Methodological architecture in artificial intelligence-mediated design, São Paulo, Brazil

Abstract

The incorporation of artificial intelligence (AI) into architectural design processes in Latin America has been predominantly instrumental and fragmented, lacking explicit methodological frameworks that articulate design decisions, disciplinary responsibility, and technological mediation. This paper proposes and qualitatively demonstrates a constructive methodological framework to integrate AI as a structural component of design reasoning, organized into three interdependent levels: theoretical, operational, and critical/evaluative. The core of the method is a matrix and an application protocol that relates project phases, cognitive tasks, and functional categories of AI, requiring the declaration of inputs, criteria, operations, and trace records at each intersection. A pilot study in an academic context revealed consistent improvements in process traceability and procedural clarity, strengthening the capacity to justify, audit, and communicate decisions without implying automatic formal optimization. Results indicate that the primary transformation introduced by AI lies not in task automation, but in the reconfiguration of the cognitive power that structures the space of design possibilities. As an original contribution, the study formulates a transferable methodological architecture that, through algorithmic design thinking, governs algorithmic mediation through explicit criteria, verifiable records, and design responsibility, shifting innovation from technical novelty toward the epistemological organization of the process. It concludes that the critical integration of AI in architecture depends on traceable methodological structures capable of making limits, risks, and zones of opacity visible, constituting an internal condition of disciplinary, ethical, and scientific validity in contemporary design practice.