Tool for Assessing and Conveying Testimony Veracity (TACT-V)

Human Intelligence (HUMINT) collectors, whether trained Military Intelligence (MI) experts or Infantrymen, face a transition in the way they collect intelligence from indigenous populations, particularly in Counter-Insurgency (COIN) operations. This transition is driven by the high frequency with which asymmetric insurgent threats are embedded in the local populace, which forces an approach to HUMINT gathering where testimony is captured along with an assessment its quality (e.g., reliability, consistency) to help determine its value in any future decision-making process. To enhance the collection and exploitation of tactical HUMINT, soldiers need a tool to identify, assess, and dynamically alert them to false and inconsistent testimony gathered during an interview. We propose to design, develop, and evaluate a Tool for Assessing and Conveying Testimony Veracity (TACT-V) that will allow HUMINT collectors to rapidly assess and convey the contextual veracity of source testimony. For this effort, we will leverage our significant experience developing intelligent systems for Military Intelligence (MI) activities and exploit our access to MI experts to create a tool that leverages Bayesian reasoning over networks representing source testimonies to associate overlapping testimony, identify inconsistencies within those overlaps, and generalize source reliability based on these inconsistencies.