Working with the tool itself

AI will happily think for you. These methods keep the hard part on your side.

Verify claims. Ask what it does not know. Refuse the full answer. Audit who owned the last session.

Treat this family as standing hygiene for every subject, not an optional advanced track.

The verification drill

Tell, reliably and quickly, when the thing you're learning from is wrong, and build that into a habit rather than a worry.

Learning impact Strong change Learning evidence Credible with caveats When to use this When checking yourself How often Worth repeating often

Deliberate disagreement

Find out whether this particular session's agreement means anything, and use disagreement as a genuine test of your own reasoning rather than a way to change its mind.

Learning impact Strong change Learning evidence Credible with caveats When to use this When checking yourself How often Worth repeating often

The refusal-to-answer request

Instruct it, up front, that it is not allowed to give you answers, only questions.

Learning impact Strong change Learning evidence Credible with caveats When to use this While you practise How often Build and keep as a habit

K17

Knowing when not to use it

Make the decision deliberately, so that using it is a choice rather than a default.

Learning impact Strong change Learning evidence Credible with caveats When to use this When something feels wrong How often Build and keep as a habit

Source triangulation

Confirm a claim across three independent sources, so one confident paragraph stops being enough.

Learning impact Strong change Learning evidence Credible with caveats When to use this When checking yourself How often Worth repeating often

K20

Session ownership audit

Mark every paragraph 'I generated' or 'AI generated' and count the ratio, so you can't argue with a number about who did the thinking.

Learning impact Strong change Learning evidence Credible with caveats When to use this At the end of a session How often Worth repeating often

K16

Transcript review

Measure it, by reading your own transcript back.

Learning impact Reliable gain Learning evidence Credible with caveats When to use this At the end of a session How often Worth repeating often

Hallucination hunting

Find planted errors in AI output: training the verification reflex before you depend on the tool for real work.

Learning impact Reliable gain Learning evidence Credible with caveats When to use this While you practise How often Worth repeating often

Citation chasing

Follow every reference to its source before believing it: killing trust in fluent fake footnotes.

Learning impact Reliable gain Learning evidence Credible with caveats When to use this When checking yourself How often Worth repeating often

Bias probing

Ask the same question several ways and compare the answers, which shows you how much of what you got back was about the subject and how much was about how you asked.

Learning impact Clear help, modest gain Learning evidence Credible with caveats When to use this When checking yourself How often Worth repeating often

K14

Context provisioning

Give it your actual syllabus, notes and marking criteria, so it's answering about your course rather than about the subject in general.

Learning impact Clear help, modest gain Learning evidence Credible with caveats When to use this Before you start studying How often Build and keep as a habit

K15

Scope discipline

Keep a session to one question rather than drifting, so you leave with an answer, not a tour of the universe.

Learning impact Clear help, modest gain Learning evidence Credible with caveats When to use this Before you start studying How often Build and keep as a habit

K22

External working-memory offload

Use the AI as queryable external memory for a long project while still reconstructing key parts unaided: offload without amnesia.

Learning impact Clear help, modest gain Learning evidence Credible with caveats When to use this While you practise How often Build and keep as a habit

Asking what it doesn't know

Require a confidence level on every non-obvious claim, then spend your checking on the low ones.

Learning impact Clear help, modest gain Learning evidence Thin or mixed evidence When to use this When checking yourself How often Worth repeating often

Prompt iteration as thinking

Treat refining the question as the intellectual work, so each rewrite sharpens what you actually need to know, not just how you boss the model.

Learning impact Clear help, modest gain Learning evidence Thin or mixed evidence When to use this While you practise How often Worth repeating often

K10

Confidence-language auditing

Notice where the AI sounds certain without grounds, separating rhetorical confidence from evidential support.

Learning impact Clear help, modest gain Learning evidence Thin or mixed evidence When to use this When checking yourself How often Worth repeating often

K12

Output-diff critique

Compare two AI drafts and say which is better and why: training criteria, not passive acceptance of the latest regenerate.

Learning impact Clear help, modest gain Learning evidence Thin or mixed evidence When to use this When checking yourself How often Worth repeating often

K11

Cross-model comparison

Put the same question to different systems and reconcile: using disagreement as a signal to verify, not as a coin flip.

Learning impact Clear help, modest gain Learning evidence Plausible, lightly tested When to use this When checking yourself How often Worth repeating often

K13

Constraint specification

State format, level, length, and audience precisely, so outputs match the job instead of a generic essay shape.

Learning impact Unlocks better later work Learning evidence Plausible, lightly tested When to use this Before you start studying How often Build and keep as a habit

K18

Deliberate under-specification

Give a deliberately incomplete prompt, then diagnose what was missing: training prompt skill by studying failure.

Learning impact Unlocks better later work Learning evidence Plausible, lightly tested When to use this While you practise How often Worth repeating often

K19

Adversarial prompting

Try to make the AI produce a known wrong answer, then analyse why it did or didn't: mapping failure modes instead of assuming safety.

Learning impact Unlocks better later work Learning evidence Plausible, lightly tested When to use this When checking yourself How often Worth repeating often