Homeland Security's 'Pre-Crime' Screening Will Never Work

http://www.theatlantic.com/technology/archive/2012/04/homeland-securitys-pre-crime-screening-will-never-work/255971/

by spathak • 14 years ago

56 19 14 years ago

jessriedel3• 14 years ago

So the obvious problem with the movie Minority Report was the false dichotomy between (a) imperfectly predicting crimes and then punishing people before they commit them and (b) not predicting crimes. The obvious third choice, which has been in use since time immemorial, is (c) imperfectly predicting crime and then applying normal crime-prevention measures to increase the chance of catching the perpetrator and/or reduce the chance that the crime is committed at all, while simultaneously using less resources. In any sane world, the technology of Minority Report would still be used to reduce murders through direct intervention and the deterrent of more effective enforcement, without taking people off to pre-crime prison.

The author's implication that FAST is an implementation of the obviously-bad idea of punishing people for crimes they haven't yet committed is just one of the many bad and misleading aspects of this article.

roc3• 14 years ago

Indeed. I read FAST as a sort of olive-branch in security theatre. It seems that it, or something like it, will be used to remove porno-scanners and grope-checks from standard security. If 70% of people are passing through security without being hassled, their PR problems nosedive.

And frankly, though it's likely to be error prone and leave airport security still largely ineffective, at least it's move toward trying to intelligently identify risks. Rather that simply sending every single person through ineffective security. And lessening the load on the scanners and gropers can only raise their effectiveness.

3pt14159• 14 years ago

I agree that that is their goal, but I still don't like rights violated without a warrant or something similar.

As an aside, I find it funny that your name is roc, since http://en.wikipedia.org/wiki/Receiver_operating_characterist... is a good tool to measure accuracy in systems like this.

kenj0418• 14 years ago

70% accuracy is pretty weak. I can predict whether a given passenger will attempt to commit a terrorist act with greater than 99.99% accuracy. Here is my code:

boolean isTerrorist(Passenger p) { return false; }

Lagged2Death• 14 years ago

...will be used to remove porno-scanners and grope-checks from standard security.

Has this ever happened? Has the TSA ever given up on a security screening practice that had become a widespread routine?

If 70% of people are passing through security without being hassled, their PR problems nosedive.

If 70% of passengers walk through the screening point without removing their shoes, without walking through a metal detector, without having their bags x-rayed, etc., and the other 30% are singled out for special treatment, won't that 30% feel extra persecuted? The TSA's PR problems are mainly related to rarer-that-30% outrages like selecting comely young women for pornoscanning, interfering with uncommon medical equipment, etc.

lukeschlather• 14 years ago

>at least it's move toward trying to intelligently identify risks.

If we were looking to intelligently identify risks we would be spending more money on car crashes, "ordinary" murders, tornadoes, earthquakes and other things that kill so many people that we've grown inured to it.

But people are more interested in a magic bullet to stop us from losing 3,000 people in another fell swoop. Which is really a pretty small number given how many untimely deaths we have every year. (Looking at it from a yearly basis, we lose maybe 300 people a year to domestic terrorism. Not a good thing, to be sure, but we have bigger issues.)

peeters• 14 years ago

The article DOES address this, you just missed it.

> Given that none of these people would have actually committed a terrorist act yet distinguishing the innocent false positives from the guilty might be a non-trivial, and invasive task.

Invading the privacy of (or refusing travel to) the 99.99% false positives IS punishing the innocent. The article does not succumb to the false dichotomy you describe.

jessriedel3• 14 years ago

That's why I carefully said "implication". He doesn't come out and say it because it would be wrong; he just tries to slander by association.

Furthermore, the FAST data will not result in the refusal of travel (without further evidence) nor will it result in an invasion of privacy which we do not already allow at random. The author is purposefully confusing logically separate issues in order to confuse the reader. The author's argument, as stated, could be directed at any form of intelligence gathering by TSA whatsoever, an obviously wrong conclusion.

kijin• 14 years ago

> In any sane world...

The problem seems to be that we don't live in a sane world.

maxerickson• 14 years ago

One of the subplots was that the technology basically required total subjugation of the precogs. That doesn't seem all that sane.

jessriedel3• 14 years ago

OK, I agree that sane people could take the principled stand that it's not right to subjugate 3 pre-cogs to save the several hundred people murdered every year in Washington DC. But first, I don't think most people actually would. And second, there has got to be a way to convince 3 pre-cogs to do this work voluntarily, seeing as that you've got about of order a billion dollars (the economic value of 200 lives in the US) to work with.

mistercow• 14 years ago

> So the obvious problem with the movie Minority Report

... was that the screenwriters weren't smart enough to write consistent time-travel logic, and therefore the plot made no sense whatsoever. Also Tom Cruise.

