There was a chance to clean up the Internet from fraudsters,as well as on the speech culture of other users. Florida State University engineers have created artificial intelligence that copes well with the role of the polygraph. He monitors the words that the user sends, and the speed of their typing. In addition, technology separates truth from lies.
As shown by test results, AI is able to understand the lies and truth with an accuracy of 85%. It should be noted that the indicators of the detectors currently available are about 60 - 70%.
To train the system on the difference of truth and liesThe authors took advantage of machine learning. A group of 40 volunteers were offered to have fun with an online game. In it, they randomly determined the roles of sinners or saints. During the 80 sessions, the players answered different questions. Artificial intelligence while carefully tracking their messages.
As it turned out, sinners and saints are useddifferent sets of words. Yes, and they print at different speeds. Liars love to answer extremely quickly and use the words "never" and "always", thereby emphasizing their confidence. Those who speak the truth often use the notion of “possible”. They are thoughtful, which means that they type words relatively slowly.
Virtual online polygraph has learned to followset speed and words used, while recognizing the truth with an accuracy of 85 - 100%. As the researchers admit, testing was conducted on too few people. However, it is clear that the development potential is quite large. She is going to improve. After that, the authors of the project want to agree with the developers of different instant messengers on the popularization of their project. So it may well become commercial. They hope that over time, thanks to their development, communication on the Web will become better.