Conversational Agents in a Pattern Oriented Training Environment (CAPOTE)

Initial pilot training programs such as Undergraduate Pilot Training (UPT) presents students with an array of complex skills to acquire and integrate in a dynamic, time-sensitive performance context. Since most of a student's hours in UPT are spent on theground, training could be fundamentally improved if students had on-demand access to flight training devices that were sophisticated enough to provide automated instruction in complex, highly interactive flight regimes. Moreover, training skills that relyexclusively on time in aircraft or in dedicated, high-fidelity simulators is put at risk given the limited time window UPT programs afford. A solution to this problem is to address two needs at the same time, providing (1) greater access to operationalflight trainers; and (2) an automated intelligent tutor that reduces reliance on human instructor pilots. We propose Conversational Agents in a Pattern Oriented Training Environment (CAPOTE), a framework that offers important, high-payoff extensions toCOTS simulation. CAPOTE employs intelligent tutoring to provide effective, instructor-less training opportunities, and synthetic agents that can assume key roles within training scenarios and interact in spoken language with the trainee. Our deepexperience in training systems and speech-enabled cognitive agents provides mitigation against the risks presented by these complex technologies. Technology and intellectual property arising from this research and development program are expected to havesubstantial commercialization potential in both the public and private sectors. The specific technologies discussed here are simulation-based training coupled with intelligent tutoring, and speech-enabled synthetic agents. In the public sector,simulation-based training is a growing concern in both military and non-military agencies. The need to effectively train military and non-military personnel is especially relevant for cross-agency collaborations such as homeland security,counter-narcotics, disaster relief, and emergency services. In such instances, the need for training teams can be met with the use of synthetic, speech-enabled agents to stand in for missing team members or to role-play other participants in a trainingscenario. In the private sector, large and fast-growing market segments to which this technology applies include operator training and distributed learning. Widespread use in industry and education of internet and telecommunication technologies areincreasing demand for delivering remote training and for aggregating teams of geographically distributed collaborators. Tailoring instruction and assessing performance of participants in remote team training will hinge centrally on a capability to providesophisticated automated tutors, and on the availability of synthetic agents to populate scenarios.