posted 09-27-2005 12:34 AM         

Take a look at the General Public forum entitled "What do these test results mean?\Computerized Scoring;

I firmly believe for the Polyscore algorithm to analyze and to obtain an accurate decision, the examiner must review the chart data and input into the software for elimination any artifacts (physiological distortions) and/or reactivity that is occurring prematurely and impacting the question being posed (sometimes referred to as channel noise). I sometimes refer to this as “garbage in (for analysis)- garbage out (for decision).”

To me Joey55, inputing this information takes as about as much time as 'immediate handscoring'-

I understand your agency wanting to tell an a pre-employment examinee a definitive they "passed or not(meaning failed)" prior to their departure; although I, myself,would not use Polyscore to get a "rough idea" just to tell the examinee "something" before they left.

A 'rough idea' does not equate to the more definitive "pass (NSR)\fail(SR)" Antipolygraph.org is just 'chocked' full of disgruntled pre-employment hopefuls who were told one thing (NSR), and then something different (SR\Incl\Countermeasures suspected) after further review or QC-

If your agency is 'adament' then I suggest an immediate 'handscore' while the next candidate waits in a 20-30 minute holding pattern. 'Handscoring' does take precedence over computerized scoring.

Curious if Sig Responses are present, does your agency allow for a follow-up Specific issue test as recommended by ASTM for a diagnostic opinion of Truth\Deception to the SigResponse issue?

posted 09-27-2005 07:28 AM            

I agree with Bob. If you MUST provide an answer to an examinee before their departure, then a hand evauation is necessary.

In my opinion, Polyscore (and the other algorythms) are cute gimmicks to show an examinee the "results", but should never be used without a normal hand evaluation.

I don't know what you mean when you say you run "several tests a day", so, I guess it's time to talk with your agency and explain the facts of the profession. Perhaps one less test per day? I'd suggest you limit testing to no more than 2 or three a day to ensure quality, not only of your technique, but also your ability to administer them...

Have a good,

Jim

posted 09-27-2005 07:40 AM         

I couldn't agree more with Bob and Sackett. NEVER LET AN ALGORITHM MAKE A DECISSION FOR YOU! Tell your Supv. that YOUR test is not complete until you have had time to hand score the charts. You can't discuss the exam results with the examinee before that anyway.

posted 09-27-2005 02:57 PM            

This is a matter of professional ethics and the empirical basis for the polygraph test - and the meaning of the test results in screening and investigative (event-specific) situations.

In prinicple, tests don't make decisions - competent professionals make decisions. Test simply give information.

It's tempting in any professional practice that is laden with anxiety about correct decisions and potential risk or harm to others, to irresponsibly and simplisticaly surrender one's professional authority to a test.

For consumers of polygraph tests, faced with making critical decisions - and living with themselves without constant self-doubt and anxiety - it becomes tempting buy "confidence" where-ever it is sold. This makes people vulnerable to being sold "confidence" by any sheister snake-oil salesman with a good pitch. Science (and predictive statistics) is always laden with a degree of uncertainty, and it is sometimes more difficult to sell science and uncertainty that it is to sell "confidence."

While its acceptable to make up a theme or use a tool to gain a confession that can be further investigated and corroborated, its not acceptable to rely blindly on our tools to make decisions for us.

I've been in court and heard supposed (opposing) expert users of psychological tests such as the MMPI-2 or MCMI-III orr an their strongest argument that their conclusions were those offered by the computerized scoring and interpretation program - and that they should not be questioned regarding those conclusions. Those "experts" have no more business testifying than any second grader who could read the computerized report. While those programs very good, this is an (explitive deleted) irresponsible shirk of professional responsibility.

Tests don't make decisions, professionals do, and professionals have to take responsibility for their decisions.

If I can't argue why I reached a conclusion that may affect another persons rights and liberties, then I should refer the matter to a professional who can. Of course, in less serious situations that do not affect rights and liberties, the stakes are lower and the intensity of this rant is less vehement.

We've all seen too many situations in which one algorithm says one thing and another algorithm says another. So far the makers of those algorithms have not offered satisfactory explainations about this, and have fully disclosed to us how the scores are derived or how decision thresholds are normed.

We're in the business of messing with people's lives here - lets take responsibility for our work.

end of .02 worth of rant

r

[This message has been edited by rnelson (edited 09-27-2005).]

[This message has been edited by rnelson (edited 09-28-2005).]

