Automatic Method to Identify E-Learner Emotions Using Behavioral Cues

In this article, we offer and test a nonsurvey-based method to characterize learner emotions. Our method, instead of using surveys, uses logs of learner behaviors in learning management systems (LMS) to reason about the emotional state of the e-learner. We use the control value theory (CVT) as the theoretical base of measuring emotions. Using this theory, learner emotions are directly tied to their achievements. We develop two fuzzy inference systems, one to measure academic self-efficacy (ASE), that we call ASEMEL, and another to measure task value, TAVAMEL. These two factors, according to the CVT, can identify the prospective outcome emotions in a learning environment. We conducted our experiment in an LMS with a sample of 30 students and validated the performance of our nonsurvey-based systems by comparing the results with the measures of an equivalent survey-based method. Finally, by linking our ASEMEL and TAVAMEL results, our system anticipated “hopelessness,” “anticipated relief” and “no emotion” with 97% accuracy, “hope/anxiety” with 77% accuracy, and “anticipatory joy” with 87% accuracy compared with the self-reports of the students.