Improving learning experience in MOOCs with educational content linking

Abstract

Since the first MOOC (Massive Open Online Course) in 2011, there have been over 4,000 MOOCs on various subjects on the Web, serving over 35 million learners. MOOCs have shown the ability to transcend time and space, democratize knowledge dissemination, and bring the best education in the world to every learner. However, the disparate distances between participants, the size of the learner population, and the heterogeneity of the learner backgrounds make it difficult for instructors to interact with learners in a timely manner, which adversely affects their learning outcome. To address these challenges, in this thesis, we propose a framework of educational content linking. By linking pieces of learning content scattered in the various course materials into an easily accessible structure, we hypothesize that this framework will guide learners and improve content navigation.

Since most instruction and knowledge acquisition in MOOCs takes place when learners are surveying course materials, better content navigation may help learners find supporting information to clear up confusion and improve the learning outcome. To support our conjecture, we present end-to-end studies to investigate our framework around two research questions. We first ask, does manually generated linking improve learning? To investigate this question, we choose two STEM courses, statistics and programming language, and demonstrate how the annotation of linking among course materials can be accomplished with collaboration between course staff and online workers. With this annotation, we implement an interface that can simultaneously present learning materials and visualize the linking among them. In a large-scale user study, we observe that this interface enables users to find desired course materials more efficiently, and retain more concepts more readily.

This result supports the notion that manual linking does indeed improve learning outcomes. Second, we ask, can learning content be generated using machine learning methods? For this question, we propose an automatic linking algorithm based on conditional random fields. We demonstrate that automatically generated linking can still lead to better learning, although the magnitude of the improvement over the unlinked interface is smaller. We conclude that the proposed linking framework can be implemented at scale with machine learning techniques.