The issue of Accuracy

[KatZahra]

This is my first time posting in this forum (or in any forum if truth is to be told, but anyways..) so please be patient...

First and foremost this site and the ebook "The Lie Behind the Lie Detector" have been really helpful and I congratulate Mr George Maschke for the extensive work he has provided on the polygraph test.

Having said that, I still remain with a little dilemma on the accuracy of a polygraph test. Those who are in favour of the polygraph test mark the accuracy at 90-98% however those who are against the polygraph state that the accuracy is better than chance but lower than perfection.

Quote #1:
These studies, almost exclusively carried out in criminal investigation settings using one particular test, the Comparison Question Test (CQT), indicate that the CQT is able to discriminate between truth tellers and liars above the level of chance, but the accuracy levels are far from perfect.

Vrij, A. (2008). Detecting lies and deceit, 2nd Edition
Pgs. 295-29

Quote #2: [You should know this ]
...specific-incident polygraph tests for event-specific investigations can discriminate lying from truth telling at rates well above chance, though well below perfection.

Maschke, G. W. and Scalabrini, G. J. (2005). The lie behind the lie detector, 4th Digital Edition
Pg. 27

____

However whilst reading a paper by Lykken D. T. (1988) The case against polygraph testing I encountered the following statement:

"We can conclude, therefore, that the CQT, while reasonably accurate (about 84 percent on guilty suspects) in detecting lying in persons not specifically trained in countermeasures, is exceedingly inaccurate in detecting truthfulness (about 53 percent accuracy, versus a chance of accuracy of 50 percent)."

Pg. 117

What does this mean? How come CQT has a high percentage in detecting lying but a lower one in detecting truthfulness? ..Aren't these percentages related with each other?

Having said that I also have an interesting quote :

Why do critics figures vary?

One of the problems in discussing accuracy figures and the differences between the statistics quoted by proponents and opponents of the polygraph technique is the way that the figures are calculated. At the risk of over simplification, critics, who often don't understand polygraph testing, classify inconclusive test results as errors.

from http://www.polygraph.org/section/resources/frequently-asked-questions

How important are inconclusive tests? As far as I have understood inconclusive tests which a polygrapher cannot give a definite verdict upon [if the person is either lying or not], so technically they shouldn't be included in accuracy rates because they are neither correct or incorrect... or am I absolutely wrong on this?

PS. I know, it's a bit foggy and some of you may find it stupid so please excuse my ignorance.

[George W. Maschke]

On the contrary, I don't know of any valid reason for excluding inconclusive outcomes when considering the accuracy rate of a diagnostic test.

But it's important to bear in mind that polygraph "tests" aren't really "tests" because the procedure lacks both standardization and control. As such, its sensitivity and specificity cannot be determined. But taking the best published field studies of polygraphy at face value, Dr. Alan Zelicoff has shown through a statistical analysis that "if a subject fails a polygraph, the probability that she is, in fact, being deceptive is little more than chance alone; that is, one could flip a coin and get virtually the same result for a positive test based on the published data."

[Drew Richardson]

KatZahra,

When a bureaucratic system exists that knowingly will allow an honest person to suffer negative consequence(s) stemming solely from an inconclusive polygraph result, that result is, at best, nothing but an error with an asterisk, and said system rests, at least in part, on a non-thinking abuse of power.  The aforementioned would be true even if the polygraph process had diagnostic validity when a conclusive result was obtained.  Lie detection has no such validity.