Adaptive Training for Real-Time Intelligence Monitoring & Evaluation

Period of Performance: 06/13/2002 - 03/13/2003

$100K

Phase 1 SBIR

Recipient Firm

Aptima, Inc.
12 Gill Street Array
Woburn, MA 01801
Principal Investigator

Abstract

We propose to develop an Intelligent Tutoring System for AF intelligence analysts (ITS4Intel) that helps student analysts to refine and apply the two cognitive skills that are fundamental to their work: categorization and inference. It will use scenario-based training to team them to assess the quality of intel products (e.g., messages); assess their relevance to current information requirements; make assessments of intent, predictions, and other inferences from the data; and select the best sources from which to elicit additional information. The system will employ three technologies: an extension of Latent Semantic Analysis for modeling human categorization abilities, an IBIS architecture for representing relations between elements of knowledge and emulating expert inference over it, and an instructional system shell that integrates these modeling engines and adaptively presents explicit instruction, practice scenarios, and feedback. A significant innovation is the explicit representation of mental models that elicits observable and measurable indicators of cognitive state and process. By working with the explicit models, students will learn how to structure and use a large, complex body of intelligence material. A key efficiency of this effort is that it will leverage knowledge acquisition research underway in an existing AFRL SBIR concerning AF intelligence analysis. When completed, the Phase I work will produce as catalog of mental models and quality assessment heuristics used by expert intelligence analysts in a specific domain, training objectives for intel analysts, measures of student performance, a prototype training system, and an evaluation of the prototype by operational personnel. The Phase II work will produce a robust system for training AF intelligence analysts, deep and formally represented knowledge of the mental models employed by analysts, and validated measures of cognitive state and cognitive processes pertaining to intel analysis. In Phase III, we will transition the system to other markets that value training in qualitative analysis skills, and address opportunities to transform the training system into a job aid for AF intel analysts.