Realism check: How accurate are lie detectors?

[nexus]

In the real world, how accurate are polygraphs at detecting falsehoods?

[Remjin]

Re: Realism check: How accurate are lie detectors?

 

Lie detectors, or more accurately Polygraphs, are reasonably accurate. It really depends, honestly.

 

They work by monitoring your respiration rate and depth via a spring and by monitoring your pulse. Thus, if you have a conscience, they work well. If you have an overly guilty one, they overdo it, and if you have none... they don't work at all.

 

When most people lie and are put on the spot they get nervous, their hearbeat increases in pace, and they may respirate more quickly or in shorter breathes. Thus, they are lying. If they feel no guilt then nothing happens. This method is considered invalid in courts because the spread can be quite wild due to the varying nature of people. However, a large majority of people do have a conscience or at least a fear of being caught, and thus it usually works... but they can be beaten, deceived, and are many times not conclusive one way or another.

 

I've taken two of these tests (required in my field of work sometimes, despite all of this). I thought I'd fail one question because I had a weird moment where I remembered stealing as a child, then relaxed when I remembered that they had asked if I'd stolen anything in my adult life. I took a toy car from the store at age 2. hehe. Oversensitive I guess.

[Bunny]

Re: Realism check: How accurate are lie detectors?

 

Why didn't you ask me this last night?

 

We had a discussion about this in one of my classes a while back. Nowdays, they have computer polygraphs that are generally considered to be nearly 100% accurate. But the expertise of the user is what makes mistakes usually. It's sort of like diagnosing a rare disease. An examiner may not be able to distinguish the results, and instead of putting "inconclusive", he puts down that there were errors in the test. This changes the accuracy rate of the test. Instead of being able to toss out the inconclusive results, they use the inconclusive results as inaccurate. They are still the most controversial tool in law enforcement by far. Correct test patters have to be established of course, to lay down what the person's bio output is when they tell the truth. They'll be nervous anyway, of course, but they'll be nervous on normal questions rather than just on the important ones, thus letting a decent groundwork be laid for any changes in their output. The test is only as good as its operator, then, I guess. We never settled on anything in our class discussion. It's effective enough in some circumstances in law enforcement, at this point in time, for it's used a good deal for the results just to be tossed out. An argument could be made I suppose on the expertise of the operator and the experience and equipment of the company issuing the test. More importantly would be, in what government (state, country, etc.) are you using it? Some put more stock into it than others. In some places, you can take one then toss it right out because it's not good for much.

[Comic]

Re: Realism check: How accurate are lie detectors?

 

In game terms, the real world technology is far more reliable than the superheroic powers get to be based on dice rolls. They don't have to hit the target's ECV or have to hit Ego +, they don't have a PER modifier, they don't always fail on a roll of 18.

 

They take time, require collaboration from the person being tested (though not necessarily full cooperation -- hence the point of detecting lies), need to be done by a (somewhat) skilled operator and must be interpreted based on the limitations of what they are: not truth or lie detectors, but biological monitors corresponded to psychological interaction.

 

Built in game terms, they combine discriminating & analyze (for the more advanced devices) detect somatic response, plus what is probably best described as the conversation skill (with the limitation that it applies only to question and answer interactions) and the skill analyze humanoid psychology.

[McCoy]

Re: Realism check: How accurate are lie detectors?

 

A lot of what Bunny said. Accuracy varies widely, and depends more on the operator's expertse more than any other factor.

[Trebuchet]

Re: Realism check: How accurate are lie detectors?

 

They did a bit about this recently on Mythbusters. If the operator of the polygraph is skilled and gets enough time fr questioning (3 - 4 hours), it's almost perfectly accurate. Most of the things that supposedly can confuse the machine, such as stepping on a tack to induce pain, in real life can't.

[Steel Legs]

Re: Realism check: How accurate are lie detectors?

 

They are stress detectors, not lie detectors. When used right they mix in innocent questions with "everyone's guilty of this" questions at the start, then get to what they want to know. If you stress out over the innocent questions and calm down on the EGOT ones, the baseline is screwed up. Keep it up during the stuff the want to know and the results are futzed up.

 

It ain't magic, it's a rough guide at best. Polygraphs are beaten all the time.

 

I don't know what you're taking classes in, Bunny, but if it's cop-stuff, they shouldn't lie to you like that.

[Toadmaster]

Re: Realism check: How accurate are lie detectors?

 

Inaccurate enough that they are not allowed in normal court proceedings and are a hot topic in employment as is psychological profiling. There may be some valid uses but what is acceptable accuracy? 90% is an A but would you be willing to tell 1 in 10 people that they are lying when they are not?

 

I guess you can put me in the sceptical camp, there are a lot of people who seem to be pretty good at telling when someone is lying without a machine too, but not all of them know what they are talking about either.

[Psylint]

Re: Realism check: How accurate are lie detectors?

 

In the real world polygraphs are 100% inaccurate at detecting falsehood, because they do not measure falsehood. They measure certain physiological responses such as respiration rate and galvanic skin response. Assuming the theory that these measures correlate to an attempt by the person measured is being deceptive, at best they determine incidences when that person is attempting to deceive. They do not measure falsehood.

 

Have you ever been so sure of something, you could've sworn, and only to find out later that you had been completely wrong? My father swore up and down, there are 7 Sacraments and marriage was one of them. I explained, but Dad, you're an Anglican, Henry VIII, Ann Boelene, the Act of Supremacy, any of this ringing a bell? But he persisted until I showed him in the Book of Common Prayer. If you had put him up to a lie detector, it would not show that he was being deceptive, even though his statement is 100% false.

 

Polygraphs are not admissible as evidence because in the view of the relevant scientific community they are not sufficiently accurate and reliable.

