大型語言模型驅動之智慧學習助教對網路科學探究之成效評估;Evaluation of the Effect of Intelligent Learning Assistant based on Large Language Models in Online Scientific Inquiry Contexts

NCU Institutional Repository > 資訊電機學院 > 資訊工程研究所 > 博碩士論文 > Item 987654321/95618

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/95618

题名:	大型語言模型驅動之智慧學習助教對網路科學探究之成效評估;Evaluation of the Effect of Intelligent Learning Assistant based on Large Language Models in Online Scientific Inquiry Contexts
作者:	許志仲;Hsu, Chih-Chung
贡献者:	資訊工程學系
关键词:	科學探究;電腦模擬;智慧代理人;大型語言模型;Scientific inquiry;Computer simulation;Intelligent agent;Large Language Model
日期:	2024-07-26
上传时间:	2024-10-09 17:06:14 (UTC+8)
出版者:	國立中央大學
摘要:	科學探究被認為是一種促進學生主動參與的學習方式，被認為是科學教育的核心方法，透過在探究學習中加入可視化與互動性的電腦模擬讓學生操作並進行知識建構，達到與傳統實體實驗相同的效果，並將其架設於網路環境以克服教學環境的限制。然而在探究教學中，教師需要花費心力為個別學生提供適當的引導以幫助學生，基於大型語言模型驅動的智慧代理人即是可能一個解決方案，透過提示詞規範的智慧代理人可以為學生提供即時且客製化的教學引導，在線上探究教學環境中輔助學習。本研究以大型語言模型GPT-3.5-turbo為基礎，透過提示詞設計用於科學探究教學引導的智慧助教，以語音以及文字訊息的方式溝通，在結合電腦模擬的線上科學教學平台中根據學生的個別學習狀況提供探究引導。本研究進行為期半天的科學探究學習活動，以臺灣宜蘭縣某國民中學的86位八年級學生為實驗對象，實驗組的44名學生在活動過程中由智慧助教引導進行探究學習，控制組的42名學生則使用傳統學習講義進行自主探究。本研究蒐集活動中之探究問題回答、科學概念測驗表現、自我效能以及探究行為與對話，以進一步分析學生的探究學習單答題狀況、概念測驗表現、自我效能以及實驗組學生與智慧助教的互動行為。結果顯示，實驗組學生在活動中的探究學習單回答表現以及活動後的概念測驗得分皆顯著高於控制組，顯示智慧助教的介入對於學生於探究活動學習成效的提升。在操作行為方面，藉由智慧助教的互動引導使實驗組的學生更加投入於探究學習活動中，擁有比控制組更多的系統操作，顯示實驗組學生對於學習動機的提升。另在自我效能方面，智慧助教的引導降低了實驗組學生遇到困難時的挫折感，呈現了較高的自我效能。此外，本研究對智慧助教與實驗組學生的互動行為進行分類以及統計，呈現智慧助教於探究學習活動時對學生進行引導的方式。並根據研究結果發現智慧助教之成效以及限制，提出未來應用大型語言模型於教育領域的發展與建議。 ;Scientific inquiry is recognized as a core method in science education that promotes active student engagement. Integrating visual and interactive computer simulations into inquiry learning allows students to manipulate and construct knowledge, achieving effects similar to traditional physical experiments, and can be implemented in online environments to overcome teaching limitations. However, in inquiry-based teaching, teachers need to invest considerable effort to provide appropriate guidance to individual students. Intelligent agents driven by large language models present a potential solution, offering real-time and personalized instructional guidance to support online inquiry-based learning. This study utilizes the GPT-3.5-turbo language model to design an intelligent learning assistant for guiding scientific inquiry leraning. The assistant communicates through both voice and text messages, providing tailored guidance based on individual student learning conditions within an online science learning platform that incorporates computer simulations. This study conducted a half-day scientific learning activity involving 86 eighth-grade students from a junior high school in Yilan County, Taiwan. The experimental group, consisting of 44 students, was guided by an intelligent learning assistant during the inquiry learning process, while the control group of 42 students used traditional learning handouts for self-directed inquiry. The study collected data on inquiry learning question responses, science concept test performance, self-efficacy, and inquiry behaviors and dialogues. These data were analyzed to examine the students′ inquiry learning question responses, concept test performance, self-efficacy, and the interaction behaviors between the experimental group students and the intelligent learning assistant. The results showed that students in the experimental group exhibited significantly better performance in their inquiry learning questions and higher scores on the concept test after the activity compared to the control group. This indicates that the intervention of the intelligent learning assistant was effective in enhancing students’ learning outcomes during the inquiry learning activity. In terms of engagement, the interaction and guidance provided by the intelligent learning assistant led to greater system usage and involvement in the inquiry learning activity among the experimental group students compared to the control group, suggesting an increase in their learning motivation. Furthermore, the intelligent learning assistant’s guidance reduced feelings of frustration in the experimental group, reflecting higher self-efficacy. Additionally, this study classified and analyzed the interaction behaviors between the intelligent learning assistant and the experimental group students, revealing the methods of guidance used during the inquiry learning activity. Based on the findings, the study discusses the effectiveness and limitations of the intelligent learning assistant and offers recommendations for future applications of large language models in the field of education.
显示于类别:	[資訊工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	40	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....