Artificial Intelligence for Spatial and Environmental Research in Architectural Sciences: Towards a Problem-Oriented AI Research Framework

This chapter proposes a Problem-Oriented AI Framework for architectural research through a literature-based conceptual review. The literature resource comprises 227 Web of Science Core Collection records published between 2022 and 2026. A preliminary matrix uses metadata, keywords, and available abstracts. Selected studies are compared by research problem, data, analytical task, AI method, and evaluation. Unavailable original search details prevent claims of a reproducible systematic review or comprehensive assessment of all records. The discussion highlights distinct evidence and validation requirements across spatial, behavioural, environmental, design, heritage, and construction applications. The framework comprises five stages: define the research problem; identify and prepare the data; specify the analytical objective; select an appropriate AI method; and validate and interpret the outputs. Comparison with existing approaches and an illustrative residential daylight scenario clarify its proposed contribution. The framework organises methodological decisions while distinguishing computational performance from architectural interpretation. Its practical value and transferability require further empirical application and comparative evaluation.