Analyzing Designer’s Design Behavior to Predict Design Intentions and Support Design Exploration

Abstract

Design entails a multitude of considerations such as form and proportion, with a combination of design elements. The design space created by these elements is nearly infinite, making it impossible for designers to explore all possible design spaces. From this perspective, the emergence of computational support systems such as generative design and generative artificial intelligence significantly eases the process of generating and finding feasible designs within the design space. However, no matter how many designs these computational systems generate, their utility is limited if they do not align with the designer's intent. Therefore, it is crucial that computational support systems develop in a designer-centric manner, providing designs that suit the designer’s needs. To this end, this thesis proposes methodologies and systems through two studies to support the design exploration process by predicting the designer's intent: 1) A methodology and system that analyzes real-time eye-gazing data of designers to probabilistically interpret changing design intents and provide design references; and 2) A methodology and system that supports the design generation and synthesis process based on the inference of design intent at the design element level. This thesis validates these methodologies and systems through experiments with designers. It confirms their effectiveness in not only accurately predicting the designers' intent but also in recommending and generating suitable designs based on these predictions. Consequently, this thesis highlight the significant role and contributions of designer-centric computational support, which is achieved by predicting and understanding designer behavior.