五步 AI 協作開發法：從初學者到 AI 指揮官的程式實作教學
講者/Speaker： 陳子皓, Tzu-Hao Chen
隨著生成式人工智慧技術的普及，程式設計教育正面臨典範轉移的契機。然而，現行 AI 輔助學習方案多將人工智慧定位為單純的程式碼補全工具，缺乏針對初學者設計的系統性協作流程，致使非資訊背景學習者在認知負荷、專案分解能力及技術決策主動性等面向仍面臨顯著挑戰。
有鑑於此，本研究提出「五步 AI 協作開發法」，整合提示工程（Prompt Engineering）與軟體開發生命週期（SDLC）之核心概念，建構一套以學習者為主體的結構化人機協作框架。該框架透過「決策前置的技術選型評估機制」、「以已知語法約束 AI 輸出範圍之限制性提示策略」，以及「模組化原子開發流程」三項核心設計，系統性地將 AI 角色從被動的程式碼產出工具，重塑為主動的專案諮詢與架構引導者，進而培養學習者具備指揮 AI 進行系統化開發之後設認知能力。
本研究以材料科學與工程、環境科學與工程等非資訊專業學生為實施對象，評估結果顯示，逾 95% 的組別成功開發出具備圖形使用者介面（GUI）之互動式應用程式，超過 90% 的組別完整達成所有核心開發步驟，且多數組別於功能完整性與介面品質之評分均達 90 分以上。上述結果表明，本教學法能有效降低非專業背景學習者的程式開發門檻，並具備跨學科推廣之潛力。
As generative artificial intelligence (AIGC) technology becomes increasingly prevalent, programming education is presented with a significant opportunity for paradigm shift. However, existing AI-assisted learning solutions predominantly position artificial intelligence as a mere code completion tool, lacking systematic collaborative frameworks designed specifically for novice learners. Consequently, students without a computer science background continue to face considerable challenges in terms of cognitive load, project decomposition capabilities, and agency over technical decision-making.
In response to these limitations, this study proposes the "Five-Step AI Collaborative Development Method," which integrates core concepts from Prompt Engineering and the Software Development Life Cycle (SDLC) to construct a structured human-AI collaboration framework centered on the learner. The framework comprises three core design elements: a decision-first technology selection and evaluation mechanism, a constrained prompting strategy that leverages learners' existing syntactic knowledge to define the boundaries of AI-generated output, and a modular atomic development process. Through these elements, the framework systematically repositions the role of AI from a passive code generation tool to an active project consultant and architectural guide, thereby cultivating learners' metacognitive capacity to direct AI in conducting systematic software development.
This study was implemented with students majoring in Materials Science and Engineering, and Environmental Science and Engineering. Evaluation results indicate that over 95% of groups successfully developed interactive applications with a graphical user interface (GUI), more than 90% of groups fully completed all core development steps, and the majority of groups achieved scores above 90 in both functional completeness and interface quality. These findings suggest that the proposed instructional method effectively lowers the barriers to programming development for learners without a computer science background, and demonstrates potential for interdisciplinary adoption.
- 論文全文：五步 AI 協作開發法：從初學者到 AI 指揮官的程式實作教學
- 影像： 相關影片