Generative policy design: AI for simulating sustainability governance frameworks

The complexity and interconnectedness of sustainability challenges demand governance frameworks that are adaptive, systemic, and scenario-resilient. This research introduces a novel approach to Generative Policy Design (GPD) using large language models (LLMs) and generative simulation to create, iterate, and evaluate sustainability-focused governance architectures. By integrating structured datasets on climate policy, socio-economic indicators, and regulatory typologies, the system employs transformer-based generative models to synthesize policy alternatives aligned with the UN Sustainable Development Goals (SDGs). Each generated governance framework is evaluated through a multi-agent systems simulation that assesses impact on carbon reduction, equity, and institutional adaptability under various future scenarios. A case study on urban climate governance demonstrates that generative policy design outperforms traditional rule-based drafting in speed, stakeholder inclusion, and alignment with long-term sustainability metrics. The findings position generative AI as a critical tool in co-creating adaptive and data-driven public policy for a just and sustainable transition.