A high-school science student can now generate a competent lab report in one prompt. If your grade book cannot distinguish that report from one written by a student who actually ran the experiment, the grade book is measuring the wrong thing.
This post walks through a semester-long design that moves the graded surface off the polished write-up and onto three weekly gates: prediction, revision, and falsifier. It extends gate design for middle school, adding one artifact: at this level the student writes, and defends, the specific observation that would prove their claim wrong.
Why a falsifier is the load-bearing high-school artifact
Middle school gate design already asks a student to predict, revise, and cite (see prediction then check in daily lessons). The added high-school move is the falsifier: a written sentence of the form, "my claim is wrong if I observe X." That sentence turns a student's answer into a scientific claim.
Active inference gives clean vocabulary for why this works. A learner's generative model produces predictions; sensory evidence produces prediction error; the model updates to reduce future error (Class E, Parr, Pezzulo, and Friston, 2022). A falsifier is the student naming, in advance, what prediction error would look like. Without it, the student can absorb any outcome as confirming their claim.
The three gates, at high-school resolution
- Prediction gate. Before the lab, before the dataset is opened, the student writes a specific, quantitative-where-possible prediction. "The current will increase" passes. "Something interesting will happen" does not.
- Revision gate. After evidence arrives, the student writes what changed in their model and what specific observation caused the change. If nothing changed, they name the observation that would have forced revision if it had gone the other way.
- Falsifier gate. The student writes one sentence naming the specific future observation that would prove their revised claim wrong. It has to be an observation the student, or a classmate, could plausibly make within the course.
Scoring is binary at the gate level and cumulative at the unit level. Polish is not on the rubric. The reasoning behind binary gate scoring is in curriculum gates that teach thinking.
A worked semester
Here is a sixteen-week physical-science sequence built around this design. Each unit ends in a written falsifier the class keeps on the wall.
Weeks 1 to 5, motion. Predict the time for a cart to roll down a ramp at three angles. Revise after measurement. Falsifier: "my model that time scales inversely with angle is wrong if the time at sixty degrees is not shorter than the time at thirty degrees by a factor I have written down." Test at a fourth angle.
Weeks 6 to 10, forces and energy. Predict which of two carts will move farther after a collision, and why, using a stated mechanism. Revise from data. Falsifier: "my mechanism is wrong if we swap the masses and the outcome does not swap direction accordingly."
Weeks 11 to 14, waves. Predict what happens to a standing wave in a string as tension changes. Revise. Falsifier: "my claim about frequency is wrong if the observed frequency at doubled tension is outside a range I have written down."
Weeks 15 to 16, integration. The student picks one prior unit and writes a new falsifier stronger than the one they wrote at the time, judged by two peer reviewers against a shared rubric.
The semester ends with a wall of written falsifiers. That wall is evidence the course produced scientists, not paragraph-producers.
What a graded falsifier looks like
A passing high-school falsifier meets three tests: it names a specific observation, not a feeling; it is decidable within the course; and it puts the claim at genuine risk. A falsifier no realistic outcome could satisfy fails the gate. A falsifier that requires equipment the class cannot access is aspirational, not gradeable.
The teacher's job at the falsifier gate is to point at each test and ask the student to show which clause of their sentence meets which. That is a short conversation, not an essay grade, which is why the design scales to a full class.
What this changes on the teacher's side
The lesson plan reorganizes around one planning question per unit: what has the student got to predict before evidence, what evidence could force revision, and what falsifier will they be graded on at the end. That is the same question the pillar post (curriculum gates that teach thinking) is built on. A high-school science teacher can adopt this without any additional software beyond a notebook per student and a shared wall.
Next steps
- Bring one unit you already teach to the /workshop and rewrite it around the three gates before touching anything else.
- Read gate design for middle school if you also teach a lower grade band and want the language to line up.
- Walk through prediction then check in daily lessons for the daily-lesson version of the pattern.
- See curriculum gates that teach thinking for the pillar frame the cluster hangs on.
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.