Conference Agenda

The aim of this research is to analyze and expose risks associated with using AI tools such as Large Language Models and Text to Image Models to create user personas, and initiate a discussion about their application in design practice. First, a model is presented comparing the traditional user research approach with an AI persona generation method. Through a case study involving the creation of personas within the context of Mexico City, a critical analysis is conducted, revealing biases. Possible causes and risks for design practice and education are discussed, as well as potential benefits. Finally, a model for scalability of AI personas and generation of design ideas is presented. The significance of these findings shows the importance for design research to question how AI tools work.

Bio-inspired design (BID) is a design methodology that employs biological analogies for engineering design, encompassing problem-driven and solution-driven BID. Solution-driven BID starts with knowledge of a specific biological system for technical design. Despite the proven benefits of solution-driven BID, the gap between biological solutions and engineering problems hinders its effective application, with designers frequently encountering misaligned problem-solution pairs and facing multidisciplinary knowledge gaps in the analogical transfer process. Therefore, this research proposes a large language model (LLM)-based concept generation method, designed to automatically search for problems, transfer biological analogy, and generate solution-driven BID concepts in the form of natural language. A concept generator and two evaluators are identified and fine-tuned from the LLM. The method is evaluated by an ablation study, machine-based quantitative assessments, and human subjective evaluations. The results show our method can generate solution-driven BID concepts with high quality.

Generative AI has greatly enhanced the production of digital content and has had a significant impact on the creative activities of designers. However, generalized generative AI falls short of designers' expectations in semantic understanding and image generation, and thus been poorly used in specific design domains (e.g., automotive design). This paper aims to explore the integration of generative AI into the design process, focusing on the generation of automotive design. We deconstructed the process of automotive design through user research, extracted the needs and pain points of designers, and transformed them into fine-tuning tasks for generative models. We trained three models in different styles based on Stable Diffusion: abstract forms, digital painting, and realistic rendering, and integrated them into the design workflow for practical automotive design. Performance evaluations and user studies indicate that our auxiliary models and generative design process can produce satisfactory automotive design and enhance efficiency.

Products imbued with traditional cultural semantic information hold significance in commerce, culture, and the dissemination of information. However, the integration of implicit cultural semantics into the design process of cultural products poses a significant challenge. Key issues include the inaccurate expression of implicit semantics and the inadequacy of semantic information retrieval and inspi-ration. Therefore, we adopt a datadriven approach to achieve symbolic semantic expression in generating and inspiring design concepts for cultural products. In this paper, we utilize the generative pretrained transformer (GPT-3.5) as the base language model (PLM). By analyzing semantic information features in layers and mapping, we identify two design concept generators, fine-tuning them for the automatic retrieval and expression of semantic information. This is undertaken to generate cultural product designs in a natural language form. The method under-goes experimental evaluation, and the results demonstrate that our approach can generate cultural product design concepts containing accurate cultural information.