Polygraphs are able to detect physiological changes in someone but they cannot determine if that person is lying. The notion that they can is based on junk science and selective memory, which are among the reasons the results are inadmissible in court.

A more disturbing use of aggressive policing based on bad data is the tiny number of law enforcement agencies that have adopted a Minority Report approach to fighting crime. They take a person’s criminal history and a secretive profiling method, then combine this with constant surveillance and harassment in the form of trumped-up charges like “grass too high,” until the supposed criminal, who is really the victim, moves to another jurisdiction.

Today we will look at a third dubious notion when it comes to crime fighting: That there exists a way to tell if someone is lying through their facial cues. BBC reported that Israeli researchers claim to have discovered a way to use electromyography to measure facial muscles to determine prevarication.

The technique employs electrodes that measure contraction in certain facial muscles of the subject. Researchers ran two trials with subjects, who heard one of two phrases, at which point the subject then told another person which phrase he or she had heard, either telling the truth or lying.

Researchers then took their data and used it to develop an Artificial Intelligence algorithm that clustered facial movements that they deduced correlated with lying. Next, they used that to develop a predictive model to detect lies based on cheek and eyebrow movements. Human receivers managed to do no better than chance, while the algorithm did 7.5 percent better than that. That’s a decent performance but insufficient to show demonstrable superiority, and certainly of no value to law enforcement, as it still results in a more than 40 percent error rate.

The problem with the entire approach is that there are no universal indicators to determine if someone is fibbing. The algorithm can determine facial movements but not necessarily deception. Another issue is that a research subject could act and feel much different than a suspect being grilled for a double murder. The concept of a genuine lie detector is attractive. Imagine being able to hook up O.J. Simpson or Donald Trump to one. Alas, it remains firmly in the realm of science fiction.