The asymmetry of the appeal

When a person sets out to appeal a decision that was made about them, the contest they are entering is not between two parties of comparable strength. It only looks symmetrical on paper, where there is an applicant on one side and a respondent on the other and a neutral process in between. In substance the two sides are not even playing the same game. The institution knows what it did, why it did it, what it looked at, and what rule it applied. The person knows only that the answer was no. Everything the appeal would need to turn on lives entirely on one side of the table.

This is the asymmetry that quietly determines the outcome of most appeals before they begin. The institution holds the inputs to the decision, the model or the policy that processed them, the threshold that was applied, and the record of how the case moved through the system. The person holds a letter. To contest the decision, the person has to reconstruct, from the outside and at a disadvantage, what was done to them on the inside with full visibility. They are asked to disprove a claim while being denied the evidence on which the claim rests. It is the structure of a rigged trial, even when no one running it intends anything of the kind.

And usually no one does intend it. The asymmetry is not a conspiracy; it is the default. Information accumulates wherever the decision is made, which is inside the institution, and unless something is deliberately done to move it across the table, it stays there. The person appealing is not being actively deceived. They are simply standing where almost none of the relevant facts happen to be. Left alone, the gap does not close. It is the resting state of the relationship.

A right to appeal that leaves the person without the facts is not a right. It is an invitation to lose more slowly, on the appellant's own time and at the appellant's own cost.

What a right to appeal actually requires

We talk about the right to appeal as though the right consists in the procedure — the form to fill in, the second reviewer, the window of days in which to file. But a procedure is only as real as the appellant's capacity to use it, and capacity here means information. A person cannot contest what they cannot see. If the appeal asks them to identify the error in a decision whose inputs and rules are withheld, the procedure exists and the right does not. The two are routinely confused, because the procedure is visible and the asymmetry is not.

What a genuine right to appeal requires, then, is not merely a channel but a transfer — the deliberate movement of the relevant facts from the side that has them to the side that needs them. The person must be given the evidence that was actually in front of the decision, the rule that was actually applied, and enough of the decision's state to see how the one produced the other. Not a summary written by the institution, which is the institution grading its own work and handing the appellant its preferred account. The actual record, legible to the party who wants to fight it. Anything less leaves the asymmetry exactly where it was and calls the leftover a remedy.

This is the work a Decision Receipt is built to do, and it is worth naming the work precisely, because it is easy to mistake for transparency in general. The point is not that institutions should be more open as a matter of good manners. The point is narrower and sharper: a legible, shared record arms the weaker party with the same facts the stronger party already holds, and in doing so it converts an appeal from a performance into a contest. With the record in hand, the person is no longer guessing at what was done. They are examining it. The asymmetry of information does not vanish, but it stops being decisive, because both sides are now arguing from the same set of facts.

Symmetry as the condition of a fair fight

There is a deeper principle underneath this, and it predates anything automated. A contest is fair only when both parties can reach the evidence the outcome turns on. We accept all sorts of inequalities between adversaries — in resources, in expertise, in time — but we have long treated unequal access to the facts of the matter as the one inequality that makes the contest a sham. It is why disclosure exists at all: not to be generous to the weaker party, but because a decision reached on facts only one side can see is not a decision anyone else has reason to respect.

Automated decisions concentrate this problem rather than inventing it. When a determination is produced inside a system the affected person will never see, by a process they could not have observed, the information gap that disclosure was meant to close is at its widest. The person is further than ever from the facts, and the facts are more completely held on the other side than ever before. If the right to appeal is going to mean anything against a machine-mediated institution, the correction has to be at least as deliberate as the concentration it is correcting. A shared, contestable record is that correction. It is the disclosure obligation, re-expressed for decisions that no longer happen in rooms.

So the test for whether an appeal is real is not whether the channel exists. It is whether the person filing it has been handed the same facts the institution used against them. If they have, the asymmetry has been corrected and the contest is a contest. If they have not, then whatever procedure remains is a courtesy extended to the losing party — a way of letting them feel heard while ensuring they cannot prevail, because the one thing they would need to prevail has been kept on the far side of the table. A right that arms only the side that was already winning is not a right. It is the asymmetry, wearing the costume of one.

— Dispatches · Summit Cognitive