The A1 at Lloyd's

In the last decades of the seventeenth century, a coffeehouse kept by a man named Edward Lloyd became the place in London where people who cared about ships gathered. Merchants, shipowners, and the underwriters who insured voyages came for the coffee and stayed for the information — who was sailing where, which vessels had come safe to port and which had not, what a cargo bound for the Levant might be worth and what might befall it on the way. Lloyd, sensing what his customers valued, began to collect and circulate shipping news. The coffeehouse became, in effect, an exchange for a particular kind of knowledge: the condition and prospects of ships a man was being asked to stake money on and would almost never see with his own eyes. That last fact is the whole problem. To insure a vessel, to consign goods to it, to lend against its voyage, you had to form a judgment about a thing you could not inspect. The instinct of the age would have been to trust the owner, or a broker who knew him. The institution that grew out of Lloyd's coffeehouse did something more durable: it built a way to trust the ship instead.

By the 1760s that impulse had hardened into a printed artifact. A register of shipping began to appear — a book that listed vessels and recorded facts about them, including an assessment of their condition. Over the following decades this developed into a full system of classification: an independent surveyor, answerable to the register rather than to the owner, would inspect a ship against a published standard and assign it a grade. The hull was rated by a letter; the ship's equipment — her anchors, cables, stores — by a number. A vessel whose hull was in the first class and whose gear was likewise in the first condition earned the top mark in each column: the letter A, the figure 1. A1. The phrase escaped the register and entered the language, where it survives as a byword for excellence by people who have never seen a bill of lading. What they are quoting, without knowing it, is a grade.

It is worth being careful about two Lloyd's. The classification society — Lloyd's Register — grew alongside, but is distinct from, Lloyd's of London, the insurance market. The market priced risk; the register surveyed and graded the ships. It is the register, the classifier, that concerns us here.

A stranger could stake his money on a ship he would never board, on the strength of two characters set down by someone with no reason to flatter her.

The grade a stranger could trust

Consider what A1 actually accomplished. An underwriter in London, deciding whether to insure a voyage, did not have to travel to a shipyard and crawl the vessel's timbers. A merchant weighing whether to load his goods did not have to take the owner's assurances that the hull was sound. Both could consult the register and read a grade produced by a third party who had inspected the ship against a known standard and had no stake in selling her. The grade was portable: it meant the same thing to everyone who read it, because everyone knew what standard it had been measured against. It converted a private, unrepeatable act — one surveyor's inspection on one afternoon — into a public fact the whole market could rely on without repeating the work. This is the quiet genius of classification. It does not ask each party to trust the owner, and it does not ask each party to inspect the ship. It interposes a single trusted survey between the ship and the market, and lets the resulting grade travel.

Notice how different this is from taking the owner's word. The owner has every incentive to describe his ship favorably; his assurance is worth exactly as much as his honesty, which the stranger cannot verify. The surveyor's grade is worth something else — it is anchored to a standard the reader can look up, produced by a body whose entire value depends on grading straight. Trust in an unseen ship was not talked into existence by the owner. It was manufactured, deliberately, by independent classification against a known standard.

Trust manufactured by independent classification

Pull the mechanism out of the timber and rigging and it has three parts, each doing distinct work. There is an independent third party — a classifier who answers to the standard, not to the owner of the thing being graded, and whose judgment is therefore not for sale. There is a public standard — a written specification of what counts as first-class condition, so that the grade is not one surveyor's private taste but a measurement against a bar anyone can inspect. And there is a portable grade — a compact, legible token that carries the survey's finding out into a market of strangers, each of whom can rely on it without re-inspecting the ship. Remove any one and the thing collapses. A grade with no public standard behind it is just an opinion. A standard with no independent surveyor applying it is just a document. An independent survey that never issues a portable grade helps no one but the surveyor.

There is a fourth property, easy to miss, that distinguishes this from marking an object once and being done. A ship is not a static thing. She ages; her timbers work; her gear wears. So classification is not a single verdict but a relationship over time — the vessel is re-surveyed as she ages, and her grade is confirmed, revised, or withdrawn as her condition changes. This is what separates classifying a ship from assaying a coin or hallmarking a piece of silver, which fix the quality of an object at a moment. Classification grades an operating, changing system and stands ready to grade it again. The A1 you read this year is a claim about the ship this year, backed by a body that will look again.

A classification society for the decision system

Now turn the lens. We are building machines that decide things — who is approved, who is flagged, what is recommended, what is denied — and we are asking a market of strangers to rely on them. The person affected by an automated decision cannot inspect the system that produced it; neither, usually, can the regulator, the counterparty, or the court. And we have, for these consequential decision systems, nothing like a classification society. There is no independent body that surveys a decision system's soundness against a published standard and issues a grade the market can read and carry. We take the operator's word that the system is sound — exactly as a merchant might once have taken a shipowner's word that the hull was sound, in the absence of any better instrument.

This is a narrower and more institutional point than the familiar complaint that assurance is only as good as the record it examines. That complaint is about the substrate — whether there is anything trustworthy to survey. The gap I mean here is about the form: even where a decision leaves a genuine record — a Decision Receipt that carries its real inputs, the rules in force, and enough state to be replayed — there is no standing classifier, no open standard of soundness against which such records are surveyed, and no portable grade that lets a party who cannot inspect the system rely on someone who did. Provenance and admissibility give us a ship worth surveying. Classification is the institution that surveys it and issues the mark.

What Lloyd's Register is the model for is precisely this: the independent classification of decision systems. A trusted body, answerable to a public standard rather than to the operator, that surveys a system's conformance and soundness and issues a portable grade — one that means the same thing to a regulator, a counterparty, and an affected person, none of whom can crawl the timbers themselves. And because a decision system, like a ship, is an operating and changing thing — retrained, reconfigured, drifting — the grade would have to be a relationship over time, re-surveyed as the system ages, not a certificate hung on the wall in year one and forgotten. We invented the answer to the stranger's problem three centuries ago in a coffeehouse full of people who could not board the ships they were staking their fortunes on. They stopped asking the owner to be trusted and built something that could be. The two characters they arrived at are still the standard we are being asked, and still failing, to earn: A1.

— Dispatches · Summit Cognitive

---