AI for StudentsLesson 28 of 73

Contents

AI for Students

Lesson 28 of 73. Sign in to keep your place.

Module 3 · Study that raises the difficulty

Proving it to yourself

8 min read

The gap between what you predict and what you get050100212Week of termMark on the unaided check (%)—— Predicted before marking– – ActualOne student’s calibration column across a term: the predicted mark written down before marking,against the mark. The average gap, here about fourteen points narrowing to five, is the most usefulsingle number anybody can hold about themselves, because it says how much to discount your own senseof being ready. Students leaning hardest on an assistant tend to over-predict most, which is thefluency trap showing up as arithmetic.
One student's calibration column across a term: the predicted mark written down before marking, against the mark. The average gap, here about fourteen points narrowing to five, is the most useful single number anybody can hold about themselves, because it says how much to discount your own sense of being ready. Students leaning hardest on an assistant tend to over-predict most, which is the fluency trap showing up as arithmetic.

Figure

The only test that counts

Every technique in this module points at one question, and it is worth stating plainly, because it is the thing the tools make easy to avoid:

With everything closed, can I do it?

Not recognise it. Not follow it. Not feel that I could. Do it (produce the derivation, write the paragraph, name the mechanism, solve the problem) on blank paper, with no window open and nobody to ask.

Everything else in study is instrumental to that. And it is measurable, cheaply, at any moment.

The unaided check

The procedure takes ten minutes and belongs at the end of every topic, not the end of the term.

  1. Close everything. Phone in another room, not face down on the desk.
  2. Set a timer for the length the real task would take.
  3. Do the thing. A past-paper question, a paragraph of the essay, a derivation, an explanation.
  4. Only then open the source and mark it honestly against the mark scheme if there is one.
  5. Write down the specific failure, not a grade. Not weak on kinetics but could not remember whether the inhibitor changes Km or Vmax and guessed.

Step five is where the value is. A grade tells you how you did. A specific failure tells you what to do on Wednesday.

The calibration column

Before you check, predict your score. Write the number down.

Over a term this produces the most useful single statistic you can have about yourself: the average gap between predicted and actual. If you consistently over-predict by 15 per cent, you now know that your sense of preparedness runs 15 per cent hot, and you can correct for it, which is worth more than any individual mark, because it tells you when to stop revising a topic and when you were wrong to.

Students using AI assistants heavily tend to over-predict more, for the reason in the fluency lesson. Measuring your own gap is how you find out whether that describes you, rather than assuming it does or does not.

Building a portfolio of things you can do

Keep a single file, appended to after each unaided check. One line each: the date, the task, and whether you completed it unaided.

12 Oct: derived the quadratic formula by completing the square, unaided, 6 min.
14 Oct: could not reproduce the proof that root 2 is irrational without a hint. Retry 17 Oct.
17 Oct: reproduced it unaided.

By the end of a module this file is a list of demonstrated capabilities with dates. It is the answer to what have I actually learned, and it is immune to the feeling of competence, because it records outcomes rather than impressions.

It is also, if you ever need one, an extraordinarily good piece of evidence in an academic-integrity conversation: a dated record of you building capability under your own steam. Module 7 returns to this.

The failure this catches early

The pattern to fear is: an excellent term of assisted work, high coursework marks, a comfortable sense of understanding, and then a closed exam where none of it is available. By the time that gap appears in an exam hall it is far too late to close.

The unaided check finds it in week three, when it costs an afternoon to fix. It is a smoke alarm, and like a smoke alarm its value is entirely in going off before the thing you were worried about.

Raising the difficulty deliberately

Once a check is comfortable, make it harder in the direction the real test will:

  • No notes if you had notes.
  • Timed if it was untimed.
  • Handwritten if the exam is handwritten. Writing by hand is slower and more effortful than typing, and if you have not practised it, you will run out of time on material you knew.
  • Cold: a topic from three weeks ago, unannounced, rather than the one you just revised.
  • Explained aloud to somebody, if any part of your assessment involves speaking.

Each of these narrows the gap between the conditions you practise in and the conditions you will be assessed in, which is the whole subject of Module 8.

The honest closing note

This is the least enjoyable lesson in the module to act on. Nothing about the unaided check is pleasant: you sit down knowing you might not be able to do it, and quite often you cannot. The assisted alternative is right there, and it feels better, and it produces a document.

But the question at the top of this lesson is the one the world will eventually ask, in an exam hall or a job or the first time something goes wrong and nobody is available. The students who came through this era able to do things are the ones who kept asking it of themselves, early, while it was still cheap to answer badly.

Every lesson here is open. Your progress is recorded once you pass the AI at Work exam.

Check your understanding

Why is writing down a predicted score before marking an unaided check more useful than the mark itself?

Pick the one you would defend. Your answer stays private.

Next lesson · 9 minMaths, and the arithmetic problem

Stuck on something?Ask Asha, your AI tutor

How to cite: Addaly (2026) ‘Proving it to yourself’, in AI for Students. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-students/proving-it-to-yourself (Accessed: 13 October 2026).