BELIEF / EVIDENCE / BAYESIAN UPDATE

Same lab result—why do two people reach opposite conclusions?

One positive slip, two opposite gut reactions. The fork is often prior belief × how the test speaks—not a plot twist in the ink. Teaching metaphor with toy numbers; not a real medical conclusion.

Bayesian update overview

Same positive—two opposite gut reads

Person A sees a positive and freezes. Person B sees the same slip and stays calm. The ink is identical; the starting beliefs are not.

Bayesian updating says: posterior ∝ prior × likelihood. If the disease is rare for you and common for them, the same evidence can land in different places. Numbers below are illustrative—not a clinic report.

Keep these three ideas

  • Prior first

    How common is the condition in this person’s world before the test?

  • Evidence is a weight

    Sensitivity and false positives shape how hard the result pulls.

  • Teaching, not diagnosis

    Toy percentages explain the shape—they are not your lab’s truth.

Step 01

Two priors: rare crowd vs higher-risk crowd

Before any slip arrives, each person carries a base rate—how often the condition shows up in people like them. Call it the prior.

In a rare-crowd prior the condition is uncommon; in a higher-risk prior it is more common. Same test later; different starting stacks of “yes” vs “no.” Illustrative percentages only.

What to notice

  • Base rate

    Prevalence in this person’s reference group—not a moral score.

  • Two stacks

    Rare vs common priors look like thin vs thick “has it” bars.

  • Toy numbers

    We use round teaching percentages so the geometry stays readable.

Step 02

Likelihood: how the test talks when truth is yes vs no

A positive result is not a pure mirror. Sensitivity asks: if the condition is truly present, how often does the test say yes? False-positive rate asks the opposite: if it is absent, how often does the test still say yes?

Those two rates form a likelihood ratio—how many times more the evidence favors “has it” over “doesn’t.” Teaching numbers; not a specific assay’s datasheet.

What to notice

  • Sensitivity

    P(positive | has it)—true-positive rate in the toy model.

  • False positive

    P(positive | clear)—how often healthy people still test positive.

  • Likelihood ratio

    Roughly sensitivity ÷ false-positive rate when both are nonzero.

Step 03

Update: prior × likelihood → posterior (tree / area)

Picture a crowd of 1,000 toy people. Split them by prior into “has it” and “clear.” Then apply the test: some true positives, some false positives. Among everyone who got a positive, what share truly has it? That share is the posterior.

Area or tree views beat formula piles: you see which positive slips came from the thin rare branch versus the thick higher-risk branch. Still illustrative—not a medical conclusion.

What to notice

  • Count positives

    True positives + false positives = everyone who tested positive.

  • Posterior share

    True positives ÷ all positives ≈ P(has it | positive).

  • Geometry > slogans

    Areas make base-rate neglect harder to miss.

Step 04

Why two people can disagree on the same slip

Hold the test fixed. Give Person A a higher-risk prior and Person B a rare prior. Run the same update. Their posteriors can sit on opposite sides of a worry threshold even though the ink matches.

That is base-rate sensitivity in action—not a personality flaw. Closing reminder: every percentage here is a teaching metaphor, not a real diagnosis or clinical advice.

What to notice

  • Shared evidence

    Likelihood is the same; priors differ.

  • Opposite guts

    One posterior can look scary, the other tolerable.

  • Not medical advice

    Talk to clinicians about real labs—this page teaches Bayesian shape only.

Put the system back together

Posterior = prior × how the evidence speaks—not the ink alone

Bayesian updating multiplies a starting belief by how diagnostic the result is. Ignore the base rate and the same positive can feel like certainty when it is still mostly false alarms—or the reverse.

Remember the dials: prior, sensitivity, false-positive rate. And remember the disclaimer: illustrative numbers, not a medical conclusion.

Reader checklist

  1. Ask what prior (base rate) you are using.
  2. Ask how often the test is positive when the condition is absent.
  3. Prefer counting positives in a toy crowd over slogan certainty.
  4. Treat this page as mechanism teaching—not diagnosis or clinical advice.

Four moving parts

  • PriorWhat starts the story?

    Base rate in this person’s reference group.

  • LikelihoodHow does the test speak?

    Sensitivity vs false-positive rate.

  • PosteriorWhat updates?

    Belief after seeing the positive.

  • DisclaimerWhat is this not?

    Not a real lab report or medical advice.

Opposite guts on the same slip often mean different priors—not different ink.

Public-education notes on Bayesian updating and base-rate neglect. Lab percentages are illustrative teaching devices—not clinical performance claims or medical conclusions. Wikipedia “Bayes' theorem” / “Base rate fallacy” provide useful background; this page is mechanism-only and not diagnosis or treatment advice.