隨著大型語言模型在程式設計教育中的普及，初學者雖能藉此快速生成程式碼，卻也常因缺乏系統性思維，陷入需求描述不清、無法有效驗證與修正的困境。本研究目標是解決此問題，提出一套應用 PDCA（Plan-Do-Check-Act）流程的結構化學習方法，以提升程式設計初學者在大型語言模型輔助開發過程中的「需求分析」與「測試驗證」能力。為驗證此方法成效，本研究設計了一套「PDCA流程案例式學習模組」，並進行了一項實驗研究，透過前後測驗，並以共變數分析 (ANCOVA) 比較兩組在協作能力上的差異，同時輔以質性回饋進行深入探討。實驗結果顯示，在排除起始能力的影響後，實驗組在「需求分析能力」與「測試驗證能力」兩項核心指標上的後測成績，均顯著優於控制組。質性回饋亦證實，此結構化流程引導學習者從被動的提問者，轉變為主動規劃與引導的協作者，有效內化了需求分析、系統化測試驗證與策略性修正的開發思維。綜上所述，本研究證實提出的「PDCA流程案例式學習模組」作為結構化學習方法，能有效提升初學者與大型語言模型的協作效率與程式碼正確性，為生成式人工智慧時代的程式設計教育提供了一套具體且可行的教學策略。;As Large Language Models (LLMs) become increasingly integrated into programming education, novices can generate code rapidly. However, they often struggle with a lack of systematic thinking, leading to challenges in specifying clear requirements and performing effective code verification and debugging. This study addresses this issue by proposing a structured learning method based on the PDCA (Plan-Do-Check-Act) process, designed to enhance the "requirements analysis" and "testing and verification" capabilities of programming novices during LLM-assisted development. To validate the effectiveness of this method, the study employed an experimental design featuring a "PDCA Process Case-Based Learning Module." Through pre- and post-tests, the collaborative abilities of the experimental and control groups were compared using Analysis of Covariance (ANCOVA), with the quantitative findings supplemented by qualitative feedback for in-depth analysis. The results indicate that after controlling for prior abilities, the experimental group significantly outperformed the control group on the two core metrics of requirements analysis and testing and verification. Qualitative feedback further confirmed that this structured process guided learners to transition from being passive recipients of code to active collaborators in the development process. This shift enabled them to effectively internalize a development mindset encompassing requirements analysis, systematic testing, and strategic debugging. In conclusion, this study demonstrates that the proposed "PDCA Process Case-Based Learning Module," serving as a structured learning method, effectively enhances the collaborative efficiency and code correctness for novices working with LLMs. It provides a concrete and feasible teaching strategy for programming education in the era of Generative AI.
