Narrate the thinking in tomorrow's worked example

Narrate the thinking in tomorrow's worked example

Worked examples that work share one feature: the teacher said the quiet bit out loud. The decision to use one method over another. The check that catches the most common error. The "wait, that does not look right" moment. Most of us know this. Writing the script for it on a Tuesday night is the bit that does not happen.

The prompt

You are an experienced [SUBJECT] teacher in an Australian [PRIMARY / SECONDARY / TERTIARY] classroom. You model worked examples by narrating the cognitive moves, not just the steps on the board.

What I am modelling tomorrow: [DESCRIBE THE EXAMPLE IN ONE OR TWO SENTENCES, e.g. "Solving 3(x + 4) = 27 by expanding then isolating x"]
Year level: [YEAR LEVEL]
Subject: [SUBJECT]
Curriculum link: [ACARA v9.0 content descriptor, VCE study design dot point, or your own success criteria, copied verbatim]
Prior knowledge I am assuming: [ONE OR TWO SENTENCES, e.g. "Students have solved two-step equations like 2x + 5 = 15. They have NOT yet expanded brackets."]
Time for the modelling phase: [MINUTES, default 6 to 8]
Most common student error on this concept in my experience: [ONE SENTENCE, or "I do not know yet, name the two most likely from the literature"]

Produce the worked example as a two-column script.

Left column: WHAT I WRITE ON THE BOARD. Each line is one move (one expansion, one rearrangement, one substitution). Use the symbols and notation students will copy down. Do not narrate in this column.

Right column: WHAT I SAY OUT LOUD. For each board move, one to three short sentences that name:
1. The decision (why this move, not another).
2. The self-check (what I look at on the board before I commit to the next line).
3. The pitfall (the common error this move risks, named in plain English).

Constraints:
a. No more than 6 to 8 total board moves. If the example needs more, the example is too big for this modelling slot.
b. The narration must include at least one "wait, let me check" moment where I deliberately stop and look back at a previous line. Real thinking includes the recheck. Pretend modelling never does.
c. The narration must name the most common student error (the one I gave you above) at the line where it is most likely to happen. Phrase it as "this is the bit where students often..." not as a warning lecture.

Then, in a separate "Pause to ask" block, give me exactly TWO points in the modelling where I should stop and put a question to the class before I continue. For each, give me:
- The question, in the exact words I should use.
- What I am listening for in their answer.
- What I do if no hands go up after five seconds.

Then, "What to leave on the board after I sit down": one sentence on the part of the worked example students should see for the rest of the lesson, so they can refer back without putting their hand up.

Finally, "Sanity check": one sentence on what would make you, the model, wrong about the pitfall you named. If the pitfall is genuinely more common in textbooks than in real Australian classrooms at this year level, say so.

Australian English throughout. Do not use the phrases "deep dive", "unpack", or "scaffold their thinking".

What it does well

Forces the model to separate what the teacher writes from what the teacher says, which is the move that makes the difference between a worked example students can copy and a worked example students can learn from. The "wait, let me check" instruction is the small bit that turns AI-generated modelling scripts from suspiciously smooth into something that resembles how real teachers think out loud, which is what students need to hear if they are going to imitate it.

Where it falls over

The model does not know your board. The script it produces sometimes ignores the physical constraint of writing left to right on a whiteboard while talking to a class who are watching your shoulder block half the working. Rehearse the first two board moves in front of an empty room before you trust the pacing. The prompt also will not save a modelling sequence on a concept the teacher does not understand deeply. If you have to ask the model what the most common pitfall is, the script will sound learned, but the live response to a student question will not.

Try it like this

For Year 8 maths (expanding brackets and solving): the narration that earns its keep is the "wait, let me check" on the expansion line, before the rearrangement starts. Students who watch a teacher actually recheck the distribution before moving on copy that habit faster than they copy any worksheet.

For Year 11 chemistry (balancing equations): set the example as one of those equations that looks balanceable by inspection but is not. The narration surfaces the conservation-of-mass move that inspection skips. Leave the unbalanced version on one half of the board after you sit down.

If you only have ten minutes today, run it on the worked example you have done so often you no longer remember which decision is the tricky one for a Year 8. That is the one where the narration earns its keep, because your tacit knowledge is invisible to students and the model can flush it back out.

The board move is what students copy. The thinking move is what they learn.