Charles River Analytics Chief Scientist Dr. Subrata Das spoke on computational approaches to situation assessment in an invited talk at the Third Workshop on Critical Issues in Information Fusion. Information fusion, the process of bringing together the masses of raw data from sensors and intelligence reports into graphical representations, alert messages, and other formats that intelligence analysts can use to make decisions, is a critical task for military and government intelligence organizations.
The workshop, sponsored by the Intelligence and Information Warfare Directorate (I2WD) of the Communications-Electronics Research, Development and Engineering Center (CERDEC) of the US Army Research, Development and Engineering Command (RDECOM), or, US Army RDECOM CERDEC I2WD, brought together 50 information fusion researchers and practitioners to identify the critical issues currently faced by their community. Their discussions focused on the challenges of understanding higher-level information fusion—the process of using information such as unit types and locations to predict what those units will do next.
Dr. James Llinas, an internationally recognized expert in sensor, data, and information fusion, organized the workshop. He summarized the challenges associated with higher-level fusion this way: “When you want to answer behavioral questions, you need to bring in new tools, such as artificial intelligence and automated reasoning methods.”
In his talk at the workshop, Dr. Das explained how artificial intelligence methods and algorithms can be applied to streams of intelligence data to provide analysts with an understanding of the battlefield and recommend responses to events. The technologies presented included static and dynamic Bayesian belief networks, symbolic argumentation, influence diagrams, decision trees, and particle filtering. Dr. Das reviewed several specific applications of these techniques, including terrorist threat prediction based on information gathered from news articles, and detection of an attack via an unknown bio-agent using emergency room records.
Other speakers at the workshop discussed automated learning methods and algorithms to help intelligence analysts understand rapidly evolving situations, and the use of ontological approaches to more rapidly fuse information from a wide variety of very different sources