Table of Contents:
Chepkemboi J. Mang'ira has spent a decade doing something most AI models still can't: telling a Samburu bead pattern from a Turkana one. As the founder of OwnYourCulture, she has built a practice out of treating East African body adornment not as costume or craft but as a design language with its own grammar — one that shifts generation to generation, community to community, and rarely survives translation into the databases now training the next wave of image generators.
Chepkemboi J. Mang’ira — known online as @missvavavum — is a Nairobi-based fashion artist, researcher, and trained journalist working at the intersection of ancestral fashion systems, digital storytelling, and cultural research grounded in East African knowledge systems. Her practice treats adornment as a living archive: a material culture through which identity, memory, and ancestral knowledge are carried, preserved, and transformed. She develops that inquiry across research, curatorial work, lectures, visual storytelling, and archival investigation, positioning fashion not as a trend cycle but as a system of knowledge production and historical continuity.
She is the founder of OwnYourCulture (OYC), a platform for research, learning experiences, and digital community building exploring ancestral fashion and African design heritage, and she engages both local and international audiences through it — including residencies, lectures, publications, and her role as a co-curator with the Global Fashioning Assembly, where she contributes to broader conversations on diverse knowledge systems. She is also a contributing author to Rethinking Fashion Globalization (Bloomsbury). Alongside her research practice, she brings a background in fashion entrepreneurship across Africa and the United States, giving her hands-on experience with cultural production, digital infrastructure, logistics, retail systems, and the operational realities of building sustainable fashion ecosystems.
Chepkemboi has spent a decade doing something most AI models still can’t: telling a Samburu bead pattern from a Turkana one. She has built her practice out of treating East African body adornment not as costume or craft but as an artistic language with its own grammar — one that shifts generation to generation, community to community, and rarely survives translation into the databases now training the next wave of image generators.
That gap is the center of this conversation. Ask a popular model for “Kenyan beadwork” and it will confidently invent a fabric — Chepkemboi has watched one attribute a nonexistent textile to the Kisii, a community known for stone carving, not weaving. Ask it about Vlisco, and it will file a Dutch print house’s fabric under African heritage without blinking. These aren’t edge cases; they’re the default, and they raise a question ADA keeps returning to from different angles: when a system is trained on documentation of a living tradition, who is the documentation for, and who gets to correct it when it’s wrong.
Chepkemboi’s answer isn’t a rejection of the technology so much as a specification for what would make it trustworthy — community custodians, artisans, elders, and AI developers working from the same table, with naming and classification treated as the first ethical decision rather than an afterthought. Below, she traces that argument from her grandmother’s earlobes to the Maasai jewelry court cases to the murky question of who owns a pattern once it’s been fed into a model and generated an ocean away.
OwnYourCulture has been a decade-long act of documentation — beading, textile, ornamentation, the visual vocabulary of Kenyan ancestral fashion. What first drew you to this work, and how has your thinking evolved since you started? And more broadly, how has the internet — social media, digital communities, the rise of platforms like yours — shifted the way African fashion and design traditions are seen, shared, and understood?
At the time I began OwnYourCulture, I had been working in the fashion, tech and music industry for about five years. I was involved in forums and conversations at the time about the Kenyan fashion identity, and I was in Facebook groups centred on embracing natural hair. I was a fashion blogger, and also creative directing at a tech-fashion startup — my work involved the overall aesthetic of the platform as well as onboarding designers. During the day I spent time with brilliant makers across the city, cataloguing products, listening to their stories, while once a week I was at a forum with fashion stakeholders trying to formulate better systems in the industry. The Kenyan design aesthetic debate has been a quagmire since independence and we still couldn’t solve it.
“Body adornment was a knowledge system, and it still is today.”
This led me to my own research, starting with my maternal grandmother, who had beautiful elongated earlobes. I was curious about what type of earrings she wore. That curiosity led me to museum image archives and books on the people of Kenya, and jewellery seemed to be the common link across communities — we all had such unique ways of designing and expressing it. Then I got curious about what stories others knew of their heritage, so I created the hashtag #ownyourculture to invite others’ stories.
My thinking since then has expanded — it’s no longer so much about the objects alone, but the people, the stories, and the design language I began to see. Within pastoralist communities who still maintain their design heritage, beadwork styles from five years ago aren’t the same styles made today. Before doing this research, I didn’t know how aesthetically rich every community was.
I think the rise of afro-centred visual archive platforms has changed how we see and express ourselves — the amasunzu hairstyle from Rwanda has made a resurgence, and I’ve seen stylists on Pinterest recreating hairstyles exactly as Maasai morans wore them, styles that had long since ceased to exist. More and more designers now are translating their heritage into fashion. These platforms have created a demand, supply, and interpretation of our heritage in a more visceral way. They’ve also opened up questions about how African history has been interpreted, by whom, and how accurately. Above all, I think the most valuable thing has been witnessing and implementing our rich heritage into day-to-day life.
As generative AI reshapes how we create — in fashion, textile, and design — where do you see African design fitting into that landscape? Is there a genuine place for African visual traditions in these systems, and if so, what would it look like when done right?
I believe there is a place for it because we are so rich in visual aesthetics, literature, and language — each serving as a vast, untapped knowledge resource, and there are so many nuances given the diverse peoples who live here. Our names and greetings alone are just a tiny fraction of what a designer would see — that alone can be a dataset.
