Revealing jerk: If someone lies, this is revealed by two lightning-fast muscle reactions in the face that are invisible to the naked eye, as measurements using electromyography (EMG) reveal. According to this, depending on the person, either a muscle twitches in the cheek or above the eyebrow. In a first test, an AI system trained on this was able to correctly identify lies 73 percent of the time – human test subjects did not get beyond random hits.
Whether harmless white lies, cheating or complex fairy tales: Lies are considered immoral and wrong, but we humans do it again and again. It is all the more important, for example in criminology, to be able to reliably identify lies and false statements. But that is not easy. Face temperature can show whether someone is lying, but this is not a reliable indicator. Technical methods such as polygraphs and brain scans can be tricked and even trained observers often fail with experienced liars.
Tell-tale muscle signals
Anastasia Shuster from Tel Aviv University and her colleagues may have found a possible alternative to conventional lie detectors. Her method is based on the subtle reactions of the facial muscles to a lie. “According to the theory, deception manifests itself through involuntary micro-facial expressions that last only 40 to 60 milliseconds, but do not match the emotions that the person wants to convey through their lie,” explains the research team.
In order to be able to detect these tiny twitches, the team has developed particularly small but high-resolution electrodes that are simply stuck on and then pick up the electrical impulses from the facial muscles. For their study, the researchers sat two test persons across from each other, both of whom were given headphones through which they each heard one of two words. You should either tell the other person the correct word aloud or lie and use the other word.
During this test, electrodes on the forehead and cheek recorded the volunteers’ involuntary muscle reactions. It turned out that while the test subjects often only correctly recognized in half of the cases whether their counterpart was lying, the electrodes on the forehead and cheek were more successful: In all 40 participants, the researchers were able to use the muscle reactions of the forehead and cheek to tell when they were lying .
“This is the first evidence of polygraph detection using surface EMG in such a situation,” say Shuster and her colleagues. In another test, the team trained an AI system to identify the specific muscle signals. “With this method we achieved a hit rate of 73 percent – not perfect, but still better than most of the technology that has existed to date,” reports Shuster’s colleague Dino Levy.
The team suspects that we humans may unconsciously perceive some of these lightning-fast, involuntary muscle reactions – and that this could at least contribute to our assessment of a lie.
Frown or corner of the mouth lifters
Interesting: The pattern of involuntary muscle tension is not the same for all liars, instead two very different variants were found: In some test subjects, only the frown (Corrugator supercilli) reacted to a lie, which is a muscle directly above the eyebrow. In other people, however, the forehead remained completely immobile, but the corner of the mouth twitched (zygomaticus major).
“Both muscles have well-known roles in the facial expression of emotions,” explain the scientists. The corner of the mouth is also known as the laughing muscle and is associated with positive feelings and smiles. The frown, on the other hand, is more associated with negative affects and facial expressions. It is still unclear why two opposing muscles react to one’s own lying.
Camera instead of electrodes
“In our study, the lies were very simple. But when we lie in real life, we often tell longer stories that usually contain both true and false components, ”says Levy. They now want to test in further studies whether the muscle reaction will still be similar. If so, this reaction could help to better expose liars in the future.
Electrodes might not be necessary for this, because special high-resolution cameras could also detect the muscle twitches. “Whether in the bank, during police interrogations or at the airport, cameras linked to an AI could then differentiate between truth and lies,” says Levy. “As soon as our technology is perfected, the algorithms are trained even better and the electrodes are superfluous, this method could find numerous applications.” (Brain and Behavior, 2021; doi: 10.1002 / brb3.2386)
Source: Tel-Aviv University