Constraint specification

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

When to use this: Before you start studying
Learning impact: Unlocks better later work
Learning evidence: Plausible, lightly tested
Where you are with the topic: Building understanding
How often to use it: Build and keep as a habit

When to use it

Every reply feels “kind of useful” and never usable as-is.
You forgot to specify level, length, audience, or format.
You are about to start a session and want the output shape fixed first.

What to ask AI

Type a constraint template before the task: format, level, length, audience, must include, must avoid, success check.

Ask it to refuse to answer until those fields are filled.

Keep the hard part on your side: you define success, not “write something good.”

How it helps

Vague asks get vague essays.

Constraint specification is session setup: say the shape of a usable answer before you ask for content.

It is an enabler more than a standalone learning method, and it prevents a lot of wasted regenerations.

How to run it

Fill the template before any content question.
Include a success check you could mark pass/fail.
Run the task under those constraints.
If the output misses a constraint, point to the field, don’t vaguely ask for “better.”
Save a filled template for recurring assignment types.

Example exchange

You

Before you write anything, lock these constraints: an 8-bullet revision checklist, first-year level, under 120 words, for me before a quiz. Include one common mistake. Don’t introduce new topics. Success means I can run the checklist without you. Topic: photosynthesis light reactions.

Here’s a checklist that stays inside those fields. Read it once, then try running it from memory.

You

You slipped in Calvin cycle detail. That breaks “must avoid new topics.” Trim to light reactions only.

You’re right, corrected. Naming the violated field is faster than asking me to “make it better.”

Copyable prompt

Before you write content, lock these constraints:
Format: [ ]
Level: [ ]
Length: [ ]
Audience: [ ]
Must include: [ ]
Must avoid: [ ]
Success check: [ ]
Task: [ ]

The Tell

Here is how you know this method has flipped: every reply feels kind of useful and never usable as-is.

You are paying the tax of underspecification over and over.

Fill the constraint card before the next ask.

Principle evidence

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

The underlying learning idea is rated speculative-to-weak for learning outcomes. Task clarification (audience, format, success criteria) improves help quality in tutoring and writing instruction, but constraint checklists are mostly enablers. Strong as session setup; thin as a named learning method.

AI delivery evidence

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

Constraint-following varies by model. No delivery trial is claimed. The learning-adjacent benefit is that you practised stating success criteria clearly.