Date of Award
Doctor of Philosophy (PhD)
Technology has become a crucial part of the way learners and teachers are navigating the educational landscape today. Yet understanding the effects various technologies are having on the preparation, planning and delivery of instruction is both difficult and elusive. As educational interactions of all kinds continue to move to online platforms and massive amounts of data are being collected on student and teacher activity within these environments, techniques are being developed as part of the incipient field of educational data mining (or more broadly educational informatics) to both make sense of and derive value from these data. This dissertation is built upon three research studies designed to apply computational techniques to analyze and understand Earth science teachers' usage of and behavior within an online curriculum planning tool called the Curriculum Customization Service (CCS). The first of these studies examines the aggregate behavior of teachers to develop a foundation for unsupervised clustering to be used to guide the exploration of the kinds of behaviors that emerged from CCS usage. By studying CCS data generated from detailed clickstream data and applicable user interface components of the system, techniques are developed that are used to meaningfully group CCS behavior. In the second study, socio-theoretic models of technology adoption and use, based on the diffusion of innovations, are utilized to understand CCS usage and adoption within the context of a use diffusion model. By applying clustering to build a typology of CCS users, patterns of usage emerge that provide deeper insights into the sophistication of their CCS usage. By triangulating these automatically derived patterns with data that emerged from qualitative survey and observation data of CCS users, a more detailed picture of what teachers were using the CCS for and how they were using it, provided tighter validation of the results of the automatic methods. In the final study, the typology categories of the second study are married with two classes of data, teacher skill and classroom dynamics, to model and predict learner gains through semi-supervised learning techniques. Since subject-area exams were administered by teachers using the CCS, exam score differences were analyzed between two consecutive years of student and teacher data, one in which the CCS was not used, and a model was built using student demographic, class size and teacher experience, that shows that CCS usage becomes a valuable piece of data in predicting average to above average learner gains. Though this research is limited in scope to a small population of users in a specific subject area, in addition to the challenges of new ground being broken with mining educational data sets, it nonetheless contributes to the fast growing field of educational informatics. The methods and research of this dissertation therefore explore the boundaries between the technology use of teachers via their online behaviors, how that behavior translates into a response to the demands of the classroom, and the computational tools that might be used to draw linkages between those behaviors and student performance outcomes.
Maull, Keith E., "Computational Methods for Analyzing and Understanding Online Curriculum Planning Behavior of Teachers" (2013). Computer Science Graduate Theses & Dissertations. 70.