Scientist Organizes Resistance To Polygraphs 405
George Maschke writes "Brad Holian, a senior scientist at Los Alamos National Laboratory, is using a blog to organize resistance to plans for random polygraph and drug testing of Lab scientists. Holian writes: 'Polygraphy is an insulting affront to scientists, since a committee of the National Academy of Sciences has declared that, beyond being inadmissible in court, there is no scientific basis for polygraphs. In my opinion, by agreeing to be polygraphed, one thereby seriously jeopardizes his or her claim to being a scientist, which is presumably the principal reason for employment for many scientists at Los Alamos.'"
As promoted by the FSB... (Score:5, Informative)
The lie behind the lie detector (Score:5, Informative)
Long story short: the polygraph is a pile of pseudo-scientific bullshit, that can be easily beaten by anyone that knows how it works. At its core, its basically just a non-standardized investigation protocol for extracting harmful confessions by deceiving the person being investigated.
After educating myself, I passed a polygraph easily the first time, without any preparation or practice, while directly lying to my investigator. For the record, what they were asking was none of my employer's business (in my opinion). I was previously warned that the average session takes an hour, and can sometimes run into 3-4 hours when there are "complications". However, by manipulating my physiological responses to a few critical control questions, and pretending to be appropriately intimidated and impressed by the investigator and his machine, I was out of there in 15 minutes, which I was later told was something of a record.
From http://antipolygraph.org/ [antipolygraph.org]:
The dirty little secret behind the polygraph is that the "test" depends on trickery, not science. The person being "tested" is not supposed to know that while the polygraph operator declares that all questions must be answered truthfully, warning that the slightest hint of deception will be detected, he secretly assumes that denials in response to certain questions -- called "control" questions -- will be less than truthful. An example of a commonly used control question is, "Did you ever lie to get out of trouble?" The polygrapher steers the examinee into a denial by warning, for example, that anyone who would do so is the same kind of person who would commit the kind of behavior that is under investigation and then lie about it. But secretly, it is assumed that everyone has lied to get out of trouble.
The polygraph pens don't do a special dance when a person lies. The polygrapher scores the test by comparing physiological responses (breathing, blood pressure, heart, and perspiration rates) to these probable-lie control questions with reactions to relevant questions such as, "Did you ever commit an act of espionage against the United States?" (commonly asked in security screening). If the former reactions are greater, the examinee passes; if the latter are greater, he fails. If responses to both "control" and relevant questions are about the same, the result is deemed inconclusive.
The test also includes irrelevant questions such as, "Are the lights on in this room?" The polygrapher falsely explains that such questions provide a "baseline for truth," because the true answer is obvious. But in reality, they are not scored at all! They merely serve as buffers between pairs of relevant and "control" questions.
The simplistic methodology used in polygraph testing has no grounding in the scientific method: it is no more scientific than astrology or tarot cards. Government agencies value it because people who don't realize it's a fraud sometimes make damaging admissions. But as a result of reliance on this voodoo science, the truthful are often falsely branded as liars while the deceptive pass through.
Perversely, the "test" is inherently biased against the truthful, because the more honestly one answers the "control" questions, and as a consequence feels less stress when answering them, the more likely one is to fail. Conversely, liars can beat the test by covertly augmenting their physiological reactions to the "control" questions. This can be done, for example, by doing mental arithmetic, thinking exciting thoughts, altering one's breathing pattern, or simply biting the side of the tongue. Truthful persons can also use these techniques to protect themselves against the risk of a false positive outcome. Although polygraphers
Wrong again.... (Score:4, Informative)
Jim Nicholson - convicted of spying for Russia
There's two. There's hundreds found... and even many more before they get off the ground- how many people could be compromised had they been given access?
Re:Polygraphs ... (Score:3, Informative)
Not the chinese. Their favorite method is to find someone of chinese origins and then convince them that sharing information is a duty of one sort another, patriotic or for the good of any family they still have back in China. That's an over simplification, but the chinese methods are much more passive and much harder to "catch" than traditional western espionage techniques.
They do not work (Score:5, Informative)
Re:Polygraphs work--sorta (Score:5, Informative)
1) Relying on people believing a lie is bad policy. Especially if some of those aware of the truth (that it's a lie) refuse to take it, and are then fired.
2) And this 'sorta working' is very unreliable. Even if you know it's a lie, you can still get nervous while taking the test - just because you know that all they are doing is checking how nervous you are. So the tester can't know what a person acting nervous during the polygraph means - could be guilty, could be innocent.
