Investigating the Effect of Personal Emotional Score Display on Classroom Learning

In this work, we present the preliminary results from an exploratory experiment in which we studied the influence of exposing students to a personal emotional score on their classroom learning. This score derived from deep learning algorithms is presented to students on a computer screen while they are learning in a classroom. With the advent of deep learning models for the classification of emotions, it is now possible to gauge emotions in real-time. Most of these algorithms depend on facial expressions to identify human emotional and attentional levels. Such algorithms are even being integrated within video conferencing platforms such as Zoom, Teams, etc. to quantify if the participants in the meeting are attentive. The assumption here is that our facial expressions reflect our emotions. The hypotheses tested by this study is a) if such algorithms accurately reflect human emotions, and b) if the very act of showing such a metric to learners (students in a classroom) will influence their cognitive learning capabilities. We conducted the experiment in a freshman-year undergraduate class where students were asked to compile a bibliography, read the research papers and annotate them, and then write a summary. We used a state-of-the-art deep neural network trained on thousands of face images in multiple orientations to detect seven basic human emotions (neutral, happiness, sadness, anger, surprise, disgust, and fear) from the students’ facial expressions. Students’ facial images were captured in real-time through their laptop’s camera. The experimental group of students were presented with their emotional scores and labels while they were working in the class. The control group of students were not exposed to this information. After each class, we asked all students to do a self-assessment of scores and asked for the instructor’s assessment of the students’ work as well. We investigated if the self-perception of their own emotional states helped students to perform better. The key outcomes of this study are threefold. First, the AI models have not reached a stage yet where they could accurately reflect human emotions using facial expressions. Second, the study shows that the self-perception of emotions provide a boost to the perception of productivity in students. Finally, this study also reveals that a higher self-perception of productivity due to being exposed to the emotion recognition AI model do not necessarily contribute to actual higher productivity.