Methods · Creation, transfer & long horizon
F1
Creative extension
Take something you've just learned and apply it somewhere nobody asked you to, which is the cheapest available test of whether you understood it.
When to use it
- You finished the problem set and still wonder whether you could spot this idea outside the worksheet.
- An exam often drops the idea into a setting you have never seen.
- You can run the procedure, but freeze when someone asks when you would use it.
What to ask AI
Type something like: “I've just learned [concept]. I think it applies to [something I chose]. Don't tell me if I'm right yet, ask me why I think so, then judge my reasoning.”
Name the target yourself before you ask. Choosing the situation is the practice.
Keep the hard part on your side: you pick the situation, state the fit, and work it through before the AI critiques.
How it helps
Most exercises are built so the method you just learned is the answer. That helps you learn the steps. It hides whether you can recognise the situation on your own.
Creative extension flips the job. You choose where to apply the idea, from your own life or interests, then you try to make the fit hold.
Picking a target forces you to say what kind of situation the idea is about. Getting the fit wrong is often more informative than getting it right.
This differs from far-transfer problems, where someone else supplies the unfamiliar setting. Here you invent the setting yourself. That is harder, and a clearer test of whether you can characterise the idea.
How to run it
- Pick the target yourself. Do not ask the AI for “applications of X” first, that removes the exercise.
- Choose from your own life or another module you care about, not from a generic list.
- Write why you think the idea applies before you check. That reasoning is what gets assessed.
- Apply it properly. Work the idea through the new situation instead of saying “it’s kind of like that.”
- Ask where the fit breaks. Real extensions are partial; the break shows the idea’s boundaries.
- Decide what you learned: something about the idea, something about the target, or both.
Example exchange
I've just learned regression to the mean. I think it applies to our school's jump in scores after a new tutoring programme. Don't tell me if I'm right yet. Ask me why I think so first.
Why do you think regression to the mean fits that jump? What would have to be true about the first measurement for the story to be statistical rather than causal?
The school was picked because scores were unusually low that year, so the next year might look better even with no real programme effect.
That risk is real. Where does the analogy break? What evidence would you need before claiming the programme caused the jump?
Copyable prompt
I've just learned [CONCEPT]. I think it applies to [thing I chose]. Don't tell me
if I'm right yet, ask me why I think so, then tell me whether my reasoning
holds.The Tell
If every extension you try “works perfectly,” the method has flipped. Real applications are partial, and the useful learning sits where the fit fails.
If the AI chose the target for you, you practised recognising a finished suggestion, not characterising the idea.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
The underlying learning idea is rated moderate. Creative extension is a transfer and generation test: you apply an idea beyond the assigned domain. It sits near far transfer and generative learning. Generating your own applications tends to reveal understanding better than worksheet items alone, with the usual caveat that transfer is hard and context-sensitive.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
Whether an AI tutor reliably judges the honesty of your extension is a separate, thinner claim. Treat AI critique of your chosen target as feedback to check, not as proof the session produced real transfer.