Exploring the Feasibility of Generative AI in Enhancing the Identification of Spatial and Regulatory Opportunities Using Urban Digital Twins

This study explores the feasibility of using generative AI to enhance spatial and legislative opportunity finding within urban digital twins. Digital Twins (DTs) integrate real-time and historical data to provide comprehensive views of built environments, potentially aiding different disciplines involved in spatial planning practices. However, the complexity of legal frameworks and their visualization in DTs remains a challenge. Leveraging advancements in large language models (LLMs), this research investigates how multimodal AI can interpret complex legislative data to improve spatial planning. The study employs a Design Science Research (DSR) methodology, focusing on tuning existing LLMs with spatial content. Key findings include the successful generation of Geography Markup Language (GML) code, enhancing interoperability with spatial planning tools, and the iterative design process that improved the model's performance. Preliminary results indicate that a multimodal approach, including text, images, and GML code, significantly enhances the model's capability. Future research will focus on improving data quality, expanding multimodal capabilities, and evaluating real-world applications. This study contributes to the development of transparent, contestable, and explainable AI solutions for spatial planning.