AI, Architecture and Indigenous Design Thinking

From servant to collaborator: when AI-generated imagery first began gaining attention, it could be viewed as a technologically advanced assistant. Now, rather than simply fulfilling instructions, AI can become a tool for dialogic exploration.

During the early 2000s, as an architecture student in post-apartheid South Africa, I was profoundly shaped by a growing discourse around what architecture ought to be in a society undergoing radical transformation. This period was marked by a surge of interest in how the built environment could reflect African cultural heritage and respond to the shifting sociopolitical realities of the time. Prior to enrolling at the University of the Witwatersrand (Wits), I watched an episode of Carte Blanche featuring architect Fanuel Motsepe, who spoke compellingly about integrating African heritage into Johannesburg’s urban fabric. The conversation extended to students from Wits, who discussed reimagining the city’s architecture to reflect the lived realities of its changing population. This moment revealed the potential of architecture not merely as shelter, but as a medium for cultural expression and historical redress.

Later, while studying at the Wits School of Architecture and Planning, I encountered Peter Rich in my second year. At the time, Rich, alongside Professor Anne Fitchett, was working on the Mapungubwe Interpretation Centre, a project that would go on to win international acclaim. Meeting Rich proved to be a pivotal moment in my education. His expressive sketches, eccentric personality and irreverent humour made a lasting impression. More importantly, he introduced me to the idea that African architecture could emerge not from mimicry of Western forms, but through a deep engagement with the continent’s cultural, material and spatial traditions. This approach to architecture, rooted in place, people, and history, reorientated my sense of what it means to design in and for Africa.

Culturally responsive architecture

By the time I commenced my master’s studies, I had developed a clear intention to pursue a deeper investigation into the cultural dimensions of architecture. I was particularly intrigued by the influence of Pancho Guedes, an iconic figure associated with the University of the Witwatersrand School of Architecture and Planning, whose unconventional approach had also shaped the thinking of Peter Rich. Guedes’ fascination with vernacular traditions, particularly his interest in Ndebele architecture, inspired Rich to explore how indigenous cultural practices could inform architectural design.

Following in their footsteps, I decided to focus my research on my own region, the town of Kuruman in South Africa’s Northern Cape Province, and the wealth of indigenous knowledge systems embedded within it. This area is historically home to the Tswana people and earlier Khoe and San communities, whose spatial practices and cosmologies offered profound insights into the human-environment relationship.

Through extended engagement with the community and landscape, I developed a methodology rooted in immersive presence, spending prolonged periods on site, allowing the rhythms, rituals and stories of the place to shape my understanding. This ethnographic and phenomenological approach became central to both my master’s thesis and later doctoral research, enabling me to treat personal experience as legitimate data within a culturally responsive architectural inquiry (Maape, 2017).

Finding the right tools

Translating these conceptual explorations into architectural principles, however, posed a significant challenge, primarily due to the limitations of the tools available to me at the time. Influenced by both Rich and Guedes, I adopted drawing as the foundational medium of my design practice, supported by the use of narrative and myth as interpretive tools for articulating the intricate relationship between human beings and their environments. This methodological framework was further shaped by indigenous epistemologies, particularly those embedded in the visual traditions of the Khoe and San ancestors from my hometown.

Their rock art and engravings revealed a worldview not grounded in objectivity or representation in the Western sense, but rather in experiential and cosmological encounters with place. These drawings, often created during trance states or as part of broader ritual practices, offered insights into the embodied, symbolic nature of spatial engagement. A similar conceptual approach is evident in one of Peter Rich’s mythological drawings of Ndebele architecture, where visual expression served as a medium for engaging with cultural memory and spiritual cosmology.

These indigenous and interpretive traditions became critical to my own architectural inquiry, prompting the development of drawings that sought to represent not just form, but the deeper relational dynamics between people and land. In this way, drawing became both an analytical and generative tool, a bridge between indigenous knowledge systems and architectural design thinking.

