Sep 24 2008
Homeland Security “Reads Minds”
Well, that’s what the headline says, anyway. Headlines are often misleading. They are often written by headline writers, not the journalists who write the articles, and they are optimized to titillate and intrigue, not to accurately represent the content of the article.
Still, such gross misrepresentation is irritating.
The technology being discussed is an experimental airport scan that would read a a passenger’s body temperature, facial expressions, and other biological markers and infer from them stress level, which will then be used as a warning sign of terrorist intentions.
There is nothing overtly pseudoscientific about such an approach. Humans are emotional creatures, and emotions have a biological component to them. We wear our emotions on our faces, body language, and even autonomic function.
This is the same basic concept as lie detectors – measuring stress as a marker for lying. But, as with lie detectors, the “uncertainty principle” of such technologies is human variability. Someone who is “cool as a cucumber” could theoretically beat such technology. Perhaps a bit of training might be enough. Also, individuals may be nervous for a host of reasons that have nothing to do with planning to blow themselves up. Also, some people are just sweaty and agitated at baseline, even without emotional stress. Maybe they have an overactive thyroid gland.
There are two relevant questions, as far as I can see: how accurate is the technology, and how will it be used.
Accuracy of screening tests, which is what this is, involved both specificity (how often are positive tests true-positives) and sensitivity (how many true-positives will be missed and called negative). You generally want screening testing to be very sensitive, even at the expense of some specificity – as long as you have a follow up test that is more specific.
Homeland security won’t say what the results of their internal tests are (it’s classified), but they did report that it was a “home run.”
Assuming the technology works to some degree – how will it be used? It seems that it would be used as an initial screening. Passengers pass through the arch and are scanned in various ways. Those outside of the parameters and looking “suspicious” can then be pulled aside for more thorough and specific assessment.
What will be the ultimate effect of using this technology on airport screening? That, then, is the final question. Will those able to pass this screening earn lax screening elsewhere? Will sweaty people forever have to endure cavity probes to take a flight anywhere? Or will it add specificity and sensitivity to the overall screening process?
I hope these kinds of assessments are made. One of the risks of shiny new technology is that it can be distracting – it can lure the careless technophile away from the real issues. Net effects of new technology are not always obvious. They need to be studied directly, not just inferred.
One thing is for sure – this new technology does not “read minds.”
15 Responses to “Homeland Security “Reads Minds””
Leave a Reply
You must be logged in to post a comment.
“We wear our emotions on our faces, body language, and even autonomic function.”
Assumes there is more to our emotions than just what are faces show, our body language, and autonomic function. That there is some inner driving force for our emotions is a precarious position to take, or be in.
X – There is more to our emotions than just what our faces show, our body language, and autonomic function. There is also our subjective experience of the emotion, which affects our thought processes and other aspects of brain function. Not all components of emotion are externally visible. And, the experience of emotion has anatomical correlates in the brain – distinct from those directly affecting facial expression, etc. Obviously they all work together to form the overall phenomenon of emotion.
Let’s say that this provides some sort of edge, the question is “what will happen if they use it?” For simplicity, let’s say that it’s 99% accurate meaning that in 1% of times it scans an innocent person it flags them as guilty, and in 1% of times it scans a guilty person it flags them as innocent (false positive & false negative). Of course it isn’t near this accurate, but let’s suppose.
I don’t know how many people fly every day but according to a Reuters report, LAX expects 19.6 Million travellers over the summer, meaning that 196,000 people will be incorrectly flagged as a terrorist, and that’s just in one airport.
How many genuine terrorists will attempt to fly during this period? One? Two? Let’s say that 20 will try to fly. Because the machine only has 1% false negatives, let’s say that every single one of these terrorists is identified. Fine.
So that means out of about 200,000 people flagged as a terrorist, 20 are real. So even if the machine is “99% accurate”, the minuscule number of genuine terrorists compared to the huge number of legitimate travellers means that one in 10,000 people it identifies are really terrorists.
This is a recipe for inconveniencing innocent people and training security personnel to discount the alarms. Cry “terrorist” hundreds of times each day and even the most alert officers will tune it out.
There are other problems that I could imagine. How do terrorists react to the scanners? Do they act angry and upset, or does their self-righteousness make them appear calm? Given the scarcity of terrorists, how was the device actually tested? What about mentally unstable people who have been the biggest source of scares lately, does their different thought processes fool the detector? How well do thoughts and emotions translate to facial features for psychotics or people with impaired emotional ranges?
