Variation tournament

Generate five to eight versions of the same piece, rank them, and articulate the criteria: training evaluation, not just production.

When to use this: While you practise
Learning impact: Reliable 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 fixate on a first version and need distance to judge it.
You can pick a favourite draft but cannot say the criteria.
You want to train taste by ranking, not by seeking praise.

What to ask AI

Type something like: “Brief: [ ]. I will submit 6 variants. Do not rank first. Ask me to rank and state criteria. Then attack my criteria for inconsistency. Finally help me plan a merge of the top 2 without averaging into mush.”

Ask it not to rank before you do.

Keep the hard part on your side: you generate variants and articulate criteria.

How it helps

A variation tournament means generating five to eight versions of the same piece, ranking them, and articulating the criteria.

Taste becomes teachable when criteria are spoken. A favourite you cannot defend is preference without craft.

Merging top versions without averaging into mush is the final skill.

How to run it

Set a clear brief.
Produce five to eight real variants (not tiny tweaks).
Rank them yourself and write the criteria aloud.
Ask the AI to attack inconsistencies in your criteria, not to impose its ranking first.
Plan a merge of the top two that keeps distinct strengths.

Example exchange

You

I wrote a ninety-word abstract brief for my heat-preparedness paper and pasted six variants. Do not rank them. Make me rank them and state my criteria first.

Rank the six from best to worst and list the criteria you are using before I comment.

You

Best to worst: C, F, A, then the others. Criteria in order: claim clarity, then specificity of evidence, then voice.

You ranked F above A, but A has the clearer claim by your first criterion. Reconcile that or revise your criteria, then we can plan a merge of C and F.

Copyable prompt

Brief: [ ].
I will submit 6 variants. Do not rank first. Ask me to rank and state criteria. Then attack my
criteria for inconsistency. Finally help me plan a merge of the top 2 without averaging into mush.

The Tell

A favourite you cannot defend with criteria means you held a preference, not a standard.

If the AI ranked for you first, you practised agreeing with a judge, not building judgement.

Principle evidence

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

The underlying learning idea is rated moderate. Generating variations and articulating criteria trains evaluative judgement (related to divergent production plus critique). Stronger in creative practice traditions than in single experimental lines.

AI delivery evidence

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

AI as ranking coach is speculative. The learning is your criteria articulation under challenge.