Correlation between Teaching Behavior and Real-Time Student Evaluations

Many universities conduct student evaluations, with the goal of encouraging improvement in teaching. Evaluations are subjective assessments by students, and do not directly evaluate specific teaching methods or behaviors as being good or bad. Instructors therefore cannot necessarily easily relate evaluations to areas for teaching improvement. To address this issue, we proposed a teaching behavior estimation model that can estimate teaching behaviors from student evaluations of each lesson. In this study, with the goal of identifying correspondences between lesson progress and teaching behaviors, we examine a model for estimating teaching behaviors from real-time student evaluations (evaluations performed during the lesson). In a basic study toward this end, we acquired student evaluations at three points -- 30, 60, and 90 minutes into university lectures -- and investigated correlations between evaluation changes as the lesson progressed and the number of occurrences of teaching behaviors during each part of the lesson. As a result, we identified relations between teaching behaviors that affect student evaluation items as lessons progress.