[unishark]

The issue is accuracy. In terms of being able to infer lies from skin response, this is an effect that actually exists for many people. The net scientific conclusion is that can work but poorly, not that is is complete nonsense like ESP or something.

Being pseudoscience or not is not quite the same question. People claim acupuncture works for them (California is now forcing insurance to pay for it), but it is clearly based on pseudoscience.

[jacquesm]

What are the chances of an actual spy being able to beat a lie detector?

I'd put those at about 100%. So you will end up with nervous people who will increase the size of the haystack with zero chance of identifying the needles, which is a net negative.

[pavelrub]

> What are the chances of an actual spy being able to beat a lie detector?

I'd say not even remotely close to 100%. What is an "actual spy"? Those people are often regular employees with zero relevant training, that are simply recruited by a foreign intelligence agency via pressure, money, or manipulation, and are targeted based on their access to information. I'd say that a huge percentage of them won't be able to beat a lie detector.

Secondly, polygraph tests are used for a lot more than weeding out actual spies.

[kls]

What you are referring to are embedded assets. Embedded assets are different than trained spies in intel or counter intel. A trained spy may be a handler of assets but they generally do intel work themselves and usually have some form of cover. A trained agency spy will undoubtedly be able to beat a poly and are trained to do so, on a regular basis. While the accuracy of a poly is too low to be usable for the discerning of truth, they do work to some degree. They are better at telling what you said was true than they are at telling what you said was a lie. So generally if a question is asked and you don't pop then it is fairly safe to assume that you told the truth. Where their completely fail, is sometimes they will pop when you are telling the truth and you can train yourself to not set off the machine when you lie. The problem is it works on the right kind of person in a clean room setting, but that is never the real world, there are just too many variables that are not handled for it to be anything other than a psychops tool, where it just becomes a loose guide on where to dig deeper.

Fun fact, lie detectors are completely ineffective on compulsive liars.

[pavelrub]

In everyday colloquial use, the term "spy" applies to what you are calling "embedded assets", whereas the term "trained spy" isn't used professionally at all. A "handler of assets" is a case officer. An intelligence agency employee who "does intel work" in the field might be a paramilitary officer, or some other name depending on the agency.

In the context of this thread, Alexander Yuk Chung is definitely not what you are calling "a trained spy", but an asset that was recruited by Chinese intelligence almost 20 years after joining the CIA. Like I said - there is no reason to believe that most of those would be have any sort of training relevant to passing polygraph tests.

[kls]

I work in counter-intel, I am well versed on what we actually use, assets are not "spies" and to get technical about it nobody is actually called a spy. But generally a "Spy" as is commonly depicted, denotes someone with governmental or diplomatic cover, an asset is an independent citizen that provides Intel usually a national of the country in which Intel is being gathered on, and generally has no cover. A case office is not an handler, a field officer is, a case office is an analyst. There is no such thing as a paramilitary officer, their are field officers, special agents and military liaisons. <- all of the above mentioned with the exception of analyst and assets would have formal training in evasion, which would include how to flatten a lie detector.

[pavelrub]

As I said - assets are routinely called "spies" in colloquial use and in the media, as is evident even in the story that this post links to, as well as countless other media references.

> A case office is not an handler, a field officer is, a case office is an analyst.

This is incorrect, as is evident by a basic google search. An analyst is -- surprise -- called an analyst.

> There is no such thing as a paramilitary officer, their are field officers

This is also completely incorrect. CIA SAC/SAD operatives are called paramilitary officers. The official CIA designation is Paramilitary Operations Officer.

You might have encountered different definitions in your work or private life, but they are not representative of how those terms are commonly used, both colloquially and in common professional contexts.

Beyond that - your replies add nothing to the discussion. The precise definitions and terms change from agency to agency, and from country to country, and are irrelevant to the original point being made.

[jacquesm]

I'll say it again: they don't work.

You can't really use them for anything. A polygraph machine might have an application though: doorstop.

[pavelrub]

That's possible, but your previous comment is still incorrect.

[unishark]

I'm not advocating their use. Just arguing for a less politically-motivated description of a technology's accuracy.

[RichardCA]

Astrology, Tarot, and Fortune Telling are all pseudoscience, but a good practitioner is going to be very good at Cold Reading.

http://skepdic.com/coldread.html

It works as long as the "mark" is naive. As James Randi explained many times.

Which is why intelligence communities want to suppress information about polygraphs that would make people less naive.

Makes perfect sense.

[unishark]

Being good at cold reading does not mean you can actually predict the future, which is what they are claiming to do. But if they could do so with sufficiently high accuracy to accomplish something useful, then this would still be interesting.

And frankly I think the "pseudoscience" label for lie detectors is a stretch. It measures a genuine signal. The issue is the relation of this signal with lying. They do a lot of strategy in their question selection to try and make this relation hold. And to me (trained in actual science) it makes a decent degree of sense. If it fails to deliver due to there being way too many variables and noise, so be it. But it's not magic or something like psychics. There is legitimate research related to determining emotion with various types of sensors. https://scholar.google.com/scholar?q=review+wearables+emotio...