Modern phrenologists "predict" terrorism with biometrics
"Scientists" at the University at Buffalo have reinvented phrenology in the form of a set of biometrics that produce a numerical score indicating the probability that you are about to commit a terrorist act.
Link (via Futurismic)Computer and behavioral scientists at the University at Buffalo are developing automated systems that track faces, voices, bodies and other biometrics against scientifically tested behavioral indicators to provide a numerical score of the likelihood that an individual may be about to commit a terrorist act.
"The goal is to identify the perpetrator in a security setting before he or she has the chance to carry out the attack," said Venu Govindaraju, Ph.D., professor of computer science and engineering in the UB School of Engineering and Applied Sciences. Govindaraju is co-principal investigator on the project with Mark G. Frank, Ph.D., associate professor of communication in the UB College of Arts and Sciences.
(Photo credit: Phrenology1.jpg, a public domain image from Wikimedia Commons)
the latest
latest episodes
I have a bad feeling about this one.
Oh sweet Jesus, I may not be able to take much more of this maggoty rot infesting the corpse of the Republic. Let's just call off the Enlightenment, roll back the Magna Carta -- write it all off as a bad experiment. Why even anticipate progress in the human condition? It's just bad for business anyway.
Speaking of which, clearly I should also be in the security theater snake oil and nostrum business. There's obviously plenty of easy money to go around.
Hey! here's a thought. How about we stop doing the shit that creates terrorists in the first place? You know, like laying waste to entire countries and sucking them dry of every natural resource while making sure the natives are still living in mud huts even after we've gone and moved on to the next one.
It's a crazy idea, I know.
Why the panic? Their behavioral indicators are scientifically tested. Science is right, by definition.
Minority Report!
this is the same stuff (or awfully similar) that kip hawley was touting in his interview with bruce schneier a month or so ago, and scneier at the time seemed uncharacteristically non-skeptical, which i thought was strange. maybe he has an updated response.
it seems to completely remove any possibility of rational appeal. i.e., how are you supposed to defend yourself against false accusations? "my eyes weren't twitching the way they said?"
So, this is "phrenology reinvented"?
Would you use the same words to describe a polygraphic test? How about functional neuroimaging?
How about putting scare quotes around the words "scientists" in those cases?
As I read the article, the investigators have received an NSF grant to *try* combine biometric tech with behavioral science. They don't claim to have actually done it yet. This kind of thing is called "research".
The formal name of the project is "Deceit Indication through Person Specific Behavioral Dynamics". Slamming it as "snake oil" and "phrenology" and questioning their standing as scientists seems a bit premature, considering they haven't actually *done* the research yet. That's what the grant is for. If they were finished and their final report was unscientific, that would be different.
The other projects on the CUBS website don't look much like "phrenology" to me either.
Facecrime.
Putting aside the whole thing about whither it is science of not,
This makes me nervous, as any system that is basically behavior based (doing normal actions in a combination) to indicate one is ABOUT to commit a terrorist attack. What I mean by normal action of course precludes obvious indicators like purchasing explosives or something like that.
Sure, catching terrorists before they do stuff is good, but this just sounds more of the terrorist paranoia craze (See TSA, Boston Police Incidents) and is far to likely to be abused.
Maggie (7):
You betcha I would.In one of David Brin's early SF novels, personality profiling results in large numbers of people -- "probationary personalities" -- having curtailed rights; a repressive regime based on the practice is overthrown when profiling reveals that the regimes' principles are all untrustworthy nuts.
"Sir, you'll need to step aside for a minute."
"What... Was there something wrong with my carry-on?"
"I'm afraid I can't tell you that right now, sir."
"What's the problem then?"
"You'll need to look over there until the eye-twitch analysis algorithms get about 10 seconds more footage of your face."
(another guard at a computer makes a signal)
"Sir, I'll need you to step into this room over here..."
Why would you need to use biometrics to find the terrorists? It is a widely known fact that all terrorists have tattoos! Check out the the phrenology guy in the photo. He has German tattooed all over his head!
Some are even more obvious - just check out this one I found....
This is like trying to predict the weather. In order to get enough information to make any prediction you'd have to interfere with the subject so much as to make any readings meaningless. It's preposterous to even consider that anyone could or would be able to record the brain responses of a single actual terrorist in action, let alone the dozens it would take to form any scientific theory as to what they're behavior is like before they commit an attack. And who says all terrorists think they same things? Surely suicide bombers looks a lot different than bombers just planting a device and expecting to get off the plane before it explodes. Is there a difference between religious and political martyrs? Christian and Muslim martyrs? Impossible.
given the vast array of possible human behaviours, and the vanishingly slight chances that someone will commit a terroristic act (at least in the US, in the fashion that is labeled 'terrorist' i.e. sensation, 24-style attacks) this software system appears overbearing. a system to detect shoplifting would seem to me to be far more valuable in terms of actual deterrence, apprehension, and likelihood of occurrence.
but, slap an 'anti-terrorist' label on any project and no doubt your grant money increases exponentially. such a shame.
aside - i think timkuo makes some excellent remarks, but I think my above comment still stands in light of his clarification.
Wanna bet that the mid-eastern facial characters is exactly what that thing would label as terrorist?
Hmm ... the "science" is of course very well tested:
Tester 1 in white lab coat [with measuring calipers in hand] to Tester 2 [taking notes]: "... nose: Nordic 2, space between eyes, a little too tight, hair color ... [measures against sample chart], dark blond ... okay this one passes as Aryan, but we'd better see who his grandparents were, you don't want him bombing our airplanes ... er, that is ... polluting the Master Race ..."
