Originally Posted by
BehindBlueI's
This. Particularly for background random "did you ever" testing. The DoE did a pretty extensive study maybe 20 years ago after they screwed the pooch real bad accusing a scientist of stealing classified materials. They realized perhaps they'd put too much faith in the poly and wanted to really look at the science behind it. What they determined is that random crime detection was as good as a coin flip. Specific incident detection was about 75%. Easy to use countermeasures worked.
They are junk and would work just as well as the operator hooking you up to a copy machine. People confess to things they ordinarily wouldn't because they THINK the machine will detect them, and that's the only real purpose they have.