3) Actually, polygraphs can be used in a scientifically correct manner, but nearly never are. The WRONG way is to ask questions like 'did you kill Mike?' - which make anyone nervous, guilty or innocent. The RIGHT way is to do a randomized statistical test, as follows: say Mike was killed by a shotgun, a fact which only the police know. You can then ask the suspect the following questions: "was Mike killed by a rifle?" "[...] a shotgun?" "[...] a handgun?" "[...] a knife?" etc. etc. Only the killer would know the true weapon, so if your suspect reacts differently to the 'shotgun' question, that would be informative. Of course other elements would also have to be statistically accounted for: you'd need to ask several controls the same questions (just to see that "shotgun" isn't a word that evokes special responses in general); to randomize the order of the questions; to have the person asking the questions not know the answers; and so forth. Basically, to do the same things you'd do in a scientific experiment.
But this is (a) hard and time-consuming, and (b) not always possible (you need information only the killer would know).
Re:Polygraphs work--sorta (Score:5, Informative)
Common polygraph use assumes that lying causes increases in vital measurements, yes. But notice that what I mentioned in my post was a difference in vital measurements. If you get enough randomized trials, you can conduct a statistical test just like of every other scientific hypothesis.
Yes, perhaps some people react more to "shotgun", and some respond less. If, out of 100 people, the suspected killer reacts in a not-statistically-significantly-different manner, then that is one thing. But if, on the other hand, he reacts in a unique way, then the odds of that occurring were he not privy to information about the shotgun would be 1% (i.e. the Null Hypothesis is that all 100 people tested are the same, so the chance that a single person has a different result by chance, and that that person is our suspect, is at most 1 in 100 - speaking in general terms).
Increases in vitals like body temperature, perspiration and heart rate correlate with nervousness, not deception.
Agreed, which is why an increase in these vitals in a single individual is not enough, by itself, to show anything.
Furthermore, a suspect reacting to the word "shotgun" is not informative in the slightest. The shotgun from Doom might've just been his favorite weapon in that game. Or he might have some other past traumatic experience with a shotgun. It means nothing.
As I said above, this is possible, yes - it can occur by chance. But by a correct statistical test, you can check whether the reaction is explainable by coincidence or not. This is exactly the same way surveys are done or experiments in medicine or the social sciences. (Of course it isn't perfect, but then nothing is 100% perfect; the law can convict above a reasonable doubt.)
Polygraphs are just another interrogation tool to make the suspect feel more powerless and make the interrogator look more powerful.
Agreed. Polygraphs, as they are commonly used, are useless or worse than useless (dishonest, easily abused, etc.). But what I wrote in the post you are responding to is something completely different.
A note about the basic science behind this stuff: there is plenty of evidence of bodily responses to familiar stimuli (for example the cognitive psychology literature on 'priming', also electrophysiology, etc.). However, the commonly-used polygraph may not use the measures proven to work. If all it does is test blood pressure and GSR (galvanic skin response), then we may be right to be skeptical (although perhaps research on GSR has improved in recent years - I don't know). However, things like EEG are also non-invasive and easy to test, and research has shown them to be informative about various things. So: even if the commonly-used polygraph is a sham, correct use of science and statistics can be used to devise a better method, and hopefully things will continue to progress in that direction.
Re:Polygraphs ... (Score:3, Informative)
Most (not all) of the charges against Wen Ho Lee were dropped; Lee plea bargained. As a physicist, I know people who know people at LANL, and usually up on the general lab gossip, but I don't actually know the reality of Lee's case. He may in fact have been spying and the government gave him a deal because the evidence was weak, or he may have been loyal and they just nailed him for protocol violations to save face. His particular case is not totally clear.
What is clear is that the PRC is running a massive intel campaign against the US, and much of it centers around getting military and high-tech secrets. The OP made it clear that Chinese patriotism was only one of the tools used; sometimes threats against family still in China will be made, and one doesn't have to be patriotic to be susceptible to that. Incidentally, Russia is also still conducting lots of intelligence ops against the US. Of course, this doesn't mean that all or most Chinese or Russians in the US are spys--just, that a few are.
Re:Polygraphs work--sorta (Score:3, Informative)
Re:Bogus statistics (Score:3, Informative)
As far as random screening is concerned, you must consider the positive and negative predictive values of the tests. A very good test might have, for example, a 1% false positive rate and a 2% false negative rate. It is commonly assumed -- falsely -- that a test with 1% false positive rate has 99% predictive value; that the subject is 99% likely to have whatever is being tested for.
From the false positive and false negative rates you have to compute the positive (and negative) predictive value -- that is, the probability that somebody who tests positive (or negative) really has (or does not have) what the test shows. To compute positive predictive value you need to know the prevalance in the population being tested. Suppose the prevalance is 1 in 1000 and you test at random. That means that for every true positive you'll get ten false positives. That is, the positive predictive value is 9%. A far cry from 99%!
Now to compute false positive and negative rates of the order of 1%, you need sample sizes of at least several hundred -- probably thousands. I don't believe there is any physiological test of truthfulness that has shown anything even resembling a statistically significant result, which is why I take great exception to your statement:
The controlled experiments for polygraphs have shown between 40% and 70% false positive and false negative rates which, for the sample sizes used, are indistinguishable from chance.
Not one? (Score:2, Informative)
Spin away!