Polygraphs are not admissible as evidence in court and civilian employers are not allowed to administer them to employees. Regardless, the US Government still uses them for security clearance.

More information: https://antipolygraph.org/

The 2009 NAS report is particularly damming, but I can't find a free copy and I guess that's over a decade now, my mistake.

Somewhere between zero and very depending on how good the person doing the interrogating is and how much the person being questioned knows about the techniques being used against them.

It's basically a good cop bad cop routine without the expense of the good cop and half as many human factors to screw it up.

With your "very effective" conditions, I'd argue that an innocent person would quite often respond the same as a guilty person, making the process unreliable in a different way.

So I don't think it's mainly window-dressing. It's not perfect but enough to scare off most bad guys.

1. You almost never go through them unless you're in the process of obtaining a security clearance (I'd be quite surprised if the UK didn't do something similar for their clearance process). In fact, I have never heard of any other instance where someone was required to have one, as they are generally not admissible evidence in courts.

2. It is my understanding that the accuracy of polygraphs is mostly down to the skill of the person reading the device and giving the test (hence why they aren't usable as evidence).

The poly is not limited to one session. The examiner will keep asking the subject to come back until they are satisfied with the results, or until the subject quits their job. Depending on the personality type of the subject, it can be extremely stressful, or just routine. Many engineers are "Type A" and that is the type most susceptible to stress.