Methods · Assessment & diagnosis
H1
Diagnostic pre-test
Sit a short test before you study, to find out what you actually need to study.
When to use it
- You are about to start a revision block and do not know what the block is actually for.
- You are returning to a subject after a break and your sense of what you still know feels unreliable.
- You keep restarting at chapter one and running out of time before the hard parts.
- You want evidence before you commit hours to a topic or pay for a course.
What to ask AI
Type something like: “I'm about to study elasticity. Before I do, give me 12 questions spanning the whole topic. Don't help me.”
After you answer, ask for a priority list of sub-topics to study, and which ones you can safely skip.
Refuse hints during the test itself. The cold attempt is the point.
Keep the hard part on your side: you answer first; the plan comes second.
How it helps
Two separate effects matter here, and both are well attested.
First, a pre-test makes revision targeted. Most revision is uniform: start at chapter one, work through, run out of time before the end.
A cold quiz can show that chapters one to three are secure and chapter five is a hole. That changes what a fortnight of work is worth.
Second, attempting questions you cannot yet answer can improve the learning that follows, the pretesting effect.
Failing to retrieve seems to prime you to notice the answer when it arrives. Getting it wrong is part of the mechanism.
A pre-test you fail badly is often working as intended. The output you want is an ordered study plan, not a mark to feel bad about.
How to run it
- Name the topic narrowly enough that twelve questions can span it, a module unit, not “all of biology.”
- Ask for twelve questions from basics to hard edges. Forbid help until you finish.
- Answer cold. Write something for every item, even guesses. Blank lines give no diagnosis.
- Only then ask which sub-topics to prioritise and which you can skip.
- Build this week's plan from that list. If the plan matches the order you always use, the diagnosis did not land.
- Optional: ask for three items designed to catch misconceptions rather than gaps, and have those labelled afterwards.
Example exchange
I'm about to revise market structures. Can you quiz me cold first, basics through the hard edges, then tell me what to prioritise?
Twelve questions spanning the topic. Answer all of them before I mark anything. Ready?
Ready. I wrote something for each, including guesses on oligopoly.
Prioritise oligopoly models and contestability, your answers fall apart there. Demand-curve shifts look secure; skip those for now. Next session: three oligopoly problems only.
Copyable prompt
I'm about to study [TOPIC]. Before I do, give me 12 questions spanning the whole
topic (from the basics to the hardest parts) that will show you where my gaps
are. Don't help me. After my answers, don't just mark them: tell me which
specific sub-topics I should spend my time on, in priority order, and which I can
safely skip.The Tell
You did the pre-test and then revised in the same order you were always going to.
The plan is the point of the session, not the score.
If your study list looks identical to last week's chapter order, the diagnosis never landed.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
The underlying learning idea is rated strong. Diagnostic pre-tests sit in formative assessment and placement testing: locate the starting point before instruction (Black and Wiliam; mastery and diagnostic traditions). Separately, pretesting can improve later learning of the tested material. Principle strength is high for diagnosis, with a plausible pretesting benefit on top.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
AI item quality and score interpretation are speculative. Treat the output as a study plan to verify against your real syllabus and past papers, not as an official placement score.