MACHINE LEARNING CLASSIFICATION OF HEMODYNAMICS TO PREDICT SCIENCE STUDENT LEARNING OUTCOMES IN REAL-TIME DURING VIRTUAL REALITY EAND ONLINE LEARNING SESSIONS

Abstract

Students' learning results in science content and practices are expected to be improved through automated interactive learning management systems and linked online video-based learning environments. The goal of this study is to see how hemodynamic response data may be used to build student-level answer predictions using machine learning algorithms in a science classroom while students are using an online learning management system. A charter school in the northeastern United States was used to recruit 40 participants (n=40), 21 females and 19 males. Students viewed a recorded film that included a 20-minute instruction and explanation of the DNA replication process. A female educator on a computer screen presented an overview of the DNA replication process during class. The findings illustrate those hemodynamic responses seen during topic presentations accurately predict student replies to subject-related questions. The results imply that hemodynamic response can be used to gauge degrees of student involvement in video-based tasks, with error rates in the predictive models below 30%. This could lead to the development of unique visual media assessment methodologies, allowing educators to assess whether students can comprehend the material.