Technology

Anticipatory Discourse Analytics (ADA) Tool

ADA employs SAE's text analysis toolkit to codify text data from social media and adversary statements, speeches, and propaganda into cognition, world-view, identity, group dynamic, and emotion and sentiment measures. Analysts can then use the interface to filter data by time, space, actors, issues/narratives, and indicators. The tool provides alerts for outlier messages to analysts based on filters and indicator selections and provides a robust plotting capability (from time series charts to network graphs) that visually draws the analyst to interesting parts of the data. The data and tool go beyond traditional sentiment and network analysis capabilities. In prior work, SAE demonstrated that including ADA indicators in models improves the forecasting accuracy for adversary behavior.

Assessing Deterrence Effectiveness and Predicting Tactical Successes (ADEPTS) Tool

ADEPTS focuses on empirically measuring the effects of courses of action (COAS) to shape and deter adversary behavior. To do so, SAE developed counterfactual, quasi-experimental methods and impact assessment models to rigorously estimate the effects of alternative courses of action. The methodology and software tool are tailorable and adaptable to many types of data, questions, levels of analysis (strategic, operational, and tactical), and domains (air, land, sea, social, cyber, and space). In particular, ADEPTS automates the causal model discovery process by employing a novel algorithm to generate candidate confounding and mediating variables within the causal chain. It then employs machine learning algorithms to create quasi-experimental data from observational data creating control and experimental groups for causal analysis. Next, the system estimates causal effects using a variety of algorithms and provides a number of metrics to assess confidence in the estimates.