Investigating intelligent generation of multimodal creative stimuli in conceptual design: strategies and implications
Zhuoshu Li, Pei Chen, Yexinrui Wu, Jiayi Yao, Hongbo Zhang, Xuanming Liu, Lingyun Sun
Abstract
Creative stimuli are pivotal sources of inspiration in conceptual design. As large generative models develop, AI has been increasingly utilised in creativity support tools to furnish designers with high-quality stimuli. However, the effects of stimuli vary across stages, rendering the certain types of creative stimuli generated by existing tools unadaptable to the dynamically changing conceptual design process. To explore the intelligent generation of multimodal creative stimuli in conceptual design, we first proposed an intelligent generation strategy to facilitate the autonomous generation of stimuli tailored to different stages. Subsequently, we designed and developed InsPilot, a system capable of proactively generating stimuli with consideration of factors including the design stage, design ideas, and the designer's input state. Through a comparative experiment with GPT-4V (N = 12), we examined the relative effectiveness of the system in supporting design creativity and delineated the relevant implications. Specifically, the creativity support provided by AI-generated stimuli include subdividing application scenarios and enhancing empathy in Rapid Divergence; refining, concretising, and transferring ideas in In-depth Divergence; identifying issues and focussing on core functions in Convergence, etc. The finding also revealed InsPilot's and GPT-4V's distinct impact on designers' behaviour and frequency of stimuli engagement.