Abstract
In recent years, the burgeoning field of garden co-design, spearheaded by urban residents, has significantly advanced public engagement and satisfied intrinsic needs. Nonetheless, conventional implementations often encounter inefficiencies due to divergences caused by power structures, stakeholder interests, and cognitive disparities between leading designers and participant residents. With recent advancements in the cognitive capabilities and reasoning powers of Large Language Models (LLMs), a novel opportunity has emerged to leverage these technologies in reshaping the paradigm of garden co-design. This paper presents an innovative bio-harmony garden co-design model, which seamlessly integrates LLMs with collaborative design knowledge. This model facilitates a transformative shift towards a resident-centric co-design approach through (i) user requirement translation, (ii) plant category solving, and (iii) horticultural-LLM reasoning. We evaluated the proposed model on the 6,970-square-meter Arconati farm with 53 urban residents participating and providing both qualitative and quantitative feedback. Experimental results demonstrate that our model significantly enhances key aspects of the co-design process—design engagement, expectation alignment, and feedback responsiveness—by over 21%, compared to traditional designer-led paradigms.
Presenters
Xuanfang WangStudent, Ph.D., Tsinghua University, Beijing, China Yan Huang
Professor, Environmental Design, Academy of Arts & Design, Tsinghua University, Beijing, China
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
New Media, Technology and the Arts
KEYWORDS
CO-DESIGN, URBAN GARDEN, RESIDENT-CENTRIC DESIGN, LARGE LANGUAGE MODELS
