Emotion-Adaptive Learning Platform

Abstract

E-learning platforms often neglect learner disengagement, and current adaptive systems that infer emotions raise privacy, ethical, and legal concerns, especially under the EU AI Act. This thesis presents an emotionally adaptive platform that pivots from automated emotion inference to a learner-controlled self-report model grounded in self-regulated learning and achievement emotions frameworks. Developed as a modular monolith using Java Spring Boot and React/TypeScript, the system enables students to self-report via an "Emotion Bar" and receive rule-based interventions (e.g., AI summaries for confusion, humorous rewrites for boredom), while teachers can create lessons via PDF parsing, generate quizzes, and communicate with students. This research demonstrates that transparent, human-centered adaptive systems can enhance learner agency, provide meaningful support, and ensure data privacy and legal compliance.