Adapting Educational Content to Student Emotions: A User Study With the Furhat Social Robot

Abstract

Emotional artificial intelligence (AI) can enhance personalization in computational social systems by enabling machines to respond to users’ affective states. In education, adapting instruction based on emotions such as engagement and stress may lead to more supportive learning experiences. This study examines whether an emotionally adaptive social robot, driven by real-time electroencephalography (EEG) signals, can improve learners’ engagement and perception of the robot. A between-subjects experiment is conducted with 30 undergraduate students during an XML lecture delivered by the Furhat social robot in one-to-one modality. In the empathetic condition, the robot adapts its behavior in response to EEG-detected engagement and stress levels; in the nonempathetic condition, the robot follows a fixed script. The empathetic robot significantly improved users’ perceptions across all Godspeed dimensions and enhanced perceived usability and reward. EEG data showed a significant increase in engagement. Learning outcomes were statistically similar across conditions. Real-time EEG-guided emotional adaptation improves user engagement and perception of social robots in educational contexts. These findings support the role of affect-driven adaptivity in enhancing user experience in intelligent tutoring systems.