An intelligent tutoring system based upon adaptive simulation

As civilization becomes more dependent on complex technology, education develops increasing importance. Computerized tutoring is a practical mechanism for making advanced education more widely available, especially for ongoing maintenance of skills in the workplace and other situations outside formal institutions of learning. We built a simulation-based intelligent system for teaching medical personnel cardiac resuscitation skills, including a novel mechanism for monitoring student performance and dynamically adapting the simulation to the student's needs, using an expert model of the domain encoded with planning representations. Domain topics were explicitly represented so the system could reason about the student's learning needs and accomplishments, while adapting the simulation context toward states where profitable learning was expected, keeping the student challenged but not overwhelmed. Knowledge of what to teach was separated from knowledge of how to teach, leading to simplified representations that support improved communication between domain and system experts as demonstrated by the fact that a major change in the medical standards adopted during system development was implemented in about a week. The system used planning techniques to make common sense interpretations of unexpected situations while reasoning about simulations of multiple, coordinated agents with distinct roles performing multiple tasks. The system combined techniques from simulation, planning and user modeling in a multimedia tutoring environment, demanding consideration of the different strengths and idiosyncrasies of these components. This approach to integrating domain models, student models, course goals and pedagogical knowledge to create an individualized study plan in a tutoring system is novel and results in improved learning times. An iterative development methodology with formative evaluations of the system involving two classes of medical students helped direct the technology to meet the practical needs of the user community and ensured that the final system was accepted in the target environment. Results of this evaluation support our architecture and development approach. This research contributes to global access to knowledge which, it is hoped, will lead to improved decision making at all levels of civilization, providing for more efficient use of resources, more effective conflict resolution strategies and an overall increase in the satisfaction of critical social goals.