| 摘要: | 科學探究是一種以問題為導向的學習方法，但教師的指導對於科學探究學習的有效性至關重要，因此本研究將結合線上科學探究平台與基於生成式人工智慧的對話機器人，建立可以引導學生高層次思維的教育對話機器人──智慧助教，擔任教師的角色，引導學生進行科學探究，探討智慧助教的協助下對學生答題表現、學習成效、探究行為與自我效能的影響。
本研究招募台灣北部的高中生共62人，在CoSci線上科學探究平台中進行探究活動，主題為高中物理的質心與動量單元，活動中包含物理模擬與多個探究問題，學生需完成活動中的所有探究問題。本研究將所有學生分為兩組，實驗組採用本研究提出之引導探究智慧助教輔助進行探究學習活動(33人)，控制組學生則搭配相關物理教材自主進行探究活動(29人)。活動前後蒐集學生物理概念測驗、自我效能問卷，並記錄學生在探究學習活動中的回答和操作行為。此外，學生與智慧助教的對話、互動紀錄也將一併蒐集。
研究結果顯示，在經過活動後，兩組學生的物理概念測驗後測成績和自我效能皆有顯著提升。然而，接受智慧助教協助的學生在探究學習活動中的答題表現顯著高於控制組，且在物理概念後測成績中的進步幅度也有顯著提升，這表明智慧助教的引導有助於學生建構科學概念。但在自我效能方面，可能由於活動時間不長，因此智慧助教的協助對其並無顯著影響。在探究行為方面，智慧助教引導的學生在進行活動時參與度較高，回答探究問題的次數顯著高於控制組，在進行探究活動時更有策略性；而自主進行探究的學生則是在進行活動時較無頭緒與方向性，在切換不同探究問題的次數顯著高於實驗組。此外，在對話互動分析中，發現智慧助教的回饋可以在學生需要支援時提供幫助，其提問也可引發學生思考。綜上所述，在線上學習環境中，智慧助教的協助可以有效幫助學生進行科學探究，能夠提供學生即時的教學輔助，以提升學生的學習表現與對物理概念的掌握程度。
;Scientific inquiry is a problem-oriented learning method, but teacher guidance is crucial for its effectiveness. Therefore, this study combines an online scientific inquiry platform with a Generative AI-based chatbot to establish an pedagogical agent called the "AI-Tutor," which can guide students′ higher-order thinking. Acting in the role of a teacher, the AI-Tutor guides students through scientific inquiry, investigating the effects of its assistance on students′ answer performance, learning effectiveness, inquiry behavior, and self-efficacy.
This study recruited 62 high school students from northern Taiwan to participate in an inquiry activity on the CoSci online scientific inquiry platform. The activity focused on the "center of mass and momentum" unit in high school physics and included computer simulations and multiple inquiry questions that students needed to complete. The students were divided into two groups: the experimental group, which was guided by the AI-Tutor (33 students) , and the control group, which conducted the inquiry activities independently using related physics materials (29 students). This study collected pre- and post-test data on both physics conceptual test and self-efficacy questionnaire. Students′ answers on inquiry questions and operational behaviors during the inquiry activities were recorded. Additionally, the interactions and dialogues between students and the AI-Tutor were collected.
The results showed that both groups significantly improved their post-test scores in the physics conceptual test and their self-efficacy after the activity. However, students assisted by the AI-Tutor performed significantly better on inquiry activity questions than the control group and showed greater improvement in their post-test physics concept scores, indicating that the guidance of the AI-Tutor helped students construct scientific concepts. However, due to the short duration of the activity, the assistance of the AI-Tutor did not have a significant impact on self-efficacy. In terms of inquiry behaviors, students guided by the AI-Tutor participated more actively and answered inquiry questions significantly more frequently than the control group. They were also more strategic during the inquiry activities. Students in the control group were more directionless and switched between different inquiry questions significantly more frequently than the experimental group. Moreover, analysis of the dialogue interactions revealed that the feedback from the AI-Tutor provided support when students needed it, and its questions prompted students to think critically. In summary, in an online learning environment, the assistance of the AI-Tutor can effectively help students engage in scientific inquiry, providing timely instructional support to enhance their learning performance and understanding of physics concepts. |
