Conditional Probability and Base Rates

New information should change your mind, but by how much? Conditional probability is the arithmetic of updating: the chance of something given that you now know something else. It is also where intuition breaks most expensively. Doctors misjudge what a positive test means; juries misread forensic evidence. The common thread is a neglected base rate, the background frequency of a thing before any test or clue, without which even excellent evidence can point the wrong way. The fix is one habit: start from how common something is, then let new information nudge that starting point rather than replacing it.

Probability given information

A conditional probability is the chance of A given B, "the probability of A, now that we know B." New information shrinks the world to the cases where B is true, and we re-measure A within that smaller world. Crucially, "A given B" and "B given A" are different questions: the chance of a positive test given the disease is not the chance of the disease given a positive test, and confusing the two is the costliest error in this area.

Base rates

The base rate is how common something is before you see any specific evidence, sometimes called the prior. One in a thousand people have a condition; most people flagged by a rare-event alarm are false alarms. The base rate is the anchor: evidence adjusts it but cannot be read without it. Base-rate neglect is the habit of fixing on the vivid new evidence while forgetting how rare the thing was to begin with.

Bayesian updating in plain language

Bayesian thinking is a three-beat rhythm: start with the base rate, look at the evidence, update in proportion. Strong evidence moves you far, weak evidence a little, but you always move from the base rate, never a blank slate. The rarer the thing, the more evidence you need before feeling confident.

A condition affects 1 in 1,000 people. A test catches 99% of real cases and wrongly flags just 1% of healthy people. You test positive. Start from the base rate. Among 10,000 people, 10 have the condition. The test flags all 10, plus about 1% of the 9,990 healthy, roughly 100 people, so about 110 positives, of whom only 10 truly have it. Your chance is near 9%, not 99%: the base rate is tiny, so the healthy majority, even rarely misflagged, supplies most positives.

Base rates cut both ways. Run the same test where symptoms have already made the condition common, 1 in 5 rather than 1 in 1,000. Among 10,000 people, 2,000 are sick and nearly all are flagged, while the 8,000 healthy give about 80 false positives. A positive result now means roughly a 96% chance. Same test, same accuracy, opposite conclusion, because the anchor moved. This is why "how common is the condition in this person?" comes before "how accurate is the test?"

In 1990 Marilyn vos Savant answered a reader in Parade magazine about the game show "Let's Make a Deal," hosted by Monty Hall. Three doors hide one car and two goats; you pick one. The host, who knows what is behind each, opens a different door to reveal a goat and asks whether you want to switch to the last door. She said yes: switching wins two times in three. About a thousand readers, many with mathematics PhDs, wrote to insist she was wrong, and she was right. Your first pick has a 1/3 chance and keeps it; because the host deliberately avoids the car, his choice pours the remaining 2/3 onto the other unopened door. It is the classic demonstration that conditional reasoning defies intuition.

• "'A given B' equals 'B given A'." They are different probabilities and can differ enormously.
• "An accurate test makes a positive trustworthy." Not when the condition is rare; the base rate can leave most positives false.
• "New evidence should replace your prior view." It should update it, from the base rate, not from scratch.
• "In Monty Hall the two doors are 50-50." The host's informed choice loads the odds; switching wins two-thirds of the time.

Work through diagnosis and Monty Hall scenarios, sliding the base rate up and down to watch how the same piece of evidence leads to wildly different conclusions.

Think Like a Maester: Before you trust a striking piece of evidence, ask how common the thing was to begin with. The base rate is the anchor; evidence only moves the boat.

1. What does a conditional probability measure?
2. Why can a 99% accurate test still make most positive results false?
3. What is a base rate, and why must it anchor the judgment?
4. Describe Bayesian updating in a single sentence.
5. In the Monty Hall problem, why does switching win two-thirds of the time?

Conditional probability is the mathematics of changing your mind: the chance of something given new information. Its master key is the base rate, how common a thing is before any evidence, which even a very accurate test cannot override when the thing is rare. Bayesian thinking makes this a habit: start from the base rate, weigh the evidence, update in proportion. From the rare-disease test to the Monty Hall doors, intuition leaps to the vivid clue and forgets the anchor, while disciplined reasoning starts from how likely the thing was to begin with.

• Thinking in Probabilities
• Risk and Uncertainty