Monday 19 May 2008

getting csi right -- UPDATE

On Saturday, I wrote about the problem of error in forensics. Grits for Breakfast notes one of the problems I didn't mention explicitly: in some states, lab reports are not treated as "testimonial" evidence (and therefore defendants do not have the right to cross examine lab technicians). This position is being challenged, and the Supreme Court is watching with interest.

Grits worries that
Forensic science isn't "objective" science, it's goal oriented.
That's not the problem; science is always goal oriented. The problem is that a life hangs in the balance, and the single case matters.

Grits also worries that certain forensic findings cannot be falsified. That seems implausible, but I could be missing something. Even the least-reliable of forensic tests will show some potential suspects to be inconsistent with the evidence found at the scene of the crime. For example, even if a fingerprint whorl cannot reliably select a particular individual from the general population, it can eliminate anyone without whorly fingers. Surely it is malpractice to exclude such evidence when found, so it seems that the only worry could be that the CSIs stop looking for evidence (or more likely, the lab technicians stop processing the evidence they have). Here, the worry isn't that the findings can't be falsified, it's that evidence is only processed until the prosecution is satisfied it can make its case--and this standard is too low.

A much larger worry is that
most 'pattern evidence' - handwriting analysis, shoe and tire print comparisons, etc., has no research-based foundation at all. Much of forensic science is 'soft' science, [former executive director of the National Forensic Technology Science Center Bill Tilstone] said, that at best has not or even cannot be comprehensively tested for accuracy.
Ouch.

No comments: