Higher Automated Learning (HAL)

The preparation of training materials is a labor-intensive process. It typically starts with manual task analysis or knowledge elicitation sessions requiring significant time commitments from both training professionals and SMEs. Those products must then be manually transformed into the building blocks for training content development and learning assessment. Clearly it would be of great benefit to automate the development of training materials, and the proposed solution Higher Automated Learning (HAL), represents a first step in that direction. HAL uses a Partially Observable Markov Decision Processes (POMDP) model based on the domain ontology and student performance data to model the learning of the target domain. HAL"s model will feature nodal representations of students"actual learning states as they progress (or struggle) through the curriculum. Additionally, HAL will automate the definition of metrics and assessments to provide trainees with a sense of where they stand, what their future performance potential is, and what experiences might support more rapid improvement. When fully developed at the end of Phase II, HAL will accelerate and deepen the development of student/trainee models and enable automatic discovery of performance metrics, which identify critical relationships among key constructs and support diagnostic assessments of student states.