What the Algorithm Doesn't Know About You

The archive is not a neutral technology. It has never been. Every decision about what enters the historical record — what is preserved, what is lost, what is deemed significant enough to keep — is an act of power dressed as administration. We have always known this about institutions. We are only beginning to understand it about ourselves.

There is a photograph I have kept for eleven years. I could not tell you precisely why. It is not technically accomplished. It documents nothing historically significant. It is a specific quality of light on a specific afternoon, and something about the way the shadows fall makes me return to it in a way I have never fully been able to explain. Every major technology platform that holds my data has registered that I kept it. None of them know why. This is the condition I keep returning to — not as a technical problem but as a philosophical one. The gap between what people do and what they mean. Between the data and the person it came from. Between the surface of a life, captured with extraordinary precision, and its interior, which remains entirely elsewhere.

We are the most documented generation in history. We are also, in a specific and consequential way, the least understood by the systems documenting us. The AI being built to know us — to predict us, serve us, reflect us back to ourselves — is trained almost entirely on behavioural residue. What we clicked. Where we paused. The metadata of our attention. It is a system of enormous technical sophistication trained on the least meaningful data available. It knows what people did. It has no access to what they meant.

I came to understand this through curating. The act of building an exhibition — deciding which works to place in relation to which others, which histories to hold up and which silences to name — is not only an intellectual act. It leaks. Every curatorial decision is simultaneously a self-portrait. The choices I have made in practice have told a story about me that I did not consciously choose to tell. The archive I was constructing was also confessing me.

When I was curating Nigeria Imaginary for the Nigeria Pavilion at the 60th Venice Biennale, this became impossible to ignore. The exhibition was born from a fragmented understanding of my own Nigerian-ness — assembled from partial memories, stories told across distance, attachments to an archive that was never fully mine to begin with. I was born in England. Nigeria, for me, has always been partly imaginary: constructed from what was passed down, what was absent, what I reached toward across the gap of diaspora. I was not representing Nigeria. I was curating my relationship to a place I had always known through fragments. And in recording the voices of Nigerians for the Incubator Project — asking them what Nigeria sounds like, tastes like, feels like in the body — I was doing something I didn't yet have a name for. I was generating intentional emotional data. I was asking people to curate their relationship to their own experience consciously, on record, as an act of witness.

That is categorically different from what surveillance captures. Not a better version of the same thing — a different thing entirely.

When someone is asked which twelve photographs from a year of their life they would keep — and they choose, and they record in their own voice why they chose — the signal that generates is not behavioural residue. It is the record of intention. It carries the grammar of how a specific person relates to their own experience. The attachments they could not explain. The corners of memory they return to without knowing why. The quiet decisions about what to hold and what to release. AI trained on that material would not just know more about the person. It would know differently. It would learn not what people do but what they mean.

Saidiya Hartman developed critical fabulation as a method for restoring interiority to those the colonial archive had reduced to administrative data — imaginatively reconstructing the inner life that institutional record-keeping had suppressed or erased. We are now living inside its inversion. Algorithmic systems are performing their own critical fabulation on all of us: constructing emotional selves, inferring psychological interiority, fabricating the inner life from our data traces — without our authorship, without our consent, and in whose interest we are rarely permitted to ask.

The question I am sitting with is not how to stop this. It is what becomes possible when it is reauthored. When the act of self-curation is conscious rather than incidental. When the training data is intentional rather than residual. When the person whose life is being read is also the person deciding what gets read — and when a machine learning from that consciousness produces something that has never existed before: an intelligence that knows not just what you did, but what you meant. Not just the surface of a life, but its interior logic.

I do not yet know what that intelligence looks like. I do not know what it would take to build it responsibly. These are the questions I am sitting with. They are not yet resolved.

What I do know is that the photograph I have kept for eleven years is not a data point. It is evidence of something. And whatever that something is — the feeling it preserves, the self it reflects, the meaning it holds that I have never quite been able to articulate — it is exactly what the algorithm doesn't know about me.

And it is exactly what I think it should learn.