在當前的教育領域中，「生成式人工智慧」(Generative Artificial Intelligence, GenAI)已被視為一項具有潛力的重要工具，並廣泛應用在各種教育情境中，可用於改善教學和學習過程並促進知識的創新。然而，儘管其應用前景看好，但對於如何最有效地運用GenAI工具進行學習的方法仍有待深入探討。因此，本研究旨在探討臺灣大學生運用GenAI工具進行學習時所採用的方法，透過現象圖學法對28名大學生的半結構式訪談資料進行分析後，找到可能使用的七種學習方法，其中包括「複製」、「關鍵字」、「嘗試錯誤」、「追問」、「情境設定」、「分治法」及「驗證」，並進一步探討「科學、技術、工程和數學」(Science, Technology, Engineering, Mathematics, STEM)和非STEM主修學生在使用相關工具的學習方法差異。主要結果顯示，這些STEM與非STEM大學生均採用了「嘗試錯誤」的學習方法。再者，本研究進一步將這些學習方法分為表層與深層學習方法，並使用卡方檢定後發現STEM大學生傾向使用表層學習方法，而非STEM大學生則更傾向使用深層學習方法。本研究透過分析不同主修之臺灣大學生使用GenAI工具進行學習的實際情況，將有助於了解不同專業背景對學習方法選擇的影響，從而對未來的教育實踐和相關研究提供建議及方向。

Generative artificial intelligence (GenAI) has emerged as a promising tool in the educational landscape, with the potential for being applied in various learning contexts. Despite its potential, how to utilize GenAI tools for learning remains largely unexplored. This study aims to explore the approaches to learning employed by Taiwanese university students by using GenAI tools. Through a phenomenographic analysis of semi-structured interviews with 28 university students, seven approaches were identified: “Copy,” “Keyword,” “Trial and error,” “Probing,”“Contextualization,” “Divide and conquer,” and “Verification.” The findings also revealed similarities and differences in learning approaches between the STEM and non-STEM students. For instance, “Trial and error” was the most common approach for both groups. These categories were then categorized into surface and deep learning approaches. Chi-square tests indicated that the STEM students were more inclined towards adopting surface learning approaches, while the non-STEM students favored deep learning approaches. These findings advance understanding of how different academic backgrounds may influence the adoption of approaches to learning and provide insights for future educational practices and research.

STEM教育、大學生、生成式人工智慧、現象圖學法、學習方法

STEM education; University student; Generative artificial intelligence; Phenomenography; Approaches to learning