The audience of one

When an organization builds the record of how it makes decisions, it almost always pictures a particular reader. The reader is professional. The reader is patient, fluent in the institution's vocabulary, and there by choice — an auditor working through a sample, a regulator checking a process, a lawyer assembling a file. The record gets shaped to satisfy that reader, and because that reader is competent and well-resourced, the record can afford to be dense, technical, and assume a great deal. The trouble is that this reader, the one the record is built for, is not the reader the record is actually about.

There is another reader, and that reader is the point. It is the person the decision was made about — the applicant who was declined, the account that was frozen, the claim that was denied, the file that was flagged. This reader did not choose to be in the conversation. They have, of everyone who will ever open the record, the most at stake in it and the least preparation for it. And in most systems they are precisely the reader the record was not designed to serve. The account is written for the professionals who oversee the institution, and the person the institution acted upon is left to overhear it.

The claim I want to make is that this ordering is backwards, and that getting it right is not a matter of courtesy but of legitimacy. A decision record has a real and proper audience of many — auditors, regulators, courts, all of whom have a legitimate reason to read it. But it has an audience of one in a deeper sense: the single person the decision was about, who is the reason the decision had any weight to begin with. Design the record for that reader, and the others are served as a consequence. Design it for the others, and that reader is served by accident, if at all.

The person the decision was about did not ask to read the record. They have the most at stake in it and the least preparation for it. Write it for them.

Who the record is really for

Consider what distinguishes this single reader from all the others. The auditor reads many records and is invested in none of them; this reader reads one record and is wholly inside it. The regulator brings expertise; this reader brings only the experience of being on the receiving end. The court has standing by office; this reader has standing by injury. Every other reader comes to the record from a position of relative power and relative detachment. This reader comes to it from the position the whole apparatus of accountability is supposed to protect — the position of the person things were done to, who now wants to understand what happened and whether it was fair.

If you build the account for the professional readers, you optimize for the qualities they value: completeness, technical precision, defensibility against expert challenge. None of those qualities is wrong, but none of them is the same as legibility to the affected person. A record can be flawlessly complete and entirely opaque to the one reader who most needs to understand it. It can satisfy every auditor and still fail to tell the declined applicant, in terms they can actually act on, what was considered and what rule decided their case. When that happens, accountability has quietly relocated. It has become a conversation institutions hold with each other about a person, conducted over that person's head.

That relocation is the failure mode worth naming, because it is invisible from the inside. From the institution's vantage, everything looks accountable: the records exist, the auditors are satisfied, the regulators sign off. The decision is well-documented by every measure the institution applies to itself. What none of those measures capture is whether the person the decision was about could ever read the account and recognize their own case in it. A system can pass every internal test of accountability and still be unaccountable to the only reader who was not in the room when the tests were written.

Designing for the reader with the most at stake

Writing for the audience of one does not mean dumbing the record down, and it does not mean stripping out what the professional readers need. It means choosing, when the two pull apart, the version legible to the person at the center. It means that the inputs that were considered, the rule that was applied, and the path to contesting the outcome are present in a form the affected person can actually use — not buried in a format that assumes the reader is a specialist who chose to be there. A Decision Receipt earns its name when the person it concerns can read it and understand both what happened and what they can do about it. That is a higher bar than passing an audit, and it is the bar that matters.

The reason it matters is that legitimacy is not conferred by oversight alone. An institution can be thoroughly overseen and still illegitimate in the eyes of the people it acts upon, because legitimacy is a relationship with those people, not a certificate held among peers. The affected person is the source of the institution's obligation to explain itself in the first place. A record that cannot be read by that person has severed the explanation from the one it was owed to, and kept the form of accountability while discarding its object. Oversight that never reaches the overseen is not the same as accountability; it is a closed loop that has mistaken itself for an open one.

So the test I would apply to any decision record is to ask who, concretely, it is written for — and to be suspicious of any answer that names only the professionals. The auditors and regulators and courts are real readers with real claims, and a good record serves them. But it serves them best by serving first the reader who did not choose to be there, who carries the most and understands the least, and for whom the record is not a document among many but the only account they will ever get of something that happened to them. Write for that reader, and accountability stays pointed where it belongs. Write for everyone else, and it slowly becomes a conversation the institution has only with itself.

— Dispatches · Summit Cognitive