Ma and colleagues argue that restorative quality resides not only in visible objects but in their positional relations. Their iconic idea is to represent streets and cities as graphs whose nodes, sequences and topological connections encode spatial structure more adequately than isolated images. The theoretical contribution extends Attention Restoration Theory from a catalogue of perceptual qualities towards a relational morphology in which being away, extent, fascination and compatibility are conditioned by spatial arrangement. Methodologically, sequential street-view imagery generates street-level graphs, while road topology and perceptual, spatial and socioeconomic features form a city-level graph processed through graph neural networks. The operation converts non-Euclidean urban structure into an analysable model of psychological restoration, demonstrating that equal restoration scores may emerge from distinct spatial configurations. Its wider bridge joins environmental psychology, urban morphology, computer vision and GeoAI, while also raising a critical question for spatial theory: when restorative experience is inferred computationally, the model does not merely describe urban form but establishes which relations become legible as psychologically consequential.
URBANISM MEETS ART