Afrointrospection is the category in my work that feels most personal and most urgent. It is where the technical questions about AI and image generation intersect with questions about identity, representation, history, and the right to be seen clearly — on your own terms, in your own complexity.

For as long as visual culture has existed in the forms we currently recognise, the representation of Black people — African, Caribbean, Afro-diasporic — has been subject to forces that flattened, distorted, exoticised, or erased. Mainstream visual media has improved in some respects over recent decades, but the defaults remain. The hero is usually this. Beauty usually looks like that. Power, elegance, and aspiration carry particular visual signatures that have historically not included the fullness of Black experience.

AI image models inherit and replicate these defaults unless actively resisted. Working with them in the Afrointrospection mode means exactly that — active, intentional resistance, combined with the use of the tools to generate images that would otherwise be inaccessible.

What the Tool Defaults To

When you write a prompt describing a Black subject without careful specificity, current AI image models tend to produce images that reflect a narrow, often Western-mediated visual vocabulary of Blackness. The faces, the styling, the aesthetic register — these tend toward what has been most visible in the training data, which reflects the biases of the internet and the media industries that shaped it.

This is not a reason to abandon the tools. It is a reason to approach them with the level of specificity and intentionality that the work demands. When I write prompts in this category, I am working with particular visual references: West African textile traditions, Yoruba sculptural aesthetics, Afrofuturist visual languages developed by artists like Wangechi Mutu and Kehinde Wiley, the bold colour vocabularies of the Harlem Renaissance. I am naming specific cultural contexts, not generic ones. And I am iterating — sometimes extensively — until the output begins to approach the actual vision.

AI as Reclamation Tool

There is something genuinely radical in the possibility that AI tools, for all their biases, can be used to generate images of Black excellence, Black interiority, Black beauty, and Black complexity at a scale and with an accessibility that was previously impossible. Not as a replacement for the work of Black photographers, painters, and filmmakers — that work is irreplaceable, irreducible, and essential. But as an additional instrument.

The Afrointrospection images in this exhibition are attempts to use the tool's generative capacity to ask visual questions I could not ask otherwise. What does regal look like, grounded in a specific West African aesthetic tradition rather than a European one? What does melancholy look like on a Black face rendered with the visual language of Baroque painting? What does belonging look like in a city that was not built for you?

These are not new questions. Black artists have been asking them for as long as there have been Black artists. What is new is having access to a tool that can generate visual responses at speed — a sketch pad that operates at scale, for a set of visual questions that matter.

The Limits and the Dangers

The tool can also get it wrong in ways that are harmful rather than merely inaccurate. AI models trained on data that includes racist imagery, exoticising photography, and misrepresentative visual media can reproduce those patterns in ways that are painful, reductive, and damaging. This requires vigilance — a willingness to reject outputs that replicate the visual violence that the work is intended to counter, and to push further until something closer to truth emerges.

There is also the danger of a false sense of resolution. Generating beautiful images of Black subjects with AI does not address the structural inequalities in the creative industries, the underrepresentation of Black artists in galleries and media, or the economic conditions that shape who gets to make art and who gets paid for it. The images are meaningful. They are not sufficient.

Cultural Knowledge as Prerequisite

The most important thing I can say to anyone who wants to use AI tools to explore African or Afro-diasporic visual culture: the tool cannot give you the knowledge. You have to bring it.

If you are working with visual traditions you do not understand deeply — their history, their symbolic vocabulary, their social context — AI will not compensate for that gap. It will generate surface aesthetics without the meaning that makes those aesthetics significant. It will produce masks without understanding what masks signify, textiles without knowing what their patterns communicate, faces without the cultural context that makes them specific rather than generic.

The depth of the output is limited by the depth of the input. And in this category of work, the input includes knowledge, respect, and a relationship to the traditions being engaged with — things that come from the person making the work, not from the machine.

An Ongoing Practice

Afrointrospection is not a category I consider finished. It is an ongoing investigation — of the tool, of the traditions it engages with, and of my own relationship to both. What I have made so far is early work in what I expect to be a long conversation.

The images in this exhibition are invitations to look — at the people in them, at the traditions they carry, at the questions they raise about who gets to be rendered, who gets to render them, and what is gained or lost in the space between a human vision and a machine's execution of it.

The mirror is imperfect. All mirrors are. The point is not the mirror's perfection — it is what you choose to look at in it, and what you do with what you see.