We can see it in your zygomaticus major.
Did you ever watch the Tim Roth-starring TV series Lie to Me?
Just us, huh.
In it a grizzled Roth plays Dr Cal Lightman, a sort of consulting lie detective. He’s so good at reading facial expressions he can spot a fib with 100% reliability, even in highly stressful situations and on people he’s never met before.
Lightman is loosely based on a real face expert, University of California emeritus professor Paul Ekman, though we doubt he would stake as much on his strike rate as his TV version.
But now a team from Tel Aviv is offering to give even Dr Lightman a run for his money.
In this study published in Brain and Behavior, the researchers describe an automated process for detecting lies using electrode arrays and algorithms, which they report to have up to 73% accuracy.
Humans can spot deception at rates no better than chance, and law enforcement is barely any better, the authors write. The polygraph test can measure the supposed physical signs of deception, but the results are open to interpretation and can be faked.
The Facial Action Coding System, developed by Ekman, relies on a theory of transient involuntary microexpressions to betray lies, but is “susceptive to biases and inaccuracies that are not necessarily related to deception”.
Facial surface electromyography, the authors say, is by contrast a reliable way to measure certain muscle contractions imperceptible to the human eye.
They developed an eight-electrode array to be worn on the face – focusing on the zygomaticus major and corrugator supercilii regions, which control smiling and frowning respectively – and used machine learning to identify telltales in individual subjects.
Forty-eight participants were paired up and took turns to truly or falsely report to each other which of two words they had heard, with the hearer deciding whether the speaker was lying or telling the truth. Monetary incentives to compensate the speaker for lying successfully and the hearer for getting it right were introduced halfway through – adding “ecological validity” to the deception model.
The machine “successfully detected lies in all the participants and did so significantly better than untrained human detectors”, the authors proudly report.
But machine and human results were correlated, suggesting both were relying on the same signals. And a good liar is a good liar: “Interestingly, individuals who were able to successfully deceive their human counterparts were also poorly detected by the machine-learning algorithm.”
One in the eye for Dr Lightman is that people don’t all lie the same, with some having a “tell” in their frown muscles and others in their smiles; moreover some people’s tells changed over the duration of the trial.
Applications? “Our findings demonstrate the feasibility of using wearable electrode arrays in detecting human lies in a social setting and set the stage for future research on individual differences in deception expression.”
We’re not sure any setting in which you ask your companion to wear an electrode array on their face counts as social, but we’ll keep an eye on developments.
If you see something that sets off your inner lie detector, send it to felicity@medicalrepublic.com.au