Fine, it can detect when dear old Aunt Sally illegally smuggles in an extra can of water or Jimmy uses a forged ticket-name to fly on his brother’s ticket but how can it help passenger security?
Stop it, Adrian! You’re not supposed to question authority like that. (Whispering) This new technology only works when everyone believes that it works, get it?
The super-high-tech MALINTENT system works a lot like a polygraph. Except of course it’s less specific, less sensitive, and less accurate. But what it lacks in efficacy, it more than makes up for in sheer volume! Besides, you really can’t expect Homeland Security to invent a technology as foolproof as the polygraph while also being fast and unobtrusive enough for application to airport security, right? Okay.
Now be a good boy, will you? Your apology is accepted. And by the way, how did you know about that thing with my brother?
Comming from Fox news, it’s no big surprise.
It always cracks me up when I hear their catch frase: “Fair and Balanced”
I think democracy is good (sorta), and capitalism is good (sorta), and freedom of expression is excelent. But put all that together and make a press company, and your mileage may vary.
Yet I have no idea on what a good solution might be. The perils of having news agencies be controlled by government would be much worse.
“A home run” means that you covered all the bases without being called out.
In other words you got to tell your story to the blindly credulous faux news without being called liars or without the slightest doubts being raised, meaning you will continue to get buckets full of money for more testing. But the details can’t be told because “full disclosure would compromise future tests.” More likely full disclosure would show how abysmally poor it was so that funding would be cut off.
Sorry if I am so cynical, but the degree of difficulty in this is so high. Humans can read emotions because humans have large parts of their highly evolved and highly trained brains configured to do so. Reading emotions is not something trivial like reading text, something that is still so difficult for machines to do that it is used to prevent access by spam bots. Humans are going to be much better than machines at reading human emotions for a very long time.
daedalus2u,
As someone involved in machine learning (with an emphasis on speech and music), I can safely tell you that emotion detection is advancing at a very fast rate. For example, there was a thesis from my school (Georgia Tech) a couple years ago on detecting clinical depression from voice characteristics. It’s true that computer may never be 100%; however, there are obvious advantages over simple human screening. First, it’s extremely expensive to have someone have ample time to judge whether a passenger is a threat. Currently, there is no judge of character unless someone is being obvious (running and screaming in the terminal). Approaches with automatic screening allow a cheap, effective first step in screening. Obviously, all any automatic approach would do is alert to the need for a closer look: much how a medical test may be followed with more tests.
Second, there are MANY involuntary actions that a person has based on their emotions. Your speech patterns change greatly when stressed, for example. This is something that’s hard to fake as well. I’m actually not surprised to see this article at all and would be interested if they could add a vocal component.
Third, no system would take a single attribute into account, but would look at several factors that are accurate and difficult to fake. The decision on whether someone should undergo further screening would be aggregated among all factors considered.
Emotion detection is advancing rapidly. Let’s assume that we can detect an emotional state using cues obtainable electronically and unobtrusively.
To use it in the application suggested, one assumes a person about to blow-up an airplane would be having a particular emotion.
What emotion would that be?
“One thing is for sure – this new technology does not “read minds.””
Ya got that right!
Sonic, how about the emotional state of religious fervor? If this machine could reliably detect religious fervor and flag the highest 1% for extra scrutiny there would be such an uproar that it would never be used.
Daedalus2u,
Penn Jillette has an ingenious device that detects religious fervor. It’s a T-shirt upon which is printed the phrase “No God”.
daedalus2u,
religious ferver doesn’t make a terrorist and not all terrorists are religious.
So what emotion are we really looking for?
From the newscientist blagpost:
‘At an equestrian centre in Maryland, 140 paid volunteers walked through a pair of trailers kitted out with a battery of FAST sensors, including cameras, infrared heat sensors and an eyesafe laser radar, called a Bio-Lidar, that measures pulse and breathing rate from a distance.
Some subjects were told to act shifty, be evasive, deceptive and hostile. And many were detected. “We’re still very early on in this research, but it is looking very promising,” says DHS science spokesman John Verrico. “We are running at about 78% accuracy on mal-intent detection, and 80% on deception.”‘
I think the strength of their testing methodology and results pretty much speaks (or rather, doesn’t speak) for itself.