My early experimentation involved combining traditional ink sketches with digital technologies (Maape, 2021). One such method was the integration of 3D models created through photogrammetry, produced in collaboration with Stephen Wessels from the University of Cape Town’s School of Architecture, Planning and Geomatics. This process enabled a critical reflection on the tools we were employing, ultimately revealing the embedded assumptions inherent in their design and use (Wessels, Maape, Schoville et al., 2023). One case study that highlighted these limitations involved a ritual site, a dark, secluded cave with rich cultural significance. This space is not merely a geological formation but a site steeped in myth and initiation rites. Local narratives describe it as the dwelling place of a giant snake, and it serves as a rite of passage for young initiates who must confront its darkness and silence as part of their transformation into adulthood.

However, photogrammetry, as a technology dependent on light for data capture, produced a visual rendering of the cave that was entirely antithetical to its cultural meaning. The resulting 3D model illuminated the cave in ways that stripped it of its ritual and mythological dimensions. The technology, by virtue of its operational logic, failed to engage with the affective and symbolic experience of the space as understood within local epistemologies.

This example underscores the extent to which digital technologies, often perceived as objective or neutral, are in fact shaped by specific cultural, perceptual, and epistemological frameworks. In this case, the inability of photogrammetry to represent darkness as a meaningful spatial and symbolic condition revealed the disjunction between Western technological paradigms and indigenous understandings of place.

These reflections led us to an important realisation: digital technology is not culturally or politically neutral. Even before widespread discussions around AI bias gained traction, it was already evident to us that the development and function of these technologies were shaped by the cultural assumptions and epistemologies of their creators.

The implications of this are profound, not only for AI, but for a wide array of architectural tools such as CAD, Revit and other digital design software. These tools are often perceived as purely scientific or mathematical in nature, grounded in objectivity and computational precision. Yet, this perception conceals the extent to which they are embedded within particular worldviews, shaped by the sociopolitical and cultural contexts from which they emerge. To treat them as neutral is to ignore the ways in which they mediate and constrain architectural thought and expression (Buolamwini, 2023).

AI and African design

These experiences deepened my awareness of the limitations embedded in digital technologies developed within the Global North, particularly when such tools are applied in socio-cultural contexts beyond Western capitalist frameworks. When AI-generated imagery first began gaining attention approximately two years ago, I observed the early outputs with curiosity but struggled to identify immediate applicability for my own design practice. Nonetheless, the core premise, that one could type a phrase, referred to as a “prompt,” and generate a corresponding image, was conceptually compelling. The resulting visuals were often highly rendered and appeared to reflect, at least superficially, the intent of the prompter.

In 2023, I gained access to OpenAI’s DALL·E, an early yet powerful generative image model. I approached it with optimism, hopeful that I could leverage its capabilities to visually manifest the African-inspired architectural ideas I had long been developing. My intention was to utilise language, one of my strengths, to articulate design visions rooted in indigenous knowledge systems. While I cannot recall the precise wording of my first prompt, it likely resembled something along the lines of: “A futuristic African architectural design inspired by indigenous African knowledge.”

As one might expect, the initial results were far from satisfactory. The image generated was, for lack of a better term, superficial, lacking cultural specificity and composed of a disjointed amalgamation of African and arguably non-African architectural motifs. This encounter quickly revealed the limitations of using simple textual prompts when attempting to extract nuanced cultural and spatial imaginaries from AI systems. While some outputs began to exhibit a degree of aesthetic intrigue, they rarely aligned with the architectural visions I had hoped to generate. It was only when newer iterations of these tools enabled multimodal prompts, incorporating images alongside text, that the outputs began to resonate more closely with my intentions. This led to a critical realisation.

Over the course of several weeks and months, I had approached the AI as a kind of technologically advanced assistant, an obedient executor of instructions capable of interpreting and rendering my mental image of what “African architecture” ought to be. I had assumed that if I could clearly articulate this vision through language, the AI would reproduce it with fidelity. However, by introducing my own visual materials, images which were themselves open-ended, reflective and exploratory, the AI responded not with passive execution, but with a kind of interpretive engagement. This marked a fundamental shift in my understanding of the AI: from servant to collaborator. Rather than simply fulfilling instructions, the AI became a tool for dialogic exploration, capable of interrogating the very assumptions I held about identity, representation, and architectural form.

