After four weeks, check if your beliefs under-react or over-react to evidence
After four weeks of belief tracking, examine whether beliefs barely moved despite evidence (conservatism) or swung dramatically on single data points (base rate neglect) to identify domain-specific updating patterns.
Why This Is a Rule
People update their beliefs in characteristically broken ways, and the pattern is often domain-specific. In domains they know well, they tend toward conservatism — barely updating even when strong evidence arrives, because their existing model feels solid. In domains they're uncertain about, they tend toward base rate neglect — swinging dramatically on single data points, because they lack a stable prior to anchor against.
Four weeks of belief tracking reveals your updating pattern. Did you encounter evidence that should have shifted your view but didn't? That's conservatism — you're weighting your prior too heavily. Did a single customer complaint make you want to redesign the entire product? That's base rate neglect — you're weighting new evidence too heavily relative to the full distribution.
The diagnosis matters because the corrections are opposite. Conservatism needs more responsiveness to evidence. Base rate neglect needs more anchoring to priors. Applying the wrong correction amplifies the problem.
When This Fires
- After maintaining a belief-tracking journal for at least four weeks
- When you suspect you're either too stubborn or too reactive in specific domains
- During periodic meta-cognitive reviews of your decision-making quality
- When building domain-specific calibration strategies
Common Failure Mode
Tracking beliefs without recording the evidence strength. A belief that didn't move might be correct conservatism (the evidence was weak) or problematic conservatism (the evidence was strong but you ignored it). Without evidence strength annotations, you can't distinguish appropriate stability from pathological rigidity.
The Protocol
For four weeks, track significant beliefs: (1) Record the belief and your confidence level. (2) When evidence arrives, record what it is and how strong it is (weak/moderate/strong). (3) Record your updated confidence. After four weeks: (4) For each belief, plot the trajectory. Beliefs that barely moved despite strong evidence → conservatism in that domain. Beliefs that swung 30%+ on single data points → base rate neglect. (5) Apply the appropriate correction: push yourself to update more in conservative domains, anchor to base rates more in volatile domains.