Tuesday, August 09, 2005

Forensic Science

This is interesting - and scary! Apparently, Forensic science isn't quite as scientific as we all thought:

"The underlying principles of the identification forensic sciences have never been rigorously scientifically proven," said Jay Siegel, director of the undergraduate forensic science program at Indiana-Purdue University in Indianapolis.

"This includes handwriting, fingerprints, and firearms, and tool marks," he said.

According to a new study, traditional forensic analysis often relies on untested assumptions and semi-informed guesswork. It can also sometimes produce the wrong results.

Drawing on data from 86 DNA exoneration cases, the researchers found that forensic science testing errors and false or misleading testimony by forensic scientists had been leading causes in the false convictions.

In other words, folks are being convicted based on the prestige value of having "scientific" evidence?

One of the researchers, Jonathan Koehler, says it's time for forensic sciences to adopt the culture of other sciences.
"This includes being more conservative, requiring empirical support for claims, adopting higher professional standards, and generally conceding the important role that the possibility of error plays in interpreting results," said Koehler, a professor of behavioral decision-making at the University of Texas in Austin.

Wonder what the intelligent design community will make of that since ID frequently claims Forensic Science as an ally?

Error rates range from 20% (fingerprint analysis) to above 60% (bite marks, voice identification):

During the past decade, scores of people who were convicted of serious crimes have been exonerated by DNA analyses of crime-scene evidence that had not been tested at the time of their trials.

Drawing on data from 86 such cases, Koehler and his colleague, Michael Saks, a law professor at Arizona State University in Tempe, found that forensic science testing errors played a part in 63 percent of the wrongful convictions. Only eyewitness error was a more common factor.

The study also found that forensic scientists are the witnesses most likely to present misleading or fraudulent testimony. In 27 percent of the cases reviewed, expert witnesses were found to have given false testimony.

Part of the problem concerns a concept called discernable uniqueness which states that if two marks appear indistinguishable, they must have been produced by the same object:

"Forensic scientists will likely say that they know it, because they've never found different objects that have identical markings across many years of forensic science. At first blush, this might seem like an impressive boast. But consider this: Suppose that exactly a hundred pairs of firearms out of the estimated 100,000 guns in a Texas town do, in fact, share indistinguishable rifling markings."

"If each of a hundred firearms experts examined ten pairs of guns from the town's gun population every day for ten years, there is about a 93 percent chance that none of the indistinguishable pairs will have come under examination. That is, despite a thousand 'collective years' of forensic science experience—and more than three million gun-pair examinations—the failure to find even a single pair of guns with indistinguishable markings would offer little basis for drawing conclusions about whether bullet markings on guns in this town are indeed unique."

So, if I understand this correctly you would get a lot of false positives - which have to due with sampling procedure.

It goes without saying that this has profound implications for how expert testimony is used by the criminal justice system (and I may do a post on how this impacts fornesic anthropology) - but that is not the point of this post. The point is that those aspects of forensic science which ID proponents argue give the most support to ID are being abandoned by forensic science. They are being replaced by a more rigorous scientific procedure that gives no support to ID.