Reflexive Multimodal Learning for Clean Energy Transitions: Causal Insights on Cooking Fuel Access, Urbanization, and Carbon Emissions

Achieving Sustainable Development Goal 7: Affordable and Clean Energy requires not only technological innovation but also a deeper understanding of how socio-economic factors influence energy access and carbon emissions. However, important open questions remain regarding the specific methods for quantifying the socio-economic drivers of energy access, the domains in which these drivers interact—such as policy, technology, and infrastructure—and the objectives of feedback processes that influence energy systems. To address gaps in understanding the socio-economic drivers of climate change, this study develops an AI-based framework, ClimateAgents, which combines large language models with domain-specialized agents for hypothesis generation and scenario exploration. Leveraging 20 years of socio-economic and emissions data from 266 countries across 70 indicators, the framework applies a machine learning–based causal inference approach to identify key determinants of carbon emissions. The analysis reveals three key determinants: (1) access to clean cooking fuels in rural areas, (2) access to clean cooking fuels in urban areas, and (3) the proportion of the population living in urban areas. These findings underscore the important role of clean cooking technologies and urbanization patterns in shaping carbon emission outcomes. The proposed framework, ClimateAgents, supports exploratory hypothesis generation, multimodal data synthesis, and interpretable scenario construction, centered on the critical paradigm shift from building independent models to designing modular, reflexive systems for dynamic social environments. This work contributes actionable insights for policy design across diverse socio-economic contexts and highlights how generative AI can facilitate equitable climate action through data-informed, context-aware reasoning.