Deliberate under-specification

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

When to use this: While you practise
Learning impact: Unlocks better later work
Learning evidence: Plausible, lightly tested
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You want to train prompt skill by studying failure on purpose.
Vague asks keep producing vaguely wrong shapes.
You need to see which missing constraint caused the mess.

What to ask AI

Type something like: “Round 1: deliberately vague: [thin ask]. After your answer, I will diagnose missing constraints. Round 2 I respecify.”

Then ask it to summarise what under-specification taught you, without flattery.

Keep the hard part on your side: you diagnose what was missing.

How it helps

You learn constraints faster by watching what goes wrong without them.

Deliberate under-specification is a drill: fail, name the missing field, respecify.

It is a prompt-skill exercise, not a core content-learning method.

How to run it

Ask something intentionally thin.
Read the failure and list missing constraints (audience, length, level, success check, forbidden moves).
Respecify and compare.
Update your templates so the same miss doesn’t recur.

Example exchange

You

I’m going to ask something deliberately vague on purpose: “Help with my essay.” Just answer that, and then wait.

Sure, here’s a generic five-paragraph essay about “your topic.”

You

That was useless because I never named the subject, thesis, word limit, citation style, audience, or what I’d already written. I’ll respecify those next. What did the vague ask teach me? Don’t flatter.

The vague ask invited a generic product. Your diagnosis list is the real template. Round two will only be as good as those fields.

Copyable prompt

First I’ll ask something deliberately thin: [thin ask].
After you answer, I’ll list the constraints I left out, then respecify and try again.
Afterwards, summarise what the vague ask taught me, don’t flatter.

The Tell

Here is how you know this method has flipped: you never update templates after failures.

Then you are collecting bad outputs instead of learning constraints.

Write the missing fields into a reusable card after each failed vague ask.

Principle evidence

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

The underlying learning idea is rated speculative. Diagnosing missing constraints from failed output is a plausible skill-acquisition drill, adjacent to problem formulation, but thin as named educational evidence. Useful as a prompt-skill exercise, not a core learning method.

AI delivery evidence

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

Any chatbot can answer a vague ask poorly. That is easy. No delivery trial is claimed for deliberate under-specification as pedagogy.