The Lie Behind the Lie Detector | Hacker News Reader

The Lie Behind the Lie Detector

antipolygraph.org

221 pointsgiles_corey96 comments

Every lie detector should begin with a game of high-low: The subject writes down a number between 1 and a 1000 and puts it in their shoe, the questioner needs to get the right answer. Since the questioner needs only 10 questions to get the right answer (+ control questions), it provides a practical calibration of whether the test is worth a damn or not.

Prediction: No one will ever use my practical calibration of whether the test is worth a damn or not, until lie-detector technology is actually worth a damn.

To play devil's advocate, a lie detector isn't really supposed to measure meaningless lies. It's about emotion and nervousness and the like. Math questions about a randomly assigned number would likely not generate much of a meaningful signal, and thus this test would not be expected to work.

Then just make it meaningful by threatening to shoot the interviewee in the head if they fail to trick you.

Yeah, that would totally be acceptable. /s

Could use blanks or a paintball gun.. would probably be enough to get a baseline stress level.

You could always offer the subject a hundred bucks (or some other significant amount) if he is able to obfuscate it.

It sounds like you're arguing that a lie detector only detects lies that the subject considers "meaningful".

So now you have two problems: Detecting whether a subject believes a particular topic is "meaningful", and then detecting whether they are lying about it.

You're kind of missing the point of the book: "Polygraphy is not science." It's an interrogation technique.

I think interrogators are missing the point: If your interrogation technique can't even produce information that is verifiably correct, then why on earth would you believe any information produced that is non-verifiable?

I have taken and "passed" multiple polygraph examinations by US state-level agencies. Providing truthful answers would have disqualified me in each test, and I have never been disqualified. This book is very accurate, and has been my only influence/guide/assistance in every polygraph I have submitted to(ie I have no "training" in this field). I have not read this new edition, but I can't say enough good things about the prior ones.

I can’t see how it would be in your interest to post this. Are you not concerned that this comment could be tied back to your working identity and used against you?

Another possibility is that you are a lying fed stirring the pot. For instance, a fed could point to your comment as evidence that antipolygraph.org is inciting people to break the law, which is a gambit they have used before.

Read my comment again please - I'm not a lying fed, a troll, or anything else malicious. I'm not in the industry I once was, which involved regular polygraph examinations. I have no plans to return to the industry. Posting my comment carries zero-to-low risk for me. If it helps even 1 person, then it's worth it.

All I'm saying is... if you are about to take a polygraph, or think you will in the near future: it is in your best interest to read this guide. On another note, after you read the guide and possibly benefit from it...you might find your attitude towards them changed. If the polygraph remains the de facto "lie detector", then we can all rest easy: because they are trivial to pass.

Why would I rest easy if they’re trivial to pass? That’s great if I ever have to take one, but it really sucks if, say, someone passes one while falsely accusing me of a crime.

Polygraph examinations aren't admissible in that way. Otherwise crime would be as simple as pass/fail not-guilty/guilty.

This is why a (good) defense attorney advises you against taking a polygraph if one is offered during a trial. If you pass, the case is not magically dismissed: it's useless. If you fail, you aren't automatically convicted, but the jury has their view of you negatively influenced by your failure of the test. You have nothing to gain by taking and passing one in court, but you have an undefined jury sentiment to lose should you fail.

That’s only because they’re not considered to be a real lie detector in that context.

Isn't the jury view of you also negatively influenced by the refusal to take it, though?

That would not come into evidence as long as you have a minimally competent lawyer. The jury would never know.

Kind of like talking to the police. It can never help you and most likely will be used against you in court.

That's why polygraphs are not allowed as evidence in the courtroom.

The real problem isn't whether they are easy or hard to pass, it's that they're easy but people believe that they're hard.

Just curious... did you developed specific techniques or did you just sailed through without even thinking/planning?

The part that resonated in the guide for me was the control questions, and identifying the expected lies. Dealing with those is all I ever did. Your goal during the test is to raise the output from the control questions as high as possible.

For example, the control question I got in at least 7 examinations was the same: "Have you ever stolen anything in your life?"

When I heard the question, I did three things at once:

1) Mentally tried to divide 1717 by 3, over and over, knowing that I would not solve it and allowing that to make me anxious.

2) Increase my rate of breathing(pretend you are inflating party balloons).

3) Tense my entire body, by imagining fire ants crawling up my legs, about to reach my groin.

A few seconds later(don't wait to long), I answered "no". The examiner pauses, and asks if I have told him the truth. I confess to petty crimes such as stealing candy from my sister as a young child, and further confess I knew it at the time I answered, but was ashamed to tell him.

