Teaching for Understanding With AI, Step by Step | Cerebrote
Methodologies for Teaching Humans (With AI as Co-Pilot)
In the previous module [link] we reached an uncomfortable conclusion: our students’ brains have changed. Module 2 of the DEXIA method [link] had to face the obvious follow-up: so what methodology do I use to teach that brain?
Short answer: active methodologies. Long answer — the good one — is this post.
What is an active methodology? Put the body to work
Every active methodology has serious research behind it, but they all share one idea, which in class we summarize like this: put the body to work, so the student discovers they hold knowledge that is actually good for something. Less student-as-spectator, more student-as-protagonist. Knowledge that never gets used evaporates; knowledge put into practice stays.
There are many active methodologies and several work beautifully. But if you’ll let us recommend just one — and by now we know each other a little — study this one: Teaching for Understanding (TfU), developed at Harvard’s Project Zero. It’s the framework of choice in schools that have thought their pedagogy through carefully. And it works.
TfU explained on one hand
At Cerebrote we teach it on an open hand, finger by finger, so you never forget it:
- Generative topics (thumb). Don’t start with “the topic”: start by connecting the topic to today’s reality. What do you need to teach? Why should anyone understand it? What is understanding it for?
- Understanding goals (index). Define what you want the student to genuinely understand — not what you want to “cover in class.” “The student interprets texts in English” is a goal. “Covering the simple past” is not.
- Performances of understanding (middle). The visible actions through which the student demonstrates — by doing — that they understood. This is where the body works.
- Ongoing assessment (ring). You don’t assess at the end: you give points and progress throughout the whole process, with constant feedback.
- Following everyone (pinky). The fifth finger closes the fist: you are checking on all of them, always. Nobody stays invisible.
Simple? Yes. Easy? No. Which is why the rest of the module was about getting it moving with help from AI.
AI as co-designer: the workflow we used in class
Here comes the heart of the module — and the cleanest application of the DEXIA thesis: AI behind the teacher. The workflow has three steps:
Step 1. Feed your assistant the good source. Don’t ask the AI to “know” TfU: hand it the original documents from Project Zero and its researchers — the papers and books any of us would use to prepare the class. Load them into your AI reading assistant — NotebookLM* or any tool that accepts file uploads — so that every answer comes from that source, not from the model’s imagination. Same rule as Module 1: evidence first, tool second.
Step 2. Hand it your real context. The magic is not in asking for “a history lesson”: it’s in the details only you know. In the live demo, a secondary school teacher from Mexico lent us his real case: the ethics, nature and societies learning field within Mexico’s Nueva Escuela Mexicana model, a two-week project, 50-minute sessions, starting from zero prior knowledge. With that level of detail, the AI simply cannot answer with a generic lesson.
Step 3. Ask for a design under the framework, not a loose lesson. This is the prompt template we used, distilled from the live demonstration; adapt it to your subject:
Based on the Teaching for Understanding documents I uploaded, design a learning experience under that framework for [grade/level], in the subject [name], within the [your country’s or institution’s] educational model. This is the first session of a [duration] project, with [minutes]-long meetings. Students have no prior knowledge of the topic, so focus on the exploration stage and on hooking them with the generative topic. Structure your answer with: generative topic, understanding goals, performances of understanding, and a session plan that includes a thinking routine.
The result? Live, the AI returned an experience titled “America, One Heart of Many Voices” for fifth grade: a generative topic on Indigenous peoples as living cultures (not dead history), clear understanding goals, and a session opener built on the “See, Think, Wonder” thinking routine in front of a collage — a totem in Alaska, a Mayan pyramid, an Andean textile. Everything aligned with the framework, everything ready for the teacher to adjust with what only they know about their group.
Look closely at the casting, because the whole method lives there: the AI brought structure and speed; Harvard brought the framework; the context, the judgment and the final call came from the teacher. That’s what AI as co-pilot looks like. You never let go of the wheel.
What we learned doing it
House honesty, in two parts. First: TfU is not the only valid methodology — at university level, for instance, others work just as well or better depending on the case — and forcing a single framework on everyone would betray the framework itself. We recommend it as a starting point because it’s solid, well documented, and quick to learn; not because it’s religion. Second: the step teachers struggled with most was not the AI — it was the thumb. Turning “the topic I have to cover” into a genuine generative topic is intellectual work that belongs to the teacher, and no tool will do it for you. Good news: it’s trainable, and it gets better with every class you design.
To close
This module was facilitated by Luis Fernando Botero Mendoza, M.Ed. (Colombia), academic director of the program, and it works the second side of the DEXIA Triangle: Pedagogy. Because a tool without a methodology is a toy, and a methodology without humanity is a form to fill out.
Next in the series: the AI toolkit every teacher should know, with the judgment (and healthy skepticism) of Prof. Gustavo Coronel from Argentina. All living resources for this module are at cerebrote.com/dexia/.
Latin America, world power of humanity.