Virtual reality (VR) technology provides a high sense of immersion and presence to users and can enhance the engagement and performance of learning. However, the VR learning environment introduces more complex audio–visual stimuli than the traditional multimedia learning environment. These excessive stimuli cause negative effects such as distraction and cognitive overload. To minimize these negative impacts and improve the learning environment, we must evaluate learners’ cognitive states under the VR environment. Cognitive states can be evaluated subjectively (e.g., through questionnaires) or objectively (e.g., using biometric signals). Subjective and objective methods must be used simultaneously, and correlations between them must be analyzed for quantifying objective measures. The accurate detection of cognitive states is challenging for traditional statistical analysis methods, necessitating the exploration of artificial intelligence (AI) techniques that can classify cognitive states. This study develops a VR learning experience evaluation system based on eye-tracking data. Cognitive states during VR learning are classified as cognitive overload, immersion, and distraction. Correlations between each cognitive state and eye-tracking metrics are evaluated, and the possibility of cognitive-state quantification is discussed. An LSTM-based model developed in this study classified cognitive states from eye-tracking data with moderate accuracy (75.60%) under a subject-independent validation setting.
Choi et al. (Tue,) studied this question.