Generative pre-trained transformers for climate scenarios: a statistical coefficient for future policy development

Abstract Environmental risks, driven by anthropogenic activities, pose critical challenges for ecosystems and human societies. Climate change, pollution, deforestation, and biodiversity loss are accelerating due to unsustainable industrial and agricultural practices, necessitating urgent scientific and policy interventions. In Futures Studies, scenario development is an essential tool in addressing these challenges, enabling policymakers to anticipate risks and develop adaptive strategies. The Delphi method, a structured, expert-based technique, plays a crucial role in scenario development by identifying emerging trends and critical uncertainties. However, a common limitation in scenario-based studies is the gap between scenario construction and actionable policy recommendations, as deriving concrete strategies remains a resource-intensive process. To bridge this gap, this study integrates generative pre-trained transformers into a spatial version of the Delphi method, namely the Real-Time Spatial Delphi, optimizing AI to assist experts in drafting policy recommendations based on scenario insights. Considering a statistical coefficient based on spatial and importance scores, this approach reduces expert workload while maintaining human oversight and refinement by automating the initial policy formulation. The proposed methodology is applied to a case study on climate adaptation strategies for Dublin 2050, demonstrating how AI-assisted policy generation can enhance decision-making in environmental planning.