Skip to content
How to ThinkIn the Age of AI
Rulev2RL-198

Write a specific pre-correction procedure for each known…

For each identified bias in your profile, write a specific pre-correction question or procedure to execute before acting on judgments in that domain.

Why This Is a Rule

Knowing you're biased doesn't fix the bias. Research on cognitive debiasing consistently shows that awareness alone has minimal effect on judgment quality — you can know about the planning fallacy and still underestimate every project. What works is procedural debiasing: specific, pre-committed actions that intervene at the moment of judgment, before the bias has a chance to distort the output.

This rule bridges the gap between bias identification and operational improvement. For each bias in your profile, you write a concrete pre-correction: a question to ask, a calculation to perform, or a step to execute before acting on your judgment. The pre-correction is externalized and domain-specific, so it fires automatically at the right moment rather than relying on in-the-moment self-awareness (which the bias itself compromises).

When This Fires

  • After completing a bias journal and identifying your personal bias profile
  • When you know you're biased in a domain but keep making the same errors
  • Setting up decision protocols for high-stakes recurring judgments
  • Any time bias awareness isn't translating into improved judgment

Common Failure Mode

Writing generic pre-corrections that are too vague to execute. "Consider whether I'm being overconfident" is awareness, not a procedure. "Multiply my estimate by 1.4 before stating it" is a procedure. The pre-correction must be specific enough that you can execute it mechanically, without needing the very self-awareness that the bias undermines.

The Protocol

For each identified bias: (1) Name the domain it appears in (e.g., "timeline estimation for engineering projects"). (2) State the direction and magnitude of the error (e.g., "I underestimate by ~40% on average"). (3) Write a pre-correction that is specific, mechanical, and executable: a question ("What would a pessimist estimate?"), a formula ("Multiply by 1.4"), or a procedure ("Ask three people for independent estimates and take the median"). (4) Store these pre-corrections where you'll encounter them at the moment of judgment — not in a document you'll forget.