Modular Annotated Learning for Instructional Authoring (MALINA)

ABSTRACT:Recent efforts have produced effective concepts for reusable infrastructure and standards for intelligent tutoring system (ITS) development, but even with these tools it remains a complex task to encode assessment knowledge for a given domain.Stottler Henke proposes to develop a set of authoring tools called Modular Annotated Learning for Instructional Authoring (MALINA) that provides a novel collaborative approach to constructing assessment knowledge that combines machine learning with instructor-provided assessment and annotations. The MALINA technology will use constraint-based knowledge representation techniques for cost-effective system development, via innovative machine learning techniques for constructing the knowledge required for automated assessments. These techniques will allow instructors to specify this knowledge by demonstrating and explaining a range of optimal and suboptimal solutions to exercise scenarios, collected either as traces from actual exercises, or instructor-created examples. The system induces performance assessment knowledge from the annotated demonstrations. Assessments of simulation performance will be used to update a long-term student model for each student using Bayesian inference. Phase II will lead to the development and evaluations of the MALINA authoring tool as well as prototype exemplar training capabilities for two domains, first to apply the authoring methodology and second to explore its generalizability.BENEFIT:The technology and products resulting from this effort have both direct and indirect transition potential.Most specifically, research products will include exemplar training applications developed for two satellite control domains using our authoring approach.Our partnership with Boeing will help in identifying transition paths for the authoring tools to related satellite control domains, including payload operations.More broadly for authoring tools, there are a variety of commercialization and transition directions for the authoring concept of using annotated machine learning to facilitate the knowledge- intensive work of developing performance assessment mechanisms for a given domain.Authoring tools that make Intelligent Tutoring Systems (ITS) more cost effective ultimately benefit the entire industry, both from an end user perspective and a developer perspective.End users benefit from the prospect of more direct participation in the process of building training systems that meet their needs.Developers also benefit when authoring tools reduce the cost proposition for marketing development services, which may have previously been a barrier to the adoption of ITS technology.Tools to facilitate the representation of domain knowledge through machine learning also have transition opportunities for emerging trends in massively open online courses and even autonomous control technologies for unmanned systems.