Application of Generative Artificial Intelligence in Creating an Adaptive Architectural Language for Changing Climates

As climate change accelerates, traditional architectural responses prove insufficient in addressing the growing need for adaptable, environmentally responsive built environments. This study explores the application of generative artificial intelligence (GAI) in developing a novel architectural language that dynamically responds to diverse and shifting climatic conditions. Employing a mixed-method approach, the research integrates conditional generative adversarial networks (cGANs) and diffusion models with climate-specific data—including solar radiation, temperature variation, and humidity—to produce context-aware architectural forms. A dataset of precedents from five global climate zones was used to train the models. The AI-generated designs were evaluated through expert panels and environmental simulations using tools such as Ladybug and Honeybee. Results indicated strong correlations between climatic inputs and spatial features, with notable typologies—such as courtyard-based or vertically stacked forms—emerging based on environmental conditions. Cluster analysis using PCA and K-means further confirmed the existence of latent climatic grammars in the generative outputs. While the findings demonstrate the capacity of GAI to generate innovative and adaptive forms, they also highlight limitations in constructability, cultural representation, and model interpretability. The study concludes by proposing strategies for responsible AI integration, including diverse datasets, hybrid design pipelines, and ethical design frameworks. These insights contribute to the growing discourse on how AI can support climate resilience in architectural design.

Keywords:

generative design , climate-responsive architecture , artificial intelligence , adaptive systems , computational design