The Economist on the CSI Effect

[L.J.Steele]

http://www.economist.com/science-techno ... d=15949089

[Pat A. Wertheim]

In reality, scientists do not deal in certainty but in probabilities, and the way they calculate these probabilities is complex. For example, when testifying in court, a fingerprint expert may say that there is a 90% chance of obtaining a match if the defendant left the mark, and a one in several billion chance of a match if someone else left it.

Oh, really?

[David Fairhurst]

That quote looks familiar.

Check the New Scientist article that Kasey put in this week's Detail.

Champod thinks that the language of certainty
that examiners are forced to use hides
this uncertainty from the court. He proposes
that fingerprint evidence be interpreted in
probabilistic terms — bringing it in line with
other forensic domains — and that examiners
should be free to talk about probable or possible
matches. In a criminal case, this would
mean that an examiner could testify that there
was, say, a 95% chance of a match if the defendant
left the mark, but a one in a billion chance
of a match if someone else left it.

[L.J.Steele]

Oh, really?

That raised my eyebrows too -- are there folks giving this sort of probability testimony in the UK? And if so, based on what?

[Pat A. Wertheim]

Hi Lisa

To the best of my knowledge, no fingerprint examiner in the world is testifying like that today. Christophe Champod (University of Lausanne) was quoted saying that hypothetically, we may testify like that some day. Christophe, Cedric Neumann (Forensic Science Service), Glenn Langenburg, and others are working on various models of calculating probabilities or likelihood ratios for fingerprint identification. All of these models are in the development stage and are being experimented with. No such model has yet been perfected, peer reviewed, or validated.

I had the privilege of working with Cedric's computer model last year. A number of groups have been exposed to it and played with it. I am encouraged by the possibility that fingerprint examiners may have such a tool within a few years that would allow us to testify to a likelihood ratio, much the way DNA analysts now testify. I want to emphasize that such a model is NOT intended to make the identification. Nothing would change in the way a fingerprint examiner goes about comparing fingerprints and making identifications. The difference would be that after an identification is made, the computer model would allow the examiner to calculate a likelihood ratio for statistical support of the identification already made.

The problem with the article in The Economist was that the reporter apparently read Champod's hypothetical statement and paraphrased it (without crediting Champod) to make it sound like we are already doing it that way.

[L.J.Steele]

The Economist is pretty good about corrections -- you or someone else who knows the area may want to drop a line to their editor.