Type:

Year:

Editor:

Author:

Ji HAN (1); Boheng WANG (2); Yuan YIN (1); Pingfei JIANG (1)

Series:

Institution:

Page(s):

Abstract:

In recent years, there has been a growing interest in developing Large Language Models (LLMs)-driven tools for supporting varying design activities, including idea generation. However, current LLMs frequently generate ideas that exhibit limited creativity, demonstrate contextual misalignment, and lack practical feasibility. Therefore, this paper proposes the Creativity-RAG framework for creative design idea generation by adopting the Retrieve Augmented Generation (RAG) technique using LLMs. The framework involves two RAG processes supported by two knowledge bases: (1) a domain-specific product design knowledge base for retrieving relevant design attributes of the target design, and (2) a creative product design knowledge base for retrieving corresponding design precedents that contain or are related to the design attributes. The framework combines design features from the retrieved creative design precedents to inform new design ideas that are creative, contextually relevant, and feasible. A case study was conducted to demonstrate the applicability of the framework by realising it through a customised ChatGPT using Wikipedia and Red Dot Design award winners as the two RAG knowledge bases. The paper advances research in design and creativity, demonstrating how LLMs can reuse existing knowledge and precedents to enhance creative design idea generation.

Keywords:

Design Creativity, Idea Generation, Large Language Models, Retrieve Augmented Generation, Creativity Tools