gets a read on the class's actual priors before the lesson, not after the quiz. If eight students predicted the graph would go down, the teacher knows what to teach next, in that period, not next week.
Gate design fits inside the routine
The prediction line and the update line are natural checkpoints. They are also gates: places where the work has to pass a visible standard before the class moves on. See gate design for middle school for how we build those checkpoints so they carry the routine instead of interrupting it. Structurally, a prediction that never gets a written update is a gate that was skipped.
A note on the theory
We are teaching UNI as a working hypothesis on an attainable path toward General Natural Intelligence, natural not artificial. That hypothesis says learning is prediction plus revision under uncertainty, all the way down. The classroom routine above is one small, concrete place where the theory becomes a Tuesday morning. Do not take the frame on faith. Run the routine for a week. Look at the update lines your students wrote. See if the theory earned its keep.
Where to next
- Read learner agency in a world of generative models for the underlying frame.
- See evidence classes as classroom vocabulary for how students learn to tag their own claims.
- Use gate design for middle school when you are ready to build the routine into the unit, not just the day.
- Book a seat at the workshop to practice the pattern with other teachers and get feedback on your unit plan.
AI-authorship fence: this post was drafted with LLM assistance, reviewed and edited by a human author, and published under human editorial responsibility. No claim is made that an LLM authored the reasoning. The framing is a working hypothesis on the attainable path toward General Natural Intelligence, natural not artificial, tested in the open.