Emotional Recognition in Affective Tutoring System: A Systematic Review

Affective tutoring system (ATS), as an extension of Intelligent Tutoring Systems, could assess learners' emotions and regulate their learning process for personalized learning. The design and evaluation of the emotional module are key elements in the design of the emotional guidance learning system, and they are of great significance for enhancing its educational effectiveness. The current research conducted a systematic literature review to analyze 78 selected papers of emotion recognition methods within ATS for enhancing ongoing research of affective computing in personalized learning. Findings revealed that current ATS mainly use unimodal facial expression data, specifically facial expression data, mostly applied in classroom teaching environments, and frequently utilize intrusive devices. And, the most commonly used emotional theory model is Paul Ekman's basic emotion model. Bayesian networks and their related models are the most frequently used algorithmic models. Emotional information in ATS is primarily utilized in four categories: personalized teaching and tutoring, emotional motivation and support, personalized learning recommendations, and emotional feedback and guidance. This research has organized the current challenges and proposed future directions for development. This is intended to help researchers better conduct research related to affective tutoring systems.