Use errors and failures to refine the mental representation, not just the technique

When something goes wrong, ask: "What was wrong with my model?" — not just "What was wrong with my execution?"

Why it works

Most practitioners respond to failure by correcting execution. Expert learners respond by also asking whether the error reveals a flaw in the mental representation — a case the model predicted wrong, a pattern it failed to discriminate, a planning error that the pre-run did not catch. Representation-level error analysis leads to improvements that generalize across instances; execution-level correction fixes only that instance.

How to do it

1. After a significant error or performance failure, reconstruct what your mental model predicted at each step.
2. Identify the point where the prediction diverged from what happened.
3. Ask: "Is this a prediction my representation would always get wrong? What would it need to be different to predict correctly?"

Evidence

Error analysis as a representation-refinement process is the mechanism underlying "deliberate practice" in Ericsson’s framework; experts systematically use failures to update their models rather than just noting the mistake and repeating. Case studies of expert learning confirm representation updating through error. (mechanistic)

Model-level error analysis requires sufficient existing representation to diagnose at that level; novices often cannot distinguish model errors from execution errors and benefit more from execution feedback at early stages.

Sources

• Ericsson & Pool (2016), "Peak" (errors as representation feedback)

Common mistake

Treating all errors as execution failures and practicing harder, when the actual problem is a flawed model that will generate the same category of error no matter how many times the execution is repeated.