The machines can be fooled at will by anyone who knows how they work or anyone who watched the Penn & Teller ep about them.
Experts (judges, policemen, psychiatrists, etc.) identify lies 50% of the time under controlled circumstances. At least according to John Cleese in his documentary series called Face. It’s an odd cite, but it matches other things I’ve read that I don’t remember nearly as well. Oh, and all of the experts, when asked, claimed that they were good at spotting lies.
Face reports that the 1% who identify lies exceptionally well are actually reading micro-expressions: facial expressions that last less than a second. Reading micro-expressions can be learned, but interpreting them can be tricky. The FBI does have a program for teaching the reading of micro-expressions. So if Laggard’s friend has taken the training, he may be more accurate than a tossed coin. If he’s saying he can “instantly tell” as if he can always tell, though, he probably hasn’t had the training. Or he thinks it’s good SOP to make the claim.
The show Lie to Me is based on micro-expressions and other bits of research that show ways to catch lies that work better than 50% of the time. It has, of course, been Hollywooded beyond all recognition. Oddly, a recent study (don’t remember which science blog reported it) showed that people who watch Lie to Me are less able to detect lies than average. So watching the show will do you no good.
Trust but verify. And don’t trust anyone waving lie detector results around.
This agrees with what Richard Wiseman said in the book Quirkology that I referenced earlier. Experts are as good as flipping a coin.
I find it interesting that Cleese talks about watching the facial movements to increase that percentage, but Wiseman did studies that showed people do better if they can’t see the face. He did things like analyzing the number of words in the answer, counting the number of times people say “I” or “me,” and other lexical statistics. That improved the percentages to somewhere in the 70’s (I don’t remember the exact number), which is still a far cry from 100% – or even 90%.
As has been mentioned studies have been done on this and the police are (IIRC) just a bit better than flipping a coin to determine if someone is lying.
Good interrogation techniques however can get you good results. Ask the perp questions. Ask them again in a different way. Look for inconsistencies and exploit them. That is not a lesson on interrogation techniques but merely me saying that with some work an FBI agent very well may be able to reveal a liar with some regularity. I just doubt they can do it on the spot reliably.
Indeed if he thinks he “knows” he can spot a liar “instantly” I would suggest this makes him a poor agent when questioning. His bias will assert itself and color his investigation from word-one.
Sigh. I may as well put that on my Amazon wish list now. I’m going to end up buying it eventually.
I think Pen & Teller recommended clenching your butt cheeks during the control questions or questions you were answering truthfully to and unclenching during your lies. Poetic, really.
One of the episodes of Lie to Me had had a great visual for the prevalence of false positives. To discredit an ‘improved’ lie detector being touted to the FBI, the main character had an assistant with significant cleavage sit on the table near the machine and re-ask the questions while leaning towards the test subject. Suddenly the machine was marking every answer as a lie.
I was in high school and going for a job at the brand new, not opened yet, T.G.I. Fridays opening near my school (back then TGI Fridays was a fairly new chain…this wasn’t the first ever but they were a lot smaller than they are today…this is like 1983-84).
I was going for a job as a bus boy and they were making every applicant submit to a polygraph (shit you not).
Even at that age I knew how a polygraph worked so when the questioner asked me the beginning control questions I intentionally held my breath or clenched or paused. Apparently he could tell he was getting readings that would throw it off so he kept repeating the control questions (e.g. what is your name, how old are you, where do you live and so on).
Eventually I stopped fucking around because I wanted the job. I just wanted to see if I could screw up the test and I could. It was trivially easy
We then proceeded through all the questions and at the end I asked the interviewer about one question. He had asked, “Are there any hidden reasons you want this job?” I recall, when the question was asked, wondering what the hell that meant. So I thought about it trying to sort it in my head. Coming to no good conclusion I answered “no”. (FWIW he told me the question was to ask if I was a spy or somesuch to learn the secrets of TGIF.)
When I asked him about that he smiled and said that was the only question I “failed”.
I got the job (which was a good one for a high school kid…we got tip sharing from the waiters). Still, it taught me the polygraph was trivially easy to mess up. If a 16-17 year old kid can manage it who didn’t know shit from shineola I cannot imagine it is useful in the larger world.
I suppose someone who is not the least bit knowledgeable about how the thing works and is honestly sitting there answering might produce better than coin-flipping results but a modicum of knowledge will ruin its reliability. They are not hard to beat.
I bring this up every time there’s a polygraph thread. I worked for years at a small company that sold medical sensors and software. I was often hooked up to a full range of sensors to show clients how to use the things.
Polygraphs stink at detecting lies. Heartbeat, respiration, galvanic skin response etc can all be controlled with proper training. More than that, screwing up the readings is easy if you feel like it. The majority of the training was on how to get a nice and accurate reading free of artifacts.
I think it was Oliver Sacks who wrote in one of his books about brain injuries about how aphasics (people who lose language) can tell if someone is lying, even though they can’t tell what the person is saying. He mentioned them finding political speeches on television to be hysterical.
I recall reading about a study some years ago where the ability of members of different branches of law enforcement were tested for their ability to detect lies. The only branch that was better than chance at detecting lies was the Secret Service; the speculation was offered that it might be because their job centers so much on watching people, observing them closely to see if they are a threat.
My girlfriend back at that same time (1983) was applying for a job at Radio Shack. They (or at least, that store) made applicants take polygraph tests, too. (She apparently passed, as she got the job.)
I met a guy once who administered polygraph tests for a living. He said the key when you thought a person was lying was to convince him that the machine proved he was lying, and then use that to get him to confess. Once you got him to confess you would have him write out and sign the confession and once you had that the polygraph didn’t matter anymore.
He said it worked great for convincing people that they were bad liars.
Besides the problem of training to control responses; false positives because people are upset but not guilty (I often enough think “I know that I’m saying the truth, but will people who don’t know me believe me because it sounds unusual?” which stresses me and makes me blush); and people who don’t feel guilt for that particular thing (sociopaths; conscience clear for that question); there’s also the problem of memory.
In a criminal or important session, the questions need to be not only “Did you kill Y?” but also “Did you do X?” “Where were you on …”
But human memory is very suspectible to suggestion. (See several cracked articles). If either the suspect has himself convinced of an innocent version of events he can pass; or if the investigator asks questions that lead the suspect to remember a different, guilty, version of events he can fail.
See the cracked article that confrontational normal interrogation gets about 1/3 false confessions (that is from innocents), but the soft approach (come on, just confess, you’ll feel better, we understand how you reacted) gets 2/3 false confessions!
Another anecdote about being able to control the output from machines – in this case not a lie detector but an ordinary BP/pulse monitor. My dad was in the hospital after some surgery or other, hooked up to one of these. The nurses had to keep going in to check on him because the machine seemed to show strange blips from time to time – sudden changes in pressure or heart rate. Finally someone noticed that the monitor was within dad’s view. He was *doing it on purpose *to see if he could make the lines change and by how much. The nurses moved the monitor. The strange blips went stopped.
That’s what I want to know. I’d like to do high-level security work at some point in my life, but I have an overactive conscience and could easily see myself getting a false positive on one of those things. Why in the world are they still using them?
WAG: It’s easier to put the blame for a false decision in the hiring process onto a machine than having the guts to make the decision of yes/no by a person who then has to bear responsibility for a possibly bad decision?
