Research on Cognitive Modes of Generative Product Design

[[abstract]]本研究的目的在於建立生成式產品設計的認知模式概念架構，用於探討生成式設計師在數位化情境中表徵、生成和評價產品目標對象的演算法規則和幾何圖形的設計認知過程。綜觀生成式產品設計、設計思維的認知過程、生成式產品的設計認知等相關理論研究，生成式設計正成為人工智慧技術趨勢下的新興研究領域，被譽為智能設計自動化的新一波浪潮。生成式設計是設計與計算、設計與認知交互演化中產生的設計新範式。基於性能數據驅動的生成式演算法，如拓撲優化、形狀文法、遺傳演算法等已然成熟，但是礙於認知過程的複雜性，人的心理、生理、行為數據暫時還不能夠直接驅動演算法，設計師與電腦仍舊是間接合作的狀態，因此，生成式設計中的認知科學問題成為了研究熱點。本研究執行三個研究子題。子題甲：生成式產品設計的視覺心像模式研究，通過問卷調查法，以語意差異量表獲取設計師對生成式產品設計結果的心理量測數據，採取結構方程模型分析，驗證三個研究假設並建構“美學-創新-風格”的認知模式，結果顯示，結構方程模型在不同設計門類和不同群組間具備良好的適配性和擬合度，“重複感”和“起伏感”為生成式設計風格提供了較高的解釋力；子題乙：生成式產品設計的選擇性注意模式研究，通過實驗法，以眼動追蹤和口語報告獲取設計師對生成式產品設計結果的生理量測數據，採取變異數分析，驗證三個研究假設並建構“搜索-注意-決策”的認知模式，結果顯示，具備適度複雜性的演算法所生成的產品造型，更有利於視覺特徵的搜索、注意和決策；子題丙：生成式產品設計的問題解決模式研究，通過質性研究法，以內容分析技術獲取設計師在生成式產品設計過程的行為量測數據，採取回歸方程分析，驗證七個研究假設並建構“內容-過程-情境”的認知模式，結果顯示，生成式設計是實踐中的反思過程，“演算法-圖形”情境是設計問題和解決方案的宏觀迴圈，“建構-評估-調整”過程是中觀迴圈，“框架-單元-變數”內容是微觀迴圈，設計師傾向採取前向增量的方式。最後，以設計師（人）、產品造型（物）、設計過程（事）及其交互影響關係作為認知模式的理論框架，綜合各子題的研究結果，建立生成式產品設計的認知模式概念架構。認知模式概念架構籍由人與事（表徵）、人與物（評價）、事與物（生成），有效地描述和解釋了生成式設計師在視覺心像、選擇性注意以及問題解決維度的認知過程。本研究是基於現有的認知科學研究範式和間接的人機交互技術條件下，對生成式產品設計認知模式的基礎性、階段性的研究成果；也是面向數位化、智能化設計，推動更多樣、更動態、更複雜的設計研究範式和更有效、更直接、更系統的電腦輔助設計技術發展的探索性、前瞻性的研究成果。[[abstract]]The purpose of this study is to construct a cognitive mode conceptual framework of generative product design for exploring generative designers' design cognitive process of representing, generating and evaluating the algorithmic rules and geometry of product goals, in a digital context. Throughout the theoretical research of generative product design, the cognitive process of design thinking, and the design cognition of generative product, generative design is becoming an emerging research field under the trend of artificial intelligence technology, and is considered as a new wave of intelligent design automation. Generative design is a new design paradigm arising from the interactive evolution of design and computation, design and cognition. Generative algorithms driven by performance data, such as topology optimization, shape grammar and genetic algorithm, have been mature. However, due to the complexity of cognitive process, the psychological, physiological and behavioral data of human cannot directly drive algorithms for the time being. Designers and computers are still in indirect cooperation. Therefore, the cognitive science research of generative design has become a current hot topic. This study carries out three sub-studies. Sub-study A: a study of visual imagery mode of generative product design, uses a questionnaire to obtain designers’ psychometric data on the results of generative product design through the semantic difference scale, and adopts structural equation modeling to verify three research hypotheses and construct a cognitive mode of "Aesthetics - Innovation - Style". The results show that the structural equation model has a good adaptability and fits among different design disciplines and cohorts. Moreover, "sense of repetition" and "sense of fluctuation" provide a high explanatory power for generative design style. Sub-study B: a study of selective attention mode for generative product design, uses experimental methods to obtain designers' physiological measurements on the results of generative product design with eye-tracking and verbal protocol, and ANOVA analysis to validate three research hypotheses and construct a cognitive mode of "Search - Attention - Decision". The results show that the product shapes generated by algorithms with moderate complexity are more conducive to the search, attention and decision making of visual features. Sub-study C: a study of problem-solving mode for generative product design, uses qualitative research methods to capture the designers' behavioral data in the process of generative product design through content analysis technique, and adopts regression equation analysis to validate seven research hypotheses and construct a cognitive mode of "Content - Process - Context". The results shows that generative design is a reflective process in practice. The "Algorithm - Geometry" context is the macroscopic loop of design problems and solutions, the "Construct - Evaluate - Modify" process is the mesoscopic loop, while the "Frame - Unit - Variate" content is the microscopic loop. Designers tend to adopt forward incremental approach. Eventually, a cognitive mode conceptual framework of generative product design is constructed by integrating the research results of the three subtopics, using the designer (human), product shape (object), design process (event), and their interactive influence relationships as the cognitive mode theoretical foundation. This cognitive mode conceptual framework describes and explains generative designers’ the cognitive process of visual imagery dimension, selective attention dimension, and problem solving dimension by "human and events (representation)", "human and objects (evaluation)", and "events and objects (generation)". This study is a fundamental and staged work, which is based on the existing cognitive science research paradigm and indirect human-computer interaction technology conditions, to think about and discuss generative product design, as well as a exploratory and forward-looking research result for digital and intelligent design to promote more diverse, dynamic and complex design research paradigms and more effective, more direct and systematic development of computer-aided design technology