Hacker News Reader

Hacker News

Comment by clouddrover | Hacker News Reader

Parent Full thread

clouddrover

> current polygraph techniques can allegedly be beaten with training

No training is needed because polygraphs don't work. Polygraph testing is about as worthwhile as astrology:

https://www.apa.org/research/action/polygraph

https://www.salon.com/2000/03/02/polygraph/

And inadmissible in court in many places:

http://www.austlii.edu.au/au/journals/MurUEJL/2000/6.html

> if trained polygraph techs with special equipment

There are no trained polygraph techs. They are, at best, con artists.

View on HN

I’m not going to vigorously defend polygraphs, but I’ll play Devil’s advocate for a bit. From your the Salon piece you linked to, “In studies, polygraph diagnoses are often wrong, with rates hovering around 80 percent correct”

Which to some people sounds terrible. I certainly wouldn’t want to hang a murder conviction on something with a 20 percent error rate. But on the other hand, 80 percent is better than all but one of the cited results from the Atlantic article, and beats the results from all experiments using untrained people. These con artists are, at best, performing on par with specific lie detection training.

Maybe the equipment is just for show, but who’s to say that further biometric measurements, such as functional MRI, wouldn’t significantly increase the accuracy of biometric lie detection (still technically within the realm of nonverbal cues from the Atlantic article)? I don’t think it could be 100% accurate, but I can reasonably see future polygraphs being legitimate tools for lie detection.

I wouldn’t want my murder conviction to hang on the results of a current polygraph, but I would be satisfied using one to catch an office supply thief. Well, except it would probably be cheaper to just let someone keep stealing supplies than to pay $8k per employee to administer the tests.

> Maybe the equipment is just for show

Yes, it's a con.

> but who’s to say that further biometric measurements, such as functional MRI, wouldn’t significantly increase the accuracy of biometric lie detection

That isn't what they're doing. The "who" that will say it is actual, statistically significant evidence that it works.

The US government still requires polygraphs as a condition of having administrative access to classified systems. They're one data point of many, since you already have a clearance and were thoroughly investigated by other means, but this is an obvious area where it is acceptable to use shitty evidence. The cost of a false negative is extremely high, but the cost of a false positive is close to nothing. Any person who can't pass just has to get a different job.

Courts are the exact opposite situation and I hope the practice of using polygraphs in criminal investigations disappears at some point.

People tend to miss that the error rate in a diagnostic procedure is not enough to evaluate the usefulness of a procedure because you need the relative cost of different types of error as well. It's the same reason FAANGs can get away with shitty interview methods. Cost of false negatives is much higher than cost of false positives.

That your post has been downvoted is disheartening. Your reply was factually correct, cited, and reasonably polite.

Perhaps instead of saying "at best" they are con artists, it would have been more polite to say at worst, they are con artists. Probably many earnestly believe in what they are doing, and are just wrong.

One thing to add:

> And inadmissible in court in many places:

I'm not aware of any place that accepts polygraphs in a court.

Then I'll quote the article:

"In the United States of America, (where polygraph testing is a growth industry) the admissibility of lie detector test results is determined by courts and legislators on a State by State basis.

In the Federal legal system, test results are inadmissible as substantive evidence. Whilst some States have allowed test results in criminal trials, States such as such as California have prohibited the admission of such evidence unless all parties consent to its admission."

> It is difficult to get a man to understand something when his salary depends upon his not understanding it.

To be precise: a lot of people (even on hn, especially on hn) are personally and practically invested in authoritarian practices being legitimate.

I’m kinda surprised. I felt this group was much more rebellious a few years ago. A lot of people might say it’s growing up, but I still don’t like it.

I look back on the people I admired, and they questioned authority.

To vehemently support the policy of the hour as moral and correct, and then to be later shown to be wrong. At best a fool and at worst a criminal. Is more than many can tolerate. They would rather maintain the lie in the face of all evidence.

It really doesn't matter how earnest they are when polygraphs have been known to be garbage pseudoscience for decades, and especially when people's lives can be derailed by the fraudulent "results" obtained from a polygraph or even just the fiction that it can get results:

https://www.bloomberg.com/graphics/2015-doug-williams-war-on...

Calling them con artists is about as polite as I can be.

There is one weird bit of logic in favor of them in niche cases: they might be good at detecting people who have been trained to "pass" a polygraph test.

Reply on news.ycombinator.com