An investigation of addressing students behavioral disengagement in technology-mediated learning environments

Abstract

There is an increasing trend of integrating digital technologies into teaching and learning practices. Contemporary techniques, such as cloud computing, can facilitate students learning without overburdening teachers. However, keeping students engaged remains challenging in technology-mediated learning environments. Studies have found that disengagement can be the early warning indicators of dropping out, which has been a major concern of online courses. Indeed, students who are disengaged in learning activities tend to have poor academic performance and communication skills. This issue is even more severe in flipped learning classrooms because limited engagement with pre-class course content, indicating poor student preparation, poses formidable threats to the effective application of flipped classrooms.

Many technical approaches have been proposed to foster student engagement. A plethora of studies suggested digital technology integration without changing teachers' instructional design. For instance, some researchers used machine learning to predict the engagement level of students, and then provided interventions to disengaged students. However, problems such as absence of proper interventions, shallow feature engineering, and inadequacy of training data are commonly found in relevant studies. Furthermore, some researchers preferred using proactive methods to engage students via integrating virtual reality, gamification, and other techniques into in-class learning activities, but students are isolated and easily disengaged outside the class. Moreover, other researchers tried to help adjust teaching design and increase student engagement through teacher-centered dashboards. Nonetheless, these studies exhibit a lack of theoretical framework, and they employ indicators that are superficial.

This dissertation aims to address student behavioral disengagement in technology-mediated learning environments in three aspects. Firstly, a multidimensional disengagement prediction model is proposed to address high dropout rates in Massive Open Online Courses (MOOCs). With in-depth Web analytics features and multi-MOOC data, this model does not have the problem of insufficient training data, shows good performance (AUC: 0.83), and enables dimension-wise interventions to at-risk students. In addition, to increase students' pre-class engagement in flipped classrooms, cloud collaboration tool is utilized to help construct an online community of inquiry. Experiment results showed that students in the experimental groups had a strikingly higher pre-class engagement. Last but not least, an online course engagement evaluation framework is built based on Moore's learning interaction theory and Web analytics knowledge. Teachers can utilize this framework to facilitate their teaching practice, adjust teaching design, and improve student engagement in the course level. Ten third-party instructional designers were invited to comment on this framework and derived insights, and most of them showed positive attitudes.