Wednesday, January 12, 2011

THE FALLIBILITY OF FINGERPRINTING

The easy truth is that fingerprints are not reliable. The deeper facts are necessarily more complicated. Whether and to what extent fingerprints - or, ultimately DNA, retinal scans, and other biometric identifications - may be putatively "infallible" there remains an underlying desire to segregate good people from bad, to control the latter for the benefit of the former. 

Suspect Identies: A History of Fingerprinting and Criminal Identification  
by Simon Cole (Cambridge: Harvard University Press, 2001).

However, an admittedly unscientific poll by CNN Money in 2004 found that 21% of Americans cheat on their federal income  tax.     A story in the New York Times for May 1, 2010, claimed that 20% to 30% of the wealth in Greece is off the books in undisclosed transactions.   That number might reflect American realities as well.  (Prof. Liqun Cao told our criminology classes that perhaps 20% of the goods on the market have no clear title.)  Furthermore, according to the social taxonomy for "anomie" presented by Robert King Merton, "innovators" are those who seek socially-approved goals in unapproved ways.  You might think that defines robbery (and it does), but it also defines the "pirates of Silicon Valley" who created new computer resources otherwise denied to them: they were hackers before they became inventors.  Even if we could tag everyone, do we want to?  What is achieved by criminalizing young people for thinking outside the box?

The basic problem here, of course, is that such tagging is not reliable.  It does not work.  Experts who study fingerprints know the problems.  Worst of all, they deny, minimize, and obfuscate their different interpretations of the facts in order to present a united front to the world, claming that fingerprints are certain, not probabilistic. Simon Cole's work has a few weak points.  He presents the same information twice, in the absence of giving new data to substantiate a new claim.  With that caveat, it remains that this book is an excellent history of the attempts to measure and identify criminals. 

Though with several other roots as well - India under British rule in 1858, for instance - "Anthropometry" began in the 1870s with Cesare Lombrosco. Again, both motives were active.  First, criminologists thought that law-breakers were physically different from normal people.  The criminal could be detected by the geometry of his skull and the ratio of his legs to this torso. Calling themselves scientists, they developed complex taxonomies for ear shape.  In theory, if they could measure everyone, they could cull the criminals from society and prevent if not eradicate crime.  Second, the police and courts (and the military) had a large body of repeat offenders. Identifying them would facilitate processing -- and prevent wrongful pleas for mercy from self-defined "first time offenders."  Similarly, the U.S. Army needed to keep out those it threw out but who came back via another recruiter. 

Anthropometry was hard work. Special calipers applied to the unclothed body provided raw numbers for the calculation of ratios and proportions.  And it only worked with males.  In the 19th century, despite hand-wringing over the evils of prostitution, few were willing to measure a woman in the same manner.  And the numerical models were lacking.  Anthropometry had limitations, of course; and Coles tells (pp. 140-142) the famous case of the two William Wests carefully measured and then confused by Leavenworth Prison.  That hallmark case is a classic justification for the reliability of fingerprinting.  It appears in college textbooks for criminal investigations.

Fingerprinting solved a lot of problems, not the least of which was differentiating the two Misters West. Fingerprints are easy to get -- from men or women, living or dead. 

Meticulous cataloguing suggested that no two people had the same set of prints.  That may well be true.  We do not know for sure.  This is an empirical truth only.  Like primitives for whom the sun has always come up in the east, but who have no knowledge of astronomy, we can take all the measurements we want, but we do not know the genomes and alleles for fingerprints; we know the what, but not the why.

We could live with that.  But in a few decades from the 1920s to the 1940s, fingerprinting devolved to the point where trained examiners claimed under oath that a partial print unabiguously identified a perpetrator.  Comparing good "native" prints from each of ten fingers was reduced to accepting claims about partial latent (secondary) prints.

Fingerprint examiners never disagree with each other in court.  They only agree that the prints identify or are "inconclusive."  Coles tells the story (pp. 264-266; 272-274; 282) of experts in disagreement all of whose certifications were revoked because that was the only way for the International Association for Indentification to maintain the fiction that experts do not disagree about fingerprints.  Today, fingerprints are unassailable.

The fact remains that fingerprints have been forged by several methods by police and by criminals.  (The FBI first found a forgery by a law enforcement officer in 1925.)  And of course there is politics.  J. Edgar Hoover and New York City Police Commissioner Richard E. Enright both had strong claims that their agency was best suited to be the central clearning house and repository.  Suspect Identities is a dense book that covers a lot of ground in 350 pages, but the index is an exemplary epistemological map to this complex presentation.


The epistemological and methodological problems in fingerprinting are not taught to law enforcement professionals.  After completing a BSc in criminology in 2008, I was watching a rerun of NUMB3RS, the CBS drama series (2005-2010) about mathematician Charlie Eppes and his brother, FBI Special Agent Don.  The first season show Identity Crisis centered on fingerprinting.  Prof. Eppes identified the key difference between DNA and fingerprinting, asking rhetorically, "But prints don't have odds?" A fuller treatment of the problem can be found in The Numbers Behind NUMB3RS by Keith Devlin and Gary Lorden.  (Dr. Devlin teaches at Stanford and is NPR's "Math Guy."  Dr. Lorden teaches at CalTech and was the technical advisor for the series.) 
The case of Brandon Mayfield may become paradigmatic.  You can google him and find the Wikipedia article.  As Mayfield is an attorney, the National Association of Criminal Defense Lawyers has an insightful summary on their website.  Mayfield was wrongfully accused of being an Al Qaeda operative, in large part because FBI fingerprint examiners refused to disagree with their boss.  The evidence - such as it was - began with a single latent fingerprint.


ALSO ON NECESSARY FACTS
Employee Theft
Junk Science in the Courtroom
Eyewitness Testimony: Popper, Wittgenstein, and the Innocence Project
Forgery and Fraud in Numismatics
William Sheldon: Psychologist, Numismatist, Thief
Is Physics a Science?

No comments:

Post a Comment