There will be several more control questions, and some of them(obvious ones) will be of that type(an expected lie, but not relevant to the topic of the examination). When you give the expected lies, you want to raise the baseline. When you give expected truths(such as "Are the lights on in this room?"), you want to answer normally. The end goal, is that any response your body has to a real conceived lie, should be lower than what you exhibited during the control questions for expected lies.

Yeah I treat it as "common knowledge" that polygraphs work about as well as the ADE 651 (https://www.nytimes.com/2009/11/04/world/middleeast/04sensor...).

It's always fun to think about how I could be totally wrong tho. Thought experiment:

Suppose you had a piece of technology which could infallibly detect whether someone thinks they are telling the truth.

To preserve the device's maximum efficacy, would you (a) keep it a secret and only use it for ultra-high-stakes investigations (as with cell-site simulators - e.g. https://arstechnica.com/tech-policy/2015/04/fbi-would-rather... ); or (b) promulgate the idea that it is worthless or can be trivially manipulated to return any result, and train people to use it for everything?

Hard for me to distinguish between a world where the polygraph does nothing and everyone knows it; and a world where the polygraph does nothing and everyone knows it (but that's just what they want you to think)

They can’t subvert everybody. In the latter world, you’d have some independent researchers saying it was highly accurate. In reality, the ineffectiveness of lie detectors doesn’t seem to be in any dispute when studied scientifically.

It's a fun idea, but even the premise points out the major fault: "whether someone thinks they are telling the truth".

Studies on memory show that human beings, as a whole, are terrible at remembering things. We edit, we re-write, we create. Almost certainly some event you remember clearly from more than 20 years ago either did not happen at all, happened to someone else and you heard about it, or happened in a completely different way than you remember. That's true for just about everyone[0].

What good is a device that can tell if you believe you're telling the truth when your beliefs and memories are so flaky anyways?

[0] Except my wife, whose near-eidetic memory is frightening.

> Studies on memory show that human beings, as a whole, are terrible at remembering things.

I always remember testifying in court about an assault that took place some months earlier. By the time I was actually there, I couldn't remember the event at all. I simply remembered describing the event previously (a bunch of times).

Did I see what I said I did? I believe so. Was the testimony in court worth anything at all when compared to the Police interview 2 hours after the fact? God no.

At that point my truth was that I was confident that I didn't lie in previous testimony. That's it. As you say - what value is there in something which tests that?

Exactly. What's worse is that society presumes that memory is faultless. If you honestly said in court that you couldn't remember very well, but that you believe your statement to be accurate (the truth), a judge and jury would both agree that you were less truthful, or trying to cover up the truth.

And even if we agree that memory is faulty, no one agrees that their memory could be faulty. Just everyone else's.

My favourite example of this is my father. He tells a story about some funny event that happened to him up at his parent's cottage when he was in his 20s. Right after it happened, he told his family, including his brother. Within a few years, his brother started telling the story like it had happened to him- and in fact seemed to believe it did happen to him. My father tells me that and scoffs, saying "See how faulty memory is? Your uncle remembers an event that never happened to him!". I asked him "How do you know it's his memory that's wrong and not yours, since memory is so faulty?" "Oh now, I know that my memory is right".

Apparently jurors generally have a wrong understanding of how memory works.[0] Sometimes I feel like it isn't really a privilege to be judged by your peers.

[0] http://www.physics.smu.edu/scalise/P3333fa14/Eyewitness/Eyew...

Penn Jillette likes to discuss this. Having worked so long with one person, they have a lot of shared stories. And he tells of listening to Teller tell a story and know that he's full of shit, getting nearly everything wrong. And also having Teller listen to him tell the same story, and thinking he's full of shit and getting everything wrong.

And then there's the third person or some sort of record of the event that shows that both of them are full of shit and getting everything wrong.

And it's especially scary knowing that memories can be suggested to people, to the point that they genuinely believe they experienced the false memories.

This American Life did an episode about a guy who was a polygraph operator; discovered how flawed it was; and has been crusading against them ever since and teaching people how to beat them. Worth an hour of your day:

https://www.thisamericanlife.org/618/mr-lie-detector

Yes. That's the story of our friend, Doug Williams, who was targeted for entrapment by federal agents in a criminal investigation that targeted polygraph countermeasure instructors. We briefly discuss his case in the section on "Operation Lie Busters" in Chapter 4 of The Lie Behind the Lie Detector.

For more on Doug Williams' case, see: https://antipolygraph.org/litigation.shtml#doug-williams

It's scary that something that is not only controversial but also has laws (e.g. EPPA) in the USA that prohibits its use in some areas (displaying the lack of faith that the nation has on it), is not completely banned.

