EXP-LA: Deceit Indication Through Person Specific Behavioral Dynamics

我们的军事和国土安全人员最重要的目标之一是在袭击我们的军事和民用设施之前抓住那些计划袭击我们的军事和民用设施的人。 该项目旨在识别，解释和开发自动测量意图攻击的人所表现出的行为线索类型的程序，特别是当他们试图隐藏或捏造他们的真实意图时。 尽管研究表明，没有一个信号或行为可以揭示一个人是在隐瞒他或她的意图还是在撒谎，但研究人员一致认为，大多数欺骗行为的线索都是因人而异的，这需要确定一个人正常行为的基线阅读，并校准基线的变化。该项目改进了先前的研究，因为它将检查在更现实，高风险场景中引发的行为，并且将在不给人附加任何工具的情况下进行，同时应用自动机器学习系统来获取每个指标行为，并确定哪些变化最能识别一个人是真实的还是说谎的。 然后，该程序将产生一个原型，该原型检查点和“坐下”审讯会话的视频，并自动产生欺骗/渎职可能性的单个综合评分。该项目更广泛的影响是，它将汇集行为、计算机和工程科学家，应用他们最先进的工具创建一个系统，该系统有可能在潜在的恐怖分子发动袭击之前识别他或她。 研究人员之间这种罕见的技能和知识结合所产生的协同作用不仅将促进这一特定主题的知识，而且还将促进社会心理学，计算机视觉和自动机器学习等研究领域的知识。最后，该项目将通过培训未来几代人，不仅是博士，和博士后学生，但也将适用于正在进行的反恐培训计划。

One of the most important goals for our military and homeland security personnel is to catch people who are planning to attack our military and civilian facilities before they strike. This project proposes to identify, interpret, and develop procedures for automatically measuring the types of behavioral clues exhibited by people who intend on attacking, particularly when they attempt to conceal or fabricate their true intentions. Although research has shown that there is no one signal or behavior that reveals whether a person is concealing his or her intentions or telling a lie, researchers agree that most behavioral clues to deceit are person specific, which require determining a baseline reading of the person's normal behavior and calibrating changes from that baseline. This project improves on prior research because it will examine behaviors that are elicited in more realistic, high stake scenarios, and will do it without attaching any instruments to the person, while applying automatic machine learning systems to take each indicator behavior and identify which changes are best able to identify a person as truthful or lying. This procedure will then produce a prototype that examines video of both checkpoint and 'sit down' interrogation sessions and automatically produces a single integrated score of deception/malfeasance likelihood. The broader impact of this project is that it will bring together behavioral, computer, and engineering scientists to apply their state of the art tools to creating a system with the potential for identifying a potential terrorist before he or she attacks. The synergy generated by this rare combination of skills and knowledge between the researchers will advance knowledge not only in this specific topic but also in areas of research such as social psychology, computer vision and automatic machine learning. Finally, this project will maintain this synergy momentum by training future generations of not only Ph.D. and Postdoctoral students, but will also be applied to ongoing counter-terrorism training programs.