運用AI學伴促進國小資優生發展批判思考能力之成效與限制

Effectiveness and Limitations of AI Companions in Fostering Critical Thinking Development among Gifted Elementary Students

隨著生成式人工智慧技術快速發展，AI融入教育已成為重要趨勢。本研究旨在探討教育部因材網AI學伴「e度」應用於國小資優學生批判思考能力發展之成效與限制。研究採個案研究法，以中部某國小五年級資優班3名學生為研究參與者，實施為期十週之AI輔助Podcast課程。資料蒐集包含課室觀察紀錄、「e度」對話紀錄、Podcast訪談及焦點團體訪談，並依據Facione批判思考六大核心向度（詮釋、分析、評鑑、推論、解釋與自我調整）進行分析，透過三角檢證確保研究信實度。研究結果顯示，「e度」在任務導向課程中能有效促進學生主動提問、觀點轉換與推論深化，其反詰式對話機制有助於引發思辨與後設反思。然而，「e度」在語意與語境理解上仍存在限制，且在缺乏教師中介時，學生易產生依賴AI回饋而弱化思考主動性之現象。本研究發現，AI學伴對資優學生批判思考之支持效果，主要仰賴課程設計與教師引導。「e度」之教育價值在於作為可陪伴學習的認知工具，而非替代思考的權威來源。據此，本研究建議未來課程設計應結合真實情境任務，並強化教師專業角色，以維持學生之學習主體性。本研究成果可作為AI融入資優教育課程設計與教學實踐之參考。

With the rapid advancement of generative artificial intelligence, integrating AI into education has become a significant trend. This study aimed to investigate the effectiveness and limitations of ''e-Du,'' an AI learning companion developed by Taiwan's Ministry of Education for the Adaptive Learning Platform, in fostering critical thinking skills among elementary school gifted students. A case study approach was adopted, involving three gifted fifth-grade students from a primary school in central Taiwan who participated in a ten-week AI-assisted Podcast course. Data were collected through classroom observations, ''e-Du'' dialogue records, Podcast interviews, and focus group interviews. The analysis was guided by Facione's six core dimensions of critical thinking: interpretation, analysis, evaluation, inference, explanation, and self-regulation. Triangulation was employed to ensure research trustworthiness. The findings revealed that ''e-Du'' effectively promoted active questioning, perspective-shifting, and deeper inference within task-oriented learning contexts. Its Socratic questioning mechanism facilitated critical dialogue and metacognitive reflection. However, limitations were identified in ''e-Du's'' semantic and contextual comprehension. Furthermore, without adequate teacher mediation, students tended to rely on AI feedback, which potentially weakened their thinking initiative. This study concluded that the effectiveness of AI learning companions in supporting gifted students' critical thinking is mainly dependent on curriculum design and teacher guidance. The educational value of ''e-Du'' lies in serving as a cognitive tool for learning companionship rather than an authoritative substitute for thinking. Accordingly, this study recommends that future curriculum design incorporate authentic task-based learning while strengthening the teacher's professional role to maintain student learning agency. These findings provide practical implications for integrating AI into gifted education curriculum design and instructional practice.