INT. LECTURE HALL — EVENING
A slide glows behind the LECTURER: a demolition notice, stamped over the facade of an old building.
LECTURER
We found another interesting project to work on, a building slated for demolition. We thought about why buildings like that should be demolished. We wanted to save it. There's a regulation, it turns out: if you can prove the community has an emotional attachment to a building, there's a strong case to save it. So we went in and tried to capture the essence of the building, the parts we felt weren't captured in the official documents. We set up sensors. Sound recording. Movement-capturing sensors. We collected data. Then we took the sound and made a track from it, and visualised it, to show the building is alive.
(She advances the slide. The room leans in, a visualisation of moving parts, luminous, unmistakably alive.)
LECTURER (CONT'D)
We thought, if we could turn this into music, we could connect people to this building. Heritage buildings are so often lost without ever being completely documented. Only a few aspects of them make it into the record. The rest is just... forgotten.
Cut to black.
INT. LECTURE HALL — CONTINUOUS
A new slide: a grid of sky photographs, the same rooftop line, taken years apart.
LECTURER
We wanted to show how data is censored. I was looking at Google Street View images, and I realised how much the neighbourhood had changed in twelve years. I was thinking, how sad that we lost all those land to new buildings? And then I noticed the blue sky and how much the built environment has taken away from it. So we gathered photographs of a single site, taken across many years, from different sources, and we processed each one to count how many of its pixels were blue.
(Beat.)
LECTURER (CONT'D)
We fed that sequence into a script that mapped the numbers to sound, each frame produced a tone, and the whole run of years became a piece of music you could sit and listen to. We showed how data can be censored. And why cities aren't taking that into account in their decision-making.
The music plays.
If these don't trouble you, we are collectively in trouble!
These were two excerpts from a lecture titled, absent data, data politics, censorship, representation. Everything in the diagnosis of these works is correct. Cities do omit. Heritage is selectively documented. Evidence is a currency and some people are issued it and some aren't. It's in the framing of this diagnosis and the response to it that these works become paradoxical and in effect destructive!
Absence is a claim about a complete set
Start with the phrase itself. Absent data. Absent from what?
The word only works if there is a set the data should have been in and isn't. To say the sky measurements are "missing" is to say that a proper record of the world exists, that the sky is a legitimate entry in it, and that the entry has not been filled in. Now put that next to the premise the same lecture opened with; that data is never neutral, that measurement encodes the worldview of the measurer, that what you count depends on who you are and what you are trying to see.
That premise, taken seriously, says there is no complete set. There is no total record from which things could go missing, because every record is a cut through an infinitely describable world, made by someone with a particular interest at a particular moment. By that logic, everything is absent data. Absence is not inherently the mark of a deliberate censorship. It is simply the nature of what a record is, an inevitable condition of measuring anything at all, not a conspiracy to be uncovered.
In a way, these works, while presenting themselves as resistance, are not actually opposed to the belief that by collecting, observing, and datafying the world we arrive at a fuller understanding of it, and therefore better control over it. The lecture's demand, to collect more and collect differently, is a legitimate one. But it is a demand addressed to the datafied state, in its own language, seeking fuller enrolment in it. It is not opposition. It is a feature request, ambitious, sincere, well-designed, filed from deep inside the very system it claims to resist.
The distinction matters because the two positions imply entirely different politics. One says, the measuring apparatus is fine, we are merely underrepresented in it. The other says, the apparatus itself decides what counts as real, and that is the problem. The first is answered by more sensors. The second is not answered by sensors at all. Very often than not, similar to this lecture, the rhetoric of the second is used along with the method of the first. It has become genuinely hard to notice, because the vocabulary of critique and the vocabulary of coverage have grown so alike that you can slide between them mid-sentence.
The Reduction Happens At The Counting, Not At The Rendering
Follow the pipeline. A photograph goes in. A rule decides which pixels are blue and which are not, a threshold, a boundary drawn somewhere in colour space. Every dusk, every polluted afternoon near that boundary gets sorted to one side or the other by a decision no one will ever see or audit. The blue pixels are counted. The sky, on that day, at that site, is now an integer.
That is the moment. That is where the sky stops being a sky and becomes a value. Everything that follows, the mapping to pitch, the choice of timbre, the arrangement into a run of years, is applied to the integer. Not to the sky. The sky left the building at the counting stage.
So when the output arrives as sound rather than as a bar chart, what has changed? The presentation layer. The reduction that a critique of datafication is supposed to object to, the collapse of an unrepeatable phenomenon into a digit that can be compared, ranked, summed, and optimised, has already fully occurred before any aesthetic decision is made. A number can be rendered as a bar, a colour, a tone, a vibration in your chair. It remains a number.
"I turned it into music" is not an escape. Music made from a count is a count. It is not a different form of knowledge; it is a different interface to the same form of knowledge. If your claim is that there are ways of knowing a place that measurement cannot reach, and I believe that claim is true, then the demonstration of it cannot be a thing built out of digital measurements.
You Cannot Oppose A Power With Its Own Tools
This is where Marshall McLuhan becomes unavoidable. His and many media ecologists' argument was people build technologies to serve a design intent, but the real cultural consequences, what McLuhan called the technology's 'message', surface later, often decades later, and usually without our noticing. The message washes over us and changes us before we are aware of it. "We shape our tools, and thereafter, our tools shape us."
The consequential thing about a technology is never the content you pour into it. It is the technology itself, the habits of perception it installs, the kinds of relations it makes normal, the definition of 'real' it quietly enforces. Content is negotiable; the medium is not. Which is why using a medium critically, against the grain, for the community changes nothing at the level where the medium operates. The sensor array in the heritage building carries the same message whether it is installed by a developer or by an artist trying to stop the developer: this building is the kind of thing sensors can know.
What Would Actually Be Different
The same structure is now everywhere. AI's core claim, not the marketing, the claim implicit in the enterprise itself, is that phenomena are reducible to computation. That the world is the kind of thing that can be represented, compressed, predicted, and generated.
AI for humanitarian response. AI for the archive. AI for endangered languages. AI as an artistic medium, used critically, used to expose the model's own biases in a gallery. Every one of these reinforce the core claim. They dispute who benefits. They do not, cannot, dispute that the domain is computable, because their existence is a demonstration that it is. A model that documents a dying language, however tenderly, has maintained that the language is the kind of thing a model can hold. That claim is not in the values layer, where good intentions live. It is in the tool.
Real opposition would not be a better use of tools, not a better dataset, a warmer interface, a more communal sensor. Real opposition is finding a new way of understanding the world beyond the digital. That is a philosophical question, not a technological one, and no innovative use of a tool can answer it. Anything short of that does not change the landscape. It only redistributes power within the same idea of what is real. And if that idea goes unquestioned while everyone congratulates themselves on resisting it, then we might be missing out on the greatest alternatives altogether.
