Using Learning Analytics to Detect Off-Task Reading Behaviors in Class

Abstract

In this paper, we aimed at detecting off-task behaviors of the students by analyzing logs from a digital textbook reader. We analyzed 47 students’ reading logs from a 60-minutes long in-class reading activity. During the preprocess, we extracted each student’s reading patterns as a single vector. Then we used cluster analysis to find the most common reading patterns. Our results indicated that there are two major reading patterns in data. The first pattern is, the students who are following the instructor from the beginning until the end of the lecture. The second pattern is, students who are following the instructor’s pattern until the first 17th minute but not during the rest of the lecture. Based on these patterns we labeled first group as on-task students while the other group as off-task students. We also investigated academic performance of students in these two groups. Obtained results can be used to design data-driven support for in-class teaching. Instructors can plan interventions when off-task behaviors occur while the lecture is in progress.