結合雙重編碼理論與CDIO教育框架──探討AI生成3D模型對學生空間能力提升之影響

Integrating Dual Coding Theory and the CDIO Educational Framework: Exploring the Enhancement of Students’Spatial Abilities through AI-Generated 3D Models

本研究探討人工智慧（AI）於3D模型生成中的應用，聚焦其促進空間能力發展的可行性與成效。隨著生成式AI技術進展，3D建模逐漸趨於直覺化與高效率，然而在傳統教學情境中，學生仍常面臨空間能力不足與軟體學習門檻高等問題。本研究結合雙重編碼理論（Dual Coding Theory）與CDIO教育架構，針對34位非3D設計背景學生進行實驗，評估其心理旋轉與空間視覺化能力，並輔以使用者經驗評估。此外，本研究透過對10位學生與5位3D設計專家的半結構式訪談，分析其對AI工具的差異觀點。研究結果顯示，學生空間能力呈現顯著提升，且使用者經驗問卷在吸引力與操作可靠性上呈現正向評價。訪談進一步揭示學生與專家之間的認知差異。AI生成模型在可控性、動態結構與幾何複雜度方面仍有待提升，但在激發學習興趣與輔助空間能力發展上展現潛力；相較之下，專業設計師更重視透過模型拆解與重構深化空間理解。本研究結果可作為AI應用於設計教育整合不同學習族群之參考。

This study investigates the application of artificial intelligence (AI) in 3D model generation, focusing on its feasibility and efficacy in fostering spatial ability development. As generative AI evolves, 3D modeling has become increasingly intuitive and efficient. However, in traditional pedagogical settings, students often face challenges such as underdeveloped spatial skills and steep learning curves for software operations. By synthesizing Dual Coding Theory with the CDIO educational framework, this study conducted an experiment involving 34 non-3D design majors to assess their mental rotation and spatial visualization abilities, complemented by user experience evaluations.
Additionally, semi-structured interviews were conducted with 10 students and 5 3D design experts to explore divergent perspectives on AI tools. The results demonstrated significant improvements in students’spatial abilities, while user experience questionnaires indicated positive feedback regarding the attractiveness and operational reliability of AI-generated tools. The interviews further revealed a perceptual gap between students and experts. While AI-generated models require further refinement in controllability, kinematic structures, and geometric complexity, they exhibit substantial potential in stimulating learning interest and augmenting spatial skills. In contrast, professional designers place greater emphasis on the deconstruction and reconstruction of models to deepen spatial understanding. These findings provide valuable insights for integrating AI into design education across diverse learner groups.