運用視覺化學習分析瞭解程序性技能學習行為：以大學統計實習課程為例

Tracing Students’ Learning Behaviors in Statistical Practice Sessions: What Do Visualizations of Learning Logs Tell Us?

對大班統計實習課程而言，觀察學習者行為非常重要，也具有高度挑戰性。為了更全面地瞭解課堂動態，本研究在大學基礎統計學實習課程中，建置採用線上課程平臺來記錄學習者觀看與註記等學習行為資料，並以視覺化方式對應課堂環境、教學內容與學習者行為進行時間序列分析。總共蒐集並分析58名大學生，在一堂統計實習課中的7,869筆行為數據。研究結果顯示，高表現與低表現學習者有不同的課程講義參照行為，高表現學習者在實習課堂和測驗中都展現更具策略性的學習行為，他們較為積極地跟隨教學者的講授，開書測驗中有計劃地遍閱講義，檢索與判斷相關資源的表現也都比較好；低表現者則不會逐頁跟隨教師的講解，測驗中也不太會參照講義，依賴聽課時的印象答題。研究結果支持觀看講義頁面的次序與軌跡等學習者行為的質性指標，可以對照測驗分數等量化指標，作為有效的另類評量方法。教師由視覺化分析結果能更快地捕捉全班學生的學習和測驗行為概況。本研究結果支持學習分析的實施應基於學習歷程檔案評量典範，可視化學習者行為作為評量結果，能夠導引教學者適時提供學習資源與支援，增進教學互動。本研究對相關課程教學與評量應用具有參考性，並對未來研究提出重要的理論概念與方法等建議。

Observing the learning behaviors of large groups of students during hands-on statistical learning activities is challenging for instructors and calls for radical new measures. This study adopted an online learning platform that recorded student behaviors as learning logs while they completed drills and practices. A visualization approach was used to analyze the time-series behavioral data with direct and detailed measurements in a real-time classroom setting. A total of 7,869 learning logs from 58 college students in a statistical practice session were collected and visualized to map student behaviors to the classroom environment, instruction, and phases of the course. The key findings suggested that students with higher test scores exhibited better strategic learning behaviors by referencing course materials frequently during class and tests. High performers actively followed lectures and instructors and strategically located and used appropriate learning resources during tests. In contrast, low performers were reluctant to use lectures and handouts referentially but relied more on the instructor’s in-person guidance. The results supported the idea that learning logs tracking student movements, trajectories, and annotations could complement the assessment of quantitative test scores. Visualized analytics allowed instructors to understand students’ learning and test-taking strategies at a group level, and visualizing student behaviors as assessment results could guide instructors in providing timely learning resources and support to enhance teaching interactions. This study filled an important research gap by connecting the abundant behavioral information to its practical applications in statistics teaching and learning by bridging the divide between the vast quantities of educational data available and its meaningful use in instructional settings. The findings also provided empirical support for the feasibility and importance of implementing learning analytics based on the paradigm of portfolio assessment, with specific implications for teaching statistics and related practice courses.