Digital Architecture and Generative Artificial Intelligence: Toward a Tri-Variable Theoretical Framework for Evaluating Visual Outputs in Architecture

Abstract

The field of architecture has undergone a significant transformation with the advent of generative AI tools like Midjourney and Stable Diffusion which enable the rapid production of high-resolution architectural images. While this technological leap opened new horizons, it also raised a fundamental question: how can these visual outputs be elevated to ensure their quality and support for the creative process? Previous literature addressed these aspects separately, technical studies mainly focused on efficiency and aesthetics while others explored simulation and performance. Some writings emphasized cultural identity or the architect’s role in its creativity, this divergence created a knowledge gap and the absence of an integrated framework that unites all of these coherent aspects into one optimal product. This paper aims to bridge this gap by proposing a conceptual framework linking three main variables: -The technical dimension -The cultural dimensions -The creative dimension. The methodology relies on a review following the historical progression from Computer-Aided Design (CAD) to generative AI, categorizing studies based on their focus. From this analysis the three variables were identified and their relationship formulated as an interconnected system where each aspect complements the others and addresses their biases. The paper concludes that its scientific contribution lies in framing the relationship between these variables so that visual evaluation in generative architecture becomes a system reflecting quality, identity and creativity. This framework also paves the way for future applied studies to test these indicators practically and develop them as reliable evaluation tools.