Drafting images: The making of AI architecture

To envision, develop, and visualize urban realities that have not (yet) been built, professionals in urban future-making draw on tools and technologies: from traditional analogue methods such as paper and pen or scale models to digital solutions including computer-aided design and virtual reality technologies. With the advent of image-generative AI tools, the ability to create visual representations of unbuilt spaces and buildings is undergoing significant transformation: These tools allow for the creation of photorealistic and at the same time speculative images early in the design process – well before elaborated plans have been developed.

While this technological shift opens up new opportunities, it also raises questions: How do AI tools shape the imagination, negotiation, and creation of urban futures? What architectures are generated when using AI tools? Who holds the power in designing AI-drafted imaginaries? This research project investigates from a sociological perspective how emerging AI-generative visualization technologies are being used in the field of urban future-making. The methodological approach is grounded in actor-network theory (ANT) and science and technology studies (STS), allowing for an in-depth analysis of the sociotechnical processes underpinning the application of image-generative AI tools.

Context

Professionals in architecture and urban design have always been confronted with the challenge of visualizing (possible) urban futures, both to support design processes and to communicate their visions to stakeholders. Images serve as an eloquent medium for both purposes, which is why the discipline of architecture has always been closely linked to imaging processes.¹ Historically, this began with drawings and paintings rendered by hand; in the 21st century, however, digital visualizations, ranging from CAD models to virtual reality, have come to dominate the field.

The emergence of visual AI tools, so-called generative diffusion models, mark a significant transformation in architectural visualization. This development raises questions for the profession of architecture while reflecting sociotechnical processes that encompass both technological advancement and societal change.² As images play a pivotal role in social processes in general and in the production of the future in particular, the creation and dissemination of AI-generated images is crucial for architecture and society, ultimately influencing how future urban spaces are envisioned and performed.

Aims

This study seeks to explore the ongoing integration of image-generative AI tools into the practices of architects, with two main objectives: (1) to investigate how these tools are used in urban future-making to address the need for urban transformation and thereby reshape professional design practices, planning processes, and their outcomes; and (2) to explore the use and effects of AI-generated images in the negotiation, imagination, and creation of future urban spaces. The goal is to show how generative diffusion models act as non-human actors in the creation of collective urban imaginaries and realities.

This research contributes to the fields of techno-sociology, STS, urban studies, and architectural theory, ultimately aiming to deepen our understanding of how digital technologies shape the processes and outcomes of architectural and urban design practices, with significant implications for how we imagine and build the cities of the future.

Research design

Central to the project are ethnographic observations and qualitative interviews conducted within architectural or urban design studios that employ visual AI tools. This approach is methodologically grounded on ‘architectural ethnography’ as described by Albena Yaneva.³ To account for the mediating role of photorealistic images, the study also examines how these visualizations circulate beyond the architectural studio, sparking negotiation in broader arenas of urban future-making, such as public discourses, political spheres, or local initiatives. This includes analysing both the reception of images created by professionals and the contributions from non-professionals who use visual AI tools for imagining urban futures.

Focusing on a specific case study allows for an in-depth analysis of the actors, practices, and power relations that structure the creation of urban imaginaries. The methodological framework is grounded in actor-network theory (ANT) and science and technology studies (STS). Following core STS premises, the study assumes that while images are always highly constructed, visual practices and artefacts profoundly shape both our world and the way in which we perceive it.

Supervisor: