Keep a two-week bias journal to build your personal error profile

Why This Is a Rule

Everyone has a unique bias profile — a characteristic pattern of systematic errors. Some people are chronically overconfident. Others underestimate duration but overestimate difficulty. Some are biased toward optimism in domains they love and pessimism in domains they fear. This profile is invisible without data because your biases feel like accurate perception from the inside.

A two-week bias journal provides the data. By recording significant judgments alongside confidence levels and then comparing against outcomes, you build a personal error map: where your judgment is calibrated, where it's biased, and in which direction.

Two weeks is the minimum viable duration — long enough to accumulate enough data points across different judgment types, short enough that you'll actually complete it. The value isn't in the journal itself but in the error categorization that follows: grouping errors by direction (over/under), domain (technical, interpersonal, temporal), and type (probability, magnitude, timing) reveals your specific bias fingerprint.

When This Fires

• You want to improve your judgment but don't know where it's weakest
• You suspect you're biased in certain domains but can't prove it
• You're about to take on a role that requires calibrated judgment (leadership, estimation, risk assessment)
• You've never systematically tracked your judgment accuracy

Common Failure Mode

Recording judgments but skipping the confidence calibration. Without confidence levels, you can't distinguish between being wrong about something you were uncertain about (normal) and being wrong about something you were certain about (a bias signal). The confidence level is what turns an error into information: being 90% confident and wrong is a much stronger signal than being 55% confident and wrong.

The Protocol

For two weeks: (1) At the end of each day, write down 2-3 significant judgments you made (estimates, predictions, assessments). (2) For each, record your confidence level (50-99%). (3) When outcomes become known, record the actual result. After two weeks: (4) Calculate your calibration — when you said 80% confident, were you right ~80% of the time? (5) Categorize errors: which domains, which direction, which judgment types. (6) The resulting profile tells you exactly where to apply correction factors and where to seek outside input.