AN EXPLORATORY STUDY OF FACTORS INDICATIVE OF AFFECTIVE STATES OF STUDENTS USING SQL-TUTOR

Abstract

The goal of this study was to model the affective states exhibited by students using SQL-Tutor. Based on current literature, we selected academic affective states of interest and measured their incidence among students during an SQL-Tutor session. We observed that students using SQL-Tutor most often exhibited engaged concentration, confusion and boredom; however, none of these states were correlated with student achievement in the final exam. Using D’Mello’s Likelihood metric, L, we found that boredom and frustration tended to persist. We then correlated features extracted from SQL-Tutor log files with these two states’ L values. We found that boredom was negatively correlated with the number of completed/attempted problems, and the number of constraints used. It was positively correlated with the average time needed to complete problems and the average number of attempts. Persistent boredom was negatively correlated with the number of solved problems and positively correlated with the mean time to solve problems and the average number of attempts per solved problem. Frustration was not significantly correlated with any of the factors, but persistent frustration was negatively correlated with the number of constraints used and positively correlated with the average number of attempts per solved problem.