The software developed by the University of Michigan examines the verbal and body language: If your intelocutore moving too hands, for example, is not good sign
Live
> (Photo: Darrin Klimek / Getty)
Knows Dr Cal Lightman (Tim Roth) Lie to Me when lie , we involuntarily a series of clues which you can smascherarci. Small gestures, the intonation, the direction of our gaze: all details that can help an expert to understand when we are saying a lie .
With this in mind, a team of researchers at the ‘ University of Michigan has developed a Machine truth new generation: a lie detection software that can detect a lie without the sensors of the old polygraphs , but simply viewing the film a interrogation .
The program, presented during the recent International Conference on Multimodal Interaction , it works thanks to the machine learning or learning Automatic . The researchers trained that the software by sending it more than 120 movies to interviews , and then establish if the people shooting were lying or not using the verdict of the relevant process.
In this way, the software was able to identify several clues that reveal a potential lie: a liar would tend for example to move their hands more often , try to seem more certain of his statements, would look most often in the eye your interrogator, and distanzierebbe from the scene that describes, using more often terms like “ he ” or “ you “, rather than “ I “or” we . “
At the end of the training, the program had at its disposal a wide repertoire of gestures and verbal clues can detect a lie. It was then put to the test on a new group of movies, demonstrating that it can identify a lie 75% of the cases. Perhaps not enough to send someone to jail, but surely we would know better than we humans do. In the experiment, in fact, people who have undergone the same movies have shown a capacity of just over 50% (basically how to make heads or tails) to recognize when these were lies.
“ Humans are very good at recognizing a lie ,” says Rada Mihalcea , a researcher who coordinated the project. “ It simply is not the type of activity in which we can well. There are many clues that humans provide involuntarily when they lie, but we do not pay enough attention to recognize them. We do not count as how many times a person says I, or how often looks up to heaven because they have focused on a higher level of communication . “
Despite the good results, the researchers point than that achieved so far is only a piece of a larger project, which aims to create a device to identify the lies that can prove to be useful to the police, medical personnel and during the trial. To do so, they are approaching the problem from different angles, to further improve the accuracy of the program. The next step, for example, will also be integrated in the data analyzed by conventional polygraphs, such as heart rate, breathing and body temperature, but collected with non-invasive methods of imaging heat.
Want to receive updates on this topic?
Enter your email address -> -> Follow
No comments:
Post a Comment