Considering the immense computational resources required to operate AI systems, both in terms of energy consumption and financial cost, it is worth questioning whether their optimal use lies merely in automating tasks we are reluctant to perform ourselves. It became increasingly clear to me that AI could serve a much deeper purpose: not as a servant to carry out instructions, but as a collaborator capable of unlocking new modes of thinking and creative exploration.

A significant breakthrough occurred when I uploaded an image of Pancho Guedes’ iconic building, the Smiling Lion, into an AI image-generation platform. My aim was to see how the system would interpret this architecture and whether it could extrapolate or generate similar designs. To my surprise, the AI reproduced the essence of Guedes’ architectural language, not by mimicking superficial stylistic details, but by engaging with the underlying formal logic. It identified the use of rhythmic vertical profiles or silhouettes that Guedes employed to introduce a sense of seriality and sculptural articulation. These elements, typically rendered in 2D section drawings, speak to a deeper design strategy embedded in the work.

What struck me most, however, was the AI’s decision to omit the decorative and symbolic motifs often featured on the building’s facade, motifs frequently interpreted as the hallmarks of its “Africanness”. Instead, the AI retained the sculpted silhouette as the primary formal gesture. This was revealing. It suggested that the AI, unencumbered by the cultural conditioning many of us inherited from architectural training at institutions like Wits, had bypassed the expected signifiers of identity. It perceived something else entirely, something deeper and perhaps more architecturally essential.

This moment marked a shift in my understanding of AI as a design tool. No longer did I see it as a passive executor of visual instructions, but as a potent analytical instrument capable of deconstructing preconceived notions and reframing architectural meaning. In this context, AI can operate at an ontological level, helping designers transcend inherited paradigms and engage with the world in radically new ways.

Deconstructing assumptions

Ultimately, my engagement with AI transformed into a method of deconstructing my own assumptions. Initially, my ambition to define a new modern African architecture was grounded in a set of unexamined beliefs: that there is a singular notion of “African” architecture, that certain precedents could serve as authentic representations of it, and that such a concept could be clearly and consistently articulated.

Through my iterative encounters with AI, and by integrating it with more traditional modalities of sketching, sculpting and embodied research, I began to view the design process less as a quest for definitive answers and more as a space of open-ended inquiry. Rather than seeking to define what African architecture is, I have shifted toward exploring what architecture might become when liberated from static cultural categories. This has led me towards what I now describe as a planetary architecture, one that recognises all humans as indigenous inhabitants of Earth, each embedded within diverse but interconnected systems of meaning.

Within this paradigm, AI becomes a powerful agent of speculation: a tool not merely for representation, but for revealing productive tensions, surfacing contradictions and imagining futures that hold space for multiplicity and ambiguity. By embracing this approach, we can begin to transcend inherited limitations in our thinking and design practices. AI, paradoxically, offers us the possibility of becoming more human, of reconnecting with the complexities, contradictions, and creative capacities that define us. In doing so, it may help us reframe not only the idea of “African” architecture, but also the foundations of architectural thought in a rapidly transforming world. www.afreetekture.com

References
Maape, S. (2017). Architecture for resilience: Dialogues with place in the indigenous communities of Kuruman during the Holocene. University of the Witwatersrand.
Maape, S. (2021). Drawing creepy places: Representing liminal ritual spaces in Kuruman, South Africa. Dialectic, 9, 2–12.
Maape, S. (2022). Afreetekture | African Architecture. Retrieved 1 September 1 2022 from https://www.afreetekture.com/
Wessels, S., Maape, S., Schoville, B.J. et al. The Drone, the Snake, and the Crystal: Manifesting Potency in 3D Digital Replicas of Living Heritage and
Archaeological Places. Arch 19, 395–429 (2023). https://doi.org/10.1007/s11759-022-09460-3
Buolamwini, J. (2023). Unmasking AI: My Mission to Protect What Is Human in a World of Machines. New York: Random House.