If it is not generally used anywhere outside of the USA then it is way too fragile to be allowed anywhere. The principles are easy to understand but also easy to falsify as many people have, no doubt, found out to their detriment.

it’s just an interrogation tool. everything about it makes more sense holding that in mind.

Similarly, SCOTUS upheld the use of drug-sniffing dogs a few years ago. The research seems to show that the dogs are effectively confirmation bias.

At least we don't use dowsing rods: https://www.theregister.co.uk/2016/07/26/iraqi_government_fi...

Many people and organizations still do. In some parts of the USA, if you hire someone to dig a well on your property, they don’t dig until someone with a divining rod tells them where.

https://www.theguardian.com/business/2017/nov/21/uk-water-fi...

I wonder if a machine-learning approach really would learn how to detect lies. There is already a huge corpus to train on: police interrogation videos. It would take significant effort to label deception that was later confirmed. But I’ll bet we could teach a computer to see through almost any lie if we decided it was an important thing to spend time on.

Maybe if you're looking at non-syntax indicators like pace of speech, posture adjustments, facial expressions, etc. But NLP is still a major unsolved problem. AI simply isn't strong enough at understanding meaning especially in broken English or slang (ex. "Nah, I ain't done it.").

I’m not suggesting that the model would infer the truth from the input. Just that it would render a reasonably reliable estimate that the subject is dissembling. I don’t think the actual words used would be the principle component of such a determination.

Julia Hirschberg gave a keynote at this year's EMNLP on this topic. Short answer was that ML can do better than trained humans in some cases, but very far from "almost any".

I don't think EMNLP has posted talk videos yet but these slides look very similar: http://www.cs.columbia.edu/~julia/talks/ICASSP2018-latest.pd...

I remember reading somewhere an idea to make the polygraph produce more objective results:

Instead of asking for example "Did you rob the bank?" and interpret the response as "lying" or "not lying" they can be asked "Was the getaway car white?" "Was the getaway car red?" "Was the getaway car blue?". By asking several of these you can statistically prove the suspect is guilty if they responded differently to several detail questions about information only they knew. Tthis doesn't help prove innocence (because people can beat the machine) but it reduces false positives and can therefor prove guilt (because it's unlikely a nervous person would randomly be more nervous when the true details are being asked vs. the decoys).

Of course to be done right the operator must be "blind" (not know the answers and subconsciously tip the suspect off) and other statistical safe-guards need to be in place, so how useful in the real-world this is I don't know.

The technique you're thinking of is called the Guilty Knowledge Test, which is seldom used. It's discussed at p. 121 of The Lie Behind the Lie Detector: https://antipolygraph.org/lie-behind-the-lie-detector.pdf#pa...

The rational wiki calls it a presupposition and mentions what I remember from logic class which is to respond "moo" although I could swear it was spelled "mu" when I learned the term. The example I heard was something like "So tell the court: have you stopped beating your wife?"

https://rationalwiki.org/wiki/Presupposition

I think the correct spelling is "無", but wikipedia lists english spellings as "mu" (from Japanese) or "wu" (from Chinese)[0]

0 - https://en.wikipedia.org/wiki/Mu_(negative)

Thank you!

Polygraph testing has nothing to do with the scientific legitimacy of the actual instrument used. It's just an interrogation technique. Using it effectively is all about the skill of the person administering the test, and their ability to judge the person being tested. It is just one part of building a profile of whether you think someone is lying or not.

The purpose of the test is to put someone through a highly stressful situation and see how they react to questioning. It's about determining whether they've been coached, and how familiar they are with details of the case, not determining a definitive "yes/no" response to specific questions. Polygraph questions can be passed with 100% accuracy by someone with extensive training, so it's entirely about building a gut feeling in the person administering the test as to whether you are lying or not.

"Casey, just appointed C.I.A. chief, told me he was going to challenge Baker to a polygraph test to show who was lying. Figuring my old pal Casey was the culprit, I wondered why he would take the gamble. He reminded me he was an old O.S.S. spymaster, and that by using dodges like a sphincter-muscle trick and a Valium pill, he could defeat any polygraph operator. Baker wisely did not take Casey up on the challenge."

https://www.nytimes.com/2002/10/10/opinion/lying-lie-detecto...

I also enjoy this: "Why Lie Detectors Don't Detect Lies" https://www.youtube.com/watch?v=nyDMoGjKvNk -- 4.5 minutes long.

