Infrastructure, agency, and cross-regional differentiation: a curriculum typology of adaptive responses to generative AI in spatial design education

Generative artificial intelligence is reshaping spatial design education worldwide. However, existing studies have largely overlooked the constraints imposed by structural differences in infrastructure and academic resources on pedagogical adaptation strategies. Based on a three-dimensional analytical framework comprising Infrastructure Sovereignty, Algorithm Intervention Depth, and Interaction Topology, this study conducted an empirical comparative analysis of 95 courses across 18 representative institutions worldwide. Four adaptive patterns were identified: (1) the Passive-Consumer pattern, which was slightly more frequent among sampled Global North courses, especially mid-tier cases and indicates a risk that accessible commercial AI may encourage low-reflection technological consumption; (2) the Sovereign-Accelerator pattern, which appeared prominently in sampled East Asian and selected Global South courses and tended to organize code-based intervention; (3) the Sandbox-Reflective pattern, which was most frequently observed among sampled Global North courses while also appearing in selected East Asian settings, and was characterized by recursive experimentation; (4) the Dependent-Compensatory pattern, which was most clearly represented in selected Global South cases and involved intensive human intervention. These findings suggest that access to infrastructure alone does not guarantee cognitive depth. Rather, pedagogical scaffolding and institutional agency play an important role in mediating how generative AI is translated into curriculum practice.