A18

Worked example with errors

Study a solution that contains a deliberate mistake and find it, which is a far better test of understanding than following a correct one.

When to use this: While you practise
Learning impact: Strong change
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 tonight and need real practice, not more highlighting.
Homework is done, but you still want one honest attempt before bed.
You got a problem wrong and want help aimed at that mistake, not a full re-teach.
You want a dress rehearsal that feels like the real assessment, not a casual chat.

What to ask AI

Type something like: “Give me a worked solution to … with exactly ONE error in it. Make”.

Ask it to generate material for you to attempt, problems, cues, or a schedule, not finished answers.

Paste your attempt and ask for critique of the reasoning, not a replacement essay.

Keep the hard part on your side: you retrieve, attempt, explain, or decide before you ask for help.

How it helps

Reading a correct solution is passive in a way that's hard to notice.

Each line follows from the last, nothing resists you, and you finish with the feeling of understanding.

That feeling is fluency, and it's why worked examples need a partner method.

Put one error in the solution and everything changes.

Now you can't follow, you have to check.

Each line has to be evaluated rather than absorbed, because any of them might be the wrong one.

The reading becomes active without any instruction to read actively, which is why this works where "read critically" doesn't.

There's a second benefit that's specific to error-finding: you learn what errors look like.

How to run it

Ask for one error, unmarked, in an otherwise correct solution. One, not several, several lets you find the easy one and stop.
Ask for it to be plausible. A random arithmetic slip is a proofreading exercise. A conceptual error, a rule misapplied in a way a student genuinely would, is the useful version.
Find it before reading any commentary. Time yourself loosely; if you're past five minutes on a short solution, that's informative.
Say what the error implies, not just where it is. "Line 4 is wrong" is proofreading. "Line 4 treats the constant as if it depended on x, which means whoever wrote this thinks…" is diagnosis.
Ask what the error would have done downstream if it hadn't been caught, would the final answer look obviously wrong, or plausibly right? Errors that produce plausible answers are the dangerous class.
Next: escalate. Once you're finding them quickly, ask for the error to be subtler, or for there to be possibly no error at all, which removes the assumption that one exists.

Example exchange

You

Give me a worked solution for balancing this redox equation with exactly one error in it. Don't tell me where the error is.

Here's a full write-up. Find the error and say what wrong belief would make someone write it that way.

You

You balanced the atoms but left the charges unequal on both sides. Someone might think atom balance is enough.

Right diagnosis. Fix only that step, then stop. Don't rewrite the whole solution.

Copyable prompt

Give me a worked solution to [problem type] with exactly ONE error in it. Make
it a plausible conceptual mistake a student would really make, not a typo. Don't
tell me where it is. I'll find it.

The Tell

Here is how you know this method has flipped: you find the error every time, fast. Ask for subtler ones, comfort here means you're practising a level you've passed.

In that moment the AI (or the schedule, or the story) did the thinking, and you only recognised a finished product.

Tighten the prompt for worked example with errors so you attempt, decide, or retrieve before anything is handed to you.

Principle evidence

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

The underlying learning idea is rated moderate. Erroneous worked examples / finding planted mistakes engage errorful learning and example-based learning variants (Stark, Kopp, Fischer and related; also educational work on learning from errors). Benefits hinge on feedback after detection. Principle moderate; AI-planted errors must be pedagogically realistic, speculative delivery.

AI delivery evidence

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

The learning principle may be moderate, but that does not prove an AI session delivers it well. Treat AI delivery as speculative unless a study of tutoring with this method is named, and none is claimed here.