The decision without a place

Every consequential decision a person ever made happened somewhere. A loan was granted at a desk, in a branch, in a town with a name. A sentence was passed in a courtroom that had an address and a clerk. A claim was approved in an office you could walk into, on a floor you could find, by someone whose chair you could stand beside. We rarely thought about the place as part of the decision. It was simply the room the decision was in. But the place was doing quiet, structural work the whole time, and we are only now noticing it because we have built a generation of decisions that happen in no place at all.

Consider what a location used to carry. It carried jurisdiction — the decision fell under some authority's law, and you knew which authority because you knew where you stood. It carried a record office, a physical or institutional somewhere that the paperwork went to and could be retrieved from. It carried, above all, a person you could go to: a clerk, a manager, a judge, a desk with a nameplate. When a decision had a place, accountability had a starting coordinate. You knew where to direct the question.

An automated decision has none of this. It is assembled in flight across a dozen services, in regions chosen for latency rather than meaning, by code whose execution is distributed by design. Ask where the decision happened and the honest answer is everywhere and nowhere — a little in one data center, a little in another, momentarily in memory that no longer exists. The decision did not occur in a room. It occurred in a topology. And a topology is not somewhere you can go.

What place quietly provided

The reason placelessness matters is that place was never just a backdrop; it was an anchor for three things we cannot do without. The first is standing — the answer to whose decision was this, under what authority? A human decision sat inside a jurisdiction the way a building sits inside a city. The distributed decision sits inside no jurisdiction in particular, which means the question of which rules governed it is not answered by where it happened. It has to be answered some other way, or it goes unanswered, and an unanswered question of authority is the beginning of every accountability failure.

The second is retrievability. A place implied a record office: somewhere the account was kept and could be found again. When a decision happens across ephemeral infrastructure, there is no office, no shelf, no drawer. There is whatever the system happened to write down, and if it wrote nothing, the decision left no place behind it at all. The default state of a placeless decision is to vanish — not to be hidden, which would at least imply something to find, but simply to leave no residue.

The third is the person. Place gave you someone to go to. The most corrosive thing about a decision distributed across regions is not that it is fast or opaque but that it offers no door. There is no desk. There is no nameplate. When the decision goes wrong, the question who do I see about this has no addressee, and a grievance with no addressee curdles into something worse than anger: resignation.

Place was never the backdrop to accountability. It was the anchor. Remove it, and the question of where to bring the complaint has no coordinate at all.

A record is the place we have to rebuild

This is the work a decision record has to do, and it is more demanding than it first appears. The record is not a convenience or an audit nicety. It is the deliberate reconstruction of everything that place used to provide for free. It has to fix a locus — a single, findable somewhere that the decision can be located at, after the infrastructure that produced it has scaled away to nothing. It has to carry its own jurisdiction: the rules that governed it, frozen as they stood, so that the question of authority travels inside the record rather than depending on a building that no longer applies. And it has to name a place the question can be brought, so that contestability has a coordinate.

A Decision Receipt, done properly, is a place. That is the most precise way I know to describe what it is for. It restores the somewhere that distribution dissolved. When you hold one, you are holding the room the decision happened in — not a metaphor for the room, but its functional replacement: the inputs that were present, the rules that were active, the basis on which the thing was decided, gathered into a single locus you can return to. The provenance the record carries is the address. Without it, the decision is genuinely placeless, and a placeless decision is one that nobody, in the end, is standing anywhere to answer for.

We did not lose accountability when we automated. We lost the place that used to hold it, and we mistook the place for scenery. The task now is to notice what the room was doing, and to build, on purpose, the somewhere the decision no longer has. The record is that somewhere. It is the address we write down so the decision is not lost in a topology — so that when the question comes, there is a door, and behind the door, the account.

— Dispatches · Summit Cognitive