Abstract

Life cycle assessment (LCA) is a well-established approach and benchmark for design for sustainability efforts, in which detailed reports are produced that can serve as decision-making guides for developing new products. However, LCA reports are typically dense and technically complex, making it difficult for many engineering design project stakeholders to appropriately leverage the information found within them. Our work seeks to understand and improve the transfer of knowledge from LCA reports during the early stages of the design process, specifically leveraging the natural language capabilities of large language models (LLMs). In this paper, we investigate how four LCA- and sustainabilitycentric prompting frameworks can extract relevant design knowledge from LCA reports, demonstrated through a case study where an LLM (ChatGPT) is prompted on a provided electric toothbrush LCA report. Key findings illustrate the prompting frameworks can establish high-level summaries and identify life-cycle specific information, but the development of specific and design-focused sub-prompts will allow for richer understanding. We envision designers can use these proposed frameworks to query an LLM to gain context and insights from relevant LCA reports. The proposed techniques serve as a basis for automatic knowledge extraction from life cycle documents, creating accessible information in a user-friendly manner for designers who look to develop life-cycle-informed products.