Everybody lies and that’s a fact. But is everyone capable of picking up on the clues suggesting a lie? The answer may be discouraging: people aren’t really good at detecting lies. However, machine learning could do that for us. A novel lie detecting software accurately reveals the truth and has been designed by researchers with the University of Michigan.
The research involved watching hours of footage depicting high-stakes court cases. The lie detecting software is unlike any polygraph currently being used. Instead of measuring a subject’s pulse, heartbeat or breathing rate, the novel lie detecting software accurately reveals the truth based on an analysis of gestures and the use of language.
We often try to pick up the same hints. However, as we’re too focused on the act of communication itself and invest some emotional resources in the act, we’re not proficient in reading the telltale signs.
The lie detecting software developed at the University of Michigan is. During tests, the software has an accuracy rate of 75%. People detect a lie in only 50% of the cases. Generally, the percentage also refers to blatant lies. So what of the high-stakes court cases? Monitoring hours of footage depicting the high-stakes course allowed the researchers to create a large database of gestures and language specific to people lying when the stakes are high. The results were correlated with the verdicts of the trials.
The gestures included too much use of a person’s hands, grimaces or scowls. Pay attention to someone using a hesitant ‘umm’ too often. It might be a telltale sign of a lie. At the same time, most of those lying in the courtroom employed pronouns like ‘he’ or ‘she’ more often than ‘I’ or ‘we’. This suggests an assumed distance between the perpetrator and the victim or committed crime. Some of those who were found to tell lies attempted to make the interviewer believe they were telling the truth by maintaining direct eye contact for a longer period of time.
The newness of these results is based on the fact that they focused on real-life situations rather than laboratory tests. According to Rada Mihalcea, researchers working on the lie-detecting software:
“In laboratory experiments, it’s difficult to create a setting that motivates people to truly lie. The stakes are not high enough. (…) in the real world there is true motivation to deceive”.
The lie-detecting software will be improved, say the researchers. In addition to analysing gestures and language, it will also look at physiological parameters through thermal imaging. No more touching with the new lie detection technology.
Photo Credits: Pixabay