本研究透過多模態學習分析系統，探討在STEM協作學習環境中實施精準教育的可能性與效果。研究發現，透過結合臉部表情分析、視線追蹤與語音識別等多模態數據，能夠更精確地識別學生的學習需求和行為模式，從而支援教師提供個性化的學習指導和及時的干預。實驗結果顯示，相較於傳統教學模式和單一數據源的學習分析支持，多模態學習分析系統能顯著提升學生的學習動機與知識建構行為，尤其在促進學生深入討論與協商意見上更為明顯。此外，研究亦指出，運用多模態學習分析系統於STEM教育中來追蹤學生的知識建構行為，不僅能提高學習效率和教學品質，更有助於實現精準教育的目標，為學生提供更加適性化的學習經驗。未來研究需進一步探討多模態學習分析在不同學習環境中的應用潛力與挑戰，以促進教育技術與教學實踐的融合與創新。

This study investigates the feasibility and effects of implementing precision education in STEM collaborative learning environments through a multimodal learning analytics system. The findings reveal that by integrating multimodal data sources such as facial expression analysis, eye-tracking, and voice recognition, it is possible to more accurately identify students' learning needs and behavioral patterns. This, in turn, supports teachers in providing personalized instructional guidance and timely interventions. The experimental results show that, compared to traditional teaching models and single data source learning analytics support, the multimodal learning analytics system significantly enhances students' learning motivation and knowledge construction behaviors, especially in facilitating in-depth discussions and consensus-building among students. Furthermore, the study indicates that employing a multimodal learning analytics system in STEM education to track students' knowledge construction behaviors not only improves learning efficiency and teaching quality but also contributes to achieving the goals of precision education, offering students a more personalized learning experience. Future research should further explore the potential and challenges of applying multimodal learning analytics in various learning environments to promote the integration and innovation of educational technology and teaching practices.