The Sniper's Nest Corner boxes in the 6th floor Museum are wrong

[Drew Phipps]

Computers aren't very good (yet) at comparing fingerprints, according to all the fingerprint experts (who I concede have a financial interest in that fact). However, the current fingerprint system in use by the government (AFIS) stores information about prints which have been scanned or inked under near perfect conditions by a law enforcement officer, usually at a booking desk, such as the black and white part of the comparative image above. AFIS has a better than 95% rate of accuracy finding a match between two sets of "tenprints" (i.e. two sets of all ten prints inked under ideal conditions compared to each other). And AFIS can search a million records in a matter of minutes. However, AFIS has a much poorer rate of success in comparing latents. The best hit rate AFIS got was 70%, with a 25% false positive rating. That means if AFIS tells you there's a match, it's completely wrong slightly more than 25% of the time!


There are many reasons why this is true, most of which fall under the general rule that latent print at a crime scene are very rarely recovered under ideal enough conditions; smudged, or swiped, or partial, or elongated, prints require the good old "Eyeball Mark 1." The latent is the red part of the black/red comparison above. As you can see, while cops rarely record material between the ridges, a real latent will often not be just the ridges. Human experts have a far better accuracy rating than AFIS on latents. However, nearly half of the IAI professional experts fail their re-certification tests in a given year, and have to retake the test.


Even under favorable conditions, sometimes the latent print match is wrong. The most recent high profile case I recall is when a Seattle lawyer (Mayfield?) was accused of being a terrorist in Spain. The FBI claims a 100% accuracy rating for its print guys, but they blew that one. The actual terrorist was a guy from Africa, and the Spanish fingerprint experts, who named the real terrorist, have far more stringent requirements for a match than the FBI. I've also heard it said that the FBI's re-certification test is "too easy", and they use the same exact prints on the test every year.


The cute little fingerprint graphic you see on TV shows and movies are just a theatrical device for depicting 100 hours of grueling detective work in a single second. Fingerprint comparison is still as much of an art, as it is a science, and even well meaning professional examiners can have widely differing opinions, and often do.