Reading Emotions: Automatic Recognition of Engagement Using Facial Expressions and Physiological Measures

Abstract

Engagement is a state of being involved, occupied, or fully absorbed in something, the advantage of engagement is widely known in the literature. Despite its importance, there are many methods for measuring engagement. This study utilizes both subjective and objective measures to investigate engagement by using facial action units, skin conductance, eye movements, and pupil diameter to understand how these measures are affected by engagement. The results indicated that there were some significant differences in measure when participants were engaged, compared to neutral and not engaged, but these changes were subtle and effect sizes were low. Similarly, multiple machine learning algorithms were failed to classify engagement using facial action units. One interpretation of these findings is that it was not possible to generate enough engagement responses with this experimental structure. On the other hand, another interpretation is that engagement itself in this specific context doesn't generate significant physiological responses.