目前已有研究預測學生的學習成效，並利用可解釋人工智慧（Explainable AI, XAI）提供解釋。然而，可解釋人工智慧所提供的僅是特徵對結果的影響程度，並非直接給出因果解釋。這對非專業人士來說可能不易理解特徵與結果之間的關聯。過去的研究表明，學習回饋對學生有幫助，但為每位學生量身定制回饋成本較高，此時，生成式人工智慧（Generative AI, GenAI）提供了解決方案。本研究旨在運用生成式人工智慧為所有學生提供個人化的學習回饋建議。研究使用LBLS-516資料集訓練預測模型，並使用可解釋人工智慧說明特徵的影響程度，接著利用生成式人工智慧提供進一步解釋與改進建議。為了評估生成式人工智慧與可解釋人工智慧生成的解釋對於學生的有效性，本研究使用系統因果性量表(System Causability Scale, SCS）進行評量。結果顯示，生成式人工智慧更能展示學生學習行為策略與成效之間的因果關係，透過生成式人工智慧生成的改善建議確實能提升學生的學習成效、協助學生改善自我調解與學習程式的能力。藉由生成式人工智慧的幫助，除了能讓學生更了解其學習行為、策略與學習成效的因果關係，還能大量生成解釋，對於教師或學習分析團隊有著重要的幫助。;Currently, studies predict student learning outcomes and utilize explainable artificial intelligence (XAI) for explanations. However, XAI only provides the extent of features′ influence on results, rather than directly offering causal explanations. Non-experts may find it challenging to grasp the correlation between features and outcomes. Past research indicates that learning feedback benefits students, yet customizing feedback for each student is expensive. At present, generative AI (GenAI) offer a solution. This study aims to use GenAI to provide personalized learning feedback suggestions for all students. The study utilizes the LBLS-516 dataset to train the prediction model, employs XAI to demonstrate feature influence, and subsequently utilizes GenAI to provide additional explanations and improvement recommendations. To evaluate whether GenAI can enhance the explanatory power of XAI, this study uses the System Causability Scale (SCS). Findings indicate that GenAI can more effectively illustrate the causal relationship between students′ learning behavior strategies and their effectiveness. With GenAI assistance, students can better comprehend the causal links between their learning behaviors, strategies, and learning effectiveness. Additionally, GenAI can produce numerous explanations, providing valuable support to teachers or learning analysis teams.
