Quote: changed my mindset. changed the way I treated subjects. Ceased with the BS stim and card trick tests. Ceased with the "My p/g is now set to detect whenever you lie" BS. After that, my average DI
Call rate went down to approx 20%.
However, I still achieved the same number (%) of confessions and that opened my eyes to the fact that the p/g was merely a prop with which to obtain confessions. In truth, those that confessed usually did not require to be polygraphed. during the Pre-Test phase I encouraged subjects to talk and step by step obtained confessions. Verified confessions where subjects undertook in writing to make reparation.
You sound like a rather weak and unsatisfied interviewer, who probably compensated by acting intimidating to your clients.
Now, if your overall % of confessions stayed the same, and the number of “DI calls” was reduced to 20%, then the number of admissions would be reduced while the volume is proportionally equivalent.
That would be hard to see anecdotally; d'you do the math on that there data? Or, d'ya just make that up?
If you have the data, then you should really write that up and publish it. Its quite interesting.
Quote:NAS stated that the p/g (system) was imperfect but achieved levels of accuracy better or greater than chance. (Not a comforting reassurance surely). What is chance? where does 'level of chance' sit?
That same statement could be said of so many other tests. Its really quite vague, if you care to think about it (not just read it.)
As for chance. That is another straw-man argument on your part, intended not to facilitate any further conversation or knowledge, but to handicap the conversation from progress.
There is a lot we don't know about base-rates. Similarly, there is a lot we do know about base rates, though not with absolute certainty. (thatsa whole 'nuther conversation)
The whole field of sex offender risk assessment struggles with this problem. Its really only an insurmountable problem if you can't comprehend anything except simplistic bayesian models. Signal detection models, for example, as used by Karl Hanson, allow us to make reasonable estimations of the accuracy of a sex offender recidivism risk assessment in the absence of clearly understood base-rates. His assessment is considered among the best, and has accuracy rates that were well above chance though well below perfection (Static 99 for example was about .77 or so).
Quote:Theoretically chance is a 50-50 situation. In reality its not. Every single situation has significantly different levels of chance - ie - a level at which one could introduce an operant condition that will produce a result greater or lesser than chance for that particular situation.
Problem: Polygraphy does not have an established, verifiable level of chance. So where does it sit?
What is it? 30%, 35%, 50% ? How do we ascertain a level of chance for p/g?
Dude. Go buy a book on Bayesian statistics, and go buy another on inferential stats. Absence of known base-rates is not an unaddressed or insurmountable issue – unless of course you want it to be one, because you can only engage in straw-man arguments.
Quote:Can we establish a level of chance for p/g by the coin toss test? Would that even out at 50% ?
Well i tried it. Take a coin and mark the sides A & B. flip it 100 times then change the markings.
Do that ten times, to achieve 1000 flips. Out of 1000 flips I achieved an average of 46%. Based on the number of times each side fell facing up and then averaged the two scores.
How convenient of you, now, to slip into a parametric/inferential example.
Quote:If we play around with numbers and accept that 86% is theoretically the accuracy as achieved in laboratory mock-crime settings, then the success over 'chance' (46%) is 40%.
That's not very close to how we actually calculate an example of this type, but that's another mathematical matter that would require many minutes of mulling to mediate the myriad of mindlessness in your example,
(you can start by doing a google search on the good old fashioned z-test - go to
www.google.com)
but if this works for your simplistic model, well, that's probably all we're gonna achieve right now.
Quote:Is that good enough?
Is that significantly good enough to prejudice 14% of the population. (100 - 86 =14 )
How many hundreds of thousands does that equate to annually?
Of the potentially incorrect 14% Calls, at least 80% of those will represent truthful people that were
prejudiced by Incorrect Calls. And that should be a major concern to society as a whole.
And now you have conveniently flip-flopped again from at discussion about math and science to a discussion about social ethics.
Quote: The FBI /CIA et al quip, "We gotta get 100 in the front door to get 1 out the back door" is very interesting and most disturbing. The inference is that 90% of the US population are liars and cheats.
I dont think so. Why anyone would want to work for organisations with that mindset is a mystery to me.
More drama.
Please try to separate hyperbole from facts.
Otherwise, we ain't getting no-where.
Quote:To argue that polygraph results inter alia produce some 10% of confessions is not a convincing argument in favour of p/g. Most skilled interviewers could likely achieve the same or higher rates of
confessions without incurring false confessions.
fact check please.
References?
Or is the the world according to 1904!
Quote:The false polygraph induced confession rate is probably very low, (maybe 2% ?) but those are still dire consequences for that innocent 2%. One can only hope that the quality of investigators, interrogators and the justice system will one day improve to the level where skills will filterout the 2% false confessions.
You are again having a very one-sided conversation about these important ethical concerns.
Your heart is perhaps in the right place, and you might not be dumb, but this does not measure up to anything more than drama and editorializing.
Too bad too, 'cause in another context it could be a lot of fun to converse with you.