Truth or Lie

January 21, 2004 5:24 AM

Devices like these or these may be able to detect lying better than a polygraph and less intrusively. If this technology really really worked and everyone had access to it, say, in their cell phones, would life be any different? How would it change anything, if at all? Sex? Politics? Hip-hop?
posted by ewkpates (36 comments total)

Doesn't a banana detect lying better than a polygraph? I thought the use of polygraphs as "lie detectors" was completely discredited.
posted by majick at 5:32 AM on January 21, 2004

I think I would become extremely direct with people, actually I'd be downright abrasive.

Me: So, why does your product not work?
Suit: blah blah blah
Me (points finger directly at their nose): Bullshit!
posted by will at 5:37 AM on January 21, 2004

I really hope the "eyeglass lie detector" company puts the rush on the development of it's lie-detector glasses and gets a number of them into American homes before the 2000 election, so Americans can appreciate the sterling honesty of their politicians.
posted by troutfishing at 5:59 AM on January 21, 2004

When technology prevents people from lying, they will use that technology to create giant robots to lie for them.
posted by cardboard at 6:10 AM on January 21, 2004

Uh, seriously, don't we kinda know most of the time when peole are lying? And where do I get one of those giant robots? Oh wait--the advertising juggernaut, I get it.
posted by mecran01 at 6:17 AM on January 21, 2004

I really hope the "eyeglass lie detector" company puts the rush on the development of it's lie-detector glasses and gets a number of them into American homes before the 2000 election, so Americans can appreciate the sterling honesty of their politicians.

Psst. It's a lie detector not a time machine!

And, yes, this was my thought exactly. I can just imagine watching a debate where all the participants, moderators and spectators had these on.

Was it Bruce Sterling that wrote somewhere that we all have to get used to the fact that "privacy" is basically over and done with? If we could no longer lie credibly one of our main tools for protecting our privacy would be destroyed.
posted by rocketpup at 6:18 AM on January 21, 2004

For a really thought provoking take on this concept I highly recommend "The Truth Machine" by James Halperin. What would the future be like if someone invented a machine that was 100% accurate at detecting lies? A very sobering idea in light of the current political situation we are facing here in the U.S.
posted by waltb555 at 6:20 AM on January 21, 2004

When technology prevents people from lying, they will use that technology to create giant robots to lie for them.

After the first 18 months on the market, the robots will become smaller and more affordable, forcing early adopters to use their robots to lie about how happy they are with the first generation technology. At this moment, these robots won't be fooling anyone.
posted by .kobayashi. at 6:20 AM on January 21, 2004

The skeptic's dictionary has a good entry on polygraphs with links to articles on related technology such as mentioned in the front page post.

"I don't know anything about lie detectors other than they scare the hell out of people."
--Richard Nixon
This quote at the beginning of the article pretty much sums up the usefulness of lie detectors, no matter what technology they use.
posted by TedW at 6:32 AM on January 21, 2004

The use of lying proxy-communication robots would immediately discredit the user as a liar. This tin-man argument is a straw-man, say I!
posted by Blue Stone at 6:32 AM on January 21, 2004

What happened to that software that detect lies only by analyzing speech?
posted by nkyad at 6:41 AM on January 21, 2004

IIRC, you can beat just about any test with enough practice and it doesn't work if you can't tell the difference between fact and fiction.

Either way, it won't work on politicians.
posted by twine42 at 6:42 AM on January 21, 2004

What are they doing messing around with this stuff when they still haven't given me the X-Ray glasses I was promised 40 years ago.
posted by mss at 6:58 AM on January 21, 2004

eyeglass lie detector - it sounds like something you'd find in the back of a comic book, along with the x-ray glasses.
posted by ashbury at 7:04 AM on January 21, 2004

mss, you weren't there when I opened this thread! :)
posted by ashbury at 7:04 AM on January 21, 2004

Liar ^ :o)
posted by Eirixon at 7:31 AM on January 21, 2004

Hmm, I can't pull up any peer-reviewed publications by Amir Lieberman, so I don't think this passes the bullshit detector. Even if it were real, a 7% increase in accuracy over a polygraph is not really much to crow about.

