[Jack Simth]

Originally posted by narf poit chez BOOM:
automatic lie detecter tests for polititions whenever they say something.

This would be great - if it were possible to make them work reliably. Do you know why lie detectors aren't admissable evidence in court? It is because 90+ percent of the time, lie dector test come out in favor of the person who is paying for the analisis of the results of the test. This applies to two conflicting parties hiring different "experts" to analize the results of the exact same test with the exact same data. Even though both analyists are experts in the field. Lie detectors are so bad, that about the only Groups that actually use them anymore are talk shows and some paranoid government agencies.

[Edit] Even if you do get a reliable lie detector, there is still the 1% problem.

It goes something like this:
Assumption 1: test has a 1% false positive rate, and a 1% false negative rate (a.k.a. 99% accurate)
Assumption 2: 1% "positive" population (in this instance, 1% liars [and it would be about that after a reasonably short time under such circumstances])
Result: fully 1/2 of the positives are false!

Don't believe me? Do the math:
Total Population: 10,000
Liars = 1% of total = 100
Truth-tellers = 99% of total = 9,900
False positives = 1% of Truth tellers = 99
True positives = 99% of Liars = 99
Total positives = True positives + false positives = 198
True Positives / Total Positives = 99/198 = 1/2

On a side note, you also miss one of the liars, but at that level, it is rather insignificant.

This is also a common argument against random or blanket drug testing.
[/Edit]

[ June 01, 2003, 09:38: Message edited by: Jack Simth ]