AI image generation and the relocation of architectural intelligence

Date

2026-9

Author

Lokumcu, Bartu

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

45
views

23
downloads

Cite This

Architecture has long worked through representation: an idea was realized or failed in the act of drawing and modelling. Generative AI image systems disrupt this continuity. From a brief description or a few images, these systems generate a convincing architectural representation before any consideration of spatial relationships. The resulting question that this thesis addresses is: what becomes of architectural intelligence when representation no longer demands it? The dominance of images over architectural thought was attained, not given, over five centuries. This was facilitated by institutions that evaluated images rather than the reasoning from which they arose. What, then, does AI image generation remove? Earlier computational methods still required design reasoning to be made explicit. AI image generation removes four things at once: the need to translate between modes of representation, a trace of design decisions, a temporal interval for such reasoning to occur, and an author capable of answering for the result. The thesis advanced here is that these developments should not be understood as a loss but rather as a relocation of architectural intelligence from image making to the governance of meaning. It names this capacity as probabilistic architectural agency and outlines its four competencies: steering the generative system with architectural vi criteria, grounding the results in spatial reasoning, assessing the consequences of what an image proposes, and making justified choices when the system's workings remain opaque. The thesis is confined to image generation, not AI in architecture at large, and is largely conceptual rather than predictive.

Subject Keywords

Architectural representation, Architectural intelligence, AI image generation, Probabilistic architectural agency, Answerability

URI

https://hdl.handle.net/11511/121028

Collections

Graduate School of Natural and Applied Sciences, Thesis