Statisticide

Very recently, I was reading a fantastic book titled Once Upon A Number by John Allen Paulos, a Professor of Mathematics at Temple University. A few observations made by Prof. Paulos sparked my interest in the (ab)use of statistics in courts. To start with, I will discuss a fairly obvious, but oft missed point, which led me to dwell more on the issue of statisticide in courts.

Prof. Paulos emphasizes the distinction between:

• probability of an innocent person having an array of evidence against him; and
• probability of a person with an array of evidence against him being innocent.

To the naive, this might look like mere word-play, but the distinction is as stark as between innocent and guilty. The former probability has often been abused by chest-thumping prosecution lawyers presenting bewildering probability numbers to jurors. The slick prosecution attorney will say something like – “the probability of someone innocent having A, B, C pieces of evidence against him is 1 in a million!” (also notice his mouth-left-wide-open look for added effect). And there goes half the jury – already decided on a conviction.

I will steal Prof. Paulos’ example to illustrate the vast difference this misrepresentation/misinterpretation makes:

Consider a murder trial in a town of 1 million people. It is agreed by the prosecution and defence that the person who committed the crime is a resident of the town and has a certain type of moustache. Now, it so happens that only 2 (X and Y) of the 1 million people have such a moustache. The prosecution presents its case against X claiming that the probability of an innocent having X’s moustache is 2 in a million! This is not a lie, but the jury interprets it as a strong indication of X’s guilt. The fact however, is that the same statement can be made for Y. It follows that the probability that X, given that the murderer has his kind of moustache, is innocent is not 2 in a million, but in fact 1 in 2 in absence of any other evidence!

The above ruse (called fallacy of the transposed conditional) has probably been pulled by lawyers in innumerable cases with substantial impact on jurors.

I found instances of statisticide of above nature and others, in several trials: the O.J. Simpson trial, the Sally Clark trial and the Nurse Lucia trial being very public examples. I will detail each of these cases in separate posts soon.