Using Data Mining to Automate ADDIE

The goal of this work is to transform informational and instructional content into adaptive and personalized training experiences. We have developed semi-automated methods to do this that parallel the traditional “ADDIE” (Analysis, Design, Development, Implementation, and Evaluation) process. The source content can include documents, presentations and manuals and existing courseware. The techniques use artificial intelligence (AI), data mining, and natural language processing and generally belong to the discipline of “educational data mining.” This poster/demo demonstrates the processes and discusses the algorithms used. 1. PROBLEM STATEMENT Today’s digital environment is rich with learning content, but much of it is purely didactic in nature. This content includes manuals and presentations not intended for instructional purposes and e-learning that consists of presentations and lectures with multiple choice questions. As online learning replaces instructorled training in corporations, government agencies, and educational institutions [10], its effectiveness can be improved by transforming this wealth of didactic content into more interactive and adaptive learning experiences [5]. Here, we address aspects this transformation problem in the context of multiple research and commercial projects. A large portion of the work we report here comes from a U.S. Army Small Business Innovation Research (SBIR) project called Tools for the Rapid Generation of Expert Models, or TRADEM, that applies data mining to (a) deconstruct existing content at a deep and granular level and (b) reconstruct it in a form that can be used to create adaptive intelligent tutoring systems. This process automates many steps in the “ADDIE” (Analysis, Design, Development, Implementation, and Evaluation) process [1] commonly used to develop instructional content.