80%! That may sound good to the layman but crunch the numbers like I was trying to do and you’ll see just how awful it is. If one in ten passengers were terrorists this might be useful but when it’s one in twenty million at best, false positives will render any machine useless unless it is 99.999999% accurate. Even if we had a magical emotion-reader, does anyone seriously believe there is some constellation of emotions that all terrorists share but no innocent people do? This whole thing sounds doomed to failure.
Stop people from getting weapons on board the plane and once aboard, stop them from taking control. That means not focusing on blocking people from the _terminal_, not pissing around with liquids and shoes, and putting in a decent cabin door. Turn to the Israeli air lines who deal with genuine terrorism instead of this fear-mongering and movie plot fantasies that American airlines have. There are so many things wrong with this scheme it’s hard to know where to start.
Jim Shaveron, all of the 9/11 terrorists were characterized by religious fervor. All women’s health clinic bombers have been characterized by religious fervor, virtually all suicide bombers are characterized by religious fervor. If you look at the list of terrorist incidents, a very large number of them have been characterized by religious fervor.
http://en.wikipedia.org/wiki/Terrorist_attack
If all individuals characterized by religious fervor were subjected to extra scrutiny and/or prevented from access to means and/or places where they might commit terrorists acts there would be many fewer acts of terrorism.
Other emotions strongly associated with terrorists acts include nationalism, xenophobia, racism, bigotry and sometimes simple greed. Of course one person’s “terrorist” is another person’s “freedom fighter”. Were the right wing death squads of the Contras terrorists as their victims said, or freedom fighters as Ronald Reagan said? If people looking at all the facts and thinking about it for days can’t figure out who is a terrorist and who is not, no machine is going to figure it out by looking at “emotions” and figure it out in seconds.
It was John F. Kennedy who said: “Those who make peaceful change impossible. will make violent revolution inevitable.” Was John Brown a terrorist for using violence to fight slavery? Or were the terrorists those who supported the institution of slavery?
If you believe as Bush does that “Either you are with us, or you are with the terrorists,” then figuring out who are the “terrorists” is pretty easy. Of course the easy answer isn’t always the correct one.
The danger of this system is that it will be used to harass political opponents exactly the way the No Fly List has been used and exactly the way the Justice Department has been used. Any machine system that is sensitive enough to detect differences in emotional state can also detect differences in ethnicity and a whole host of other differences that people traditionally use to discriminate. Will the algorithms used to detect a terrorist suspect be open and transparent? Of course not, that would make them easier to defeat and less useful to those in power. Merge that data with all other electronic data on individuals and any individual could be picked out and harassed for anything.
Making peaceful change more difficult does not reduce the likelihood of terrorism, it increases it.
The key purpose of this new system is to “screen” passengers. Where it get’s interesting is “what does it mean to be screened”?
As Adrian points out, a lot of people will get screened. A screening is the violation of a persons right to privacy and basic freedoms. We have (as a society) accepted this violation as a necessary condition for public safety.
But TSA doesn’t stop with a simple on-the-spot screening process. A screened traveler may be conducted to an isloated interogation room, forced (no option) to strip in front of members of the opposite sex , undego cavity searches, and have their luggage and computer equipment searched, damaged, and even destroyed.
If perscription drugs are found in the luggage, or even worse, needles for insulin injection, a prolonged process may occur where TSA demands proof that the traveler does not have any unprescribed prescription drugs (for example a prescription that has expired). Any confusion in such matters may result in several days of incarceration while the matter is cleared up.
Hopefully, no one is so foolish as to carry pornography of any sort, as this will be confiscated and immediately indentifies the traveler as a serious terrorist threat.
Even if the TSA is totally satisfied, they may deny you access to a specific flight or place you on the no fly list leaving you stranded thousands of miles away from home.
Any protest on your part means that the TSA gets to use its tasers to subdue you. A good time is had fo all.
So if you find yourself rolled up in a ball in excruciating pain from being tasered, stripped, subjected to cavity searches, repeated interogations (water boarding is acceptable if TSA decises you are a terrorist), and denied your medications to deal with a rapidly approaching diabetic blackout, all because some machine thinks you sweat too much, just remember that you are just accepting this for “the greater good”, and Homeland Security thinks they have “hit a home run”.