杏吧原创

STATISTICIANS frequently argue that their discipline is misrepresented in
court. There is clearly a potential for judges and juries to be confused by big,
impressive-sounding stats. But are courts often misled by the subtle use of
numbers?

There are no published figures, but statisticians say such instances are
common in both civil and criminal trials. 鈥淲hen I was in court a lot in the
mid-1990s, it used to happen all the time, in all kinds of cases,鈥� says David
Balding, professor of statistical genetics at Imperial College, London.

When O. J. Simpson was on trial for murder, for example, the jury in
California heard that only 1 in 1000 wife-beaters go on to murder their wives.
The prosecution didn鈥檛 mention that when a beaten wife is murdered, the chance
that the husband did it is as high as 1 in 2
(New 杏吧原创, 13 December 1997, p 18).

The most common error is the 鈥減rosecutor鈥檚 fallacy鈥�. For example, for DNA
found at a crime, there might be a match probability of 1 in a million. 鈥淚t
means that out of the population at large, 1 in a million people would have a
matching profile,鈥� says Philip Dawid, professor of statistics at University
College London.

But juries can make the mistake of concluding that the defendant, who already
has a matching profile, therefore has only a 1 in a million chance of being
innocent. To make sense of a DNA match probability, courts need to take into
account whatever additional information is available.

That doesn鈥檛 always happen. 鈥淎 single 1-in-a-million number in a court has a
mesmerising effect on all present,鈥� says Dawid. 鈥淛uries think they can ignore
all the other evidence and that something rare can鈥檛 happen by chance.鈥�

In one British landmark DNA case following a rape in 1991, Dennis Adams
appeared to be convicted primarily on a DNA match. 鈥淎ll the other evidence
pointed to his innocence, including the victim, who said it wasn鈥檛 him,鈥� says
Dawid. 鈥淭he key is how the DNA fits with the other evidence.鈥�

At the heart of the problem is the tendency for judges, juries and barristers
to think they understand probability when they don鈥檛. 鈥淭his is difficult and
subtle,鈥� says Peter Green, president of the Royal Statistical Society. 鈥淏ut
judges tend to suggest that no special expertise is needed.鈥� Dawid says judges
need to learn to be wary of statistical evidence, so they don鈥檛 leap to mistaken
conclusions when instructing the jury.

Another possibility, raised by Lord Justice Auld in a recent review of
Britain鈥檚 criminal justice system, is to appoint an independent assessor to
weigh up statistical evidence from both sides and point out any glaring
omissions or fallacies. It鈥檚 an idea that has support in the US. 鈥淓vidence from
both sides can sound equally persuasive,鈥� says David Kaye, a legal statistics
expert at Arizona State University. 鈥淎 court-appointed advisor can say if it is
really one-sided, either to the judge or the court.鈥