Data Science

The path towards a more democratized learner success model for MOOCs has been hampered by a lack of capabilities to provide a personalized experienced to the varied demographics MOOCs aim to serve.  Primary obstacles to this end have been insufficient support of real-time learner data across platforms and a lack of maturity of recommendation models that accommodate the learning context and breadth and complexity of subject matter material in MOOCs. In this paper, we address both shortfalls with a framework for augmenting a MOOC platform with real-time logging and dynamic content presentation capabilities as well as a novel course-general recommendation model geared towards increasing learner navigational efficiency. We piloted this intervention in a portion of a live course as a proof-of-concept of the framework. The necessary augmentation of platform functionality was all made without changes to the open-edX codebase, our target platform, and instead only requires access to modify course content via an instructor role account.

The organization of the paper begins with related work, followed by technical details on augmentation of the platform’s functionality, a description of the recommendation model and its back-tested prediction results, and finally an articulation of the design decisions that went into deploying the recommendation framework in a live course.