I can't agree with the sensationalist spin of this post, phrenology purported to predict personality based on skull contours, which have never been known to have any correlation to just about anything, except your parent's, sibling's and children's skulls.
Now if you give me a few seconds (or better yet minutes) of video of someone walking around (as long as they don't know they're being observed), I could probably tell you something about them, an approximate age, how they're feeling, etc.
Of course, the terrorists probably train to avoid a similar kind of surveillance anyway, they know that human guards are going to be watching them and train to behave 'normally', sure, it's going to take longer to train how to behave to not trigger an algorithm (assuming they have some information about it), but it's not impossible.
That being said, while this may or may not actually be a good way of catching crooks/terrorists/people who dislike the government (anyone remember how they wanted to track crooks by body odor), it's Slightly (heavy sarcasm) more advanced and a bit better science than phrenology.
I think you're going to get too many false positives, how do you know that a person is nervous about catching an important flight/upcoming board meeting/wedding proposal/first plugging (anal sex) session/(insert really big event here).
And yes, I Do think that having anal sex during a board meeting whilst making a wedding proposal and on board an airplane would be downright Sexy! *goes to write in his Big Book of Things To Do*
wow, feel my head for bumps before going through the xray machine - now THAT is technological advance! :-)
@Maggie Lebber: Yes, polygraphs are junk science. The best are accurate only ~80% of the time, and most do little better than chance. That's why they are not permitted in court. And neuro-imaging, while I definitely believe it has more credentials than phrenology, is (currently) vastly over-rated as a science. The neuroscientist I share an office with agrees.
Much of my work involves creating statistical models to predict behaviour, which is not too much different from what this guy is doing. It is great for understanding relationships between variables in a general population, but it is terrible at trying to predict what a specific individual will do, especially if the stakes are high. The problem is compounded when the behaviour is very rare. For example, let's say there are 300 terrorists in the USA right now, and with this new technology, we can assess whether someone is a terrorist with 99% accuracy (which is very optimistic.) We submit everyone in the USA to the test - and have ~3 million false positives, and ~300 true positives - meaning although we caught nearly every terrorist, it was at a cost of catching 10,000 innocent people for every actual terrorist in the process (do you smell the freedom yet?) This is the same reason disease tests are not given on a mass scale - the false positive flood the true positives into meaninglessness.
There's nothing this system can do that can't be done better by a human being. We're not going to install this system at sensitive points and get rid of the human watchers. If the proposal is to use this system to add surveillance to lots of places where we don't have it now, I find it objectionable.
Polygraphs: I wouldn't think well of a witness that only lied one time out of five.
If it were possible to objectively measure behaviour, which by definition is a subjective phenomenon, this could be considered science.
But a century of research (and the benefit of all the hindsight that brings) says that this idea is as bankrupt as psychoanalysis turned out to be.
It's amazing how willing these researchers are to pervert science in the pursuit of grant money.
I have to join those who are skeptical of this. Doesn't sound very scientifically based to me, especially given that it's been done by a professor of computer science and engineering and a professor of communication, neither of which fields have much if anything to do with evaluating or predicting behavior. If it's "biometrics" then shouldn't a biologist of some sort be involved? And behavior clearly falls more within the bailiwick of the psychological arts and sciences. A psychologist should be in charge of this type of research. Sounds like someone's looking for (or has already landed) research grants from the feds.
Mr Doctorow, it seems to me that you're making an a priori condemnation of this based upon nothing more than a press release and what appears to be your own personal bias against any form of determinism. How scientific is that?
Michael (26):
Excuse me?First: This entry is about bad science, bad technology, and the invention of yet another "security" system that if used will make no one more secure, but could potentially make life a great deal less free and more troublesome. That doesn't amount to an attack on causality itself.
Second: You have no evidence whatsoever that Cory's writeup of this entry proceeds from personal bias.
Third: You yourself have failed to address any of the scientific issues involved.
Fourth: Determinism isn't science. Determinism is an untestable faith-based philosophical proposition.
@Michael-
Welcome to the Internet.
We get all worked up trying to prevent a repeat of September 11, thinking that increased security will save us.
The reason that things worked in 2001 is because the hijackers studied airline procedure and security and found workarounds.
We can't rely on "robots and science!" to save us from terrorism- in fact, nothing can. Someone who's dedicated the remains of their life into bringing down (what they see as) a diabolic evil will devote the extra weeks, months, or years, to circumventing whatever high-tech (or low-tech) security measures my tax dollars have paid for.
Something tells me that this is going to catch kids playing video games, people watching movies, and others who will be in a situation where their facial expressions change while on camera.
one minute you're eating a denver omlette^H^H^H^H^H^H^H^H^H^H^ playing nint3edo DS, the next minute theres a gun in your face^H^H^H^H^H^H^H^H^H^H^H^H airport security in your face
>This entry is about bad science,
The science hasn't been done yet. No one can say authoritatively whether the science is bad at this point. You can call it bad science if, after the research has been conducted, it proves unverifiable and unreliable. But until then, all you have are your own biases.
#7, #22:
OK, you think polygraphs are junk science. And I think there'd be widespread support for putting phrenology on the junkier side of polygraphy. But I notice nobody's slammed functional neuroimaging as junk yet...I was deliberately proposing a slippery slope, and pointing out what a smear "phrenology" is in this context.
It's just wrong to label this as "junk science" or "unscientific" *before* there's been any chance for any science to be done. It's at the proposing research stage; saying "that could never work" is every bit as unscientific.
Yeah, seems unlikely to catch terrorists, but really likely to catch me looking for an open power outlet next to the soda machines so I can recharge my laptop when stuck in the airport for hours.
Not biases, Chevan, just clean procedure. If the science isn't known yet, you don't base security systems on it.