Bayesian Updating in Everyday Life UK 2026

Three components:

• Prior probability: your belief BEFORE seeing the new evidence. Based on base rates, history, intuition.
• Likelihood: 'if the belief is true, how likely is this evidence?'.
• Posterior probability: your belief AFTER combining prior + likelihood.

Plain-English formula:

• New belief = (How likely the evidence is, IF my belief is true) × (My old belief) / (How likely the evidence is overall).

Why it's intuitive once you see it:

• Strong evidence relative to belief = update significantly.
• Weak evidence relative to belief = update mildly.
• If new evidence is equally compatible with multiple beliefs = barely update.

Scenario: You believe a particular UK FTSE 100 stock has 40% probability of beating market over next year (your prior).

New evidence: Earnings report is significantly above analyst expectations.

Bayesian update:

• Prior: 40% that the stock will beat market.
• Likelihood of strong earnings IF stock beats market: 70%.
• Likelihood of strong earnings overall (across all possible outcomes): 35%.
• Posterior: (70% × 40%) / 35% = 80% probability stock beats market.

What this tells you:

• Don't conclude 'stock will definitely beat market' (95%+).
• Don't dismiss the earnings ('stocks always bounce back').
• The probability genuinely shifted - act accordingly (size position more aggressively but not all-in).

Where humans go wrong:

• Overreact: 'This is great news, I'll buy ALL the shares' (treating posterior as certainty).
• Underreact: 'Earnings doesn't matter; market moves on macroeconomic forces' (ignoring likelihood signal).
• Right reaction: significant but not extreme position adjustment.

Scenario: You have a chronic headache that started yesterday. You believe (prior) there's 95% probability it's a tension headache, 5% probability it's a more serious condition.

New evidence: A specific symptom appears (e.g. blurred vision).

Bayesian update:

• Prior: 95% tension headache, 5% serious.
• Likelihood of blurred vision IF tension headache: 10%.
• Likelihood of blurred vision IF serious condition: 60%.
• Overall likelihood of blurred vision: 10% × 95% + 60% × 5% = 12.5%.
• Posterior probability of serious condition: (60% × 5%) / 12.5% = 24%.

What this tells you:

• Probability went from 5% to 24% - significant 5x shift.
• Tension headache STILL most likely (76% of the time).
• But 24% serious is high enough to act on - see GP urgently.
• Avoid: 'I'm dying' (treating new evidence as certainty).
• Avoid: 'Just my usual headache' (failing to update).

Scenario: You believe (prior) you have 30% probability of getting a particular type of role (e.g. senior product manager at a UK tech startup).

New evidence: Rejected after final-round interview for one specific role.

Bayesian update:

• Prior: 30% will get role this year.
• Likelihood of rejection IF you'll get role: 50% (you might get rejected from this one, succeed at another).
• Likelihood of rejection IF you won't get role: 95% (you'd be rejected from most attempts).
• Overall likelihood of rejection: 50% × 30% + 95% × 70% = 81.5%.
• Posterior probability of getting role: (50% × 30%) / 81.5% = 18%.

What this tells you:

• Probability went from 30% to 18% - moderate downward shift.
• Still possible (18%, not 0%).
• Continue applying; possibly tweak strategy.
• Avoid: 'I'll never get a job' (treating rejection as certainty about all roles).
• Avoid: 'They were just wrong; my odds are the same' (failing to update).

1. Updating too much on weak evidence: one anecdote isn't strong evidence. Don't change your view of an entire population based on a single case.
2. Updating too little on strong evidence: confirmation bias makes us discount evidence that contradicts existing beliefs. Statistically-significant evidence deserves significant update.
3. Ignoring base rates entirely: 'I scored well on this test, so I have 95% probability of acing the next' - ignores that the test was easier than average.
4. Treating posterior as certainty: 80% probability still means 20% wrong - keep that in mind.
5. Failure to update conjunction beliefs: 'I think A and B and C are all probably true' - if one is contradicted, you need to update on the conjunction, not just the contradicted piece.
6. Anchoring on initial estimates: writing down a number early then over-anchoring on it is common - check whether your posterior really moved.

Rule of thumb #1 - Strength of evidence vs belief:

• If new evidence is MUCH stronger than your prior belief: significant update.
• If new evidence is EQUALLY strong: moderate update.
• If new evidence is WEAKER: small or no update.

Rule of thumb #2 - 3x test:

• If posterior should be 3x more or less than prior: this is a major belief change. Pause + verify.
• If posterior should be less than 50% off prior: minor adjustment OK without deep analysis.

Rule of thumb #3 - Likelihood ratio approximation:

• Likelihood ratio = (probability of evidence IF belief is true) / (probability of evidence IF belief is false).
• If LR > 5: strong evidence; significant update.
• If LR 2-5: moderate evidence; modest update.
• If LR 1-2: weak evidence; tiny update.

Decision journals:

• Write down: your prior, the new evidence, your reasoning for the update, your posterior.
• Reviewing 3-6 months later reveals systematic biases (over/under-updating, ignoring evidence).
• See our guide on decision journals.

Calibration training:

• Make probabilistic predictions about events; track accuracy over time.
• Apps like Calibrate.app (UK + global) help structure this.
• Better calibration = better Bayesian updating.

Bayesian software / spreadsheets:

• For complex decisions: use Excel / Google Sheets to compute formal posteriors.
• BayesFactor R package for statistically-minded users.

Community + practice:

• Forecasting platforms (Manifold Markets, Metaculus) help develop Bayesian intuitions through repeated exposure to probability questions.
• UK rationalist meetups (London Rationalist Group, Cambridge LessWrong) discuss these techniques.

High-stakes single decisions:

• Investment in property / pension / business.
• Major career change.
• Medical treatment choice.
• Educational pathway selection.

Repeated decisions:

• Portfolio rebalancing (each trade is a Bayesian update).
• Hiring decisions (each interview updates predictions).
• Marketing / sales (each customer interaction).
• Diagnostic medicine (each test result).

Less useful for:

• One-off simple choices ('which restaurant tonight').
• Time-constrained snap decisions (no time for explicit calculation).
• Decisions where outcomes are immediate + obvious.

The compounding benefit:

• Practicing Bayesian thinking on small daily decisions trains intuition for the high-stakes ones.
• Over months/years, this compounds into significantly better judgment.