Adaptive, individualized training assessment capability (AITAC)

The revolution in simulation and virtual technologies for training is enabling learners to practice skills and learn in realistic environments. Researchers are also seeking to develop algorithms that can tailor learner practice to the estimated abilities and needs of individual students, offering the promise of individual adaptation to each learner"s dynamically evolving zone of proximal development. Such systems will likely improve learning outcomes, reduce learner frustration, and mitigate the need for artificial (out of domain) scaffolding. However, such adaptive training requires that a learning environment estimate student proficiency and need as learning progresses and most approaches assume that the primary assessment dimensions are known in advance, which limits the ultimate effectiveness of adaptive training systems. What would be preferable is the ability of the system to assess progress and stumbling blocks within the environment and diagnosis and measure these deficits in terms of dimensions"outside"of the a priori trainee representation. We propose a direct questioning approach, building on domain ontologies and prior work in machine learning of student proficiency modeling, along with dynamic assessment and tailoring, to build a more accurate and more extensible representation of individual trainees.