Bayesian Updating

Revising a belief incrementally as new evidence arrives, moving it by an amount that fits how diagnostic the evidence is, rather than holding fixed or lurching to a new conviction. The best forecasters make many small updates, neither anchoring on their first take nor overreacting to the latest headline.

Why it Matters

Beliefs formed once and defended go stale as the world moves; beliefs that swing with every new datum are noise. Updating in proportion to the evidence is how a forecast tracks reality over time, and frequent small revisions are a measured hallmark of accuracy.

Signals

• A first estimate that never moves no matter what comes in (under-reaction).
• An estimate that jumps on each new story and reverses next week (over-reaction).
• Updates driven by how recent or vivid the news is rather than how much it actually tells you.

Benefits

Forecasts that stay current, errors that get corrected early instead of compounding, and a discipline that separates real signal from the pull of the latest, loudest evidence.

Risks

Under-updating from anchoring or commitment to a public position; over-updating on weak but salient evidence; mistaking motion for improvement.

Tensions

Changing your mind often can look like having no convictions, and updating too readily invites manipulation by noise; the balance is to move with the weight of evidence, not its volume.

Examples

Nudging a forecast from 65% to 60% as one assumption weakens, rather than holding 65% or dropping to 30%; resisting a dramatic but uninformative news item that does not actually change the odds.