In Japan, there are currently over 1,000 higher education institutions, encompassing both universities and junior colleges. Private universities alone comprise approximately four times the number of national and public universities, giving rise to an oversupply of higher education institutions. Moreover, Japan’s declining birthrate has resulted in nearly 60% of private universities failing to meet their enrollment quotas. Many students who enter these institutions do so despite not having selected them as their first choice, which often diminishes their motivation to study. Consequently, a significant proportion of these students are unable to fulfill the academic requirements for graduation, ultimately leading to elevated dropout rates.

To address these issues, many universities have introduced tutoring programs for advanced students. Usually, upperclassmen or graduate students with excellent grades are appointed as tutors. However, it is difficult to find enough tutors with the necessary qualifications. For this reason, in some universities, faculty members often take on the role of tutor or mentor. However, for faculty members who already have many duties to perform, taking on these additional responsibilities can be a significant burden. One possible means of alleviating this burden is to detect abnormal student behavior based on semester-by-semester credit acquisition and learning history, and to provide appropriate intervention through tutoring. Using such an approach can enhance the effectiveness of tutoring and provide adequate care to students at an earlier stage.

In this study, we designed and built a tutoring assistance framework to assist teachers in tutoring and its implementation. A detective function utilizing generative AI evaluates several indicators, including the student's learning history, cumulative credits, and potential withdrawal risk. If an increased risk of dropping out is detected, the framework automatically notifies the teacher, enabling early remedial instruction for at-risk students. This active approach is expected to address problems at an early stage and reduce the dropout rate. Note that the framework does not directly intervene with students, but aims to provide teachers who also serve as tutors with the insights necessary for effective classroom management.

Keywords: Generative AI, educational technology, tutoring assistance, teacher support.