Toward Intelligent Instructional Handoffs Between Humans and Machines
2016; NIPS, Neural Information Processing Systems; Ritter, S.; Fancsali, S.E.; Yudelson, M.; Rus, V.; Berman, S.
We describe preliminary results of the Integrating Human and Automated Tutoring Systems (IHATS) Project, the goal of which is to leverage a unique dataset containing information about student usage of both an automated tutoring system for algebra as well as transcripts of chat sessions between these same students and human tutors. We seek answers to questions about what affective, behavioral, and cognitive factors predict that students will seek human assistance (and/or factors that drive them to human assistance) while using an automated tutoring system and what characterizes when such tutor tuttee interactions will be enhance learning, using data from the automated tutoring system to measure such learning. The project leverages a variety of statistical and machine learning techniques to answer these questions. Longer term, answers to these questions will be vital to developing systems that can intelligently guide students (i.e., make instructional handoffs) between automated and human sources of assistance as and when necessary.