I think with ethical practice, AI can make so much of our information available to more of us, because not everyone has the patience to sift through physical archives. The work of Delphine Diallo, who uses AI to reimagine African worlds and mythology, is an example of what’s possible.
“African culture goes beyond aesthetics — there is knowledge, spirit, real people, and history woven into everything.”
I think there’s space for collaboration with community elders, our grandparents, archaeologists, artisans, artists, and tech experts to make AI that is grounded in our history. When done right, it would look like involving community custodians and local experts to create a database that is accurate, respectful, and treats heritage as the knowledge system that it is.
Have you experimented with generative AI tools — image generators, pattern tools — in your own work? What did you find?
” Afro-centred AI developments led me to American institutions and their work — never local or African creations, unless I kept pushing for fine-tuning. Other popular generative AI tools kept coming up short on East African ancestral fashion, or pulled generic, mislabelled images from the internet.”
Honestly, it was hopeless. I used ChatGPT, and it made up its own textiles and then admitted it was incorrect — so my current relationship with it involves a lot of AI self-correction. While developing my latest project, an Africa-centred fashion career guide, I asked it to help narrow down my research, but even when listing well-known African researchers and personalities, it wouldn’t give me the answers I know from the ground. When I fed it information about local fashion experts, the hallucinated results that came back were almost comical.
I also noticed that searching for Afro-centred AI developments led me to American institutions and their work — never local or African creations, unless I kept pushing for fine-tuning. Other popular generative AI tools kept coming up short on East African ancestral fashion, or pulled generic, mislabelled images from the internet.
I also tried to use it to find the meanings of names from my grandmother’s generation, but again it made up a generic explanation, when a simple search on Facebook surfaces public pages with that actual information. Honestly, I’d have much better luck walking through the markets in person and taking my own photos. That said, I’ve come across Afrofashion AI and a few other platforms in development with better-trained, Afro-centred datasets — Afrofashion AI seems to understand nuances like Nigerian gele tying and fabric styles, and locally, Undameta, a textile-selling platform, has created AI fashion reels and virtual fabric mock-ups. A lot of these tools are still in developmental stages, but they show real promise.
Most large AI image models are trained predominantly on Western visual data. When you prompt them for "African textiles" or "Kenyan beadwork," what do you get? How culturally coherent is it?
West African textiles which have been widely documented and written about show up fairly accurately but not perfectly, one must double check the details and patterns generated. When it comes to Kenyan beadwork, pastoralist communities are the first images that show up but are not fully accurate and it doesn’t capture the nuances in colour differences or design aesthetic and a lot of mislabelling. Once, during one of my research voyages, it made up its own fabric, called it the soap-stone fabric of the Kisii community when in reality, this community are just excellent carvers. It also misnames the origins of Kenyan fabrics. My current relationship with AI consists of a lot of me correcting it or questioning its information sources.
There is growing work on documenting African design systems — capturing the underlying principles of African visual traditions for use in digital contexts. Do you see that as preservation, or could it also be a form of extraction?
This is a question I like to answer alongside community custodians, because there’s so much that goes into it — who is sharing, the accuracy of naming or description, and what value the original community actually receives. Personally, when I read up on and see visuals of design traditions from different parts of Africa, I feel proud, and in awe that we get to inherit this beauty. Discovering how our ancestors styled their hair helped me love and work with my own hair. Seeing archival images of ancestral fashion across East Africa helped me develop my own personal style rooted in my heritage, and the same has been true across the OwnYourCulture community, where people have used archival images as inspiration for their traditional ceremonies.
There’s also the real reality of design extinction — designs that live in museums abroad, in rare books, or behind closed doors in our own museums. In those cases, making ancestral design resources accessible becomes genuinely valuable. The difficulty comes when there aren’t custodians alive to tell us the respectful processes behind wearing or making a piece.
“The crux comes when there aren’t custodians alive to tell us the respectful processes… misnaming or misclassifying — this becomes erasure.”
There have been court cases from the Maasai community battling the extraction of their culture by international design houses and advertising corporations — this tension predates AI entirely.
If an AI system is trained on documented ancestral Kenyan patterns, and a designer in London uses it to generate a textile print — who owns that? What does cultural credit look like in this context?
I think there has to be a framework that is owned, or at least designed, by the originating community or country. In this case, the pattern belongs to the Kenyan originator, but we currently have systems where data-feeding and financial exchange end up equalling ownership. That’s where it gets murky, and we have to go back to first principles and redefine design and copyright.
This exact pattern has been playing out across the global fashion industry for years — a major fashion house “inspired” by Maasai jewellery producing and selling it, with ongoing court cases addressing it, long before AI entered the picture.
“Some ancestral designs have communal symbolism, spiritual meaning, or family heritage — it’s not just aesthetic vibes. The framework must be a multi-stakeholder participation.”
We have to remember that some ancestral designs carry communal symbolism, spiritual meaning, or family heritage — it isn’t aesthetic vibes alone, which is exactly why any framework needs multi-stakeholder participation: community level, AI developers, and copyright law together.
What would a responsible, community-led approach to training AI on African textile traditions actually look like in practice? What's your vision for it?
Personally, the most important thing is accurate and respectful recording and classification. This has been fashion and art for centuries — it isn’t always reductive costume or craft. I still believe it has to be a multi-stakeholder collaboration: the community, the makers, the museums, the AI developers, and the lawyers, all at the table together. Respectful naming and labelling is essential for this to work, along with holding the nuanced context of each tradition.
