The you in the data

When a system decides something about you — approves or declines you, prices you, routes you into the slow lane or the fast one — it feels, from your chair, as if it has judged you. It has not. It has never met you. What it judged is a figure assembled from records: addresses you have held, payments you have made or missed, categories you have been sorted into, attributes inferred from the company your data keeps. That figure walks around the world under your name, and for every consequential purpose it is treated as you. But it is not you. It is a double — a reconstruction — and the whole of your experience on the receiving end turns on the distance between the two.

This is a different complaint from the one about a score that follows you, and a different one again from being judged by a pattern learned off strangers. Those are about portability and about groups. This is narrower and, in a way, more intimate: it is about the misrepresentation of a single individual by their own data. The map has your name on it. It is still a map, and you are still the territory, and no one built the map to be walked on.

The double that stands in for you

Consider how the double is made. It is stitched together from whatever records happen to exist — never everything about you, only what was captured, retained, and made available to the system at the moment it looked. That is the first thing to understand about it: it is not a portrait, it is a sampling. It knows what was written down. It does not know what was not, and it has no way to feel the difference between a fact that is missing and a fact that is false, because to the system an absence is simply a value it does not have.

The double has a strange shape. It is confidently precise exactly where it is wrong — a mistaken entry, a stale address three moves out of date, an inference that fit someone with your profile but never fit you — and it presents these with the same flat authority as everything true about you. There is no tremor in a wrong record. And it is silent exactly where you are most yourself: the context behind a fact, the reason a number looks the way it does, the change you have made since the data was written. A late payment appears; the month you spent settling a parent's estate does not. A gap appears; what you were doing in it does not. The double keeps the fact and discards the meaning, and then it is handed to a decision as though the fact were the whole of you.

And the decision treats it as the whole of you. This is the quiet substitution at the center of automated judgment: at no point does anyone assert that the double is the person, and at every point the double is used as if it were. You are approved or declined on its record. You inherit its consequences. You are never introduced to it.

Inheriting a stranger's verdict

The harm does not live in the data being bad, exactly. It lives in the gap — the space between the person and the double — and the gap can be wide even when every individual record is technically correct. A true fact, stripped of the context that gives it meaning, becomes a kind of lie that no one told. The record says you defaulted; it does not say the debt was fraudulent and later cleared. The record says you were flagged; it does not say the match was to someone else and the flag was never resolved because no one was watching the queue. Each fact would survive an audit. Together they compose a person who does not exist and who is, in the ways that matter here, worse than you.

What makes this specifically an injury of the receiving end is that you are held responsible for that person's actions. The double defaults and your rate rises. The double resembles a risk and your application slows. You are, in a precise sense, on trial for the conduct of a stranger — one assembled in your name, from your records, whom you have never seen and cannot cross-examine. The ordinary protection against being judged by a caricature is that you get to meet the caricature and say, that is not me, and here is why. Against your data double you get no such meeting. The verdict arrives; the defendant never appears.

The system was scrupulously fair to your data and indifferent to you, and it will never notice the difference — that is your job, and it will cost you.

Notice what the system is not doing wrong, on its own terms. It may have followed its rules exactly. It may have weighed the double with perfect consistency and no malice. Fairness to the data and fairness to the person are simply not the same property, and a system optimized for the first has no instrument that even registers the second. The indifference is structural, not personal, which is why it will never correct itself and why the cost of closing the gap falls, always, on the one person standing outside it.

The right to confront your double

This is where standing enters — and where it means something more demanding than a right to complain after the fact. To have standing against your data double is to have the power to confront it: to see what the data says you are, to contest the parts that are wrong, and — the part most often forgotten — to supply the context that makes a true fact mean something different. Not just the right to correct an error, but the right to re-contextualize a fact that is accurate and misleading. The default should be visible; so should the reason it does not describe your reliability. Both are part of the record; only one usually reaches the decision.

Contestability of this kind is not a courtesy bolted on after the judgment. It is what converts a decision from an assertion into an account. A Decision Receipt that shows which facts about you the decision actually rested on — the inputs, their provenance, the version of you the system read — is the instrument that makes confrontation possible, because you cannot argue with a double you are not permitted to see. Provenance matters here for a reason people underrate: knowing where a fact came from is often the fastest way to know why it is wrong, or why it stopped being true, or why it never applied to you at all.

A decision made about a data double the person cannot see or correct is a trial in absentia where the accused is a caricature of the defendant. That is not a rhetorical flourish; it is a fair description of the procedure. Someone stood in for you, made of your records, and the case was decided on their showing while you waited outside the room, unaware there was a room. The remedy is not to make the double perfect — no reconstruction of a person from records will ever be complete, and pretending otherwise is its own error. The remedy is to give the person a standing seat: the right to be in the room, to see who is standing in their name, and to say, before the verdict lands and not only after, that is not me, and here is what the record left out. The claim is modest and it is the whole thing. You should get to meet the version of you that the world is about to judge.

— Dispatches · Summit Cognitive