Closing the Knowledge Gap in an Online Learning Community: Network-Analytic Discoveries, Simulation and Prediction

Learning communities have been shown to positively impact their participants' learning experiences. Research on learning communities and the associated peer effects have influenced decisions in many knowledge-spheres, from education to policy. However, little effort has been devoted to experimental studies that examine the impact of online interactions in characterizing how the knowledge of individual participants and the community as a whole shape up temporally. We introduce a framework called ROC Speak that we have developed towards collecting a novel dataset of six online public speaking learning communities; and use graph signal processing to elucidate the mechanism and implication of knowledge propagation therein. Using a novel total variation analysis framework, we show that participants converge to a similar increased performance level as a function of increased volume of online interactions; and that initially heterogeneous communities gradually approach knowledge homogeneity in their learners' speaking qualities. We formulate a simulation framework via consensus-based diffusion dynamics to model the observed knowledge accumulation and diffusion phenomena. At an individual level, we show that the learners' speaking qualities are impacted by the nature of their interactions. This is established by demonstrating that our proposed consensus-based prediction framework outperforms baseline network-agnostic regression models in estimating learning outcomes.