The coroner and the inquest

Among the oldest continuously held offices in English law is one whose entire purpose is to investigate death. The coroner is usually dated to the Articles of Eyre of 1194, which set out the duties of the custos placitorum coronae — the keeper of the pleas of the Crown — and by the thirteenth century the office had settled into a recognizable shape. Its central task was, and remains, to inquire into deaths that were sudden, violent, unnatural, or simply unexplained. Not every death; a person who dies old and expected at home occasions no inquiry. But a death that arrives without an obvious lawful account triggers something remarkable: a duty, vested in a public official, to find out what happened and to do it on a record that belongs to no one and to everyone. The coroner does not wait to be asked. The asking is the office.

It is worth pausing on how strange, and how deliberate, that arrangement is. The medieval Crown had fiscal motives mixed into the original office — forfeitures and fines flowed from certain findings — and the modern coroner is a very different creature from the twelfth-century one. But the enduring principle survived the centuries of reform: certain outcomes are grave enough that the community mandates an investigation by default, not merely as an option to be exercised by an aggrieved party. A society could, in theory, have said that if a family wishes to know how their kin died, they may hire someone to look into it. Instead it said the opposite. The inquiry happens because the death happened, and it happens whether or not anyone with standing or resources demands it.

The coroner does not establish who is guilty. The coroner establishes what is true — and that turns out to be the harder, and the more durable, thing to guarantee.

The death too important to ignore

The instrument of the coroner's inquiry is the inquest: a public, fact-finding proceeding. Its purpose is narrow and precise, and the narrowness is the point. An inquest answers four questions — who the deceased was, and how, when, and where they came by their death. It does not determine criminal liability or civil blame; in the modern coronial function it is explicitly barred from framing its findings so as to appear to decide a question of guilt, which belongs to other courts. This can seem like a limitation, even an evasion. It is neither. It is a separation of two things that our instinct constantly conflates — the establishment of what happened and the assignment of who is at fault — and the recognition that the first can be secured cleanly only if it is not held hostage to the second.

Because once you fuse them, the record contracts. If the only reason to investigate a death is to convict someone, then investigation happens only where there is a plausible defendant, a willing prosecutor, and a case worth bringing. Deaths that implicate no one convictable, or that would embarrass an institution more powerful than the deceased's family, quietly go unexamined. The genius of the inquest is that it decouples the standing duty to find out from the contingent question of whether anyone will be punished. The community gets its account of the facts regardless. The truth is placed on the record as a public good, and the separate machinery of blame can then draw on that record, or not, without ever being the reason it exists.

Inquiry that does not wait for a plaintiff

This is the feature I want to hold up against the present, because it is the one we have almost entirely failed to reproduce for the decisions that now govern so much of ordinary life. The inquest embodies a proposition that markets and litigation do not: that some outcomes matter enough that the investigation into them cannot be left to the appetite and the wallet of a private party. A wrongful death does not go unexamined because the widow is poor, or unlettered, or too exhausted by grief to file. The examination is owed to the fact of the death itself. What a legal system organized around private complaint produces is investigation proportional to the complainant's resources; what the inquest produces is investigation proportional to the gravity of the outcome. Those are radically different distributions of who gets the truth, and the difference falls hardest on exactly the people least able to demand it.

And the record is the point — not the verdict, the record. An inquest can end without anyone being blamed and still have done its whole job, because it has produced a durable, public account of how a life ended that will outlast the memories and the interests of everyone involved. That account can be cited, contested, reopened, and built upon. It converts a private catastrophe into a matter of public knowledge. When we say a death was "never properly investigated," we are not usually complaining that no one was punished. We are complaining that no reliable account was ever established — that the truth was allowed to depend on whether someone had the standing and the stamina to force it into the light, and no one did.

A coroner for the algorithmic harm

Now consider the grave outcomes produced by machine decision systems: the care that is denied, the benefit that is terminated, the person wrongly flagged and detained, the catastrophic error propagated at the speed and scale that only automation allows. These are examined, when they are examined at all, on precisely the pre-inquest model the coroner was invented to replace. Investigation happens only when an affected party has the resources, the sophistication, and the legal standing to compel it — which is to say, rarely, and unevenly, and least of all for the people most exposed. There is no coroner for the algorithmic harm. There is no office whose standing duty is to convene an inquiry into a serious adverse automated decision by default, on a public record, whether or not the person harmed can afford to fight. The default is silence, and the silence is heaviest exactly where the outcomes are gravest.

What the inquest offers is not a technology but an institutional form worth deliberately copying: a mandatory, default, on-the-record inquiry into grave outcomes, structurally separated from the question of blame so that the facts get established regardless of who is eventually held to answer. This is a different thing from a flight recorder. A recorder is the survivable device and the account it preserves; it is a precondition. The inquest is the standing duty to convene and examine — the obligation to ask, vested in an office that does not wait to be prompted. You can have the record and still have no one obligated to read it. The coroner is the answer to that gap.

Such an inquiry has a precondition of its own, and it is the one thing an automated decision can be built to guarantee where a human death cannot: that the decision arrives already able to be reconstructed. A body must be examined after the fact, its story inferred from what it left behind. A consequential machine decision need not be so mute. If each such decision produced a Decision Receipt — the evidence actually before it, the rules in force at the moment, enough preserved state to replay how the outcome was reached — then the inquest into a grave automated harm would not begin, as coroners have always had to begin, by reconstructing from fragments. It would begin with an account already fit to be examined. The historical office supplies the standing duty; the modern record supplies the material the duty needs. Neither is sufficient without the other, and we have, so far, built neither. We decided eight centuries ago that some outcomes are too grave to leave to whoever happens to complain. We have simply not yet decided it about the decisions that now, quietly and at scale, end things too.

— Dispatches · Summit Cognitive

---