| Item type |
紀要論文 / Departmental Bulletin Paper(1) |
| 公開日 |
2025-03-31 |
| タイトル |
|
|
タイトル |
学生の振り返りから見る学習成果：生成AIを活用した分析事例 |
|
言語 |
ja |
| タイトル |
|
|
タイトル |
Identifying Learning Outcomes from Student Reflections: A Case Report for Using Generative AI |
|
言語 |
en |
| 言語 |
|
|
言語 |
jpn |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_6501 |
|
資源タイプ |
departmental bulletin paper |
| 著者 |
高橋, 博美
Hiromi TAKAHASHI
関田, 一彦
Kazuhiko, SEKITA
|
| 抄録 |
|
|
内容記述タイプ |
Abstract |
|
内容記述 |
This study was conducted to examine the usefulness of a method that uses generative AI to efficiently conduct qualitative data analysis, and to assess the effect of a reflection-based instruction from the perspective of first-year education. The study employed three generative AIs（ ChatGPT, Claude, and Gemini） to analyze the same text data using the following prompts: “Extract frequently appearing keywords and phrases, identify important concepts that express the significance of this class among the extracted keywords and phrases, and set specific categories around these important concepts.” When the generative AIs were asked to analyze students’ reflections, they extracted and categorized self-evaluations of not only comprehension and retention of course content, but also improvement of learners’ learning skills and attitudes, and enhancement of their collaboration skills. These results are consistent with the objectives of first-year education, suggesting that regular courses can function as first-year education classes. The reflection-based instruction originally took shape in the course development from the perspective of cooperative education. The PDCA cycle of learning throughout the semester is implemented using the beginning, mid-term review, and end of semester reflection sheets, and by sharing each other’s learning with their classmates, a sense of being active participants in learning is fostered. Characteristics were observed in the level of abstraction of categories depending on the AI engine. At the time used in this study, Gemini1.5 was the most straightforward, with a classification organization that used frequently occurring words as they were. On the other hand, the GPT4 and Claude3 categories had no words that evoked class content, indicating an increased level of abstraction. Therefore, it is recommended to use different AI engines depending on the purpose of the analysis, or to have multiple engines perform the same task and use the one best suited for the purpose. |
|
言語 |
en |
| bibliographic_information |
教育学論集
号 77,
p. 191-200,
発行日 2025-02-28
|
| 出版者 |
|
|
出版者 |
創価大学教育学部・教職大学院 |
| item_3_source_id_8 |
|
|
収録物識別子タイプ |
PISSN |
|
収録物識別子 |
03855031 |
| item_3_source_id_10 |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AA1238438X |
| 出版タイプ |
|
|
出版タイプ |
VoR |
|
出版タイプResource |
http://purl.org/coar/version/c_970fb48d4fbd8a85 |
kyoikugakuronsyu0_77_12.pdf (795.4 KB)
.png)
