The article comes from: love Fan children , author: Fang Jiawen, Picture from: American drama “Lie to me”

American drama “Do n’t lie to me (Lie to Me) male lead Cal Lightman can be called a” human flesh polygraph “. By observing people’s facial expressions, you can see whether the other party is lying or hiding.

The power of Lightman is not only that he can discriminate and analyze the meaning behind different expressions, but also how fast he captures “micro-expressions.” Some data indicate that the average expression of human beings will last from 0.5 seconds to 4 seconds, but the time for those “micro-expressions” that are believed to allow experts to “see the sky” is usually less than 0.5 seconds.

However, now Dr. Lightman has “rival enemies”.

More and more researchers are beginning to incorporate artificial intelligence into the lie detector to detect “micro-expressions” and body language at a speed that humans may never reach, and even search for new methods of lie recognition.

Will they become stronger polygraphs?

Modern polygraph: from speeding up to finding new models

As early as 2013, the University of Arizona began testing a self-service immigration machine called Embodied Avatar at multiple airports in the United States. The research team claims that this machine can pass the “non-intrusive credibility test” to single out inbound passengers suspected of providing false information.

Embodied Avatar, the picture is from “Connect”

Simply put, this machine will ask passengers “Are you an American citizen?” “Have you been arrested?” and so on, and use the high-definition camera and microphone on the machine to obtain the passenger’s expression, body language and The tone information is analyzed with algorithms in real time to find suspicious points.

According to Aaron Elkins, one of the members of the research team, Embodied Avatar ’s recognition accuracy rate is between 60% and 75%, and sometimes it is as high as 80%, which is higher than the 54% to 60% accuracy of human judgment. degree.

The US Department of Homeland Security funded this research project, but after testing and evaluation, they abandoned it. Some insiders said it was because the technology studied at that time was not mature enough, while another person familiar with the matter said that the problem was speed: “We have to screen it in seconds, we can’t spend a few minutes to do this.”

Whether it is Embodied Avatar or the American drama male lead Dr. Lightman, they are based on the theory of psychologist Paul Ekman.

The prototype of the protagonist of “Do n’t Lie to Me” is Paul Ekman, the picture is from Wikipedia

Ekman is a pioneer in emotion research. His integrated “emotional set” contains tens of thousands of different human expressions. He believes that certain emotions trigger involuntary micro-expressions that are difficult to control. For many years, he has been a consultant to the US government department for a long time, and refuses to disclose some research data on the grounds of security and confidentiality.

Silent Talker ’s artificial intelligence can also track facial expressions, but it puts aside Ekman ’s research and directly uses artificial intelligence to make a new theory (Although the algorithm does not explain to you what the theory is) .

Psychologists often say that we have to find a certain model of system operation, but we do n’t have a viable model, and we do n’t need it. We just let AI think about it.

Silent Talker researcher O’Shea said.

According to him, Silent Talker’s algorithm will simultaneously monitor the testee’s 40 body “channels”, including various details from blinking speed to the angle of head movement.

picture from The Intercept

Through the analysis and learning of the training data, they claim that Silent Talker finds not the regular “what actions will you do when lying”, but more like the “regular law” —