The check
A check stops. No letter, no phone call, no explanation. Somewhere upstream, a threshold was adjusted — a calculation revised, a category redefined, a percentage recalibrated. The person knows only the effect: the money is not there.
The machinery that produced this outcome is not secret. The formula is published. The adjustment is documented. The regulation is available to anyone with the expertise to read it, the time to find it, and the institutional literacy to understand what it means. But availability is not access. The machinery is opaque — not because it is being concealed, but because its complexity exceeds what a person living paycheck to paycheck can decode.
A camera does not fail to photograph infrared. Infrared is outside what the camera was built to capture. The limitation is in the instrument, not in the world. What the camera cannot see is not what is hidden from it. It is what the camera's way of looking cannot reach.
The same is true of every system that looks. What it cannot see is determined by how it is set up to look — not by what is there.
The gradient
A flower. To the layperson, it is something lovely and nameable — a daisy, a rose, a thing with petals. The botanical classification — the genus, the morphological criteria, the reproductive structure — is inaccessible. To the botanist, the same flower is transparent. Its classification is legible, its features parse into a system of distinctions that the layperson cannot see. Same flower. Different opacity. The difference is not in the flower. It is in who is looking.
The poverty threshold. The administrator who draws the line sees the machinery: the formula, the household size adjustment, the federal percentage that determines eligibility. One hundred thirty percent of the threshold for food assistance. One hundred thirty-eight percent for Medicaid in expansion states. The administrator sees the structure because the administrator built the structure.
The person living at fourteen thousand five hundred dollars sees the cliff: eligible or not. The machinery is invisible — technically public and practically unreachable. The complexity, the institutional context, the regulatory history that produced this particular number — all of it available in principle, all of it opaque in practice.
But there is a third case. The barista with a PhD in economics, working minimum wage, living under the threshold. She reads the apparatus. She knows the formula, the adjustment, the percentage. She also lives under it — feels the cliff, navigates the consequences, experiences the category from the inside. Opacity and transparency coexist. She can see the machinery and be subject to it at the same time.
Opacity is not a switch — visible or invisible, transparent or opaque. It is a gradient. And the gradient tracks access to how a system is set up to look, not position in the system's output. You can be inside the category and still see the machinery. You can be outside the category and see nothing at all.
The instrument
A camera is a way of looking. Its frame determines what is included. Its lens determines what is sharp. Its exposure determines what is bright enough to register. Change any of these — wider frame, different lens, longer exposure — and you change what is visible. But you also change what becomes invisible. A longer exposure captures motion as blur and stillness as clarity. A shorter exposure freezes motion and loses the dark. There is no neutral setting. Every way of looking is also a way of not-looking.
Language works the same way. The constraints of grammar are opaque not because someone conceals them but because the apparatus is complex and historically layered. Centuries of use deposited into structure that most speakers navigate without seeing. The rules are too deeply embedded for most people to access explicitly.
A facial recognition system sees a face as a vector — a set of measurements in a feature space. What it cannot see — intention, context, history — is outside the system's way of looking. The system was never set up to see them. The system is succeeding at exactly what it was built to do. And what it was built to do is narrow.
Changing the instrument gives you a different view with different blindnesses. A medical imaging system that can detect tumors cannot detect grief. A credit scoring algorithm that can assess default risk cannot assess circumstance. Each instrument sees what it is built to see, and each instrument's clarity is purchased by its particular form of blindness.
The demand
Some systems do not simply look. They require that what they look at be legible — classifiable, reducible, mappable in their terms.
Facial recognition requires that a face resolve into a vector. Predictive policing requires that behavior reduce to a risk score. Algorithmic hiring requires that a person compress into a feature set. The system does not ask whether you want to be seen this way. It demands that you be seeable. Legibility is the price of being processed.
But the demand runs one way. The system sees you. You cannot see the system. The method by which you were classified — the training data, the feature weights, the threshold that placed you in one category rather than another — is structurally inaccessible. Not because it is secret in every case, but because understanding it requires expertise, institutional access, and legal standing that most people do not have. The system requires your transparency. It does not offer its own.
This asymmetry is not incidental. It is the architecture. Those who build the system retain access to how it works. Those who live under it do not. The operator sees the machinery. The subject sees only the output — the score, the classification, the decision that arrives without explanation, like a check that stops.
When this asymmetry persists long enough, something else happens. The categories start to look natural. National borders. Racial classifications. Insurance tiers. Gender binaries. Each was drawn — by someone, at some point, for some purpose. None was discovered. But when you cannot see how something was set up, it begins to feel like it was always there. What was imposed begins to feel like geography — just the way things are. Complexity, time, and distance between builders and inhabitants do this work quietly, without announcement.
The right
Edouard Glissant argued that a fully transparent Other is not encountered but absorbed. Transparency, taken to its limit, is a form of consumption. You understand the other completely — and in doing so, you have replaced them with your understanding. You have made them legible on your terms. You have not met them.
Opacity, in Glissant's sense, is the condition of encounter. The resistance of what you cannot fully access is what makes genuine engagement possible. Not ignorance — the productive limit of your way of looking, the place where your categories stop and the other person continues beyond them. The encounter begins precisely where legibility ends.
Secrecy is deliberate concealment — a choice to withhold. Opacity is something else. Opacity is structural. It exists because every way of looking has edges, and what falls beyond those edges is not being withheld. It is simply outside what the instrument can reach.
Two kinds of transparency should not be confused. Institutional transparency — audit the algorithm, examine the training data, publish the methodology, hold power accountable — is necessary. It is a political demand, and a just one. But there is a second sense of transparency: the dream of total access, frictionless comprehension, a world in which nothing resists being seen. This is a metaphysical demand, not a political one. And it treats every opacity as a problem to be solved rather than a condition to be respected.
The error is collapsing these into a single demand — as if the right to examine how an algorithm classifies you is the same thing as the right to make a person fully legible to the state. One is accountability. The other is control.
The edge
Every way of looking is also a way of not-looking. No instrument escapes this. No expansion of scope eliminates it. A wider lens does not see more of the same world — it sees a different world, with its own edges.
The demand for total legibility — total transparency, total access, the elimination of everything that resists being seen — is not neutral. It is a claim about what kind of relationship you are entitled to have with whatever you encounter. It assumes that the correct stance is full comprehension. But full comprehension may be the surest way to miss what you are looking at, because it leaves no room for what your categories cannot hold.
Not everything that resists being seen is hiding. Some of it is outside what your instrument was built to capture. And some of it has a right to stay there.
