You have ten minutes with your department head. You want a small pilot, not a purchase order. This guide gives you the sentences that hold up when they push back.
Open with the problem, not the product
Say what your students are actually struggling with. Reading a source and knowing why to trust it. Updating a belief when the evidence changes. Naming what would prove them wrong. Those are teachable, and the department head knows it.
Then say what you want to try. One unit. One class. A pretest, the unit, a posttest. A short write-up at the end.
Do not lead with vocabulary. Lead with what the students will be doing differently.
Name the framework once, plainly
When they ask what the approach is, say this: it is a teaching pattern grounded in active inference, a research program in computational neuroscience that models learners as prediction-updating agents (Class E: Parr, Pezzulo, Friston, 2022). We teach students to make a prediction, meet the evidence, and update. We tag every claim with an evidence class so the reasoning is visible.
That is the whole pitch of the framework. Do not overclaim it. It is a working hypothesis on an attainable path toward General Natural Intelligence, natural not artificial. The evidence is growing, evidence-classed, and tested in the open. You are not selling certainty. You are proposing a careful test.
Anticipate the objections
"Is this an AI tool?" No. Students do not talk to a chatbot. The classroom artifacts are written by humans, and any AI-assisted artifact carries an authorship fence that says so (Class C: authorship fence is a configuration requirement in the pilot rubric). The framework is about how students reason, not about a piece of software.
"How is this different from what we already teach?" The difference is that every claim in the unit is tagged with an evidence class, and students learn to tag their own. That is a small, concrete change to the rubric. It is not a curriculum replacement.
"What if it does not work?" Then the pretest and posttest will show it, and we stop. The pilot is designed to be falsifiable. That is the point of it.
"Do I have to learn the math?" No. The teacher-facing materials use plain language. The math sits underneath for the people who want it. You do not need to teach Bayesian updating as equations to teach it as a habit.
"What does it cost?" For a single-classroom pilot, the cost is your time and a short set of materials. Cohort pricing exists if a department wants more than one classroom, and it is structured so a school is not locked in.
Ask for the smallest thing that would count
Do not ask for a department-wide adoption. Ask for one unit in one class, with a written debrief at the end. If it goes well, the debrief is your evidence for a second classroom. If it does not go well, the debrief tells you what to change or whether to stop.
Say the sentence out loud before the meeting: "I want to run one unit, in one class, this term, and write it up honestly."
Close with what you will send them
Send the department head three short things after the meeting:
- A one-page description of the unit and its evidence tags.
- The pretest and posttest you plan to use.
- The authorship fence language you will use on any AI-assisted materials.
That is enough for a yes or a no. It is also enough for a counterproposal, which is often what you actually want.