CIF21 DIBBs: PD: Enhancing and Personalizing Educational Resources through Tools for Experimentation

Worcester Polytechnic Institute, Worcester MA

Investigators

Abstract

This project would automate the creation and data analysis of randomized controlled experiments (RCEs). RCEs are question sets designed and delivered to teachers and students in a classroom setting, and can be used to compare alternative educational strategies. This effort builds on an existing educational platform (ASSISTments) developed by the Principal Investigator, and uses a template-based approach to increase the efficiency and reliability of conducting educational research. The goal is to lower the barriers to creating and learning from randomized controlled experiments. The project has the potential to facilitate large-scale learning in education research, reaching hundreds of schools and thousands of students. The project builds upon two prior developments by this team. - ASSISTments is an online learning platform originally designed to provide students with assistance and teachers with assessments (establishing the moniker).  The system is used primarily as an online tutoring system for middle or secondary education, supporting the delivery, collection, and grading of classwork and homework, providing immediate feedback for students and explicit reporting for teachers. To date, 24 randomized controlled experiments comparing educational strategies have been published using this platform. - In addition, AssistmentsTestBed.org is a testbed developed by the PI and his group under a separate NSF grant (#1440753), to identify best practices in education and allow other researchers to propose and run their own studies leveraging ASSISTments as a shared scientific instrument through this testbed. Beneficiaries include education researchers, teachers, and students, with the existing tool being used in over 500 schools and in the education of over 50,000 students. The current project improves two components of the infrastructure that have been resource-intensive bottlenecks in prior research. One task automates the process of study creation and data analysis, through development of a Template Tool that enables studies within ASSISTments. A second task automates statistical analyses and improves usability of the existing data reporting tool (Assessment of Learning Infrastructure, or ALI). The improvements will be achieved, in part, by applying educational data mining algorithms (i.e., deep knowledge tracing) on student data collected before, during, and after experimentation. These analytics will provide researchers with covariates that will significantly improve the agenda of personalizing education. The resulting capability will assist researchers as they design and deliver question sets to teachers and students in a classroom setting, increase the efficiency and reliability of conducting educational research at scale, and streamline the research processes. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the NSF Directorate for Education and Human Resources, Division of Research on Learning in Formal and Informal Settings.