Fallacy identification

Find the flaw in arguments built to contain one, naming the failure mode precisely enough that you don't just shout "fallacy" at everything you dislike.

When to use this: While you practise
Learning impact: Clear help, modest gain
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You have twenty minutes and want drills that sharpen spotting broken moves.
You keep hearing “fallacy!” in debates and want precision, not a weaponised label.
Homework is done and you want one honest set of planted arguments before bed.

What to ask AI

Type something like: “Give me 6 short arguments. Some contain exactly one classic flaw; some are clean.”

For each, say clean or flawed, name the flaw, quote the broken move, then offer a repair.

Ask it not to reveal scores until you submit each attempt.

How it helps

Fallacy lists are easy to memorise and easy to misuse as debate weapons.

The learning value is narrower: given an argument, locate the broken move and say what repair would look like.

AI can plant a single intended flaw, or plant none, so you do not develop false positives.

Labels are hypotheses. If you cannot say how the conclusion fails to follow, you do not have an identification yet.

How to run it

Warm up on planted items: “exactly one named flaw” or “clean.”
For each: reconstruct briefly, name the flaw, and quote the locus.
Propose a repair (add a premise, restrict the conclusion, define a term).
Graduate to messier real texts where flaws overlap.
Ban drive-by labelling without a locus and a repair.

Example exchange

You

Give me one short argument that may or may not contain a classic flaw, don’t spoil which.

“Every top athlete stretches before competing, so stretching must cause elite performance.”

You

Flawed, it treats correlation as causation. To repair it I’d need evidence that stretching changes performance, not just that it co-occurs.

Correct locus. Ready for another item under the same rules?

Copyable prompt

Give me 6 short arguments. Some contain exactly one classic flaw; some are
clean. For each I will: (1) say clean or flawed, (2) if flawed, name it and
quote the broken move, (3) offer a repair. Score me; don't reveal until I
submit each.

The Tell

If you can list twenty fallacy names and cannot point to the exact sentence where this argument fails, the method flipped into trivia.

If you never offer a repair, you are collecting labels, not understanding failures.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Fallacy identification is common critical-thinking curriculum content. Recognition skill can improve with practice; far transfer to messy real discourse is often weaker in the literature.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

AI planting and detection of fallacies is uneven. Verify labels against your course list. Delivery is speculative.