Verifying Evidentiary Science

The “problems with fingerprinting” have been highlighted in one recent essay on Slate (Printing Problems – The inexact science of fingerprint analysis by David Feige, 2004/05/27) and one article in the NYT (Can Prints Lie? Yes, Man Finds to His Dismay by Benjamin Weiser, 2004/05/31). The problem with the problems with fingerprinting, though, is that these articles aren’t about problems with fingerprinting — these articles are about problems with evidentiary science.

Specifically, the real problem is that attorneys continue to present, and judges and juries continue to hear, some kinds of evidentiary science as foolproof — absolute — and weigh them inappropriately highly. Of course any piece of evidence is supposed to be weighed with all the other evidence to determine whether all the evidence reaches a particular standard of guilt (no reasonable doubt, for instance, or preponderance of the evidence). But some kinds of evidence — fingerprinting, DNA — are, in practice, treated as nearly infallible. Unfortunately, this kind of evidence is completely fallible. All evidence exists in a context, and evidence derived from analysis exists in an analytic context — in other words, a human-mediated context. And humans are nothing if not fallible.

One way to ameliorate this problem might be to require cross-checks for the use of these kinds of evidence — sort of peer review of the results. In the case discussed in the NYT, the problem was a database problem — suspect A’s prints were filed in suspect B’s record. Once suspect A’s prints were lifted from a new crime scene, suspect B’s name matched. Suspect B begged for a photo ID match but the judges told him that the fingerprint records were infallible. So what’s the problem here? A database problem. What would be a useful cross-check? A database cross-check of some sort; for example, a photo ID; or use of a separate database.

Hmm. This suggests another good reason for not having one giant identifying database … because if those records are screwed up the right way it would be really, really difficult to prove. … But that’s another rant.