[This message has been edited by rnelson (edited 09-28-2005).]

posted 09-28-2005 09:01 AM            

We all hear and no your dilema. The biggest problem, or at least one of them, is convincing chain of command that it is imperative to do things right from the very beginning.

This may not be brain surgery but you can't rush either one if you want a proper outcome. We could all do three or four tests a day every day and some people do, but my personal opinion is that when you start to get tired you start to make mistakes you are not even aware of.

Your solution is to try, always difficult, to educate them in proper testing procedures. You can site APA, AAPP, and ASTM but don't let up on preaching quality over quantity. It is a battle most of us have fought for a long time.

Always hand score before giving any decision and get your work Qc'd whenever possible.

Jack

posted 09-28-2005 05:37 PM            

Jack

posted 09-30-2005 02:04 PM            

We do use the LEPET. We have set it up for three relevant questions for people never working in law enforcement or security and four relevants for those who have.

I think the NO's are still a little high but getting better. You do have to really set the comparisons well. We have also tweaked those and no longer use time bars. That did help more.

We do re-tests on NO's and even NSR's if they give any creditable reason and sometimes even if they don't.

I think sometimes we focus so hard on comparisons that NO's are really SR's. Alot of them come back and make new admissions.

I personally think 8 relevants on a PE test is to many. If forced to ask that many I would do two series with four relevants in each series. Takes longer but more reliable.

I don't know about in your area but here the applicant pool is becoming to small and weak to give up on potentially good applicants to soon.

Jack

posted 09-30-2005 08:03 PM         

Did you know the entire federal LEPET manual is available on Gerorge's site. It is complete with control/relevent questions and test format.

posted 10-01-2005 09:13 AM         

You ARE a virgin. BEWARE! This is the site of the infamous Boy George-Ruler of the "I Hate the Polygraph Kingdom". Before you go to this site, you should go back and read every post in our private forum about this clown. He and I are very close-we even shared some fruitcake at Chritsmas time a few years ago.

If is fairly safe to surf his site on your home computer but don't download or post anything. Jack, Dr. Lou and myself have all been the victims of a nasty Trojan Horse from this jerk. If you download or post anything, do it from a secure computer at the office that has a thick firewall and good anti-virus program.

He has all kinds of good stuff on the site including the entire DoDPI manual. Just be carefull when you visit it. You should also make it a point to download "The Lie Behind the Lie Detector". It is long-180 pages but it is a MUST READ for any examiner-new or experienced.
www.antipolygraph.org

Happy surfing

posted 10-01-2005 10:34 AM            

When I first inquired to DODPI about LEPET, they actually sent me to Georgie's site, saying it was easier to download it from there. I downloaded it from work, so I know it is clean. I emailed you the file.

Jim

posted 10-04-2005 05:38 PM            

jack.ogilvie@phoenix.gov

posted 10-08-2005 08:18 AM            

First, Polyscore will not render a valid decision on such tests. None of its data (which is uses to make decisions) is based on screening exams, and none has 8 relevant questions. So polyscore is just "guessing."

Second, an 8 relevant question test is essentially worthless. With each question you ask, you reduce the chances of calling anything correctly. We used to use such a test here, and I argued against it. Since re-testing a few that had undergone the old format, I've had admissions to drug dealing and sex offenses, to name only a couple.

ASTM sets the limit at 5, and I wouldn't do any more than that. APA is working on a screening exam policy, but people can't agree on it, which is a major problem. I prefer two 3 or 4 question tests to one 6 to 8 question test.

You want a test that is sensitive to deception, and every question you add makes that less probable. I suspect you let a lot of liars through such a test. You'll catch some, but you'll miss most of them. Do you make DI calls based on that test alone? You really shouldn't be (another ASTM violation).

posted 10-09-2005 11:57 AM            

I use questions mandated by our police Academy (so I don't really care for them all), but they don't really matter as it's the format you're looking for (I think).

posted 10-09-2005 12:14 PM         

The Backster pre-employment screening test is nothing more that the standard Backster Exploratory format. It has been defended in court many times. If you want more info, just let me know.

posted 10-13-2005 09:25 PM            

3 to 5 RQs per test

1st question always a Neutral - different one for each chart

2nd Q always N or "overall truth" Q, e.g., "Do you intned to lie to any question on this test?"

