PNNL

Abstract

In the role of detecting and preventing strategic surprise, intelligence analysts rely heavily on history as they refer to past cases and events to find similarities and distinctions to a current situation or threat. Analysts’ ability to identify similar cases and events, however, is very limited because the intelligence data available for searching are largely disconnected and limited in context. As a result, analysts searching intelligence data often resort to searching by keyword or timestamp. To make better use of intelligence data, analysts should not only be able to construct hypotheses of a current situation or threat but also reconstruct the full context of past events such that those events may then be effectively compared to the current situation and hypotheses. In constructing the context of a situation, case, or hypothesis, analysts decipher patterns and relationships among many different kinds and sources of information. Analysts identify these patterns and relationships through different kinds of analyses. For instance, analysts may conduct link analysis to examine different kinds of relationships, geographical analysis to look at spatial positioning and grouping, timeline analysis to lay out chronological events, and visual analysis to identify clusters of similar information. What these different kinds of analyses provide are analytical frameworks or models for organizing intelligence data along key properties. Unfortunately, these higher-level analytical models are transient: they disappear when their associated analysis tools stop executing. Consequently, analysts cannot easily associate their analyses with their intelligence data in a way that facilitates the integration of different analyses into a larger picture or the comparison of analytical models to other cases. On the Scenario and Knowledge Framework for Analytical Modelling (SKFAM) project at the Pacific Northwest National Laboratory, we are researching and prototyping a framework that allows analysts to naturally construct decision-making analytical models, apply them with their analysis tools, and save them along with its associated data into analytical scenarios. These scenarios may later be recalled, compared against current and hypothetical situations, and shared with other analysts.