Humans are not good at lie detectors, even when staring at liars.
A machine learning tool that learns to detect lie in less time than the average person has been found to do better.
The system was developed by researchers at Tel Aviv University, Israel, and was able to detect two types of liars.
It's not perfect, but better than any existing facial recognition technology.
Wearable electrodes were used to measure the movements of facial muscles in 40 volunteers who either lied or told the truth to feed a machine learning program.
Lie-detector technology relies on a number of responses, such as heart rate, blood pressure, and breathing rate, to determine the polygraph's validity. polygraphs are considered to be inaccurate despite their use by various areas of law enforcement.
There are other ways to tell if someone is being dishonest.
The idea of genuine emotions leaking onto the face of a liar is nothing new. As far back as Charles Darwin, it was involved in psychology experiments. In 1872, he noted that the muscles of the face were the least obedient to the will.
These micro-expressions only appear for a short time, vanishing after 40 to 60 milliseconds.
sEMG is a technique used to locate facial muscles that contort to form expressions. It can measure the electrical activity of facial muscles and register subtle expressions that are too subtle for humans to detect.
A machine learning tool was trained to read facial expressions in video footage, and a new type of Wearable electrodes was tested to be more sensitive and comfortable.
The lie was very simple since it was an initial study.
Two people are facing each other. One person wore headphones and said something different to their partner who was trying to catch them.
The researchers recorded the activity of the facial muscles between the eyebrows and cheeks of the participants as they listened to the audio signals.
People didn't hesitate when lying, as you might expect.
The study found that people displayed different indicators. Some people twitched their muscles when lying, while others activated their cheek muscles.
Levy and colleagues wrote that the lie-detecting system they used did so much better than untrained human detectors, who spotted lies anywhere from 22 to 73 percent of the time.
The study found that people's telltale muscles are prone to changing over time, and that the experimental algorithm still needs a lot more work.
The researchers say that people who were able to successfully deceive their human counterparts were also poorly detected by the machine-learning algorithm.
In real life or high-stakes situations where repeat liars recount longer stories with lies and half-truths, detecting lies is more difficult.
There are other types of deception, such as omission, evasion, and the use of ambiguous language to conceal the truth, which might make things more complicated.
There are many reasons why someone might be nervous but not lying. Time will tell if this technique can tell the difference.
Levy told The Times of Israel that their hope is that after development and thorough testing, this could be a serious alternative to polygraph tests.
The team wants to train their software to detect flash facial expressions with greater accuracy, so that they can eventually do away with the electrodes.
They expect that testing their set-up with people telling more substantial lies could reveal a whole spectrum of micro-expressions associated with lying. Levy and colleagues suggest that the image analysis tool could be improved by incorporating other emerging technologies that focus on changing the tone of voice.
The researchers have uncovered two of the possible manifestations of deception.
The study was published in a journal.
