「光」單元之選擇題與建構反應試題之線上測驗研發

Online Assessment with multiple choice and constructed response items for the “Light” unit

莊峰魁(CHUANG FENG KUEI)

指導教授 ： 郭伯臣

亞洲大學/資訊學院/資訊工程學系碩士在職專班/碩士(2012年)

摘要

I 中文摘要 傳統紙筆測驗或電腦化測驗，往往受到侷限而無法給予教師更多的資訊來幫 助學生學習：選擇題型，只能針對學生最後的答案，予以計分，無法確認學生是 猜對的或是真正學到正確的概念；開放題型，閱卷曠日費時，且計分不夠客觀。 因此，本研究欲設計理化科之電腦化建構反應題，詳細紀錄學生的解題歷程，並 根據解題過程之規則，對學生的錯誤類型進行自動化診斷與分類，並自動計分， 使學生測驗完畢，可以即時獲得回饋。 本研究依據文獻教學經驗及教材特性，設計適合的建構反應試題後，成功地 建立自動化分析模型及建置診斷系統，透過診斷系統可詳細記錄學生解題過程， 並藉由電腦診斷出學生的錯誤類型，教師閱卷無負擔，學生也可獲得立即的學習 回饋，並使教師獲得學生的迷思概念及學習狀況等完整資訊。 本研究之研究結論分述如下： 一、由學生在本單元之建構反應題作答反應所獲得的錯誤類型資訊，比歷來 研究所獲得之錯誤類型更詳盡、多元。且電腦判別之錯誤類型平均正確 率為100％，顯示該研究所建置之建構反應題之自然科診斷測驗成效良 好。 二、選擇題之錯誤類型最佳辨識率平均值為90.22 ％，子技能最佳辨識率平 均值為98.91％；含建構反應試題之錯誤類型最佳辨識率平均值為 92.09％，子技能最佳辨識率平均值為95.96％。 三、含建構反應題之適性省題率為19.86％，比僅選擇題的省題率稍高，可 見適性測驗亦可發展建構反應題。

關鍵字

電腦適性測驗 ； 建構反應題 ； 光

並列摘要

Traditional writing tests or computerized tests usually fail to provide teachers
enough information to help students learn: multiple choices tests, which are graded
by students’ final answers, cannot confirm whether students have learned what they
are supposed to; open-ended question, which take much time to grade, are not
objective. Therefore, the research is to design computerized constructive response
tests for science, which record the students’ detailed problem solving process and
automatically grade by diagnosing and classifying students’ bugs, so that students can
get real time feedback.
The research successfully builds automatized analyzing models and construct
diagnosing systems after designing appropriate constructive response items.
Students’ bugs are diagnosed by computer systems, so students can get learning
feedback immediately and teachers get the complete information about students’
thinking mistake conception without spending much time grading.
The conclusion of the research is as follows:
1.The information about the bugs acquired by doing the constructive response
items is more detailed and diversified than the ones done before. The average
correctness rate of the mistake type judged by the computer system is 100%, which
shows that the nature diagnosed tests constructed according the research is valid.
2.The average optimal bug identifying rate of multiple choice tests is
90.22%,while the sub-skill average optimal mistake type identifying rate of that is
93.91%; the average optimal mistake type identifying rate of constructive response
items is 92.09 %,while the sub-skill average optimal bug identifying rate of that is
95.96%.
3.The question-economic rate of the constructive response items is 19.86
%,which is a little higher than that of multiple choice tests.
Adaptive tests are can be developed into constructive response items.

並列關鍵字

Computerized Adaptive Testing ； Constructed-response items ； light

參考文獻

王晉基、郭重吉（1992）。利用選擇題的方式來探求國中學生對「光」的迷思概

吳俊毅（2002）。整合式主動推播學習平台模型—以課程知識為基礎。南華大學

技術及職業教育研究所碩士論文，未出版，屏東縣。

陳忠志、許有亮（1998）：國中生平面鏡成像的另有概念之探討。物理教育, 2(1),

陳均依（2004）。光反射與折射的另有概念診斷工具之發展與研究。科學教育學