If you readers haven't clicked on the link and read the whole article, you owe it to yourself to do so. I did, and found this little nugget. ------------------------------------------------------------------------------- How accurate is it? The Defense Department has paid for three studies of the handheld lie detector, PCASS. Here are brief descriptions. Follow the links to read the full study reports. Test one In the first test, Army basic trainees at Fort Jackson, S.C., were put through what's called a mock-crime drill. Some recruits were told to place a fake bomb by a roadside, while others remained inside. After a detonation was heard, each was questioned. They were given an incentive to tell the truth: If they were found to be deceptive, they would have to give a 10-minute speech to their unit on the subject of honesty. (That threat was itself a lie.) The examiners were all polygraph instructors, and the interviews were in English. The test was performed in three rounds: first with a traditional polygraph, then twice with PCASS. In the first round, with a polygraph, the examiner was correct 79.7 percent of the time, or 55 out of 69. The Pentagon prefers to cite an accuracy rate that sets aside the "yellow" results, or inconclusives; that yields a success rate of 85.9 percent, or 55 out of 64. (The polygraph was uncertain in 5 cases out of 69, or 7.2 percent.) In the second round, with PCASS, the device was correct in 63.0 percent of the cases, or 46 out of 73. The Pentagon cites an accuracy rate, setting aside the inconclusives, of 86.8 percent, or 46 out of 53. (The PCASS was uncertain in 20 out of 73 cases, or 27 percent.) In the third round, with PCASS, the device was correct in 62.2 percent of the cases, or 51 out of 82. The Pentagon cites an accuracy rate, setting aside the inconclusives, of 73.9 percent, or 51 out of 69. (The PCASS was uncertain in 13 out of 82 cases, or 16 percent.) Test two In the second test, in Columbus, Ohio, civilians who answered a classified ad for a scientific study on deception. Their incentive: Participants were paid a $50 bonus if the machine showed them to be truthful. Battelle Memorial Institute, a defense contractor, was paid $305,000 to perform this test. The instructors were experienced law-enforcement polygraphers, and the participants were all American, English-speaking and college-educated. Again, this was a mock-crime test, with the fake theft of a ring from a secretary's desk. Both those who did and didn't steal the ring were then questioned. The PCASS was correct in 78.9 percent of the cases, or 56 out of 71. Setting aside the inconclusive, the Pentagon cites an accuracy rate of 91.8 percent, or 56 out of 61. (The PCASS was uncertain in 10 out of 71 cases, or 14 percent.) Test three In the third test, the algorithm that makes the decisions was tested in the lab by its creators at the Johns Hopkins University Advanced Physics Lab. The university was paid $1.2 million for its work. The PCASS was still being developed while they wrote the software, so they had no PCASS exam data to work with. They used a set of polygraph exam records, part of the same set of records that were used to develop the device. All of these records had been independently verified, such as using urinalysis to prove whether or not someone was lying about using drugs. For every 100 deceptive people, the researchers reported, the device would detect 86 (red), with two false negatives (green) and 12 uncertain (yellow). For every 100 truthful people, they said, it would detect 50 (green), with eight false positives (red) and 42 uncertain (yellow). ------------------------------------------------------------------------------- With experienced polygraphers, administering polygraph tests, the polygraph procedure was only accurate 79.7 percent of the time! Which, really confirms what the NAS has said all along about polygraph, it is better than chance. Additionally, this mini-polygraph is accurate a whopping 50% of the time when used on honest being truthful! George, you should be getting a medal for exposing this fraud, not being called a traitor!
|