Using LLM to Predict Motivation in VR Language Learning

As society becomes more globalized, foreign language education has become essential, and some users find it challenging to maintain motivation for learning. We developed a Chinese vocabulary learning application using a VR environment and investigated the relationship between the effectiveness of virtual rewards and learning motivation. The results revealed that differences in motivational factors between experienced and inexperienced language learners. Therefore, to further improve user motivation and usability, it is necessary to provide users with an appropriate rewards and gamification elements suited to their ability. To achieve this, predicting learners’ motivation levels and automatically adjusting rewards and gamification elements can be an effective approach. To provide a more comfortable learning experience for users, this study predicted user motivation levels using an LLM based on trial data from the developed Chinese vocabulary learning application. The results showed that past prediction data accumulated at Level 3, leading to improved prediction accuracy. However, the accuracy rate by level remains low, necessitating the exploration of methods to improve accuracy. For future development, the study aims to evaluate multiple prediction models and implement the most accurate one into the application based on the predicted learning motivation values.