A human–AI collaborative design and evaluation model for product Kansei engineering

Generative AI (Gen AI) is revolutionising product design as industry explores human–AI co-design. However, academia still lacks a systematic collaborative model and application cases. This study proposes a three-phase human–AI collaborative design and evaluation model for Kansei engineering. Phase 1 uses GPT-5 to synthesise imagery adjectives and define the target imagery. Phase 2 selects representative shapes from Midjourney outputs that align with the target imagery, decomposes them into shape elements to build a morphological chart, and estimates element weights and the fuzzy membership degrees of shape types to derive novel shape combinations. Subsequently, fuzzy comprehensive evaluation (FCE) combined with a Kansei questionnaire identifies the optimal shape. Phase 3 constructs a 3D model of the optimal shape, extracts its feature curve and uses Stable Diffusion to generate color schemes. Finally, the analytical hierarchy process (AHP) with a Kansei questionnaire selects the optimal design scheme. Two case studies on electric motorcycles validate the model. Results show that the model excels at analysing user imagery needs, translating imagery into form and generating color imagery, thereby advancing Kansei design. The model provides a Gen-AI-based methodology for Kansei engineering and establishes corresponding quantitative evaluation procedures.