The company said that a state police agency in the Midwest found the lie detector 89 percent accurate,

So, a single trial that was probably not rigorously constructed found that it is wrong 1 time out of every 10? Plus, what kinds of errors are these: false positives or false negatives?

Instead of color-coded LEDs, a bar graph on the display indicates how much the caller to whom you are speaking "loves" you. V Entertainment claims the love detector has demonstrated 96 percent accuracy.

96% percent accuracy in measuring the degree of love someone is exhibiting? This company has no credibility.

You might as well get one of these tchotchkes.
posted by toothless joe at 8:11 AM on January 21, 2004

The facts, when available, are 100% accurate at pinpointing liars -- but interestingly enough, when it comes to politicians it doesn't make a damn bit of difference. The problem isn't and has never been that we can't tell when they're lying, it's that too many of us don't want to know.
posted by George_Spiggott at 8:17 AM on January 21, 2004

How does one test the accuracy of a lie detector? Do you tell the subject, "OK, now tell a lie." However, if the subject know the examiner wants a lie, is that an actual lie?
posted by mischief at 8:22 AM on January 21, 2004

Just hire an aphasic to hang out with you.
posted by donpardo at 8:24 AM on January 21, 2004

When I was given a polygraph exam, the machine was calibrated. I selected a card (it had a number on it) from a mix of cards with numbers or letters. The examiner hooked me up to the machine and went through all of the available cards, including the one I had selected, asking "Does your card contain the number/letter x?" (x being one of about a dozen). I was to answer no to all of them. He knew right away which was the number that I had chosen, and told me I was like an open book.
posted by kelrae3 at 8:55 AM on January 21, 2004

After the first 18 months on the market, the robots will become smaller and more affordable...

...and design flaws will be uncovered that permit anybody with access to the robonet to execute arbitrary lies on your robot. RoboSoft will launch a massive FUD campaign promising that henceforth security will be job #1, but it'll turn out to be a lie.
posted by quonsar at 9:31 AM on January 21, 2004

Lie detectors are witches brew. There's no way of proving they're right ... the only confirmation has to come from a "known" liar.

First somebody should make a lie detector that lies. Then you'd have something reliable to calibrate the other lie detectors from.

Some human beings actually know the truth some times. Some actually tell the truth sometimes. The rest aren't even sure. Completely unreliable.

posted by Twang at 10:04 AM on January 21, 2004

someone once said about a device constructed to detect holes in the ozone layer that of course it had detected an ozone layer hole. that's what it was designed to do. you couldn't know if it worked or not until it detected one, and if it never detected one, you couldn't be sure whether there weren't any ozone holes, or whether the machine simply wasn't detecting them.
posted by quonsar at 10:30 AM on January 21, 2004

Thanks, Twang, for your well-informed opinion.

BTW, it's called a Polygraph Exam, not a lie detector.
posted by kelrae3 at 10:51 AM on January 21, 2004

I was to answer no to all of them. He knew right away which was the number that I had chosen, and told me I was like an open book.

The interesting thing about that is usually the cards are marked. It gets the subject to believe in the power of the polygraph exam, while the operator really hasn't even used it.
posted by eschewed at 11:13 AM on January 21, 2004

So, a single trial that was probably not rigorously constructed found that it is wrong 1 time out of every 10? Plus, what kinds of errors are these: false positives or false negatives?

Let's give the company the benefit of the doubt: assume a 10% rate of false positives and 0% false negatives (i.e. all liars are caught, along with 10% of non-liars). Consider now an application of the device: screening crime suspects. Let's say that for a given crime, we have 100 suspects, one of whom is the actual criminal. The lie detector will finger 11 people, one of whom is the target individual. In other words, it will be wrong over 90% of the time.

Let's consider now an application in screening terrorists at airports. If we're very generous about the abilities of terrorist organizations, we can assume that about 1 out of every 10 million airline passengers is a terrorist (I would think that the actual number is much lower: there are over 600 million airline passengers annually in the US alone). For every 10 million passengers screened, the device will identify as "liars" 1 terrorist and 1 million innocent airline passengers. It will therefore be wrong in identifying terrorists 99.9999% of the time.

