Methods · Generative practice & retrieval
B18
Exam-condition simulation
Sit a timed, closed-book paper in the real format so exam nerves and timing stop being a surprise.
When to use it
- The real exam is soon and your practice is still open-book and untimed.
- You know content in calm revision and blank under clock pressure.
- You want a paper that matches section structure, not a random quiz.
- You need one full dress rehearsal, not another content review.
What to ask AI
Type something like: “Generate a practice paper matching [exam format/length]. Closed book. I will time myself. Do not give answers or hints until I submit the whole paper.”
Paste a past-paper structure or rubric if you have one.
Forbid mid-paper hints, delayed feedback for the whole script.
How it helps
Exam performance is content plus timing plus retrieval under constraint.
Open-book calm practice does not train the constrained version.
A simulation makes timing and blankness visible while you can still fix them.
Marking afterward should teach patterns, not rewrite your script into a model answer you never earned.
How to run it
- Specify format, duration, and allowed materials (usually none).
- Generate the paper and print or move it to a distraction-free view.
- Start a timer; answer without AI until time ends.
- Submit the whole script for marking and pattern notes.
- Schedule targeted practice on weak sections, then simulate again later.
Example exchange
Can you run an A-level maths mock: 60 minutes, calculator allowed, two short and two long questions like our last paper? Closed book, and no hints until I submit everything.
Paper ready, four questions. I will stay silent until you send the full script or call time.
Time. Script attached.
Timing pattern: question 3 took about 28 minutes. Content: you missed chain-rule selection on question 2. Want a 20-minute mini-sim next on mixed selection, still closed book?
Copyable prompt
Generate a practice exam for [COURSE/LEVEL] matching this format: [STRUCTURE/TIME].
Closed book. I will time myself. Do NOT hint or mark until I submit the full
script. Afterward, comment on timing patterns and recurring errors, not a full
model paper rewrite.
The Tell
You will know the method has flipped when you peek for a hint at minute twenty, or generate the paper and study the mark scheme before sitting it.
Then it was content review in costume, not a simulation.
Seal the chat until the timer ends, and mark only afterward.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
Practice under conditions that resemble the test improves transfer to the test. Timing and closed-book constraints are part of the skill, not optional flavour.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
AI papers may not match real difficulty or mark schemes. Use official past papers when stakes are high; use AI sims as extra rehearsals, not as the only standard.