posted 02-08-2010 09:08 PM
This project began as an afterthought, when we were completing OSS-3, when we realized that the computer outperformed 9 out of 10 experienced humans while using 3 simple physiological features (compared to the complex features that humans use). Mark and I wanted to study the Kircher features further, along with 3-position manual scoring. So, you can find a description of the first experiment on this manual scoring model in the Nelson, Krapohl & Handler (2008) Brute Force paper on OSS-3.It was Barry who started to appreciate it, and then Ben who propelled the replication to completion. I helped with some of the data analysis at the end.
The participants were Chuck Slupski's trainees - and all of them scored the data well enough to meet the requirements for Marin certification (something which some experienced folks can't do).
One thing important to remember is that we invented NOTHING new. Three position scoring has been around for a long time. Kircher features have been discussed for over 10 years now, and have been described repeatedly in our published scientific literature for over 22 years going back to (Raskin, Kircher, Honts and Horowitz (1988). Two-stage decision policies (Senter Rules) have been described in several published studies beginning with Senter & Dollins (2003).
The notion that the EDA is more important that the other components is not a matter of opinion, but is evidenced through data from a number of studies.
- Capps & Ansley, (1992)
- Harris & Olsen (1994)
- Kircher & Raskin (1988, 2002)
- Raskin, Kircher, Honts, & Horowitz (1988)
- Krapohl & McManus (1999)
- Harris, Horner & McQuarrie (2000)
- Kircher, Krisjianson, Gardner & Webb (2005)
- Nelson Kraphol & Handler (2008)
and of course, now
- Blalock, Cushman & Nelson (2009)
We simply did what the science says, in the simplest way we could - which is to just double all EDA values while using the Kircher features, and follow some common sense rules.
What is exciting for us in the polygraph profession is that we have advanced to the point at which we can begin to see methods and build models that increasingly and confidently emphasize what the data tell us - without excessive distraction from simply adding new and untested fancy ideas, and without getting locked into personality based power struggles around parochial schools of thought
Of course, all of this was possibly ONLY because of the previous creative, studious and systematic work of the real giants like Cleve Backster, who seems to have singlehandedly developed a numerical scoring method for CQTs that has withstood nearly half a century of hard use and criticism (it will be difficult to find another example of a practical and scientific idea of equal durability in any profession), and John Kircher and the others at Utah (and others who did earlier work in feature development), who took the time to do tedious regression and discriminate experiments and told us the correct answer nearly 30 years ago (earliest date I can find is 1982) - when many field examiners may not have had much of an idea about how important that would become.
Anywho, it is interesting to see it work.
As always,
.02
r
------------------
"Gentlemen, you can't fight in here. This is the war room."
--(Stanley Kubrick/Peter Sellers - Dr. Strangelove, 1964)