Update beliefs incrementally, not all at once

New evidence should shift your probability somewhat — rarely from 5% to 95% in one step.

Why it works

Bayesian updating is proportional: strong evidence produces large updates; weak evidence produces small ones. Most people update either too little (anchoring) or too dramatically (representativeness: treating one vivid case as decisive). The proportional update — multiplying your prior by the likelihood ratio the evidence provides — is what the math actually prescribes, and it is almost never as large as an all-or-nothing swing.

How to do it

1. When new evidence arrives, ask: "How much more likely would this evidence be if my belief were true versus false?"
2. Use that ratio to move your probability proportionally, not to flip it.
3. If the evidence is weak or ambiguous, move a little; if strong and surprising, move more.
4. Track two or three important beliefs over time and log each update — this builds calibration intuition.

Evidence

The conservatism bias — updating too little relative to Bayesian norms — is documented in multiple judgment studies. Its counterpart, overreaction to vivid single cases, is equally well documented. Both are departures from proportional updating. (observational)

Experimental studies use simple, controlled probability problems; real-world beliefs involve more ambiguity about what the evidence actually establishes.

Sources

• Edwards (1968), conservatism in human information processing, in Formal Representation of Human Judgment

Common mistake

Treating any piece of confirming evidence as proof and any piece of disconfirming evidence as an exception — which is confirmation bias expressed as selective updating.