Title: Enhancing Science Education Assessment Through Vision Language Model Integration A Mixed-Methods Study on Automated Essay Evaluation
Authors: Irfan Ananda Ismail, Mawardi Mawardi, Lufri Lufri, Usmeldi, Festiyed
Volume: 8
Issue: 11
Pages: 28-33
Publication Date: 2024/11/28
Abstract:
This study explores the potential of Vision Language Model (VLM) technology to support massive and objective assessment of student work in secondary education. An Explanatory Sequential mixed methods approach (Creswell & Creswell, 2018) was employed, combining quantitative data analysis to test VLM effectiveness with qualitative analysis of teachers' perceptions. The research involved five key stages: 1) collecting data on teachers' essay assessment challenges, 2) developing a VLM-based system, 3) testing effectiveness through comparison with manual assessment, 4) gathering teacher feedback through questionnaires and interviews, and 5) integrating data analysis. The findings demonstrate that VLM integration successfully enhanced both learning quality and assessment efficiency. The system provided consistent, objective evaluation while significantly reducing teachers' administrative workload. Results indicated high satisfaction levels among teachers regarding assessment objectivity, accuracy, and time efficiency. This educational innovation effectively integrates cutting-edge technology with established teaching principles, offering a practical solution for large-scale student assessment challenges. The study contributes to the growing body of research on automated assessment tools in education, while providing valuable insights into their practical implementation and impact on teaching practices.