Damascus House Exploring the connectionist embodiment of the Islamic environmental intelligence by design

Past studies have demonstrated that connectionist artificial intelligence (AI) has superior capabilities for style-based generative design because it automatically searches, extracts, and applies features according to the data-represented probabilistic profile of an architectural style. To further navigate its architectural affordance, this practice-led research project explores employing connectionist artificial intelligence to produce Islamic-style architectural forms that have historically revealed environmental intelligence by embedding sociocultural factors in response to the physical and human environmental design heritage. The project applies the Pix2Pix model and inverts the logic of some existing studies to predict the building plans from daylight maps. Use multi-objective optimisation algorithms to iteratively optimise factors such as building porosity, spatial quality, and microclimate, and use it as a condition to apply a Pix2Pix to generate a corresponding porosity model that is parametrised for the further design process. The model was trained on 120 augmented, paired images based on the 30 selected examples of Islamic architecture from the Damascus Atlas to capture the relationship between the massing distribution of walls and the arrangement of major elements in an Islamic courtyard house and its thermal performance. This study seeks to test if connectionist AI can be used as a generative design tool to understand the historical development of spatial relationships in Islamic courtyard houses. It focuses on non-repetitive style metrics, embedding physical and cultural factors into data representation. The resulting environmentally intelligent model adapts to the context, with optimisation being a pragmatic design guide rather than the ultimate goal. Although the inference is based on objective probabilistic facts, the influence of the informational framework interpreted by the designers must be acknowledged.