No more than 3 RQs in a row (I limiit it to two), i.e., stick a N between them

Repeat at least one RQ on each chart (this is the part that confuses examiners, and be sure to tell the examinee Qs will be repeated)

1st repeat is always a N

1st repeated RQ should be one examinee reacted to only minimally

No back-to-back RQ repeats

At least 3 presenations of each RQ over at least 2 charts (4 okay, 5+ discouraged)

Usually 8-11 Qs (total) per chart

No inter-chart stimulation

Example:

Chart 1
N1
Overall Truth
R1
R2
N2
R3
N1
R2
R3
N3
R1

Chart 2
N3
N2
R2
R3
N1
R1

Two or more SRs for any (single) RQ can't be called NDI (question and follow-up testing, per ASTM)

Don't just use a format and stick to it. Vary it so an examinee - even if he's heard of this test - can't know what to expect.

It gets easier the more you use it. I like three questions because it's harder for liars to get through a more narrow test. Keep in mind, the more sensitive you make a test to deception (and less likely to get a false negative), the greater the chance of a false positive, which is why you don't make a DI call on a screening exam. Follow-up testing (with a single-issue format) should clear the truthful and ID the liars with much more confidence.

You can run a 4 or 5 question version in the same fashion. If you have 6 RQs you've got to ask, then run two 3-RQ tests, which requires 4 charts, but takes only about as long as would a 6 RQ (but less reliable) R/I run over three charts.

I've also played with the GQT since it came up on this board as the two disguised CQs are the same questions we ask on our "mandated" screening exam, except some consider the two questions RQs. (Go figure.) You can run four or five in that plus the two disguised CQs to help make sense of things. (Sometimes R/I tests just leave you scratching your head, but having CQs makes one a lot easier to reach a conclusion.)

I like R/Is, which I've said several times, but sometimes you need an alternative method. For example, you might have an examinee who's on a bunch of meds, reducing all signals, so you're not sure if the person doesn't respond much because of the meds or is just truthful. Meds shouldn't matter with a CQT as they'd have to supress one question type only, which can't happen. My point is that you've got to have at least a few tools in your proverbial toolbox. Cookie-cutter polygraphy isn't going to work for everything. You've got to be prepared for anything.

I hope this helps.

posted 10-13-2005 10:07 PM            

I am now prepared with my old version, the lepet and R/I. I have 1 or 2 pre-employments tomorrow and I am going to run both to see which one I like best. I agree, we have to be prepared for the ANTI folks and everyone is different.

posted 10-19-2005 11:22 PM            

I'm not familiar with the 8-relevant, 5-control question MGQT or 7-point scoring. I use the MGQT, relevants at 3-5-8 & 9, controls at 6 & 10. Scoring on each question is (+) or (-) 3 although I very seldom scored a response at 3. If I did it would be the GSR response where ratio is the criteria, even then I seldom used the max score.

When you have a minute, or anyone else familiar with the 8-relevant MGQT, please send me the format.

Thanks.....

posted 10-20-2005 07:33 AM            

Do you mean you score each channel from -3 to +3, meaning you could (theoretically) get a -9 to +9 per question per chart? If so, that is seven-position scoring because there are seven choices of scores, i.e., -1, -2, -3, 0, 1, 2, 3. If you only assign 0s or 1s, that is three-position scoring.

I've never heard of an "validated" or recognized eight relevant question MGQT, but it would violate ASTM standards, not to mention there is no reason to believe it was any better at detecting deception than CVSA. I'd stay far away from such a test.

posted 10-22-2005 10:10 PM         

1a Is your last name Smith?
SR2 Regarding your pre-employement paperwork, do you intend to answer truthfully each question about that?
C3 Before this year, did you ever lie to someone in authority?
R4 Did you withhold or falisfy any information on your application for this department?
R5 Did you ever steal anything from a place where you worked?
C6 Before the age of __, did you ever lie to avoid getting into trouble?
R7 Did you ever possess any illegal drugs?
R8 Did you ever steal anything valued over $20?
C9 Prior to 2005, did you ever lie about someone just to get him/her into trouble?
R10 Did you ever participate in any activities you could have been arrested for?
R11 Have you ever been terminated from any place of employment?
C12 Before this year, did you ever do anything, that if it became known, might be an embarrassment to you or this department?
R13 Have you ever been involved in any demestic violence situation?
R14 Have you ever engaged in any criminal sexual acts?
C15 Between the ages of 6 and __, did you ever lie to someone who trusted you?
1b Are you a U.S. Citizen?
1d Are we now in the State of ____?
1e Is is the month of ____?