Error-frequency targeting

Drill the specific mistakes you actually make most, so practice time hits your personal fossilisation points, not a generic grammar book order.

When to use this: While you practise
Learning impact: Strong change
Learning evidence: Credible with caveats
Where you are with the topic: Strengthening memory
How often to use it: Build and keep as a habit

When to use it

Your error log shows the same few patterns again and again while the textbook marches through everything.
You waste practice on grammar you already control and still stumble on two fossilised mistakes.
You finished conversation debriefs and need drills that hit your top patterns only.

What to ask AI

Type something like: “Here are my recent errors with context: [paste]. Cluster into patterns, rank by frequency and how much they hurt understanding. Design a 5-minute daily drill for #1 only. Do not teach all of grammar.”

Paste real lines with context. Do not ask the AI to invent your error frequencies.

Keep the hard part on your side: you produce the drill answers, then re-check in free speech.

How it helps

Class curricula chase coverage. Your mistakes are uneven: a few patterns do most of the damage.

Error-frequency targeting builds a personal list from your chats and writing, ranks what hurts communication, and drills the top one or two hard.

That is deliberate practice on weaknesses, not another random worksheet.

Weekly free production is the honesty check. If the pattern still appears in open chat. The drill is not done.

How to run it

Collect twenty to fifty errors with the surrounding line.
Cluster them into patterns (agreement, aspect, articles, and so on).
Rank by how often they appear and how much they hurt understanding.
Run a five-minute daily drill on the top one or two patterns only.
Once a week, check free conversation or free writing for those patterns.
Retire a pattern only after it stays clean for about two weeks.

Example exchange

You

Here are about ten lines from two weeks of my Spanish chats. Can you cluster the errors, rank them by how often they show up and how much they hurt, then give me a five-minute drill for just the top one?

Biggest cluster: ser/estar with location and emotion, six times, high damage. Gender on articles is second, four times, lower damage. For #1 I'll prompt ten location or emotion sentences; answer aloud; I'll only flag ser/estar.

You

Estoy en la biblioteca. La película es aburrida hoy, wait, emotion?

Emotion → estar. Try that sentence again, then we continue. We're not touching gender today.

Copyable prompt

Here are my recent errors with context: [ ].
Cluster into patterns, rank by frequency and how much they hurt understanding.
Design a 5-minute daily drill for #1 only. Do not teach all of grammar.

The Tell

Here is how you know this method has flipped: you complete random worksheets while the same chat errors stay intact.

Coverage feels responsible and misses your personal fossilisation points.

If drills never return to free speech for a check, you practised exercises, not change.

Principle evidence

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

The underlying learning idea is rated moderate. Targeting your highest-frequency personal errors is deliberate practice on weaknesses, the same logic as error catalogs and focused corrective feedback. It is credible practice design; it is not as heavily replicated as spacing or retrieval as a universal law.

AI delivery evidence

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

AI inventing error-frequency statistics without your data is speculative and often wrong. The method requires your real lines. Use the model to cluster and build a tiny drill, not to guess what you “probably” get wrong.