[alswart | Adobe Stock]

Can Machines Detect Lies?

March 24, 2020

Most everyone lies from time to time. In most cases, lies (deception) are no big deal. College students admit to lying twice a day on average, and others in the community admit to lying once a day on average (Lilienfeld et al. 2010). Knowing whether or not someone is lying is hard, but what if there was a machine that could accurately detect lies? Polygraph (lie detector) administers and some media outlets claim there is a way to accurately detect lies.

In the early twentieth century, researchers began to consider the idea that physiological measures may be used to distinguish truth from lies. Psychologist William Moulton Marston invented the first lie detector, which measured systolic blood pressure (the top number on a blood pressure reading). He later created Wonder Woman, who could force villains to tell the truth by wrapping them in her magic lasso. In Marston’s view, the lie detector was the equivalent to Wonder Woman’s magic lasso: a perfect detector of the truth (Fienberg and Stern 2005). Marston’s blood pressure measuring device led to the development of the modern polygraph.

Polygraphs measure physiological activity such as sweating (skin conductance, galvanic skin response), respiration, and blood pressure. This activity is plotted on a chart and provides a continuous record of autonomic arousal levels. The polygraph examiner asks the examinee questions, interprets the polygraph chart, and decides whether the person is lying. Physiological activity is associated with anxiety, so the assumption is that it could be used as a clue to indicate if one is lying. For example, being nervous increases sweating and how well our skin conducts electricity. Interpreting a polygraph chart is not clear cut but instead a difficult task. Physiological activity varies tremendously among people; a truthful person who sweats a lot might be labeled as a liar, while a deceptive person who doesn’t sweat much might be identified as truthful.         

A format often used with polygraph testing involves asking control questions (example: “Are you a resident of the United States?”) and asking investigative/relevant questions (example: “Did you steal $100 from the cash register?”). The idea is that if an individual is lying, the physiological responses will be much higher when the person is asked the investigative questions than it will be when the person is asked control questions. With the use of countermeasures, it is relatively easy to alter physiological responses. I have demonstrated how countermeasures work in the lab with many students. Things that can be done to increase physiological responses include increasing the breathing rate, tensing muscles, biting the tongue, performing taxing mental multiplication, and so on. These countermeasures can be learned in just a few minutes. The countermeasures are applied when being asked the control questions, so when comparing physiological response to control and relevant questions, even if lying, there will be little difference—or in many cases a bigger physiological response to the control questions. 

Is there really a Pinocchio Response? The “Pinocchio Response” is defined as a unique physiological profile that represents deception. The search for that response has proven unsuccessful. If the polygraph chart shows a higher physiological response when the examinee responds to relevant questions versus control questions, the difference indicates the examinee was more aroused or nervous during that time. But being nervous when responding to questions that could have a negative impact on your future isn’t unusual. This level of heightened arousal is not evidence of lying; among other things, it could be due to the shock of being falsely accused or the realization one could be jailed or fired based on how one answers the questions. Polygraph tests show a high rate of false positives—people telling the truth that the test has determined are lying (Iacono 2008).     

Confirmation bias is another limitation in polygraph tests. Administers generally have access to outside information, which may influence the way the test is administered and interpretation of the outcome. As in other contexts, there may an emphasis on confirming certain beliefs, while deemphasizing the value of information that may not support those beliefs. To illustrate the role of confirmation bias in polygraph testing, Gershon Ben-Shakhar describes an episode of 60 Minutes that aired in 1986 (Ben-Shakhar 1991). The producers of the show hired three polygraph companies to determine who stole a camera from the photography magazine’s office. In reality, there was no camera stolen; however, each of the polygraph testers identified a different employee as the thief. It had been suggested prior to testing that the wrongly identified perpetrator was probably the thief.

Michael Shermer, founder of The Skeptics Society, decided to test the accuracy of the polygraph. Shermer made a video of his experience with the lie detector (Shermer 2007). Shermer was taught countermeasures (rapid breathing and tightening muscles) that could be used to beat the lie detector. Shermer lied during the test, but the polygraph indicated he was truthful. Shermer went through the polygraph five times, and lied five times, but produced a so-called truthful chart each time. Dee Moody, a polygraph administer featured in the video, claimed the polygraph is one of the most accurate ways to detect lies, with 95–99 percent accuracy. In Shermer’s case, it was 0 percent accurate. Moody says that polygraph examiners are taught to detect countermeasures and once they have been detected to stop testing. The ease at which countermeasures can be learned and used present a major problem for the validity of the test. Moody doesn’t view this as a problem. Others view it as a major flaw and thus a key reason the use of the polygraph is problematic. After all, Shermer learned to beat the lie detector in a single session. The inaccuracy of the lie detector led to the Polygraph Protection Act of 1988, which prohibited most employers from administering lie detectors. According the Dee Moody “the polygraph protection act occurred not because the polygraph is flawed or ineffective, it was taken away because it is too effective” (Shermer 2007). Apparently, Moody views the polygraph as a 95–99 percent accurate way of telling whether or not someone is lying, no matter what the evidence says. Shermer concluded that the polygraph does not come close to detecting lies 95–99 percent of the time.

Conclusion

The earliest version of a lie detector used a single physiological measure to detect lies, and as time passed more measures were added. None of these measures accurately detect deception. There are no specific physiological profiles that are indicative of deception. The so-called Pinocchio Response has not been found. There is much variability in response to control and relevant questions. Researchers haven’t developed the equivalent of Wonder Woman’s magic lasso, and “For at least the foreseeable future, the promise of a perfect lie detector remains the stuff of science fiction and comic book fantasy” (Lilienfeld et al. 2010). When referring to the lie detector, a more appropriate label is arousal detector, as it is a measure of nonspecific psychophysiological arousal (Lilienfeld et al. 2015).

References

Ben-Shakhar, G. 1991. Clinical judgment and decision making in CQT- polygraphy. Integrative Physiological and Behavioral Science 26: 232–240. 

Fienberg, S., and P. Stern. 2005. In search of the magic lasso: The truth about the polygraph. Statistical Science 20: 249–260.  

Iacono, W. 2008. Effective policing: Understanding how polygraph tests work and are used. Criminal Justice and Behavior 35: 1295–1308. 

Lilienfeld, S., et al. 2010. Great Myths of Popular Psychology: Shattering Widespread Misconceptions about Human Behavior. Malden, MA: Wiley-Blackwell. 

Lilienfeld, S., et al. 2015. Fifty psychological and psychiatric terms to avoid: A list of inaccurate, misleading, misused, ambiguous, and logically confused words and phrases. Frontiers in Psychology 6: 1100. doi 10.3389/fpsyg.2015.01100.

Shermer, M. 2007. Michael Shermer Tests the Polygraph and Lie Detection Part 1 (video). Available online at https://www.youtube.com/watch?v=GLL3wtgBiFA.