Modeling Off-task Behavior of Learners Using Minecraft

Abstract

This study presents an attempt to develop a detector of off-task behavior using the interaction logs of the learners in Minecraft. Text replays will be used to obtain ground truth labels of off-task behavior. Detectors will be built using a Latent Response Model, which predicts whether an action or event is off-task behavior and assess how frequently each of the students is off-task. The relationship between off-task behavior and students’ learning outcome will also be examined. After developing the detector, a WHIMC module implementation will be conducted to apply intervention to keep the students from off-task behavior and to investigate how the learning outcomes will change after the intervention.