This is the standard way human discourse happens when talking about something for which there is no evidence of efficacy. You back up and try to rationalize some other reason for it's use, even when it makes no sense. Like in this case, using it to get people to admit things, can lead to false confessions or false convictions, especially when some of the people you threaten to use it on know it doesn't work! Do you think the agent who knows it doesn't work, and just uses it as a threat, is going to drop the case when you take the polygraph and fail? Nope!
It remind me very much of when people discuss some supplement that doesn't have evidence: "Oh well those studies didn't use enough/too much/wrong schedule" or acupuncture : "oh well even if it doesn't work the placebo effect is valuable"
you just can't get through to people
I'm not saying that makes for an actually effective system. I think it probably does more harm than good in that it leads to people thinking "well this guy passed the polygraph, so I can definitely trust him".
And of course, the calculus is totally different in criminal proceedings, where we absolutely should not be willing to make that recall-precision trade. I can at least understand why someone would think they're a helpful tool for counterintelligence screening, but any use in criminal investigations just strikes me as ludicrous.
I think the parent poster is 100% right. The whole conceit of an intelligence agency is deceit. The placebo effect is probably responsible for a big portion of their overall power, in a lot of different ways, and it makes sense they'd utilize it when hiring people. The only danger is using it in any other setting, or if the examiners or people who rely on it somehow start believing the placebo isn't actually a placebo.
Polygraphs are used for two investigative purposes:
1. To attain leads for investigation. This is basically tool-assisted https://en.wikipedia.org/wiki/Cold_reading. You take a person, add some probative questions, see what makes them nervous, and then dig in further on those topics. Nothing they say can be used as proof of anything, but it can sure guide you to go have a closer look in their bag/at their ID documents (TSA); or to ask the friend they mentioned to come in as a witness; or to try to get a warrant to search their vacation home. (The statements made under polygraph aren't admissible to get you the warrant, either, so you'd need to find some other probative evidence before doing that last one. Usually this is when police would start stalking someone to obtain a DNA sample from a gum wrapper.)
2. To eliminate potential suspects (or potential witnesses!) when your pool of People of Interest is huge (e.g. when a murder occurred in public in Times Square.) A polygraph is a very weak form of exculpatory evidence: it allows you to say that some people are just less suspicious than others, without letting stereotypes guide you in that judgement. If you have 400 PoI and you need to decide who to look into first, batching them through a five-minute polygraph session is an easy heuristic for ranking them (especially when combined with other ranking factors, e.g. a criminal-record check.) It's also pretty good at weeding out "witnesses" who didn't really see anything but just like attention.
Polygraph findings can't really be "wrong" in the #1 use-case. At worst, a polygraph-obtained lead will be zero-information noise, and you'll waste time investigating it. But a polygraph used this way will never cause the wrong person to be arrested/convicted.
Polygraphs can be wrong in the #2 use-case, but only in the sense that they might cause an investigation to be derailed for months/years because someone was eliminated early when they shouldn't have been. (This is why it's important to enforce the idea that it's weakly exculpatory: i.e. polygraph findings can be used to rank the "interest" of your Persons of Interest, but not to entirely eliminate them from the PoI pool.)
Where you see polygraphs really "go wrong" is not per se a failing of the polygraph itself at all, but rather it's the common failing of all police investigation: the wrong-headed thinking that you can convict someone by playing Guess Who, i.e. that if you eliminate all but one person, then that person has to be guilty (and then doing a bunch of motivated reasoning to build weak circumstantial evidence into a case against them.) Polygraphs are often used as one of the tools of this wrong-headed pursuit, allowing a biased investigator to justify a winnowing-down of a PoI pool to just the person or set of persons they personally believe are guilty.
But you don't fix that by taking the polygraph away. You fix that by firing those people.
I bet a lot of people also crack and admit to things they never did. You're essentially using the stress associated with the "magic" of the polygraph to coerce a confession.
Flash forward to three police statements and one polygraph later.
My experience with polygraph tests is they are a way for police to coerce a confession. I "failed" the polygraph test, and "should come clean now so they would go easy on me." The problem is, I didn't do it, I didn't know who did it, and I didn't want to know who did it. I was 17, and they had no leads. Just an absolute joke.
This is a bug not a feature, if everyone strapped to it are stressed out there's nothing for even a trained interrogator to read. Everyone is going to look like they're lying.
But in terms of the base "object level" of the actual graphing of multiple biometrics, the theoretical idea is to construct a baseline within the confines of the test, not between you inside and outside of the test. They let you settle in, ask you questions they know the answer to, and compare measurements during known truthful responses to measurements during or near other responses. So they're (theoretically) accounting for the added stress in the baseline.
Obviously this object-level portion of detecting honesty vs. deceit via comparison with a baseline of physiological characteristics is pseudoscientific and extremely unreliable, but it's part of the overall mind game. And even though it's pseudoscientific, it's also not like it's palm reading or something. This sounds a bit ridiculous to say, but pseudoscience is a spectrum. (For example, astrology is more pseudoscientific than MBTI.) If someone has totally normal responses to every question but a bizarrely elevated response to one single seemingly harmless question which no one else has an unusual response to, then that's something to look into and prod. So the object level of it can offer some utility.
1. there is a lot of stress around the fact that if you don't pass the poly, you don't have a clearance and so you don't have a career
2. you will "fail" at least one poly in each cycle so the stakes are raised in subsequent interviews RE: #1
The trouble comes from the fact that in order for all of this to work, you need to say the polygraph works. If you admit it's fake, or even if your procedures imply it's fake, it loses its magic. And so you naturally develop procedures that put a lot of stock in the polygraph. People start to believe in it. You build up a culture that relies on the magic of the polygraph. And then things like this happen.