From Algorithmic Design to Generative AI: Rethinking the Locus of Generative Logic in Architectural Design

سال انتشار: 1405
نوع سند: مقاله کنفرانسی
زبان: انگلیسی
مشاهده: 41

فایل این مقاله در 12 صفحه با فرمت PDF قابل دریافت می باشد

استخراج به نرم افزارهای پژوهشی:

لینک ثابت به این مقاله:

شناسه ملی سند علمی:

ICAICS02_020

تاریخ نمایه سازی: 3 آبان 1405

چکیده مقاله:

The integration of computational technologies into architectural design has transformed not only the tools of representation and drafting but also the ways in which design problems are formulated and design alternatives are generated. Within this development, algorithmic design based on explicit rules, parametric design based on relationships and dependencies, and learning-based generative systems represent distinct approaches to organizing and producing design alternatives. These approaches, however, differ not merely in their underlying technologies or modes of output generation, but also in how and where generative logic is formed and represented. With the emergence of generative artificial intelligence (AI), part of this logic shifts from explicit structures defined by the designer toward representations learned from data during model training. This raises a central question: How does generative AI alter the computational logic of design generation in architecture in relation to algorithmic and parametric approaches?This paper addresses this question through an analytical review, examining the transition in terms of the origin and representation of generative logic, its relation to the design space, and the architect’s role in the generative process. The analysis shows that a key distinction among these approaches lies in the locus and mode of formation of generative logic. In algorithmic design, generative rules are primarily defined explicitly; in parametric design, generative logic is organized through parameters, relationships, and dependencies; and in learning-based generative models, part of the patterns underlying generation is learned from data through model training. This shift also changes the architect’s role, from directly defining generative mechanisms toward problem formulation, process guidance, output evaluation, and selection. Generative AI, therefore, should not be understood merely as a new tool for producing architectural forms, but as a shift in the relationship between the designer, generative logic, and the computational system through which architectural design is produced.

نویسندگان

Alireza Parvizi

Faculty of Architectural engineering, Sepehr Danesh Moaser Institute of Higher Education, Isfahn, Iran.

Sepideh Nasresfahani

Department of Pure Mathematics,Faculty of Mathematics and Statistics, University of Isfahan, Isfahan, Iran.