OnDiscuss: visualizing asynchronous online discussions through an epistemic network analysis tool

Abstract

Asynchronous online discussions are common assignments in both hybrid and online courses to promote critical thinking and collaboration among students. However, the evaluation of these assignments can require considerable time and effort from instructors. We created OnDiscuss, a learning analytics visualization tool for instructors that utilizes text mining algorithms and Epistemic Network Analysis (ENA) to generate visualizations of student discussion data. Natural language processing and text mining techniques are used to generate an initial codebook for the instructor as well as automatically code the data. This tool allows instructors to edit their codebook and then view the resulting ENA networks for the entire class and individual students. Through empirical investigation, we assess this tool's effectiveness to help instructors in analyzing asynchronous online discussion assignments. Our findings highlight several key insights regarding the implications of this tool for enhancing the accessibility and usability of ENA as a learning analytics visualization tool. OnDiscuss is helpful to those unfamiliar with ENA since it abstracts many of the intricacies of ENA by providing an easy interface to manipulate a codebook and thus the resulting ENA networks. Future refinements, such as the addition of a baseline ENA model, can make it more helpful to those familiar with ENA. Despite the tool's automated keyword generation capabilities, it is clear that instructor intervention remains crucial for refining the codebook. Therefore, while automated techniques like Latent Dirichlet Allocation (LDA) provide valuable insights given a large amount of data, these processes must be complemented by expert guidance.