2019_I/ITSEC_Enhancing Learning Outcomes Through Adaptive Remediation with GIFT (Best Paper Education Subcommittee)
Adaptive instructional systems (AISs) are envisioned to play a vital role in the Army’s future training environment.
A key feature of AISs is the capacity to automatically tailor instruction to fit the needs and skills of individual learners.
Leveraging recent advances in artificial intelligence and machine learning, training and educational experiences can
be tailored to the goals, learning needs, and preferences of individual learners and teams of learners. Tutorial planning,
a critical component of adaptive training, controls how scaffolding and instructional interventions are structured and
delivered to learners to create personalized learning experiences. Devising computational models that effectively
scaffold learning experiences is a critical challenge for the field. For example, AISs need to determine when to
scaffold, what type of scaffolding to deliver, and how scaffolding should be realized, all in real time. In this paper, we
describe our work using the Generalized Intelligent Framework for Tutoring (GIFT), an open source framework for
creating, deploying, and evaluating adaptive training systems, to create a web-based adaptive short course for teaching
fundamental principles associated with counterinsurgency (COIN). The course presents students with a series of
videos, integrated assessments, and remediation materials about doctrinal COIN concepts. The course’s adaptive
remediation features are based on the ICAP active learning framework to deliver constructive, active, and passive
forms of remedial feedback to students. We report the results of a recent study in which 500 participants completed
the adaptive training course along with pre- and post-training knowledge tests. The paper provides an analysis of
learning gains and factors that moderated these gains and concludes with a discussion of future research as it pertains
to the goals of the broader research program investigating applications of machine learning, and reinforcement
learning in particular, to automatically generate policies for instructional remediation in AISs.