A major factor influencing learning is students' emotions and their general affective state. Given the pivotal role that affect plays in learning activities it is not surprising that there has been a good deal of interest in developing affect-aware technologies. The overwhelming majority of this work, however, has focused on modeling affect, i.e., designing computational models capable of inferring how students are feeling while interacting with an Intelligent Tutoring System (ITS). While modeling of affect is a critical first step in providing adaptive support tailored to students' affective needs, very little work exists on systematically exploring the impact of affective interventions on students' performance, learning, affect and attitudes, i.e., how to respond to students' emotions such as frustration, anxiety, boredom, and hopelessness as they arise. The research fills this gap by analyzing the value of tailoring different types of interventions to negative affective states for individual students and groups of students.
The project has two main goals. First, it addresses how to respond to negative student emotion (e.g., frustration, anxiety, boredom) in computer-based learning environments through a variety of interventions. Some correspond to specially-designed digital characters, integrated into the learning environment, which are intended to act like students' learning companions. These agents support students through (a) non-verbal behaviors (e.g., having the characters express empathy in response to student frustration), (b) messages targeting students' cognitive and meta-cognitive skills, as well as motivation and affect. Other interventions involve supporting collaboration between students to mitigate negative emotional states when detected. The impact of these interventions are investigated through a series of eight experiments with a total of 800 students. These experiments help to unveil general prescriptive principles to address student affect.
The second goal accomplished by this research is that the experiments provide valuable data to continue to extend and validate existing models of emotion. Specifically, the project triangulates and integrates a complex space of partially overlapping models and constructs of affect in learning (i.e. emotions, attitudes, incoming moods, motivation, engaged use or misuse of software). The project refines several well-established models, in particular the control-value theory of emotions to provide a more stable theoretical framework for the field of emotions in educational software.
This research is unique and ground breaking, as few researchers have targeted students' emotion in classrooms, gathered fine grained data on emotions during learning, or assessed the impact of specific affective treatments on a moment-to-moment basis. Students using the tutoring systems have already shown statistically significant gains and learning outcomes, as well as increased positive affect and attitudes. The new affective interventions will greatly increase the broad impact of these systems. This research is developing: (1) prescriptive principles about how to respond to student affect; (2) new understanding about the impact of cognitive, affective, and meta-cognitive interventions on emotions and learning; (3) new understanding about individual differences in learning, unveiling the extent to which emotion, cognitive abilities, and gender impact learning; (4) instruction that is sensitive to individual differences; and (5) refined theories of student emotion.
This research is also: (1) increasing participation in mathematics of underrepresented populations (women and minorities) who often avoid STEM careers; (2) creating broad access to web-based technologies that help to engage more students by addressing their affective and social needs; and (3) addressing the one-size-fits-all approach to education, by responding to individual student needs with alternative representations of content and pathways through which material is presented.
