Predicting Gaps in Usage in a Phone-Based Literacy Intervention System

Educational technologies may help support out-of-school learning in contexts where formal schooling fails to reach every child, but children may not persist in using such systems to learn at home. Prior research has developed methods for predicting learner dropout but primarily for adults in formal courses and Massive Open Online Courses (MOOCs), not for children’s voluntary ed tech usage. To support early literacy in rural contexts, our research group developed and deployed a phone-based literacy technology with rural families in Côte d’Ivoire in two longitudinal studies. In this paper, we investigate the feasibility of using time-series classification models trained on system log data to predict gaps in children’s voluntary usage of our system in both studies. We contribute insights around important features associated with sustained system usage, such as children’s patterns of use, performance on the platform, and involvement from other adults in their family. Finally, we contribute design implications for predicting and supporting learners’ voluntary, out-of-school usage of mobile learning applications in rural contexts.