生成式人工智慧對設計師於設計思考中發散與收斂過程之影響

The Impact of Generative Artificial Intelligence on Designers’Divergent and Convergent Thinking Processes in Design Thinking

設計思考過程中，發散思考與收斂思考是推動創新過程的核心要素，前者負責探索多元創意可能性，後者則篩選並深化最具價值的解決方案。生成式人工智慧（generative artificial intelligence, GAI）透過機器學習，並以自然語言與人類互動，生成文本或圖像內容，能為創造性思考提供嶄新的輔助方式。現有研究多關注生成式AI在發散思考的應用，對其在收斂思考中的影響仍缺乏系統性探討。本研究以設計思考模型為理論框架，藉由工作坊中文字類與圖像類生成式AI的操作，探討設計師在構思產出、篩選與協作模式上於不同設計階段的差異。研究結果顯示，生成式AI在定義以及發展階段中，並未提升設計師的構思數量；但在交付階段中，AI的操作能有效增強概念的精煉，並為最終設計方案提供有力的支持。未來研究可進一步探討如何在發散階段引入AI收斂輔助機制，並追蹤新一代生成式AI在不同設計情境下的潛在應用與影響，以實現兼顧創意多元與概念精煉的最佳平衡，促進人機協作於設計思考中的有效應用。

In the design thinking process, divergent and convergent thinking serve as the core drivers of innovation. Divergent thinking involves exploring a broad spectrum of creative possibilities, whereas convergent thinking entails the filtration and refinement of the most valuable solutions. The advent of generative artificial intelligence (AI), underpinned by machine learning and natural language interaction, has introduced a new dimension to this paradigm by facilitating the generation of text and image content, thereby offering novel forms of support for creative thinking. However, extant research has predominantly centered on the application of AI in divergent thinking, with limited systematic investigation into its role in convergent thinking. This study adopts the Double Diamond model of design thinking as its theoretical framework. Recognizing the distinct cognitive processes and resources involved in verbal and visuospatial creativity, this research incorporates two types of generative AI: text-based AI (GPT-3.5) and image-based AI (Midjourney and Niji) into a series of design thinking workshops. Through analyses of idea generation, selection, and collaboration among professional designers, the study examines the impact of generative AI at different stages of the design process. Generative AI did not significantly increase idea quantity in either the definition or development process. Conversely, in the delivery process human-AI teams showed a significantly higher modification-link index, indicating enhanced concept refinement that supports the final design solution. Future research may explore the integration of AI-assisted convergence mechanisms in early divergent stages and examine the evolving role of next-generation generative AI across various design contexts to achieve a balance between creative diversity and conceptual precision, ultimately supporting more effective human-AI collaboration.