Database Measured 'Terrorism Quotient'

Before helping to launch the criminal information project known as Matrix, a database contractor gave U.S. and Florida authorities the names of 120,000 people who showed a statistical likelihood of being terrorists ? sparking some investigations and arrests.

The "high terrorism factor" scoring system also became a key selling point for the involvement of the database company, Seisint Inc., in the Matrix project.

Public records obtained by The Associated Press from several states show that Justice Department (news - web sites) officials cited the scoring technology in appointing Seisint sole contractor on the federally funded, $12 million project.

Seisint and the law enforcement officials who oversee Matrix insist that the terrorism scoring system ultimately was kept out of the project, largely because of privacy concerns.

However, new details about Seisint's development of the "terrorism quotient," including the revelation that authorities apparently acted on the list of 120,000, are renewing privacy activists' suspicions about Matrix's potential power.

The MATRIX system (and who picked that name, anyway?) collects basic information on all citizens within its implementing state and makes that information available to law enforcement within all participating states. It's one of those projects where the devil is inevitably in the details. Information sharing helps catch terrorists and criminals but threatens the privacy of the law-abiding, so where do you draw the line? Especially when the database is designed to contain public records which the government has some right to access.

I've been nervous about the system on general principles. I don't mind making law enforcement put in some effort to get information on people; it shouldn't be an at-the-fingertips thing for the average citizen. Keep a database of people with arrest warrants, people with prior felony convictions, and suspected terrorists. That's 99% of the benefit right there. The cop who pulls you over to warn you that your taillights are broken shouldn't need to know anything more than that; it's not an investigation, after all, just a casual encounter.

But this article changes my outlook. Why? It's simple math -- math that I predicted when TIA first hit the news, and that also applies to the MATRIX system now that we know about the "terrorism quotient". In case that's not clear, the "terrorism quotient" is some kind of probability factor for a given individual being a terrorist. Given the information they put into MATRIX, pulling out a probability for such an extremely rare quality is no better than voodoo. It's like taking someone's income tax return and trying to determine if they are a homosexual.

It's possible to look at the tax return and make some educated guesses, but it's not possible to get it right with anything close to 100% accuracy. Given the low incidence of homosexuality in the general population (2-3%), and the relatively low correlation with traits visible on an income tax return, you're going to have a substantial problem with both false negatives and false positives.

False negatives are the cases where there is a terrorist, but he doesn't quite fit your profile, and gets away; the system doesn't notice him because he doesn't have any of the behaviors that would tip you off about him. To use the analogy, he's the conservative, happily married business owner. From everything the system knows, he's safe. One of the things the system knows is that the actual ratio of terrorists/homosexuals to the general population is very low -- if you pick any particular example, the chance that you have picked one of your target group is miniscule. So, in order to overcome that, there have to be some fairly strong data points. So far, I haven't seen any evidence that such data points exist (for either target group). That means it's very hard to overcome the negative presumption.

False positives are the cases where an innocent person is identified as a potential terrorist. In order to reach this level, the person must have possessed a significant number of positive factors. That seems comforting, until you realize what sort of positive factors are available in the data set. How many Muslim males between 20 and 30 are living in the United States under false or expired visas who attend a mosque regularly? Surely thousands. How many of them can you arrest on suspicion of terrorism?

(Answer: None; you have no evidence. You have to put them all under surveillance, using national security letters rather than warrants due to the evidence problem, which is much more expensive. After nearly 3 years, we've arrested probably under 50 "terrorists" and have yet to convict a significant number of those.)

There's an important third category, too -- people who just aren't in the system. These are the folks who pay a Mexican coyote to smuggle them across the border, or who try to sneak in over the Canadian border, or who board their flights in Europe. MATRIX can't spot those, because it doesn't know about them, even if it works perfectly in all other respects.

So, with our expensive MATRIX system that assigns everyone a terrorist quotient, we've managed to compile a list of people that we think might be terrorists. We know there are some terrorists who aren't on the list at all, and we know that the system won't show us all the terrorists even if they are in the system, but the cost of a successful attack is very high, so we set the parameters fairly loose -- so we can be pretty sure that most potential terrorists are on the list. We'll just have to investigate the people on the list, one by one, and see who warrants further attention.

How big is the list? Well, thanks to this article, we know that the list of potential terrorists provided to the Florida authorities contained one hundred and twenty thousand people. Based on population statistics about Florida, that's about 1% of the population. Matrix would have you believe that one person in every hundred in Florida is a potential terrorist.

If you believe that, you should seek medical help for your paranoia immediately.

That's 120,000 people who will be investigated, probably interviewed, and some certainly arrested, based on nothing more than a computer program saying "Hey, you have some traits which vaguely resemble traits that terrorists also had."

But maybe we've acclimated to this. We do take off our shoes at the airport, these days, in case anyone is tempted to repeat the amazing stupidity of Richard Reid (the "exploding shoe terrorist"). Millions of people flying on airplanes, millions of our best, brightest, and richest, cheerfully remove their shoes at the direction of TSA agents -- just in case the terrorists are stupid enough to try it again. The cumulative cost in annoyance, humiliation, and wasted time (not to mention lost shoes) must be staggering.

And yet we do it without complaint, just as we discard our nail clippers and knitting needles and box cutters. How many terrorists have we caught because the airline screeners found his box cutter, or the bomb in his shoe, or the biological weapon in her breast milk?

There's a point where employing certain strategies against even a very expensive threat has a cost far, far higher than the results justify. The excessive searches at airports fall into that category, especially when a much more effective solution exists to stop the problem -- by which I refer to the Armed Pilot program. Now that this information about MATRIX has been revealed, it falls into the same category: expensive and pointless. Why investigate 120,000 people from scratch when you can investigate known Al-Qaeda members and their contacts, with a much lower cost and a much better chance of actually finding terrorists?

Why use homosexuality? It's a trait that occurs in a small number of people across a very broad range of backgrounds, and it's a socially-hidden trait -- it's not something obvious, although there are likely correlations with some otherwise innocuous behaviors. I'm not trying to imply any kind of link between homosexuality and terrorism!

This entry was published Sat Sep 24 10:43:35 CDT 2005 by TriggerFinger and last updated 2005-09-24 10:43:35.0. [Tweet]

comments powered by Disqus

Subscribe to Atom Feed

I am not a lawyer, and nothing on this site should be taken as legal advice.

This site is run on custom blog software and is being actively developed. Please be forgiving of errors.

This website is an Amazon affiliate and will receive financial compensation for products purchased from Amazon through links on this site.