It's never a good idea to base your feelings about jurisdictional specific things on this, but the fact that few other modern western economies think they're worth it bears thinking about: in so much else, law nowadays relies on everyone else's law to help define how to make it work. So, a new case in the USA can and will be cited everywhere else as a pertinant example of good modern law. But.. not in respect of polygraph evidence. Nobody else seems to want to include it as a routine praxis. Doesn't this worry American students of the law?

You could apply this logic to the gun control question too..

Through evolution we have become very good liars. Don't count on your natural ability to lie though, it's better to just shut up. While lies are next to impossible to detect, they are easily checked with facts.

What about the "lie detector" they had on the recent update to Making a Murderer? I assume it is also bullshit (and the lawyer who used it seemed to think that too).

Yeah it seems there are few new ideas for using brain imaging. I haven't looked into it too much yet, but I strongly suspect they are also garbage intended to trick people now that a fair number of us are aware of the polygraph scam

You can get some info here: http://noliemri.com/

Let's be real though. With brain imaging, we will eventually be able to accurately detect lies. This is one of many ethically charged issues where dismissing it as unreliable/pseudoscience will only work for so long before one that actually works is built.

I'm not convinced this will ever be the case. Even if we were able to somehow discover a part of the brain that only ever activates when lying what happens when someone says something truthful while also thinking of a lie? What happens when someone technically tells the truth but omits critical information?

> With brain imaging, we will eventually be able to accurately detect lies.

That assumes that just because we have a distinct word for “lie” that that word has a clean direct correspondence with a detectable biological process.

Discussed in 2013: https://news.ycombinator.com/item?id=6307479

I was surprised - and dismayed - that the polygraph was still being used and given credence with something as high profile as the Ford/Kavanaugh hearing.

I've long been under the impression that it was commonly understood that the polygraph was fallible. Has this changed?

Nothing changed, but I’d estimate >50% of Americans still believe in “lie detectors”. They provide easy points when you don’t have actual evidence to bring forward

Over half of Americans believe cloud computing is affected by weather patterns [1]. So ya, I'm sure at least those same people believe in lie detectors.

[1] https://www.businessinsider.com/people-think-stormy-weather-...

Side note: the study you cite had the question as "is cloud computing affected by weather patterns?" This finding leads one to believe that nobody actually knows what the cloud actually is, but that is a bit sensationalist: technically data centers can be down due to weather-related power outages.

A charitable reading: if you're aware that "cloud computing involves computers all over the world", but aren't aware that undersea cables carry 99.99% of international Internet traffic, then you might assume that cloud computing involves bouncing your Internet signal off a satellite at some point. Weather can interfere with satellite signals.

Same goes for cell service: a lot of people tend to believe that, while your phone just talks to a cell tower, the cell tower is talking to a satellite (rather than to wired backhaul or a microwave link.) A satellite is depicted in all those diagrams of "how cellular networking works", so I don't blame them for this at all!

Cell towers listen to GPS satellite signals to set their clocks. If GPS stops working, time division multiplexing will soon have problems.

True. Though, how soon is "soon" for TDM desync?

I think most people, if you asked them to describe how weather affects a cell tower, would describe them as working in basically the same way that Iridium handsets do: "clouds in the sky, signal goes bye." I.e., the signal would be "choppy" on the interval of milliseconds, with packet retransmissions covering it up up to some tolerance level, but failing entirely and instantaneously if the weather was bad enough.

I imagine they would think this, even though they know intuitively that cell signals are more robust than that in practice. They don't know why they know that, and they've also been taught something that conflicts with that knowledge, so depending on which order their jumble of thoughts comes to mind, it might go either way.

(Also, something I'm curious about, as a tangent: now that atomic clocks are getting cheaper, are cell towers using them to add fault-tolerance to their precise timekeeping requirements, in the way that e.g. undersea-cable head-end switches do?)

Colloquially, data centers also have "weather", ie, sporadic events that cause random slowness, network congestion, or scheduling problems on your cloud.

When there's bad weather in a data center, you sometimes just have to wait it out and run your jobs another day.

And... clouds: https://www.theregister.co.uk/2013/06/08/facebook_cloud_vers...

This is a good example of being too smart for a test question.

To your observation: Amazon had a major outage due to hurricane related flooding and a lightning strike back in the mid 2000s.

Um.. I would answer yes to that. Azure was out due to lightning for an entire day when I finally fully gave up on the disaster they call their cloud.

Nothing has changed, I don't think laymen ever "commonly understood" the polygraph was fallible.

Especially since they are required for some government jobs.

The U.S. government actively promotes public belief in the pseudoscience of polygraphy, and the press largely plays along.

And Hollywood as well.

