Using Predictive Modeling to Optimize Student Outcomes in Online Courses based on Learner Analytics Patterns of Student Interactions with Course Contents in Blackboard Learn Learning Management System

Concurrent Session 5

Session Materials

Brief Abstract

My research looks at learner analytics on online graduate-level students' interactions with contents in various online courses between 2012 and 2016 in Blackboard Learn, and how using predictive modeling can help determine and optimize what combination of these interactions (that is, posts, hits, access, and time spent on contents—together referred to as PATH (Posts, Access, Time, and Hits) leads to the greatest variance in overall students’ outcomes such as grades, student retention, and persistence in online learning environments.

Presenters

I am Distance Learning Researcher and a K-12 and Adult Education Science Teacher. My research looks at learner analytics on online graduate-level students' interactions with contents in various online courses in Blackboard Learn Course Management System, and how using predictive modeling can help determine and optimize what combination of these interactions (that is, posts, hits, access, and time spent on contents—together referred to as PATH (Posts, Access, Time, and Hits) leads to the greatest variance in overall students’ outcomes such as grades, student retention, and persistence in online learning environments.

Extended Abstract

Exponential increases in the number of learners receiving instruction through online learning, and institutions using various online modalities for instructional delivery, make salient the need for distance learning policies in higher education, as well as K-12 domains, that focus beyond teacher presence (or student/learner-instructor) and learner-learner (student-student) interactions to other forms of interactions; most notably student interactions or engagement with online course contents (that is, learner-content (L-C) interactions).Most distance learning policies do not sufficiently address the impact of learner-content (L-C) interaction in learning environments even though a substantial proportion of students’ interactions are spent on course contents. For example, in New York State, Commissioner’s Regulation Section 50.1 (o) stipulates that for every 45 hours (one semester credit) of learning time,30 hours (70%) should be spent on student work/study out of class, that is, course contents, which includes: “reading course presentations/‘lectures,’ reading other materials, participation in online discussions, doing research, writing papers or other assignments, and completing all other assignments (e.g. projects).”Understanding the impact of L-C Interaction would help decrease perceived distance and increase psychological closeness between learners and instructors and among learners in online learning environments.

To bridge this gap, this study captured various patterns of learner-content interactions (a form of dialogue outlined in Moore (1993) Transactional Distance Theory) in severalgraduate-level onlinecoursesoffered by The School of Professional Development (SPD) between 2013 and 2016. Data/learner analytics for learner-content interactions includedTime Spent on Content, Frequency of Access to Content, access to lecture capture and archived lectures, time spent viewing archived lectures, frequency of written discussions, discussion board and other media (such as web conferencing and collaboration tools) posts, and frequency of Socratic posts (written). Predictive modeling was then used to develop a robust Learner-Content interaction model that optimized specific learner outcomes, such as student assessment scores and final course grades. Policies are proposed for analyzing the implications of learner-content interactions on learner outcomes and perceived distance; and the use of data analytics in online learning environments. Policies are also proposed for ensuring quality assurance mechanisms in this environment.