Cognitive Biases · CB-17
The mathematically correct way to revise a belief when new evidence arrives — and the way almost nobody does it by default.
Bayesian updating is the formally correct method for revising the probability of a belief given new evidence: starting from a prior probability, and updating it in proportion to how much more likely the new evidence is under that belief being true versus false. Most people's intuitive belief-revision deviates substantially from this normative standard, typically underweighting the prior (base rate) and overweighting recent, vivid evidence.
Named for Reverend Thomas Bayes, whose theorem was published posthumously in 1763; its modern application to human judgment and decision-making was substantially developed through 20th-century probability theory and popularized in behavioral economics via contrast with observed human reasoning failures.
The Mechanism
How much a new piece of evidence should actually move your belief
The correct update (green) accounts for both how strong the new evidence is and how likely the belief already was — the typical intuitive update (red) often ignores the prior almost entirely and reacts mainly to how vivid or recent the new evidence feels, producing systematically different, usually less accurate, conclusions.
01 · THE PRIOR MATTERS AS MUCH AS THE NEW EVIDENCE
A rare hypothesis needs much stronger evidence to become likely
Bayesian reasoning formally requires weighting new evidence against how probable the hypothesis already was — a surprising, dramatic piece of evidence for an inherently rare hypothesis should move your belief by less than the same evidence would for an already-plausible hypothesis, a distinction intuitive reasoning frequently misses.
02 · IT'S A SKILL THAT IMPROVES MEASURABLY WITH DELIBERATE PRACTICE
Unlike some biases, explicit training genuinely helps here
Unlike biases rooted in perceptual or motivational distortion, poor Bayesian updating is substantially a skill deficit — research on calibration training and 'superforecaster' methodology (notably Philip Tetlock's work) finds that people who explicitly practice structured probabilistic updating measurably improve their forecasting accuracy over time.
03 · IT PROVIDES A CONCRETE ANTIDOTE TO SEVERAL OTHER BIASES AT ONCE
Base rate neglect, confirmation bias, and anchoring are all partially addressed by the same discipline
Explicitly running a Bayesian update — stating your prior, stating how diagnostic the new evidence actually is, then calculating the revised probability — simultaneously counters base rate neglect (by forcing you to state the prior), confirmation bias (by forcing you to consider the evidence's likelihood under both hypotheses), and anchoring (by making the update process explicit rather than an unconscious adjustment).
Where It Fails / Inversion
Where it fails / inversion
Formal Bayesian calculation requires numbers (a prior probability, a likelihood ratio) that are often genuinely unavailable or highly uncertain in real-world judgment — forcing false precision onto a genuinely ambiguous situation can produce a spuriously confident-looking number that's no more reliable than an honest qualitative judgment, and demanding formal Bayesian math for every decision is itself impractical overkill.
How To Use It
Worked example · evaluating a surprising claim or piece of news
When encountering a dramatic, surprising claim, explicitly ask two questions before updating your belief: how likely was this to be true before I heard this claim (the prior), and how much more likely is this specific evidence if the claim is true versus if it's false (the diagnosticity) — a genuinely rare and implausible claim needs correspondingly strong, specific evidence before it should shift your belief very far, however dramatic or vivid the evidence feels.
How to use it
Before letting a single new piece of evidence substantially change your mind, explicitly state your prior belief and ask how diagnostic the new evidence actually is for distinguishing the hypothesis from its alternatives — not just how dramatic or surprising it feels. This simple discipline catches a large share of common reasoning errors at once.
See Also