| C. Troussas, A. Krouska, Ph. Mylonas, C. Sgouropoulou, I. Voyiatzis |
| EmotAdapt: Emotion-Aware Personalized Adaptive Learning Through Fuzzy Learner Profiling and LLM-Generated Content |
| 11th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM 2026), Thessaloniki, Greece, September 18-20, 2026 |
|
ABSTRACT
|
| Existing Personalized Adaptive Learning (PAL) systems strive to personalize learning to each student, however most of them ignore the importance of emotions. Educational psychology suggests that emotions such as confusion, frustration and interest are factors that can affect learning. So far, few systems use affective signals to adapt in real-time. In this paper, we propose EmotAdapt, a full framework of emotion-aware PAL that includes multimodal emotion detection, fuzzy learner modeling, Bayesian knowledge tracing, and LLM-based adaptation. EmotAdapt monitors emotions in real time through analyzing facial expression and text sentiment, and then fuse these and cognitive information with learning style classification through fuzzy reasoning. This comprehensive learner model is then used to adaptively generate explanations, hints, and practice problems from a fine-tuned LLM. We performed an experiment on 312 college students learning through four Computer Science courses throughout a 12-week semester. The proposed system EmotAdapt attained a normalized learning gain of 0.349, which is better than a non-emotional adaptive system (0.300), non-fuzzy adaptive system (0.282), and static LMS (0.221). It's also especially powerful to low performers (0.412). Learner engagement increases consistently throughout the semester in EmotAdapt, whereas in all baselines learning engagement tends to decrease. Students' frustration decreases from 18% to 8%, whereas engagement state increases from 31% to 48%, following adaptation in EmotAdapt.
|
| 18 September, 2026 |
| C. Troussas, A. Krouska, Ph. Mylonas, C. Sgouropoulou, I. Voyiatzis, "EmotAdapt: Emotion-Aware Personalized Adaptive Learning Through Fuzzy Learner Profiling and LLM-Generated Content", 11th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM 2026), Thessaloniki, Greece, September 18-20, 2026 |
[
BibTex] [
Print] [
 Back] |