The short story didn't really have a problem of precrime being imperfect per sé.

tptacek• 14 years ago

This is a good article marred by, I think, a poor story angle. FAST sounds like a terrible idea, but not because it's Orwellian.

Homeland Security is not primarily a law enforcement agency. Here are it's departments; departments with a principle law enforcement focus are starred:

    Immigration Services 
    (+) ICE
    (+) Customs
    (+) TSA
    Coast Guard
    (+) Secret Service
    FEMA
    Homeland Security Advisory
    National Infrastructure Advisory
    Homeland Security Science/Technology
    Critical Infrastructure Partnership Advisory
    Interagency Coordinating Council on Emergency Preparedness
    Task Force on New Americans
    Domestic Nuclear Detection Office
    Federal Law Enforcement Training Center
    National Protection and Programs Directorate
    (+) Federal Protective Service
    National Communications System
    Directorate for Science and Technology
    Directorate for Management
    Offices of Policy; Immigration Statistics; Health Affairs
    Offices of Intelligence and Analysis; Operations Coordination
    Office of the Secretary
    National Cyber Security Center

(Yes, DHS is a monster, everyone knows that; we should definitely slay it).

People who have worked on the intrusion detection problem in computer systems are well familiar with the Base Rate Fallacy (what this article appears to be talking about when it mentions the "false positive fallacy"). To anyone who has actually monitored a deployed IDS, the base rate problem is a killer. Most of the events generated by systems like this are mathematically certain to be bogus.

So it's great to see an Atlantic article trying to tackle the numeracy issue here: DHS policymakers looking for research funding are bragging about a terrorist detection system with a 70% accuracy rate, and 70% sounds great when you leave out the base rate of terrorism.

On the other hand: of the LEO functions of DHS, most of them fall into one of two categories:

* Departments that already have total discretion to ignore civil liberties, like Customs: it's hard to imagine any new program making Customs more intrusive, because Customs operates under the border search rule (since at least 1976, Customs can set up any search program it wants to protect the integrity of the US border).

* Departments that operate under administrative search regimes, like TSA, where "behavioral monitoring" is if anything less Orwellian than the systems already in place.

DHS research projects do not automatically imply that local and state police will be deploying "behavioral" technology to "pre-crime" people. Again the theme: when you argue against something, you need to be both right and correct. "Pre-crime" is an inaccurate framing device.

bobbydavid• 14 years ago

From the article:

"Nature reported that the first round of field tests had a 70 percent accuracy rate. From the available material it is difficult to determine exactly what this number means. There are a couple of ways to interpret this, since both the write-up and the DHS documentation (all pdfs) are unclear."

I'm glad he pointed this out.

"This might mean that the current iteration of FAST correctly classifies 70 percent of people it observes -- which would produce false positives at an abysmal rate, given the rarity of terrorists in the population."

But maybe this isn't so bad if it means that 70% of people aren't molested in the airport, at the cost of missing 30% of terrorists we would otherwise catch. I'm very okay with this: I see this as (1.73 * 0.7)-million happier citizens per day [1] and (0 * 0.3) additional terrorist incidents [2].

"The other way of interpreting this reported result is that FAST will call a terrorist a terrorist 70 percent of the time. This second option tells us nothing about the rate of false positives, but it would likely be quite high. In either case, it is likely that the false-positive paradox would be in full force for FAST, ensuring that any real terrorists identified are lost in a sea of falsely accused innocents."

The author is too smart for his own good here. The second interpretation _should_ be "FAST will call a terrorist a terrorist 70 percent of the time. What's a false positive?"

Those unfamiliar with statistics may incorrectly see it this way. Seventy percent of terrorists will be caught, so that means we'll be more than 3 times safer than we already are! Such statistical incompetence is also strongly correlated to politicians. I can't back that up, though, due to my statistical incompetence.

[1] http://www.quora.com/How-many-people-fly-domestically-in-the...

[2] http://www.gather.com/viewArticle.action?articleId=281474978...

read_wharf• 14 years ago

Assuming honest research can achieve any of this in the foreseeable future, does anyone expect DHS to conduct or oversee research honestly and competently?

Homeland Security's 'Pre-Crime' Screening Will Never Work

Related Stories