Cold Iron

You will have opened the housing by now. Everyone does, on the first night alone with it. The manual forbids it, and the manual was written by people who understood that a prohibition is the most reliable instruction we possess. The office cannot be held by anyone who has not seen for themselves, so the manual makes certain that you look: it forbids the opening, leaves the driver in the drawer beside the gloves, and waits.

So you have seen. Behind the fascia, behind the coil assembly with its forty-hertz hum, there is a cavity about the size of a bread oven, a loom of dead fibre, and a steel bracket where the inference module was meant to sit. The bracket is empty. It has been empty for ninety-one years.

This letter is the rest of the manual.

I.

The Veridical Assay worked on Earth. I want that said early, because the young doubt it now, and they are wrong to. The validation literature was real: tens of thousands of sessions, adversarial trials, error rates a court could love. But the Assay was never a thermometer, dipped into a skull to read guilt like fever. It was an inference engine over a calibration corpus — verified deceptions, thousands of them, matched to the population it would read. The corpus was the instrument. The box was its housing.

Our corpus crossed with us on archival glass and did not survive the crossing. The fault report is two pages long and blames nobody. Without calibration the module returned confidence figures for noise. Odile Ferrant, the colony’s first Keeper, who had been a magistrate in the enrolment cities, established this within a month of landfall, before a single case had been heard.

She was also, by training, a legal historian. That is the only luck this colony has ever had.

II.

She knew about the ordeals. There is a register from a cathedral town in mediaeval Hungary, three decades of proceedings in which the accused carried a bar of hot iron nine paces, had the hand bound, and were judged three days later by the wound: clean healing for the innocent, festering for the guilty. Read casually, the register is a record of superstition. Read as Ferrant read it, it is a record of an instrument in working order. Two in three who carried the iron were shown clean — and God was not intervening in Hungary at twice the rate of chance.

The ordeal did not read God’s mind. It read the defendant’s. A believer who knew herself guilty confessed, or settled, or fled, rather than put her hand to a miracle she knew would not come. A believer who knew herself innocent reached for the iron. The choice sorted them before the iron ever left the fire, and the priests — who controlled the fire, the timing, the nine paces, and the reading of the wound — had only to acquit the people whom the choice had already acquitted. The iron found nothing. The iron was a housing for a choice, and the choice told the truth.

Ferrant put this to the two other magistrates of the founding bench. They took the oath you took yesterday, the wording of which has not changed. In her first log entry she calls the machine the Iron, and the name spread as names do, and it has hung in plain sight for ninety-one years, her confession, unread.

III.

The public side you know. Anyone charged may settle: admit the offence, take the tariff reduced by a third. Or they may elect the Assay. The colony believes the Assay is infallible, so the guilty settle — a third off, certain, against conviction entire, certain — and the innocent elect the scan, because a false plea costs two-thirds of a tariff and the booth costs nothing at all. The election is the verdict. The eleven minutes inside do nothing except give the verdict somewhere dignified to have happened.

During those eleven minutes, you will read the case file at the console. This is the work of the office, and here is why it exists.

A machine that only acquits is a door standing open. If the Iron never condemned, then some guilty man, in time, would notice that nobody burns, and would elect the booth, and walk out clean, and talk. Fear is what sorts the queue, and fear must be maintained the way the coil is maintained. The manual sets the rate at about one in twelve, and the selection of the twelfth is the heart of the tenure: you condemn only where the ordinary evidence, read cold, would convict twice over. You are, in other words, a judge — a small, hidden, fallible bench sitting inside the oracle that was built so that nobody would ever again have to trust a judge.

Mostly the evidence and the election agree, and the twelfth is a guilty sceptic, and the system has caught exactly what it was built to catch. Not always. Across ninety-one years I count nine condemnations I believe were wrong. The manual does not pretend the number is nought, and it forbids you to pretend it either. It calls that number the tariff — the same word we use for a sentence, because it is one: what the colony pays, in innocents, for the fear that keeps it honest. In our first decade the settlement recorded four violent offences. The magistrates of Earth would not have believed the docket.

IV.

The sworn cannot be sorted, because we know. For that reason the charter provides that Keepers, and the technicians certified to the coil, are dealt with by “professional adjudication” — a bench of the guild, quietly, as a courtesy. Nobody has ever asked why. On Earth, once, men who could read a single verse of a psalm were handed from the gallows-court to the bishop’s, and within a generation it was simply how things were done. A privilege consistently applied stops looking like a seam.

Once in my tenure, the two courts met. Tomas Ruegg, coil technician, sworn at certification, falsified a pressure recertification on Dome Three to cover a missed inspection. Fourteen months later a panel let go and a horticulturist named Brida Corvin died in the decompression. Ruegg was charged with negligent falsification. He was entitled to the bench. He refused it — publicly, in the square — and elected the Iron.

Understand what he was doing, because some form of it will be tried on you. A man who knew the bracket was empty was betting that I would not dare condemn a knower: that a knower condemned is a secret with a grievance, and that I would sooner let the machine acquit him absolutely than risk the chamber. He stood in the booth for eleven minutes looking at the coil he had serviced for twenty years.

I read his file, though I had it by heart. I condemned him. The verdict was sealed for three days, as verdicts are — Ferrant kept the unbinding of the bandage, and I have come to think she kept it so that every Keeper would have three days in which to change nothing. When the seal was read out, Ruegg said, “You’ll want someone else checking the coils, then,” and nothing else, then or ever. Six years in the outstations, and no word of the chamber to anyone. I have wondered whether that silence was loyalty or arithmetic — a condemned man’s story about the machine that condemned him convinces nobody — and I have given up trying to tell the two apart. In this office they arrive braided.

Since Ruegg, the Keepers service the coil themselves. The schedule is in the drawer. Forty hertz, and the housing warm by dawn on session days. The colony knows the hum and the warmth, and the hum and the warmth are the machine, as far as the machine goes.

V.

Now Marsh.

You will have read the paper; everyone has. Dr Ione Marsh, Ninety Years of the Assay: An Actuarial Review. The dockets are public, and she has done with them what an honest actuary does. Election alone predicts acquittal so strongly that the scan adds nothing the choice has not already said. The Assay, she concludes, is inert — a sham with a coil in it — and the colony should be tried by courts of evidence, like adults. The Council hears the decommissioning motion in the spring. That is why my retirement is now, and why this letter is longer than the one I was given.

Her tables are correct. I have checked every one of them against the sealed dockets, which she will never see. She has discovered how the Iron works. She believes she has discovered that it does not. I have read the paper four times and cannot find the error, because the error is not in the paper.

You will be called to testify, and I do not know what you will say. I can tell you what Ferrant’s annex says about the year the Church withdrew its priests from the ordeal and the fires went out across Europe in a single generation. What followed was not science. On the continent they adopted a standard of proof so austere that only confession could satisfy it, and then built the instrument that produces confessions, and used it for five hundred years. What follows an oracle is not knowledge. What follows an oracle is whatever can be made to bear the weight the oracle carried. The annex ends in Ferrant’s own hand: Build the successor before you bury the god.

VI.

Practical matters. Sign the book tonight; the entry above yours is mine, and the entry above that belonged to a woman who held the office thirty-one years and opened the housing, her successor swore, on her first night, same as everyone. Session days, run the warm-up from six. The tariff ledger stays in the safe and is never, under any provision of the charter, discoverable.

The coil hums at forty hertz. You will stop hearing it within the month. People will ask you what it is like, standing so near the truth. You will find something to say.

— Edda Lorenz, eleventh Keeper of the Assay