Identifying Opportunities to Repurpose Decommissioned Products Using Large Language Models

Abstract Extending a product's life cycle after decommissioning through strategies such as repurposing is important for conserving natural resources. However, repurposing is challenging because it is often unclear what the decommissioned product can be repurposed for. Therefore, this study aims to investigate how to utilize Large Language Models (LLMs) to identify novel and feasible repurposing opportunities. We use a set of shared attributes among products and actions that can facilitate repurposing, which is called the Repurposable Attribute Basis (RAB). We combine it with existing repurposing examples to prompt an LLM to generate repurposing ideas for a specific product. Then, we evaluated the ideas for their novelty, technical feasibility, and scalability using both rater-based and natural language processing based metrics. We identified that using the RAB phrases with repurposing examples in a one-shot prompt format supports the LLM in generating novel and scalable ideas that move away from the traditional low-value repurposing ideas (e.g., novelty items) compared to a base prompt. Furthermore, successively using RAB phrases helped the LLM better understand the product attributes to be repurposed and the transforming actions to be utilized, generating more novel and scalable ideas while maintaining a similar level of technical feasibility to a base prompt. This approach sets the foundation for exploring and developing support tools in early concept generation for repurposing activities.