講演情報
[8K-03]Development of Generative Agents as virtual classmates to support truant students in VR learning
*Shazman Aiman Haziq1、She Wan Jou1、桑原 教彰1、Siriaraya Panote1 (1. 京都工芸繊維大学)
発表者区分：学生
論文種別：ショートペーパー
インタラクティブ発表：あり
論文種別：ショートペーパー
インタラクティブ発表：あり
キーワード：
Generative Agents、VR、E-learning、Futokou
The increasing number of truant students, referred to as futokou students, in Japan’s schools has become a growing concern in recent years. While previous studies have examined the use of VR as an educational tool to facilitate home-based learning through online classes, these typically involve interactions between only the teacher and the student, primarily due to concerns surrounding bullying among student peers. This study explores the use of generative AI agents as peer supporters and virtual classmates to provide learning and emotional support for futokou students within a virtual classroom environment. A prototype system was developed using Unity to create the virtual environment and the Convai sdk (which uses the GPT model) to develop the generative agents. A major challenge in designing these agents lies in ensuring that they avoid discussing sensitive topics while fostering a supportive atmosphere for the students. Two experiments were carried out. The first experiment focused on determining how variations in the knowledge levels and learning abilities of the agents affect user satisfaction and the second evaluated whether students feel a sense of social presence and engagement within the virtual classroom populated with generative AI classmates.