The situation gets worse when you factor in false negatives (especially since devices like this are easily gamed). This sort of device is, for all practical purposes, useless, even when you grant the validity of the numbers provided by the device's manufacturer.

BTW, it's called a Polygraph Exam, not a lie detector.

Actually, most Americans do call it a "lie detector", despite the existence of a psuedotechnical name. I would bet that were you to use the word "polygraph" in a standard conversation, your most typical immediate response would be: "You mean a lie detector?".
posted by mr_roboto at 11:17 AM on January 21, 2004

The examiner showed me the graph representation of my polygraph calibration. It was clear to me (having never taken one, or seen results of one) at which point I had 'lied' about which card I had chosen. I don't believe that the cards were marked.

I'm actually Canadian, not American, and haven't had anyone say "Oh you mean a lie detector?" when I have discussed the polygraph exam.
posted by kelrae3 at 11:25 AM on January 21, 2004

Let's ask Google!

Searched the web for "polygraph exam". Results 1 - 10 of about 4,350. Search took 0.16 seconds.

Searched the web for "lie detector". Results 1 - 10 of about 102,000. Search took 0.07 seconds

I think we can certainly conclude that the sentence "It's not called a lie detector." is untrue; many, many people are, in fact, calling it a lie detector.
posted by mr_roboto at 11:35 AM on January 21, 2004

First of all, mr_roboto, I'm not disputing it's referred to as a "lie detector". Yup, I've heard it called that too. Due to having family/friends in law enforcement and involved in the justice system, it's most often called a Polygraph exam in my social circles.

Second, I was more or less being snarky in response to Twang's comment that they are 'completely unreliable', relating his knowledge of the polygraph and his referral to it as a 'lie detector'.

Gotta love Google. Thanks!
posted by kelrae3 at 11:57 AM on January 21, 2004

Oh yeah, and one more thing:

Searched the web for polygraph. Results 1 - 10 of about 266,000. Search took 0.10 seconds

posted by kelrae3 at 12:02 PM on January 21, 2004

"look into my eye"
posted by clavdivs at 12:05 PM on January 21, 2004

A polygraph by any other name is still just as unreliable.

Seriously folks, a reliable "lie detector" is impossible. At some point, it will rely on a human being to interpret its results, or to provide input to it, or to maintain it, or to calibrate it, or at least to design it. The interesting thing about deception is that it isn't always obvious, even sometimes to the deceiver. Then there are problems with defining "reliable" itself. Who defines the metrics? It is easy to construct problem scenarios.

As a final note, regarding the love detector: Let's assume that it is reliable, to some extent as yet undefined, by metrics as yet undefined. All it could possibly tell is that the person behind a voice is experiencing some love-like emotion. What it cannot tell you is if that affection is for the person holding the detector, for some other person in the vicinity, or for a person temporarily but recently held in the memory of the person talking. My voice may well reflect my amorous intent, but it will not identify the specific object of that intent unless I should utter a name.
posted by yesster at 12:31 PM on January 21, 2004

Oops. I meant the 2004 election there. I guess that makes me an angry liberal, stuck in the past.

clavdivs - That's pretty good too, especially if your subject glances up and to the left. Then you can assume they are lying.

BUT.... if you pair up an aphasic (Donpardo's "lie detector") with an agnosic (That's agnosic. An agnostic simply will not do) you've got the finest lie detector anyone could possibly want.
posted by troutfishing at 1:44 PM on January 21, 2004

The lie detector will finger 11 people, one of whom is the target individual. In other words, it will be wrong over 90% of the time.

Good stuff. I recommend more people check with the Bayesian calculator when confronted with these sorts of 90% correct type stats.
posted by vacapinta at 3:27 PM on January 21, 2004

rocketpup: Was it Bruce Sterling that wrote somewhere that we all have to get used to the fact that "privacy" is basically over and done with?

That's probably David Brin's book The Transparent Society. Good thought-provoking read.

He's an aphasic with a checkered past. He's an agnosic with a heart of gold. Together, they fight crime!
posted by hattifattener at 9:04 PM on January 21, 2004