Lying to people proven easier than lying to computers

Though the world's best poker players might claim they can call a bluff ten out of ten times, research suggested experienced investigators only get it right 65 percent of the time. So researches at the University at Buffalo, the State University of New York went ahead and wrote some software that supposedly identifies lies 82.5 percent of the time.

Most interesting thing about the software, though, was the study conducted to test it.

Forty participants were asked for a cause they strongly supported. Then, they were given the opportunity to "steal" a check written out to whatever organization opposed the one they'd chosen.

After the opportunity, they were questioned, using the software. First, they were asked simple questions to create a baseline of sorts of their eye movement, which the software analyzes.

Then, the violin-inducing climax were this a dramatic courtroom drama: they were asked if they had stolen the check.

Those who lied and got away with it received money as a reward for lying well, and the organization they had chosen also received money. Those who were caught had nearly an opposite experience: their money was given to the organization they opposed.

Interestingly moral quandary-inducing studies aside, the software has been criticized for having such a small sample size in its testing and also for not registering whole body language when seeking out liars. Though, the researchers claim they will attempt to include this in future programs as well as create faster algorithms to allow the program to work in real time.

Curious how this would work with the "ultimate lie blocker."

Via Scientific American

For the latest tech stories, follow DVICE on Twitter
at @dvice or find us on Facebook