Understand the base rate fallacy through the “paradox” of a disease test: why even an accurate test for a rare disease is often wrong.
The forgotten base rate
Why an “accurate” test can be wrong
The base rate fallacy (base rate neglect) is when we judge a probability while forgetting how rare the event is in itself.
The classic example is a test for a rare disease. Say 1 person in 1,000 has the disease. The test is very good: when someone is sick it almost always says “sick”, and with healthy people it errs (a “false alarm”) in only 5% of cases. You get a positive result. What's the probability that you're actually sick? Intuition shouts “95%!” That's badly wrong.
Let's count for 1,000 people.
— 1 is sick — the test will almost certainly show “sick” for them: 1 true signal.
— 999 are healthy — but for 5% of them the test will be wrong: that's about 50 false “sick” results.
In total, “sick” came up for about 51 people, and only 1 of them is really sick. So with a positive test the probability of having the disease is ≈ 1 in 51, or about 2%, not 95%.
Why? Because there are VERY many healthy people, and even their rare errors outnumber the one truly sick person. The base rate (the disease is rare) decides everything — and that's exactly what people forget.
Lesson notes
Why an “accurate” test can be wrong
The base rate fallacy (base rate neglect) is when we judge a probability while forgetting how rare the event is in itself.
The classic example is a test for a rare disease. Say 1 person in 1,000 has the disease. The test is very good: when someone is sick it almost always says “sick”, and with healthy people it errs (a “false alarm”) in only 5% of cases. You get a positive result. What's the probability that you're actually sick? Intuition shouts “95%!” That's badly wrong.
Let's count for 1,000 people.
— 1 is sick — the test will almost certainly show “sick” for them: 1 true signal.
— 999 are healthy — but for 5% of them the test will be wrong: that's about 50 false “sick” results.
In total, “sick” came up for about 51 people, and only 1 of them is really sick. So with a positive test the probability of having the disease is ≈ 1 in 51, or about 2%, not 95%.
Why? Because there are VERY many healthy people, and even their rare errors outnumber the one truly sick person. The base rate (the disease is rare) decides everything — and that's exactly what people forget.
Rare stays rare
The base rate fallacy is everywhere we deal with rare events and “alarm signals”.
— Medicine: a positive screening for a rare disease is more often a false alarm than a real finding — which is why the next step is follow-up tests, not immediate treatment.
— Security: a facial recognition system with “99% accuracy” turned on a crowd where there's one criminal in a million will produce mostly false positives on innocent people.
— Stereotypes: even if some trait is more common in a rare group, a particular person with that trait most likely does NOT belong to the group — simply because the group itself is small.
The antidote rule is simple: before you panic at a signal, ask how rare the event itself is. The rarer it is, the larger the share of false alarms among all alarms. A test's “accuracy” without the base rate tells you almost nothing.