The transformation of architectural discourse into tangible design artifacts remains a central concern within architectural theory and design research. Historically, social phenomena have continuously shaped architectural discourse, which in turn has informed the emergence of built forms. However, the nature of the relationship between discourse and artifact remains ambiguous. Typically developed during the preliminary design phase, discourse encapsulates the intangible values of a project, often articulated through textual and visual media. Design cognition research investigates this transformation, focusing on the interaction between hand and mind via sketches, though such studies are limited by the number of iterations they can analyze. This study explores whether artificial intelligence, specifically text-to-image generation models, can offer a novel approach to evaluating the consistency between architectural discourse and its resulting artifact. By enabling numerous iterations, the proposed AI-supported method seeks to quantify a fundamentally qualitative research problem in architectural theory. Two case studies were selected based on contrasting discursive and formal characteristics: Rem Koolhaas’s Charette Submission for the Expansion of MoMA, and Brian Cantley’s Syntaxonome. The discourse associated with each architect was input into a text-to-image model to generate visual outputs. These were then compared with the original artifacts to evaluate the alignment between discourse and design. Results were analyzed through quantitative ratios and visual charts, offering a new perspective on the discourse- artifact continuum and highlighting the potential of AI tools in early-stage architectural design.
Uyan et al. (Tue,) studied this question.