Description
Amidst the global surge of highly sophisticated computational systems, there is a heightened interest in conceiving a collaborative system in architectural design that can harness the strengths of both human intuition and algorithmic precision. Realising this vision necessitates a platform to support profound interactions or dialogues between architects and computer algorithms. One suggested avenue for fostering such interaction involves utilising evolutionary algorithms, which can potentially optimise both quantitative (objective) and qualitative (subjective) goals. Despite their promise, a noticeable research gap exists concerning their effective deployment. In response, this dissertation investigates an interactive human-algorithm design method wherein human designers can guide algorithmic processes to enable the capture of subjective design criteria. The overarching aim of this research is to understand how human designers can interact with genetic algorithmic processes to accommodate an architect’s intuition or subjective preferences. This aim is divided into four secondary aims. These are: (i) to develop a better understanding of interactivity in genetic algorithms, (ii) to investigate interactive genetic algorithms’ capacity to accommodate combined quantitative and qualitative design aims, (iii) to formulate a robust framework for the different interaction modes in genetic algorithms that support human subjective choices in architecture and, (iv) to create a toolset that facilitates different types of interactions for the accommodation of human subjective choices. These aims, coupled with two formal research questions to address knowledge gaps in the field, are pursued using bespoke interactive software developed specifically for this research, alongside a three- stage empirical testing process. The first empirical stage collects data from a design experiment with two variations involving both novice and experienced architects (n = 26). The second stage involves semi-structured interviews and survey responses of participants in the design experiment. The last stage entails evaluating the experiment’s design outcomes through a survey of professionals and academics (n = 68), synthesising viewpoints from both participants and assessors via a mixture of quantitative (statistical) and qualitative analysis methods. The key contributions developed in this dissertation include: (i) a refined understanding of interactivity in genetic algorithms and its capacity to direct the evolutionary process towards an architect’s subjective aims, (ii) a comprehensive interaction framework that encompasses existing, newly created, and future interaction modes, serving as a roadmap for further research, (iii) the development of a novel software tool, Snowflake, that supports varied types of interactions, empowering architects to influence the evolutionary direction and, (iv) the provision of previously unavailable pseudo code for a significant genetic algorithm (SPEA2), thus broadening its accessibility to architectural researchers with a complete set of functions. Through its theoretical and experimental sections, this research augments the ongoing discourse on evolutionary algorithms in design. Specifically, it generates new knowledge about ways to harness the power of interactive genetic algorithms to enable a design process that blends quantitative and qualitative design goals while preserving the unique contributions of human creativity and expertise in design.| Period | 1 May 2024 → 31 Jul 2024 |
|---|---|
| Examinee | Zayad Motlib |
| Examination held at |
|
| Degree of Recognition | International |