Not really though. Every time it's used as a plot device to create suspense but in the end the hero or villain has been trained to beat the polygraph and get away with it. So they actually perpetuates the idea that it's fallible / can be fooled.

> So they actually perpetuates the idea that it's fallible / can be fooled.

By doing that, they also perpetuate the idea that it does work somewhat reliably and to beat it, you need special training.

Why is the government actively promoting that?

As for the press, they only play along if it fits their narrative. During that hearing Fox News framed polygraphs as pseudoscience while CNN was promoting it as evidence. Both sides will just as easily do the opposite the next time depending on what their audience wants to hear.

It's a useful interrogation technique, but only if the public believes it works.

Example in fiction: https://www.youtube.com/watch?v=DgrO_rAaiq0

LD tests are an unreliable technique if the subject has a clue of what LDs do measure, namely stress, and how to control those forms of stress measured by the device. After a little behavioral training, most anyone should be able to significantly reduce their stress response when lying, making the test results much harder to interpret effectively.

LDs don't catch spies. They catch novices. Aldrich Ames beat the test repeatedly. But he would, since his job was counterespionage.

They use it to ferret out untrustworthy people at government contractors. This catches some moles.

The real problem is when this hocus pocus enters the legal system and people get convicted based on examiner testimony or face the presumed guilt applied to anyone who refuses a polygraph.

There is no documented example of the polygraph ever catching a mole. On the contrary, all of the known moles who were subjected to polygraph screening beat the polygraph.

Saying that they "beat" the test gives the test too much credence. The test is a ploy to make the subject confess. The only step to beating the test, is the simple knowledge that it's a ploy in the first place.

Point well taken. But the U.S. government does consider the polygraph to be a pass/fail procedure. Applicants and employees suffer severe career harm for "failing."

Wasn't Philby caught by polygraph but argued his way out of the results? (He was, in fact, a spy all along).

No, Philby was not "caught by polygraph."

Exactly as 3JPLW mentioned, government agencies value the polygraph as an aid to interrogation. Naïve and gullible persons, be they applicants for government jobs, criminal suspects, or criminals released on probation or parole, will sometimes make admissions that they might not otherwise make, because the fear that the polygraph can read their mind convinces them that their best choice is to confess.

But the polygraph can only have such utility so long as the general public believes that it can detect lies. We're at the point now that applicants for public jobs who face polygraph screening are often being specifically instructed not to research polygraphy.

it can give an avenue for questioning based on raised tension or whatever. its bot a complete farse but is not very dependable and can be decieved as well as false positives.

Polygraph is a placebo "eye of God watching you" to encourage confessions during interrogation. No more, no less. Like any placebo, it doesn't work (as well) if you know it's a placebo.

Either way, it's inadmissible in court. It's a technique for aiding in interrogation (like torture would be, if it worked), not a forensic technique for producing evidence.

I thought the placebo affect can still work (to some degree) even if you knew you were given a placebo.

https://www.health.harvard.edu/blog/placebo-can-work-even-kn...

Torture does work in cases where the information is verifiable. Consider for example that the interrogator has an encrypted file and a subject that knows the key.

Torture is morally abhorrent, but it's not necessarily ineffective.

"What's the decryption key?"

"I don't know."

It is weird watching US tv shows when they seem to take them seriously. I could never work out if the they are a reflection of real US government practice or if the writers in Hollywood liked the pseudoscience. They are one of those weird things like fan death that doesn't seem to translate out of a particular culture.

What's shocking to me is that in investigations involving the use of polygraph someone(higher officials) were keen on getting desired results out of it, in hope to wrap up investigation the way they want.

Why are you shocked? Haven't you read the Mark Twain quote suggesting that statics is used in the same way. This naive reaction is boosting the credibility of the polygraph. I hope the anti-polygraph book isn't full of same reasoning.

And it was being used as "evidence" by the side with the much shakier stories.

The US department of defense still uses the polygraph as an integral part of security clearance screening.

I am also dismayed, but not surprised.

> I was surprised - and dismayed - that the polygraph was still being used

Keep in mind that it was being used here by Ford, not by the FBI or Congress. I think the people who mattered are well aware they don't really work, but is also aware of their potent PR value (thanks to years of bad Hollywood science and cliches police procedurals).

I suspected something fishy was up with that the moment Ford mentioned she'd taken a polygraph. Even if it's not commonly understood by the general population (and I'm fairly sure it is) Ford is a psychologist and the hearing was full of lawyers - I'd expect that milieu to be even more informed on the shortcomings of lie detectors than the general population.

Reply on news.ycombinator.com