METHOD FOR ESTIMATING ENGAGEMENT IN DIGITAL EDUCATIONAL GAMES
Analyzing student engagement during learning is an essential factor for achieving pedagogi cal objectives. Student engagement in a traditional classroom environment can be observed by their body language and facial expressions. However, automatically assessing student engagement in Digital Educational Games (EDGs) has been a challenge. The objective of this research is to propose a method to evaluate student engagement in EDDs, which combines objective measures (automatically detected facial and behavioral traits) with subjective measures (personal report). A literature review raised the main concepts, terms and definitions related to engagement and identified computer vision and machine learning techniques commonly used in studies of this nature. Two experiments were carried out with JEDs to validate the method and verify the relationship between student engagement (presence of "positive" emotions and/or attention) and the games. The first experiment was carried out in an “RPG” type game and considered facial expressions. The second experiment was carried out in a “Battle Royale” game and added to the engagement estimation, the head pose and gaze direction. In both experiments, the captured data were classified and combined with the students’ personal reports to try to predict engage ment. Our findings suggest that determining the level of student engagement during a digital educational game is still a challenge due to the subtlety and diversity of visible and non-visible characteristics that may be related to this affective state, requiring the integration of many elements and data and , to some extent, a certain degree of experience, subjectivity and abstraction of the "evaluator". We believe that studies in this direction can support game designers to design more engaging games and encourage teachers to use educational digital games in their pedagogical practices.