Dan Mangan wrote on Oct 12
th, 2016 at 12:25am:
Aunty, in your world model, are the APA's home-grown researchers and statisticians legitimate scientists?
If by "legitimate scientists" you mean "persons who are legally permitted to label themselves as scientists", then yes, I'm sure there are a few BS and MS degrees among them. People take academic degrees for a variety of reasons, most of which have to do with living indoors and not starving.
But if by "legitimate scientists" you mean "persons who truly practice science", then no, I've seen no evidence of that. (Uness you insist on allowing "scientist" to include "sometimes a scientist", that is, "person who might be practicing science somewhere else but is definitely not using science to study the polygraph as a lie detector".)
Science is a rigidly defined and highly disciplined way of examining the world and making predictions about it. A necessary and important part of its discipline is not caring what answer you get. The instant you conceive a preference for one fact over another, you rip the foundations out of the whole process.
But proponents of the polygraph as a lie detector need certain facts to be true; therefore there can be no true scientists among them.
As far as experiments go, I'm sure any scientist could tell you how to answer the question, "Can the polygraph be used to tell if a person is lying?"
The method is simple. Using data gathered in the field (where liars suffer real consequences), tally the number of subjects who pass and the number who fail. Independently, tally the number who were lying and the number who were honest. This would yield four categories:
Liar who passed (false negative)
Liar who failed (detected liar)
Honest who passed (detected honest)
Honest who failed (false positive)
(If the count of 'false negative' and 'false positive' is small, and the count of 'detected liar' and 'detected honest' is large, the test is good. If the ratio "false negative"/"detected honest" nearly matches the ratio "detected liar"/"false positive", the test is worthless.)
The problem with this experiment is that the second independent observation, the number who were lying vs. the number who were honest,
has never been done. Millions of public sector applicants have been subjected to "lie detector" based interrogations. No one --
no one -- knows how many of them were lying.
Many polygraph proponents have put a number to the effectiveness claims for their methods, and attempted to support that number by correlating the pass-fail observation with some other observation. Invariably they are referring, one way or another, to the number of confessions obtained during the post-test interrogation.
But the post-test confessions are an indicator only of the skill and rapaciousness of the examiner; the best interrogators can make their victims say anything. They are in no way an independent observation of whether the subject was lying or honest. These confessions
are part of the test.
I believe that building a picture of how many polygraph subjects have actually lied would be extremely difficult and would require a well-funded long-term study by some very patient and dedicated scientists. (Real scientists, not merely legitimate ones.) So far, no such body of data exists.
So we have the first observation, the number who passed vs. the number who failed, but we have nothing to correlate it with. No claims about the effectiveness of the polygraph lie detector can be made without such a correlation. Any scientist would know this and would not venture to make such a claim.
Anyone who says the effectiveness of the lie detector has been measured is
wrong. Anyone who says he is a scientist and also says the effectiveness of the lie detector has been measured is
a liar.
Now, another necessary and important part of science discipline is not lying. So to answer your question: No, there are no scientists among the APA's home-grown researchers and cargo-cult statisticians.