Maintain a belief-tracking log with timestamps

Record your beliefs and confidence levels before evidence arrives, so you can see whether you actually updated.

Why it works

Confirmation bias affects not just new information intake but memory: people misremember their prior beliefs to be more consistent with current beliefs, making it feel like they updated when they barely moved. A timestamped written record is the only reliable check: if the written prior belief and the post-evidence belief differ, the difference is real; if they don’t differ, neither did you.

How to do it

1. For any significant belief or prediction, write it down with a date and a confidence level (e.g., 75%).
2. When relevant evidence arrives, write your new confidence level before reading anyone else’s analysis.
3. Periodically review old entries and compute your actual update size.
4. Track whether your updates are systematically small (anchoring/conservatism) or systematically large (overreaction).

Evidence

Memory for prior beliefs is demonstrably distorted by current beliefs — people reconstruct their past views to be more consistent with what they now believe. Pezzo (2003) and others document this hindsight bias effect, which the written record directly prevents. Fischhoff (1975) originally established the effect experimentally, showing that outcome knowledge inflates how predictable people judge an event to have been — the precise distortion a timestamped log defeats. (observational)

A belief log requires honest recording; if beliefs are recorded post-hoc rather than pre-evidence, it reproduces the bias it was meant to prevent.

Sources

• Fischhoff (1975), hindsight ≠ foresight, Journal of Experimental Psychology: Human Perception and Performance
• Fischhoff, B. (1975). Hindsight ≠ foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288–299.

Common mistake

Writing beliefs at a granularity too coarse to detect real updating ("I thought this was likely" → "I think this is likely") — use numbers to capture what words obscure.