Cheating the Lie Detector

In its war on terror, the U.S. government uses handheld lie detectors for fast screening of terrorism suspects. It has been shown that the pocket polygraphs have high error rates and are susceptible to successful faking, so the autobiographical Implicit Association Test (aIAT) was developed as an alternative lie detection tool. Initial assessments of the aIAT reported extremely high accuracy rates. However, in a new paper published in Psychological Science, a journal of the Association for Psychological Science, psychologists Bruno Verschuere and Jan De Houwer from Ghent University in Belgium and Valentina Prati from Padua University in Italy caution that the aIAT is not immune to the inaccuracies and successful faking associated with other lie detector tests.

The aIAT is a very simple-to-use lie detector test, measuring reaction times of responses and requiring only a standard computer. During a test session, which is made up of two sections, participants indicate if a statement is true or false. In the Confession-true section, participants press the same button for statements that are true and those that indicate guilt (e.g., “I stole the CD-ROM). In the Denial-true section, participants use the same button for statements that are false and those that suggest guilt.  Guilt or innocence is determined by the speed of responses — guilty participants will response faster to statements in the Confession-true section and innocent participants respond faster in the Denial-true section.

In these experiments, a group of students participated in a mock-crime scenario and then completed an aIAT session. For the scenario, they were assigned to either the guilty (they were to steal a CD-ROM from a professor’s office) or innocent (they read about the crime in the newspaper) group.

The results indicate that guilty participants were found to be guilty more often than innocent participants, supporting the validity of the aIAT. However, the accuracy of the aIAT in the current study was much lower (61% in innocent and 67-86% in guilty participants), compared to the 91% accuracy seen in previous work. The researchers note that laboratory research tends to overestimate accuracy due to the nature of the experimental design and as such, “these accuracy figures should be regarded as the upper boundaries of the aIAT’s potential in forensic settings.”

In addition, the researchers conducted separate experiments to investigate how susceptible the aIAT is to faking. The results showed that successful faking occurred for guilty participants who had taken the aIAT previously, and even more interestingly, guilty participants taking the aIAT for the first time were also able to obtain innocent outcomes. The researchers conclude that “the aIAT is subject to the same shortcomings as other lie detection tests.”