Prosecutor’s fallacy
Inversion of conditional probabilities, Sophisme du procureur
Confusing the probability of observing a piece of evidence if the accused is innocent with the probability that the accused is innocent given that evidence.
The essentials
Mechanism
P(evidence | innocent) and P(innocent | evidence) are two different quantities: the second also depends on how many people could have produced that evidence, that is, on the base rate. The two formulations look alike and are easily confused.
In practice
Example
“This genetic profile appears only once in a million. So there is only a one-in-a-million chance that the accused is innocent.” In a country of 60 million people, about 60 people have that profile.
Famous case
The Sally Clark case (United Kingdom, 1999). She was convicted of the deaths of her two babies, notably on the word of an expert who put the probability of two cot deaths in the same family at 1 in 73 million. That figure wrongly assumes that the two deaths are independent, and it was presented as if it were the probability of her innocence. The Royal Statistical Society protested publicly; she was freed on appeal in 2003.
Defending yourself
How to protect yourself
Separate two questions: “What is the probability of this evidence if the person is innocent?” and “What is the probability that they are innocent, given this evidence?” To get from one to the other, you need the base rate (Bayes’ formula).
Don’t confuse
Not to be confused with: Base rate neglect
Base rate neglect is the general error; the prosecutor’s fallacy is its courtroom form, which consists of inverting a conditional probability.
Going further
Source
Thompson and Schumann (1987) gave this fallacy its name.
Practise
In the app, each entry offers ways to practise recognising the concept. No sign-up: your progress stays on your device.