pailryder,
The passage you quoted from (apparently the 1st edition of)
A Tremor in the Blood is retained in the 2nd edition, where it appears at p. 68. For proper context, here's the entire section in which the passage appears:
Quote:How dependable is the lie detector, then? Here is a straightforward question for which there is no simple answer. Since "dependable" is vague and the "lie detector" does not exist, I must start by rewording the question. There are several different types of polygraphic examination, each based on different assumptions; one cannot assume that all these types will have the same degree of accuracy. To provide some initial perspective, remember the purpose of a polygraph is to diagnose the individual respondent as deceptive or truthful with greater accuracy than one could achieve without the examination. I can classify subjects as truthful or deceptive and be correct half the time merely by flipping a coin; the chance accuracy of this type of dichotomous classification is 50%--if 50% of the subjects are actually lying. If my subjects are all defendants who have been brought to trial on criminal charges, and if the statistics show that 80% of this group, on average, are in fact guilty, then I could attain 80% accuracy just by classifying everyone as deceptive. For a test to be useful, its accuracy must be obviously higher than one can achieve by chance, and it should usually be higher than the base rate of the more frequent classification in the group tested.
Finally, we must ponder the deeper question of how accurate a lie test ought to be for particular applications. If one is hiring policemen or CIA operatives, then it might be thought that any additional clues, any improvement over chance at all might be worthwhile. These are sensitive positions in which the wrong person can do great mischief, and it may be in the public interest to use a screening procedure that reduces the number of undesirable candidates hired, even if this means also excluding a large number of perfectly acceptable people, wrongly called deceptive by the test. As we shall see later, however, there is reason to believe that many honorable people, the very sort of "straight arrows" we should like to see in these sensitive positions, are especially vulnerable to failing and being eliminated by these screening tests. Moreover, cases like that of Aldrich Ames indicate that false negative errors (classifying a liar as truthful) not only occur but do great harm when reasonable suspicions are quieted by unjustified faith in the polygraph. In the United Nations fantasy that we considered earlier, it would be disastrous to settle for even 90% accuracy in the Truth Verifier. If so much weight is placed on the test result that one makes less effort to seek other information or is lulled into a feeling of great confidence in the result, then one should make sure that the test result is very dependable indeed.
I don't see how you can, in good faith, suggest than Lykken lacked objectivity in his foregoing assessments. Unlike polygraph operators, Lykken's judgment was unclouded by the self-interest involved when one derives income from giving lie detector tests.