Beyond the basic emotions: what should affective computing compute?

One of the primary goals of Affective Computing (AC) is to develop computer interfaces that automatically detect and respond to users' emotions. Despite significant progress, "basic emotions" (e.g., anger, disgust, sadness) have been emphasized in AC at the expense of other non-basic emotions. The present paper questions this emphasis by analyzing data from five studies that systematically tracked both basic and non-basic emotions. The results indicate that engagement, boredom, confusion, and frustration (all non-basic emotions) occurred at five times the rate of basic emotions after generalizing across tasks, interfaces, and methodologies. Implications of these findings for AC are discussed