New Lie Detectors Are On the Way, But Are They Better Than the Old One?
by Anthony Accurso
New “lie detectors” are being marketed as viable replacements for the aging, debunked polygraph and are being tested in environments where the polygraph never penetrated. But questions remain whether such devices are any improvement on the old one.
Many people are familiar with the classic “lie detector,” aka the polygraph. It has been a mainstay in popular fiction, especially police shows such as Law & Order, where it is used to determine whether someone is guilty or innocent. It also was popular in the U.S. for employment screening until several studies called its results into question in the 1980s, and the government passed the Federal Employee Polygraph Protection Act of 1988 preventing its use by private employers.
The polygraph — “poly” meaning “many” and “graph” meaning “writing” — measures activity of the sympathetic branch of the autonomic (involuntary) nervous system as expressed by fluctuations in heart rate, respiratory rate, blood pressure, and skin conductivity (perspiration).
The problem is that, while these vectors are often influenced by the fear of being caught lying, the reactions associated with lying can be quite variable, and no scientific studies have demonstrated that the emotional response linked to lying can be measured.
Despite being featured often in television and movies almost since its creation, it has been deemed inadmissible in court because of its unreliability. While members of the American Polygraph Association often tout the test as being 95 percent accurate, research shows otherwise. Several studies have shown that the “sensitivity” of the test is around 76 percent — meaning that of 100 liars, only about 76 will be detected. If that wasn’t alarming enough, the “specificity” of the device is around 52 percent — meaning that of 100 people telling the truth, only 52 will be identified as having done so while 48 will be branded as liars. That’s a whole lot of false positives.
Two companies are now touting products that they claim are more accurate. Converus is marketing “EyeDetect,” which measures subtle changes in pupil size and eye movement. Discern Science is marketing “Avatar,” a device that has a microphone, an infra-red eye-tracking camera, and an Xbox Kinect sensor to measure body movement. These devices have been tested at the borders of the U.S. and Europe to try to detect terrorists and drug smugglers. They’re being used in the U.K. to monitor sex offenders, and they’re used by FedEx, Uber, and Experian to monitor employees outside the U.S.
The problem is that these devices were not rigorously tested before deployment in these contexts and may be causing an immeasurable amount of harm. The companies tout “algorithms,” which employ “AI” and “machine learning” to achieve a claimed accuracy of between 83 percent and 85 percent, according to Discern Science’s marketing materials.
But tech buzzwords like these may simply mask a bigger problem that plagued the polygraph: that such physiological variations are only partially correlated with deception, and people respond differently under pressure. It’s also difficult to establish an accurate baseline for deception in a testing environment where the emotional stakes aren’t so high. The tech industry has long used the phrase “Garbage In, Garbage Out” to denote the faulty results achieved from faulty inputs. Does aggregating larger data sets, which may be terribly flawed in the beginning, create better results? Only time and rigorous testing will tell.
According to sociologist Andy Balmer, such technologies pop-up at “pressurecooker points” in politics, where governments lower their requirements for scientific rigour and seek certainty in scientism. Historian Ken Alder cautions that these devices are almost always deployed against the most politically vulnerable, such as dissidents and homosexuals in the 1960s and asylum seekers and migrants today.
Any solution, like the U.S. legislation passed in the 1980s, will necessarily be born of politics. According to Alder, the test “cannot be killed by science because it was not born of science.”
Sources: theguardian.com, Skeptical Inquirer