Automated Determination of the Veracity of Interview Statements from People of Interest to an Operational Security Force

In deception detection research validity issues have been raised when subjects are used in controlled laboratory experiments. Studying real-life deception detection is a complicated endeavor because researchers do not have the control in field studies that exist in laboratory experiments so determining ground truth is challenging. This study reports the findings of the combination of some successful previous attempts at automated deception detection in computer-mediated communication results of a study of real-world data from an operation security force. Message feature mining is used to evaluate the effectiveness of technology as an aid to deception detection in actual stressful situations with unpleasant long term consequences. The study analyzes 18 statements (9 truthful, 9 deceptive) from a military service’s investigative unit using message feature mining. The analysis resulted in a 72% rate of accuracy in correctly classifying the messages.