Repair adjusts an element.
Replacement changes the approach.
The 20-minute workflow from A1 identifies what to change. This article covers a prior decision that must be made first: how to change it. The distinction between repair and replacement is not primarily about effort — it is about diagnosis.
Repair is the response to a calibration failure. The right approach was taken but the execution missed the target: the explanation was correct but too abstract, the activity was appropriate but ran too long, the assessment was well-designed but asked the wrong question at the wrong moment. The core approach is sound — it just needs adjustment.
Replacement is the response to a structural failure. The approach itself was wrong: the explanation required prior knowledge students didn't have, the activity produced the wrong cognitive process by design, the sequence made the concept harder to understand rather than easier. No amount of calibration fixes a structural failure, because the problem is in the design, not the execution.
The 2-minute diagnosis:
three questions.
The repair-or-replace decision takes 2 minutes when you apply three diagnostic questions in sequence. Each question has a binary output. The sequence of answers routes you to the correct response.
If a version of the explanation worked — even partially, even for a subset of the class — then the approach itself is viable. The issue is calibration: you used the wrong entry point, the wrong context, or moved too fast. A repaired version of the same approach, better calibrated, will produce the understanding you were targeting.
Some approaches contain the seed of their own failure. An explanation of osmosis using arrows pointing from high to low concentration structurally invites students to attribute agency to the direction — they will say the concentrated side 'pulls'. The explanation created the misconception it was trying to prevent. Repair cannot fix this. A replacement that removes directionality is necessary.
A third failure type that is neither repair nor replacement: the lesson itself was sound, but it was delivered in the wrong order. Students needed to understand X before Y made sense, and X hadn't been taught. Repairing or replacing the Y lesson doesn't fix this — addressing X does.
Repair: adjust the element.
Replace: swap the approach.
Replacing when you should
have repaired.
Replacement feels more productive than repair. Rewriting an explanation from scratch feels like real work. Adjusting an existing explanation by adding one counter-example feels minor. This psychological bias — toward action that is visible over action that is targeted — is the most common mistake in lesson iteration.
The cost of unnecessary replacement is high: it takes significantly more time, it loses the institutional knowledge embedded in the original lesson, and it introduces new failure modes. A repaired lesson has a narrow, testable change that will either fix the problem or generate a clear next signal. A replaced lesson has many new variables and makes it hard to tell whether the original problem was solved or simply obscured.
When the manual 20 minutes
becomes 3 with AI.
A3 covers the AI-powered version of the iteration workflow. The repair-or-replace decision in A2 is still yours — AI cannot make that judgment. But once you've made it, AI can execute Step 3 (the writing) in under 3 minutes for both repair and replacement tasks, with a specific prompt structure that produces targeted output rather than generic rewriting.