A single line laid down on still water — a measure that holds while everything around it moves.
In short — A sealed record proves what happened; it can’t tell you whether what happened was fair. Fairness needs a standard — and the standard has to be the group’s own values, said out loud and settled before the deciding starts. The tool’s job is narrow: draw your values out of your own words, never add one; hold your disagreements open, never average them; and never hand down a single grade — that judgment is yours, made part by part against the record.
Reading on — Piece 8, the last: one room holding its own standard is only a pilot — what happens when many rooms each hold theirs, drawn into a map that never flattens them into one national grade.
Everyone has been marked against a standard they never got to see. The performance review with the criteria decided after the year was over. The exam graded to an answer key nobody handed you. The sinking part of that feeling is that you cannot tell whether you fell short or were set up to, because “good enough” was never written down anywhere you could read it before the marking started.
The piece before this one made a promise: don’t trust us, check the working — a sealed record you can read yourself, that cannot be quietly rewritten. That promise holds, and it matters. But it has a gap in it, and this piece is about the gap. A record tells you what happened. It cannot, on its own, tell you whether what happened was fair. Fair measured against what?
A decision earns the word legitimate when it honoured what the group said mattered. Read that slowly, because the whole thing turns on it. Not what an expert said mattered, not what the software worked out mattered — what the group itself said. Which means the group has to have said it: out loud, written down, and settled before the deciding began. A standard named after the result is not a standard. It is an excuse with better handwriting.
So the values a group will be judged by have to exist first, and they have to be the group’s own. Here is where it quietly goes wrong.
Picture a group that has never written down what it stands for — which is most groups — sitting in front of a capable machine that will happily draft the list in seconds. The temptation is obvious and the draft would read well. Don’t. This is the one move the tool must never make, and the reason is worth being exact about. If the machine writes the values you will later be measured against, then every later check of “did we live up to our values?” is secretly asking a different question: did we live up to the machine’s? Your name sits on the standard, but the standard is not yours. That is not a small slant in a debate. It is a rigged ruler handed over as a fair one — and because it arrives wrapped in your own name, nobody can see the switch.
Which is why the tool’s job, done properly, is deliberately narrow. It does not tell you what to value. It reads back what your group has already said and done — a decision you are proud of, a time you split down the middle, a line you would not cross whatever it cost — and it draws the values out of your own words. It sets two of your own statements side by side and asks whether they pull against each other. It marks where you have gone quiet on something. What it never does is add a value you did not say. Every value on the finished list traces back to a sentence one of you actually spoke or wrote. The machine holds up a mirror; it does not paint the portrait.
That includes holding your disagreements open. Real groups do not share one tidy value that settles everything. They hold several at once, and those values pull against each other — fairness against unity, openness against safety, keeping faith with the past against doing the new thing. The standard you build is not a clean agreed list with the tensions ironed flat. It is a truthful map of what you hold and exactly where it strains, kept whole. A tool that smooths that into a tidy slogan is lying on your behalf — and a decision measured against a lie only looks legitimate.
And the tool never grades you. This one is easy to get wrong. So, plainly: legitimacy is not a single mark out of ten. There is no fairness score. You judge it yourselves, on more than one count — was the room fairly made up, was the reasoning any good, can the outcome be answered for, were the voices that should have been there but weren’t at least named — and each of those judgments is pinned to something you can point to in the record. Rolling all of it into one number, a “legitimacy: B+”, would be the machine making the single call it must never make: deciding, for you, whether you were fair. It can lay out the evidence. Only you can weigh it.
The order is the last of it, and it does not bend. You name the standard before you decide, not after, and once named it is sealed — so that no one, including you on a bad day, can slide it sideways to fit the result you wanted. Fixed for this decision. Then open again, out loud, to be rewritten for the next one, because groups grow and what you stand for in five years is yours to change. What you cannot do is change it quietly, halfway through, to make a loss look like a win.
One room can do this: hold its own standard, and check its own working against it. The obvious next question is what happens when many rooms do it at once — a whole community, or a country, each room holding its own standard, none of them flattened into a single national grade. That is the last piece.
The yardstick has to be yours.
Read more — the working, seams showing: how the tool stays on the floor (what it may and may not do with your words), where the values come from when a group is starting from nothing, what a sealed value-record actually holds, and why the moment you set the standard is the one place a machine’s borrowed instincts have to be shut out hardest.
The citizens’-assembly tool described here runs behind a per-Village switch that is off by default; the run described predates the platform gate opening on 20 July 2026. The composite panels it uses to rehearse an argument are not real people and never stand in for them; seats that belong to tangata whenua are left for tangata whenua, not filled by a synthetic voice.