Title: A Benchmarked Experiential System for Training (BEST) for Optimizing Instruction

Abstract We address the problem of optimizing instructional strategy for team command and control training exercises in simulator environments. In the first phase of this work, we developed model-generated, near-optimal solutions to complex C2 scenarios, as well as animations and presentation techniques that supported their use as feedback. Experimental results demonstrated a reliable advantage for the group receiving this treatment, an advantage that could theoretically halve training time. In the second phase of this work, we are combining three computational models to optimize the order of presentation of DDD C2 practice scenarios. The team developed (1) a communications model, used to assess communications content; (2) an optimization agent that generates near-optimal solutions to scenarios, used as a benchmark for human solutions; and (3) a POMDP model that recommends the next practice scenario (among many available) to accelerate team performance towards mastery of three competencies. Experimental validation is underway to validate this multi-model approach to optimizing team learning. This work advances the science of training by developing models to assess and guide team learning. In addition, this work is producing training content for air command and control teams, specifically those in AWACS and the Air Operations Center (AOC) Dynamic Targeting Cell (DTC).