Citation Context

... facial expressions andsspontaneous facial expressions, achieving accuracy ofs90.1% and 79.9%, respectively, when discriminatingsbetween instances of the AU present vs. absent [29].sGrafsgaard et al. =-=[20]-=- used CERT to recognize the level ofsfrustration (self-reported on a Likert scale) in a learningssession and achieved modest results (R 2s= .24).sAdditionally, they achieved good agreement between the...

Citation Context

...when the student stopped working on the task in order to make a comment, which is consistent with loss of focus on the task. Mouth dimpling (AU14) was also involved in negatively predictive features in both models. This facial action unit has been associated with frustration and mental effort in prior analyses [8, 14]. Open questions from the tutor (QOTUTOR) with no student brow lowering (NoAU4) were positively predictive of engagement in the model with the combined bimodal feature set (Bimodal Union). Brow lowering has been previously associated with confusion, frustration, and mental effort [5, 8, 9, 14]. Thus, this may highlight moments when the student was not perplexed by a tutor question. In the model with the trimodal feature set, student informational questions (QI-STUDENT) with outer brow raising (AU2) and running the program were positively predictive of engagement. This may indicate some interest in how the program operates, with the student actively engaging in discussion with the tutor. A sequence of tutoring events related to engagement is shown in Figure 4. The left image shows the student coding the program. The middle shows when the student had stopped coding and read a yes/no ...

Citation Context

...s of nonverbal behavior in tutoring relied on manualsobservations of affect and nonverbal behavior [2, 8, 14, 38].sMore recently, automated techniques have been leveraged tostrack nonverbal behaviors =-=[1, 11, 15, 17, 22]-=-. Most studies havesexamined individual modalities in detail, such as facialsexpression [17, 35], posture [10, 18], or gesture [18, 28].sHowever, a much smaller set of studies has examined multiplesmo...

Citation Context

...eport of how insecure, discouraged, irritated, or annoyed she felt during the session [18]. Interestingly, no statistically significant predictive model could be built for Frustration Level in students scoring below the median in Extraversion. The predictive model of frustration for extraverts was composed of four features. A higher likelihood that a movement backward (MoveBack) was followed by an inner brow raise (AU1) or mouth dimpling (AU14) was correlated with lower frustration in the student. Mouth dimpling (AU14) has been correlated with learning gain in a prior study on the same corpus [15]. A mouth dimpling (AU14) following a backward movement (MoveBack) suggests a confident student, relaxing in her seat while working through the task. Extraverts tend to express more energy in postural movement [2], so an exaggerated motion such as this may be more likely to be observed with an extravert. Similarly, if a student was more likely to mouth dimple (AU14) after eyelid tightening (AU7), the less likely that she was frustrated. Interestingly, the final feature, a hand-to-face gesture (OneHTF) following a successful compilation (CompileSuccess), was negatively correlated with frustrati...