National Academy for Educational Research

　　本研究是一兩年期計畫，以評量鐵三角為評量設計之理論，數學科四年級分數為例，結合認知診斷理論編製認知診斷測驗，開發能偵測錯誤類型之無參數認知診斷模式，建置以認知診斷模式為基礎之電腦化認知診斷評量與補救教學系統。本研究主要貢獻是以認知診斷模式自動偵測學生之學習落後點；針對每位學生之評量結果提供個別化補救教學；運用科技輔助評量與補救教學以減輕教師負擔。第一年的研究目的是1) 設計數學科認知診斷測驗;2)開發診斷錯誤類型之無參數認知診斷模式;3)以模擬研究與實證資料探討認知診斷模式之成效。第二年的研究目的是1)開發以認知診斷模式為基礎之電腦化認知診斷與補救教學系統;2)開發以認知診斷模式為基礎之補救教學教材；3)實地實驗評估無參數認知診斷與補救教學系統。研究成果包含編製完成能診斷認知屬性之分數認知診斷測驗，並收集完成593名學生資料，分數認知診斷測驗的信度是0.88，效度是以專家為校標，計算一致性，模式估計的八個學生認知屬性與專家判讀的認知屬性一致性皆在0.90以上，顯示良好的效度；模擬資料和實證資料皆顯示以參數型的認知診斷模式具有較佳的診斷正確率，故在建置以認知診斷模式為基礎之電腦化認知診斷評量與補救教學系統時是採用參數型認知診斷模式。本研究設計八個認知屬性之補救教學動畫，經由實地實驗，樣本人數是83人，結果顯示實驗組和控制組有顯著差異，實驗組的平均分數是84分，控制組的平均分數是80分，但其效果量並不高，可能是觀看補救教學的時間只有一節課，部分學生沒有看完全部的教學影片。多數的學生均同意這種補救教學的模式能幫助學生學習數學。

　　This study, which is proposed as a two-year program, develop a computerized cognitive diagnostic assessment and remedial instruction system based on the cognitive diagnostic model. The main contribution of this study is to use cognitive diagnostic model to automatically detect students’ learning weaknesses, prescribe customized remedial instructions accordingly, and reduce teachers’ workload with the assistance of technology to conduct assessment and remedial instruction. The objectives for the first year of this study are 1) to design cognitive diagnostic assessment for mathematics 2) to develop the non-parametric cognitive diagnostic assessment capable of error pattern diagnosis 3) to determine the effectiveness of the cognitive diagnostic mode with simulation studies and evidence, and for the second year 1) to develop the computerized cognitive diagnostic assessment and remedial instruction system based on the cognitive diagnostic model 2) to construct remedial instruction materials based on the cognitive diagnostic model 3) to evaluate the cognitive diagnostic assessment and remedial instruction system through field experiment. The results of the study include a fully developed cognitive diagnostic assessment for fraction problems that is capable of cognitive attribute diagnosis, and 593 student samples collected that translate to a reliability of 0.88 and validities based on the consistency measured between the assessment result and expert judgment on eight cognitive attributes, which all exceed 0.90. The simulations and evidence have shown that parametric cognitive diagnostic models have higher correctness rate of diagnosis, so parametric cognitive diagnostic model is adopted during the construction of the computerized cognitive diagnostic assessment and remedial instruction system. The field experiment, which has a sample of 83 students, then shows significant difference between experimental group and control group that score 84 and 80 respectively on average, and yet the effect size is not large which may results from the limited time for the remedial instruction where some students cannot finish all of the videos. But, in the end, most of the students agree that this mode of remedial instruction can help when learning mathematics.