How Do Students Feel in Online Learning Platforms? How They Tell It: How Does Artificial Intelligence Make a Difference?

Abstract

This study aims to investigate the effectiveness of an artificial intelligence (AI) model in determining students' emotional states during online courses and compares these AI-generated results with traditional self-report methods used in educational sciences. Conducted with 66 students from three different departments of a public university in Eastern Turkey during the 2021-2022 academic year, the study involved capturing facial images of students every 10 minutes during online lectures to analyze their emotional states using a deep learning-based CNN model. In addition, students provided their emotional states through a mood analysis form, which included personal information and subjective feelings such as happiness, sadness, anger, and surprise. The AI model achieved a high accuracy rate of 90.12% in classifying seven different emotional states, demonstrating its potential for real-time emotion recognition in educational settings. However, the study also found a 39% overlap between AI-determined emotional states and self-reported emotions. This finding emphasizes the need for a multifaceted approach to emotion measurement, integrating both advanced AI techniques and traditional self-report tools to more comprehensively understand students' emotional experiences. The results highlight the challenges and opportunities in combining technology with educational assessments and suggest directions for future research in improving emotion detection methodologies and their application in online learning environments.