Match correction type to error type: procedural for execution, epistemic for knowledge, calibrational for judgment

Why This Is a Rule

This is the correction companion to Classify errors as execution, knowledge, or judgment failures before correcting — each type needs a fundamentally different fix (error classification). Once you've classified the error type, this rule maps each type to its matching correction strategy. The matching matters because mismatched corrections are the primary reason error-correction efforts fail: checklists can't fix knowledge gaps, and information can't fix implementation sloppiness.

Execution errors → Procedural corrections: the person knew what to do but failed to do it reliably. Fix by reducing the gap between intention and action: checklists ensure steps aren't skipped, automation removes human execution from error-prone steps, and environmental forcing functions make the correct action the default (Subtract unwanted affordances before adding desired ones — elimination beats competition for attention). Atul Gawande's surgical checklist reduced complications by 36% — not by teaching surgeons new knowledge but by ensuring known steps were executed consistently.

Knowledge errors → Epistemic corrections: the person didn't have information they needed. Fix by improving information flow: new data sources, expert consultation, better documentation, cross-training. The intervention targets the information gap, not the decision-making process.

Judgment errors → Calibrational corrections: the person had the information but misjudged it. Fix by improving assessment quality: prediction tracking (Decision journal entries need six fields captured before outcomes: date, decision, reasoning, prediction, confidence %, and your current state-458), external review (After irreversible commitments, schedule external reviews with pre-defined criteria — escalation of commitment corrupts self-assessment), pre-mortems, red teams. These interventions don't add information — they improve how existing information is weighed and integrated.

When This Fires

• After classifying an error using Classify errors as execution, knowledge, or judgment failures before correcting — each type needs a fundamentally different fix's three-type taxonomy
• When designing corrections and wanting to ensure the intervention matches the error type
• During post-mortem action item design — verify each action matches its error type
• When previous corrections haven't worked — check if they targeted the wrong error type

Common Failure Mode

Applying procedural corrections to all error types: "Let's add a checklist step for this." If the error was a knowledge gap (you didn't know the API had changed), a checklist of the old API steps makes you more efficiently wrong. If the error was a judgment failure (you underestimated complexity), a checklist doesn't improve your estimation ability — it just documents the same flawed judgment.

The Protocol

(1) Classify the error (Classify errors as execution, knowledge, or judgment failures before correcting — each type needs a fundamentally different fix): execution, knowledge, or judgment? (2) Select the matching correction class: Execution error → procedural: add a checklist step, automate the error-prone step, redesign the environment to make the right action default. Knowledge error → epistemic: identify the missing information source, subscribe to updates, establish expert consultation for this domain, add a "check for changes" step to the process. Judgment error → calibrational: start tracking predictions in this domain (Decision journal entries need six fields captured before outcomes: date, decision, reasoning, prediction, confidence %, and your current state), add external review before consequential judgments (After irreversible commitments, schedule external reviews with pre-defined criteria — escalation of commitment corrupts self-assessment), run pre-mortems for high-stakes decisions. (3) Implement the matched correction. (4) Monitor: does the error recur? If yes → either the classification was wrong (re-diagnose) or the correction within the class needs adjustment. If no → the error-correction match was right.