Many goods, growing together, and no single one in charge.
In short — Lasting disagreement between good people over good things is normal, not a failure — so a good room hosts it rather than faking consensus. A room built for that must do five things: keep the disagreement verbatim, show its working, mark every AI touch, leave absent voices empty, and decide nothing itself. Design for disagreement.
Reading on — Piece 6: can such a room actually be built? Here’s one real attempt, with its seams — and one unfinished wall — on show.
Everyone has sat in the meeting. The body corporate, the family sorting out the estate, the committee that has met four times about the same thing. The disagreement is real and it is not going away, and in the end it isn’t reasoning that settles it — it’s who has the most stamina, the loudest voice, or the least to lose by outlasting everyone else. You leave knowing a decision got made and that it wasn’t a good one, and that the people who “won” mostly just endured.
Now hold that against the one deliberative room a lot of us have actually been inside: a jury. Twelve strangers, no expertise required, given the evidence and the time and a genuine decision to make. It is slow and awkward and it works — well enough that we trust it with someone’s liberty. The difference between the jury and the committee is not the people. It’s the way the room is built.
Most of our public rooms are built like the committee. This piece is about building them like the jury — and about one mistake that has to be cleared away first.
We tend to assume that if reasonable people still disagree at the end, something went wrong — someone was being difficult, or the process broke, or nobody tried hard enough. So we reach for tools that make the disagreement go away: a vote that buries the losers, a “consensus” statement everyone is too tired to contest, a facilitator who sands off the edges until it reads as agreement.
But some disagreements are not failures. They are between good people, over things that are truly good, and truly collide.1 Liberty and equality both matter and they pull against each other. Mercy and justice both matter. Feeding a district and leaving the river alone both matter. There is no master-value sitting above them that ranks them once and for all, and — this is the part that matters for the age of clever machines — no amount of better reasoning dissolves the clash, because it was never a mistake in the reasoning. It’s the shape of a plural world.
Once you see that, a certain kind of disagreement stops looking like a problem to be solved and starts looking like a sign of seriousness. Some goods are worth fighting over.2 A room where nobody disagrees about anything important is not a wise room; it’s an empty one, or a frightened one.
So the goal of a good decision-room is not “what does everyone agree on?” That question, asked of a real value-clash, produces either a lie or the loudest person’s view in everyone else’s mouth. The real question is different: what settlement can different people actually live with, and on what terms? That’s not defeat. It’s how a plural society has always held together — not by everyone wanting the same thing, but by building arrangements that the people who want different things can each stand inside. Negotiated coexistence, not consensus theatre.
The room that does this on purpose has a name.3 A citizens’ assembly takes a group of ordinary people — often chosen at random to mirror the wider population — gives them real evidence and real time, and asks them to reason together towards a considered judgement that protects the people who end up in the minority. It is the jury’s method turned on a public question. Unlike a referendum, which extracts a yes or no with none of the reasoning attached, the assembly keeps the reasoning — including the reasoning of the people who ended up disagreeing.
But a room full of the right people is not automatically fair. The confident, the fluent, the already-powerful will fill the air unless the room is built to stop them.4 “We reached consensus” too often means “the usual voices agreed and the rest went quiet.” A room built for disagreement has to actively correct for that, or its agreement is worth nothing.
Which lets us be concrete. Strip it back, and a room built for disagreement — not for the appearance of agreement — has to do five things:
The fourth is the one people skip. Where a voice belongs at the table but can’t rightly be stood in for — a mana whenua seat, say — you leave it registered and empty, and route it to real people, rather than let anyone, or a machine, speak in its place.
The obvious question is whether such a room can actually be built — with the machine helping and still kept in its place — or whether this is just a wish-list. That’s the next piece, and it comes with its seams showing.
For now, the design principle, which is also the whole point:
Design for disagreement.
Next in the series: “Don’t Trust Us — Check the Working” — one real attempt at building the room above, with its limits on full display: software that rehearses the argument, seals the record, keeps the dissent, and decides nothing.