Scenario-Based Conversational Intelligent Tutoring Systems for Decision-Making Skills

U.S. warfighters are being asked to work in ever more complex operations and environments, where everyone from the general officer to the “strategic corporal ” must engage in critical reasoning and creative thinking. Advanced military professional schools often address such skills, adopting a virtual apprenticeship or mentorship approach: students analyze past cases and work through problems while an expert practitioner probes their thinking and models good practice. As more of our military requires such skills, it is desirable to make this labor-intensive form of education readily available to a broader military audience. The emerging technology of Intelligent Tutoring Systems (ITS) has successfully provided computer-based training with automated individualized assessment and instruction across a range of procedural and reasoning tasks. Scenario-based techniques can potentially allow ITS construction even when computerized expert performance models cannot be built. However, it remains difficult to simulate interaction with a cast of simulated characters, or to duplicate the Socratic probing of an experienced instructor. We describe an approach to construction of scenario-based ITSs that (1) supports a range of interaction styles, including simulated discussion with scenario characters and an automated Tutor, encompassing both student-initiative and agent-initiative dialogs, (2) is supported by an authoring style that lowers per-scenari