| 摘要: | 生成式人工智慧（特別是ChatGPT）的快速普及，已深刻改變高等教育的學習模式，為提升學生參與度與學術成就帶來新契機。然而，缺乏結構化的ChatGPT使用常導致學生過度依賴人工智慧生成的答案，造成表面化參與及反思性思維的局限。為應對此挑戰，本研究探討如何設計與實施基於ChatGPT輔助的蘇格拉底式提問策略，以促進高等教育學生在學習情境中實現深層次參與並提升學術表現。
本研究透過兩個教學階段於高等教育商科相關課程中展開：初步探索性實施階段（N = 32）與引導性主實施階段（N = 110）。量化數據通過學業成就前後測量，以及涵蓋行為、認知、情感與主體投入的多維度參與問卷收集；質性數據則透過反思日誌、半結構化訪談、焦點團體討論，以及按蘇格拉底問題類型編碼的ChatGPT互動記錄進行分析。
量化研究結果顯示，持續實施ChatGPT輔助的蘇格拉底式提問策略後，學業成就顯著提升，具有較大的效應值及中等至較高的標準化學習成效。相關性分析表明，行為、認知、情感與主體投入參與均與學習成效顯著相關，而單純使用ChatGPT的頻率則無此關聯。質性研究進一步揭示，學生的參與模式從以澄清為導向、尋求答案的互動，逐步深化為具探索性的互動特徵，包括質疑預設、評估證據、考量替代觀點及反思影響。
綜合而言，本研究證實當ChatGPT嵌入結構化蘇格拉底提問框架並輔以教師引導時，能有效扮演對話式學習伙伴的角色。與其將人工智慧定位為自主教學，本研究強調教學設計在調節學習成果中介參與過程中的重要性。本研究提供對人工智慧輔助探究的過程導向理解，並為生成式人工智慧融入高等教育提出實證基礎的見解，以促進負責任、具反思性且有意義的學習體驗。
;The rapid adoption of generative artificial intelligence, particularly ChatGPT, has transformed learning practices in higher education, offering new opportunities to enhance student engagement and academic achievement. However, unstructured use of ChatGPT often leads students to rely on AI-generated answers, resulting in superficial engagement and limited reflective thinking. Addressing this challenge, the present study investigates how ChatGPT-supported Socratic Questioning can be designed and implemented to foster meaningful student engagement and support academic achievement in undergraduate learning contexts.
Adopting a sequential explanatory mixed-methods approach, the study was conducted in undergraduate business-related courses through two instructional phases: a preliminary exploratory implementation (N = 32) and a guided main implementation (N = 110). Quantitative data were collected using pre-test and post-test measures of academic achievement and a multidimensional engagement questionnaire encompassing behavioral, cognitive, emotional, and agentic engagement. Qualitative data were gathered through reflective journals, semi-structured interviews, focus group discussions, and analysis of ChatGPT interaction logs coded by Socratic Questioning types.
The quantitative findings demonstrate significant improvements in academic achievement following sustained implementation of ChatGPT-supported Socratic Questioning, with large effect sizes and predominantly moderate-to-high normalized learning gains. Correlational analyses reveal that behavioral, cognitive, emotional, and agentic engagement are significantly associated with learning gains, whereas the frequency of ChatGPT use alone is not. Qualitative findings further show that students’ engagement evolved from clarification-oriented and answer-seeking interactions toward deeper inquiry characterized by probing assumptions, evaluating evidence, considering alternative perspectives, and reflecting on implications.
The convergence of quantitative and qualitative evidence indicates that ChatGPT can function effectively as a dialogic learning partner when embedded within a structured Socratic Questioning framework and guided by instructor scaffolding.
Rather than positioning AI as an autonomous instructional authority, this study highlights the importance of pedagogical design in regulating engagement processes that mediate learning outcomes. The study contributes a process-oriented understanding of AI-supported inquiry and offers empirically grounded insights for integrating generative AI into higher education in ways that promote responsible, reflective, and meaningful learning. |
