Monday, January 12, 2004

Henry v. Vucetich:

The oft-told story of re-inventing wheels, competing standards, and generally un- or at best ill-informed decisions.

At lunch today I met Julia Rodriguez, a researcher from the University of New Hampshire who is looking at criminology in late 19th century Argentina. She is looking at the implementation of a fingerprinting program to document and uniquely identify all immigrants, and how science played into the adoption of this particular policy. Answer: not very well. And by now you can probably guess where she went with this, given the recent news about fingerprinting at U.S. points-of-entry and possible changes in immigration policy. Follow the parallels where they exist. There are bright spots, I promise.

In the 1870's the British invested a great deal of capital in Argentine railroads, allowing a large boom in the agricultural export industry. A period of enormous immigration flux followed, and the population grew more than ten-fold within 20 years. As might be imagined, the social and physical infrastructure could not manage this, and a great deal of social unrest occurred. Within the flood of immigrants, there were all sorts of bad seeds (to be read as 'anarchists, communists, and other criminals'), and the Argentine government wanted to winnow them out.

Social criminology theories of the day based on the work of Cesare Lombroso (nature) and Enrico Ferri (nurture) in Italy were popular, as was the evolving work of physiological measurements ("anthropometry") under Alphonse Bertillon in France. Although much maligned by later mis-use and outright quackery, the fundamental motivation was simple enough: can measurements of human physiognomy be used as identifiers, are they inherited, and can they also be used as predictors of behaviour? Many thought they could, and Argentina (among most other advanced nations of the day) set up an official system to collect measurements from criminals and immigrants. The data was to be used to help with resolving identity in criminal cases, and perhaps to identify potential troublemakers before they were allowed entry.

Among the beaurocracy was Juan Vucetich, a statistician given the task of seeing if there was anything useful in fingerprints, since recent work by Francis Galton (a cousin of Darwin's) in England pointed heavily towards fingerprints as unique identifiers. Galton had initially tried to tie fingerprints to race and to behaviour, but was not able to establish a scientifically rigorous correlation (however, Galton never renounced the belief that the correlation was there). Vucetich took Galton's fingerprint uniqueness theory, and devised a rigorous classification system for fingerprints that allowed for the searching of the growing set of records. Argentina had the first set of indexed fingerprint records, and the first criminal conviction in 1892 based on fingerprint identification (incidentally, an Argentine (Gori) also published the first professional criminological journal). The Vucetich system is still in use today by many police, judicial and correctional systems throughout the world. Vucetich's method is still taught at an institute bearing his name in Mendoza, Argentina.

However, the system as set up in the late 1800's and early 1900's in Argentina had a lot of holes. Only seaports took immigrant fingerprints. First-class passengers and Argentine citizens were exempt. Plenty of bad seeds came overland. Plenty were already in the country (...many of them home-grown), and plenty of bad seeds were in the moneyed classes. It was a classic case of implementing a policy based on a presupposed, and long-term untenable position that any unrest was caused by poor outsiders, and that border control was the answer. There were of course two fatal flaws: a porous border, and the presupposition that the threat came from an easily classified set of people. It was a short-term political fix designed to quell public fear. Given that Argentina was composed of more than 50% immigrants, riots soon ensued, and the fingerprinting program was eventually dropped.

To continue the fingerprint history, about ten years after the Vucetich system was set up in Argentina, Edward Henry devised a similar system in India. He was later transferred to London, and began using his system in Scotland Yard, from where it migrated to most of the rest of the English speaking world. In the U.S. there were several false starts, based on variants of Henry, and many incompatible systems were set up based on decisions by local police superintendents. By the 1940's most of this had been ironed out, and a massive card file began to grow in Washington.

The growth of computers and pattern recognition theory was a great incubator. The digitization of the U.S. card database and the eventual merging of international databases was contemplated under a common system. The competing Henry and Vucetich systems had come to a head, under the need for the FBI's International Automated Fingerprint Identification System (IAFIS).

They both lost, and Vucetich and Henry systems were discarded. IAFIS actually went back to Galton to look at even finer levels of detail to provide necessary levels of evidentiary confidence.

So, what is to be made of this in the present?

One. Massive as it is, the IAFIS database is generally only composed of the convicted criminal population, all federal civil servants (yours truly included), and military personnel. Most civil, unconvicted criminal, and foreign fingerprints are still on cardboard, and are not in the system. The system therefore is re-active rather than pro-active.

Two. Current immigration fingerprinting is only required for certain countries, and so would miss U.S., Canadian and British citizens, as well as current holders of green cards. And there are certainly plenty of bad seeds among those groups.

Three. Fingerprinting is only implemented at certain ports of entry, not all. ...and there is additionally always immigration that avoids ports-of-entry, anyway.

Four. Terrorists supposedly come from a classifiable set of people. Not by measuring their skulls, but by looking at their electronic habits.

Five. We are a nation of immigrants, and depend on immigration for labour and population growth.

The 1800's Argentine ingredients are there, but of course the context is very different.

The level of effort required to make this system work is daunting, and it will require giving up freedoms. At the moment, those freedoms are those belonging to the outsiders, and so there is political short-term capital to be gained from the policy. When the freedoms impinged upon belong to regular U.S. citizens, watch out. Those stories will start to emerge, as dual citizens get caught, U.S. citizens in other countries' databases get caught, and just plain errors are introduced. Godel's omega incompleteness will have its way.

Good books I found on this issue are:

Simon Cole, Suspect Identities: A History of Fingerprinting and Criminal Identification (Cambridge, MA: Harvard University Press, 2001; 369 pp.)

and the bible for fingerprinting technology:

Maltoni D., D. Maio, A.K. Jain, and S. Prabhakar, Handboook of Fingerprint Recognition (New York, NY: Springer Verlag, June 2003; 360 pp., 180 illus., DVD included, ISBN: 0-387-95431-7)

I will post another amusing story about Argentine science later. The Huemul incident is not a proud moment for a country that has a strong intellectual history.