|
Joe,
You and I and others, including Dan, were probably all taught in polygraph school that we should say that the polygraph is nearly infallible if you have a competent examiner. This is what was taught back in the day. It feels good because it both glorifies our expertise and also gives us our personal marketing angle.
Dan is simply adding confusion again. The need for restrictions on the number of exams was necessary to prevent the impulse toward practices in which the examination procedure is short-cut in time so that more exams can be completed each day. This is historically an issue for private examiners - and we've all heard the stories from the 1980s (Doug Williams era) in which examiners ran numerous exams each day, and the commodity of interest was the confession and not the test result. Government examiners today will often conduct only 1 or 2 exams per day, with sufficient time for each.
If the solutions embedded in the details and language of our published standards for examination scheduling seem to look odd or not completely satisfying then it is probably an reflection of the social and political difficulties involved in putting such a standard and restriction into place.
But yes, in old-school anti-science polygraph practice the basis of expertise is the persona of the expert.
The problem for us today is this: what we can actually describe and replicate, in terms of test precision and error rates, does not seem to agree with the historical claims of infallibility.
Dan has a unique angle on this because he has published a study showing ~100% accuracy - a study that others have panned as unscientific and a failure of the peer review process. To put this in context, none of us believe that Dan actually wrote the study - its reads exactly like the written language of Matte - and the journal that published it at the time was allowing authors to suggest their own peer reviewers.
Now look at Dan's procedure, and notice that there are 23 scoring feature in Dan's/Matte's model, along with 23 rules. This is a manual scoring protocol for which Matte reported a reliability coefficient in excess of .99. Which means manual scorers almost never disagree while using those 23 features and 23 rules. (For my part, I cannot remember 23 things let alone 46 things.) So it is suspicious. And we have the problem that those 23 features and 23 rules cannot be organized into a logical flow-chart or algorithm for which we could program a computer to achieve automated reliability. This highly complex model is in fact a subjective and unstructured clinical model disguising itself as an analytic model.
So what we have in Dan's old-school polygraph model is in fact a clinical process - not unlike the historical tradition in the greater polygraph profession of the 1980s.
Examiners who were trained before 2006 probably had to learn and memorize 23 or 25 scoring features - most of which were without scientific support - whereas today we tend to focus mainly or only on the things that are statistically significant discriminators, and these are smaller in number.
Also, notice that automated computer algorithms tend to make absolutely no use of dogmatic rules that have no scientific support.
Again, what we can actually replicate using structured and even automated procedures seems a lot more conservative than the ~100% accuracy reported by the clinical model of Dan and Matte.
In the end, it may be our choice: old school polygraph the way Doug Williams accuses - in which the test result is simply "tool" for gaining confessions - and for which examiners are secretly embarrassed about the test result because they cannot realistically quantify the margin of uncertainty around an old-school clinical process, unless they get a confession. Or we can have evidence-based 21st century polygraph, in which we attempt to realistically quantify the level of precision and uncertainty with which we should regard the test result.
At the present time we seem to be observing both old-school clinical polygraph (in which we can only adopt a form of blind faith that the examiner is in fact an un-biased expert with no subjective interference) the and new-school practices (in which we emphasize and evidence-base, norm-referenced, and standardized protocols for both test administration and test data analysis so that the analysis can be replicated). The tension we observe today is sometimes a product of the dynamic and dialectical process between these two professional practice paradigms: old-school polygraph would be called an "expert-practice model" (or the even older "experimental practice model" wherein the expert observer is simply experimenting and learning on each new case - whereas most professions today would view experimental practice with a lot of ethical caution unless we have both no existing solutions and the informed consent of the individual that will be subject to the experimental procedure). Most professions today will also look with caution at the continued use of an expert-practice clinical model - for which analytic procedures have been repeatedly shown to be potentially vulnerable to confirmation bias (just see the 1986 Diane Sawyer event), and for which the analysis is largely unreproducible and dependent on the persona of the expert - at a time when there does exist some structured and replicable test administration and analytic procedures that do not depend on selling false-hope in an "infallible" conclusion as a basis for instilling public confidence.
Its our choice.
I believe the existence and interest in this particular website is some evidence or indication of a public desire for a replicable and accountable analytic solution for the lied detection and credibility assessment needs of our communities and governments.
Finally, Dan's confusion can been seen more easily when we consider that even an evidence-base, norm-reference and standardized test administration and analytic model will still require that we take the time to do it correctly. In fact, things like standardized practices become even more important when we agree or decide not to be satisfied with a test result simply because we are impressed with the CV or persona of the expert - unless that expert uses an evidence-based norm-referenced and standardized protocol for which the analysis can actually be replicated.
As always,
.02
rn
|
Pages: 
Texas sex offender & mandatory polygraph (Read 316386 times)