 

Besides, think about it, if polygraphs were so sure-fired reliable, why would the CIA and other folks be using water-boarding? Just hook 'em up to polygraph.

[Hyper-Man]

Re: Realism check: How accurate are lie detectors?

 

..

 

Besides, think about it, if polygraphs were so sure-fired reliable, why would the CIA and other folks be using water-boarding? Just hook 'em up to polygraph.

 

 

Making sure that a person's answers to questions are truthful and actually compelling that person to give ANY answers at all are 2 completely different things. Plus, Polygraphs only use yes/no questions. Interrogation frequently uses more open ended questions (who, what, where, when, why, etc..).

[megaplayboy]

Re: Realism check: How accurate are lie detectors?

 

what about voice stress analysis? Any accuracy, or is that just pseudoscience?

[BoneDaddy]

Re: Realism check: How accurate are lie detectors?

 

While we're at it, neurolinguistics should get a nod. It's helped me spot a few of the less adept liars in the world, and I'm a coffee table amateur at it.

 

Sicilians are great liars. The best in the world. I'm Sicilian. My father was the world heavy-weight champion of Sicilian liars. From growing up with him I learned the pantomime. There are seventeen different things a guy can do when he lies to give himself away. A guys got seventeen pantomimes. A woman's got twenty' date=' but a guy's got seventeen... but, if you know them, like you know your own face, they beat lie detectors all to hell. Now, what we got here is a little game of show and tell. You don't wanna show me nothin', but you're tellin me everything. I know you know where they are, so tell me before I do some damage you won't walk away from. [/quote']

[Bloodstone]

Re: Realism check: How accurate are lie detectors?

 

I think I'd get banned if I posted Dennis Hopper's response to that

[Gadodel]

Re: Realism check: How accurate are lie detectors?

 

I've taken one in order to get my current job. In my opinion, it was an interesting little mind game. It was a worthwhile experience to obtain.

[Basil]

Re: Realism check: How accurate are lie detectors?

 

Plus' date=' Polygraphs only use yes/no questions. Interrogation frequently uses more open ended questions (who, what, where, when, why, etc..).[/quote']

 

Polygraph users can and will ask any question with a definitely true answer. Who, what, etc., will all be used.

 

If the victim can show there's room for a "not utterly true" answer, the question will be thrown out.

[Wyrm Ouroboros]

Re: Realism check: How accurate are lie detectors?

 

Polygraph and voice stress analysis both work best if there are plenty of constant 'control' questions -- and you don't look for lies, you look for guilty knowledge. The example I was given went something like this:

 

The 5/3 Bank on Louis St. was just robbed. You have a suspect in custody. Your questions have the pattern along these lines:

 

"Did you rob the Bank of America on Galway Rd.?"

"Did you rob the Hometown Bank on 7th St.?"

"Did you rob the 5/3 Bank on 7th St.?"

"Did you rob the Bank of America on Louis St.?"

"Did you rob the 5/3 Bank on Louis St.?"

 

Typically those get mixed in with other ones -- plenty of control questions, some with partial information (5/3 Bank OR Louis St.), some with full information. The polygraph reads the increase in stress (or lack thereof) when the questions with 'guilty knowledge' roll around. This, again, works best if the person isn't told the full details of who, what, etc. -- they could only know if they committed the crime.

 

And there are few people who have the skill or the lack of conscience to totally avoid these sort of responses.

[Karmakaze]

Re: Realism check: How accurate are lie detectors?

 

My father tells a story about getting himself in a bit of trouble when he was in the Navy during a routine security check because his mind was wandering and so his reactions weren't quite in synch with his answers. (I suspect folks being given criminal interrogations are less likely to be bored, so it was an idiosyncratic event, though.)

[Vondy]

Re: Realism check: How accurate are lie detectors?

 

Like the Emperor's New Clothes, eh?

 

Lie detectors (the traditional polygraph) are interpretive-subjective and not admissable in court. They rely entirely on the intuituion, perceptions, and art of the operator. They're about as good as an ink-blot test in diagnosing mental illness. They don't identify fallacy - they identify nervousness. As such, they can help you identify areas for further investigation/questioning, but cannot certify a lie by themselves.

 

The congressional general accounting office published a report on polygraphs (as they are used extensively in screening) and noted no espionage agent has ever been identified (or caught in a lie) by a polygraph. They also noted their lack of efficacy in accurately identifying suspects or qualifying candidates when used by federal law enforcement agencies. Indeed, the GAO report questioned federal law enforcement and intelligence agency's reliance on the polygraph and noted there are more effective technologies (though still not infallible, and still requiring further investigation of a subjects answers).

 

Additionally, many polygraph experts (incl. a few I have known personally) are critical of its normative application (not video taping sessions, not providing subjects with copies of the results, not allowing subjects to explain nervous results or retest). Also, there is a tendency to assume the polygraph can do more than it really does (and to therefore use it incorrectly). People tend to think it sees into your soul. It doesn't - it just says you were nervous when you answered question X and might have been deceptive. Its only as good as the interviewer - and what if he screws up?

 

In general, and this was the conclusion of the GAO report, a polygraph is primarily useful as a means of leveraging confessions of guilt or admissions during an interview process because 1) the subject is convinced of its infallibility, or 2) results indicating nervousness cause an admission. A second use it identifying areas to focus on in a REAL interview/interrogation. Even then, the answer may be associative as opposed to deceptive.

 

I'm sure there are some cases where a suspect was clearly lying and the polygraph results showed intense nervousness and the interviewer called a spade a spade, but detecting a lie takes more than that: correct questions, correct interpretation, and follow up by way of good old fashioned questioning. There's a reason a polygraph can't send you to jail.

 

Other technologies: voice stress analysis, retinal response measurment, and the like, are more effective, but still rely on the skill of the evaluator - and further questioning.