Elemental potential of streets: A self-organizing model of the geological conditions of the province of Quebec using investigative and generative AI

Current urban planning treats the subsurface as a passive solid or void, constrained by a “needs to resources” paradigm that marginalizes geological knowledge. Standard mapping methodologies, particularly those using the Analytic Hierarchy Process, rely on manual expert solicitation, which is logistically burdensome and prone to cognitive blind spots. This research proposes a theoretical shift from target-oriented systems to an obstacle-oriented framework. Designating the street segment as the fundamental unit of analysis across the province of Québec, this research presents a prototype that implements a static knowledge-retrieval pipeline where a Large Language Model serves as an informational model for preliminary planning, translating hydrogeological constraints into qualitative assessments. These textual descriptions are then encoded into a continuous partition spectrum using a one-dimensional self-organizing map. The prototype demonstrates high computational robustness in standardizing qualitative data, successfully mapping the geological dichotomy between the stable crystalline basement of the Canadian Shield and the conflict-prone sedimentary basins of the St. Lawrence Lowlands. By converting qualitative data into relational contextual numbers, the model identifies latent synergies without predefined target functions. The article concludes that this pre-specific modeling approach validates the utility of artificial intelligence in mitigating cognitive blind spots, demonstrating how computational models can complement human engineering by repositioning geological insights at the very beginning of the urban design process. By establishing a topos of communication capable of circulating around the inherent opacity of the geological volume, this method offers a scalable instrument for sustainable underground urbanism that treats the earth not as an anecdotal resource, but as an active constituent of the built environment.