動態想法地圖：探討輔助分散式想法連結之研究

Abstract

設計本身是一種設計者與想法之間的互動性行為。在早期設計階段中，此互動性行為常透過二人以上的參與者共同討論而產生多樣性的設計想法，而想法聯想的連結（linking）與分散性（distribution）問題是了解互動性行為的關鍵。針對此分散性的想法連結問題，本研究建構一個名為「動態想法地圖」（Dynamic Idea Maps, 簡稱DIM）的運算模型，試圖建構及了解上述問題的連結機制與連結有效性，進而在分散式的設計環境中，輔助多重參與者產生多樣性且關聯性的設計想法。

本論文的研究步驟，經由三個階段進行：文獻回顧與模型建立，系統實作，及實驗檢驗。在第一階段，經由背景研究與文獻回顧，探索與建立分散式想法連結的主要機制：呈現、回憶與溝通，這些機制結合案例基礎推理與軟體代理人技術，提供本研究建構DIM的基礎。在第二階段，DIM架構在一個多重代理人平台（DARIS），以具體發展成一個輔助分散性想法連結系統。在第三階段，本研究實際執行一個設計實驗，予以驗證DIM之運算的可行性與連結的有效性。

經由上述的研究方法，最後的結果證明想法聯想的分散性連結問題，可以經由運算的方式了解並實踐，同時，其所發展的輔助系統有助於參與者在分散性的設計環境中，產生多樣性的想法與想法之間的有效性連結。此外，在研究過程中的發現，也幫助了解分散式想法連結所伴隨在設計運算上的複雜性問題，而這些問題將提供作者未來之後續研究。

Design is an interactive behavior involving the evolution of ideas between two or more participants in discussion, especially during the early design stage. Idea association is an important behavior for generating diverse ideas through the dynamic exchange of varied knowledge possessed by the participants. Linking and distribution within idea association are the key issues for understanding the interactive behavior. To understand the mechanism and effectiveness of the distributed linking, this research proposes a computational model named Dynamic Idea Maps (DIM), which supports participants to generate correlated and diverse ideas in a distributed design environment.

There are three research steps in this thesis. By applying the technology of case-based reasoning and software agent, the first step is to establish the mechanisms of the distributed idea linking, which includes representation, recalling and communication. Furthermore its computational model is proposed. In the second step, some computational components within DIM are computed for supporting distributed idea linking by using a multi-agent framework (DARIS). In the third step, the computational feasibility and effective linking of DIM are evaluated through a design experiment.

Consequently, this research indicates that the distributed idea linking of idea association can be understood through DIM. Also, the DIM system can support participants to generate diverse ideas and effective linking in a distributed design environment. Following the development of DIM, future studies such as agent learning, case adaptation or supporting the creative problem solving can be pursued.