Who Counts: The Long of It

How we decide who has rights, why we have always got it wrong, and why it matters to you personally.

By Geoff Fane. August 2026.

This is the full essay, with the legal history worked through and twenty-four primary sources. For the compressed version, read Who Counts: The Short of It. The principles drawn from it are set out in The Reciprocity Manifesto, which you can sign.

1. A boy who cannot vote

Somewhere tonight there is a seventeen-year-old who works four shifts a week, pays tax on what he earns, looks after a younger sibling while his mother is at work, and follows politics more closely than anybody else on his street. He cannot vote. Next year, having done nothing in particular to change, he can.

Nobody believes he becomes wiser at midnight on his birthday. We use an age because we need a line, and lines have to go somewhere. The United Kingdom moved its line from twenty-one to eighteen in 1969 [1] and much of the world followed. Austria and Brazil put theirs at sixteen. Scotland and Wales use sixteen for some elections and eighteen for others, so a Welsh teenager can help choose a government in Cardiff but not in London. None of this tracks anything real about the individual. It tracks our need for a rule that can be applied at a polling station in four seconds without an argument.

But watch what happens when the stakes rise.

In the early 1980s a mother named Victoria Gillick went to court to stop doctors giving contraceptive advice to girls under sixteen without telling their parents. She lost, and the reason she lost changed English law. The House of Lords held that a child’s right to consent to their own medical treatment does not switch on at a birthday. It depends on whether this particular child has sufficient understanding and intelligence to grasp what is being proposed [2].

The consequence is a rule that looks untidy and is in fact precise. A mature fifteen-year-old can consent to a procedure. An immature seventeen-year-old may not. Two people the same age, in the same clinic, on the same afternoon, can get different answers, and the law thinks this is correct rather than embarrassing.

Hold onto that, because it is the thread running through everything that follows. When the stakes are low we ask what someone is. When the stakes are high we ask what they can do.

2. Four questions

The clearest version of this appears at the other end of a life.

When an adult in England or Wales is thought to have lost the ability to manage their own affairs, the Mental Capacity Act 2005 sets out exactly what has to be shown [3]. There are four questions. Can the person understand the information relevant to this decision? Can they retain it long enough to use it? Can they weigh it? Can they communicate a choice?

That is the entire test. And the Act adds something remarkable: an unwise decision is not evidence of incapacity. If it were, capacity would just mean agreeing with the assessor, which is not a safeguard but a licence.

Consider what that means in a real house on a real afternoon. A woman with moderate dementia may lack capacity to sell the family home, because she cannot hold the figures in mind long enough to weigh them. On the same afternoon she has full capacity to decide she does not want the operation, that she wants her daughter to visit and not her brother, and that she would like fish for dinner. Her capacity is not a licence that has been revoked. It is assessed decision by decision, and it is presumed to be there until somebody shows otherwise.

Now look closely at the four questions, because something is missing from them.

Understand. Retain. Weigh. Communicate.

Not one of them mentions being human. They describe what a mind does, entirely in terms of what it does. The test was written for people and has only ever been applied to people, but that is a fact about who has turned up to be tested. It is not a fact about the test.

3. Helping someone decide, rather than deciding for them

There is a development here that almost nobody outside disability law has heard of, and it is the single most useful idea in this essay.

The old model was called substituted decision-making. A guardian is appointed and the guardian decides on your behalf, in what they judge to be your best interests. It sounds benign and it has an ugly history, because best interests is a standard set by the person applying it and there is no obvious point at which it stops.

Article 12 of the United Nations Convention on the Rights of Persons with Disabilities pushed the world towards something different, called supported decision-making [4]. Here the supporter does not decide. The supporter helps you reach and express your own decision. Their duty is to ascertain your will and preference, act on it, and give reasons in writing if they depart from it.

Picture the difference. A man with an intellectual disability wants to move out of a group home and live with a friend. Under the old model, a guardian weighs it up and decides it is not in his best interests, and that is the end of the matter. Under the new model, the supporter’s job is to work out what he actually wants, help him understand the practicalities, help him say it to the people who need to hear it, and if the supporter is going to override him, put the reasons on paper where they can be challenged.

The change sounds administrative. It is not. It changes what a guardian is for. The guardian stops being a substitute and becomes an interpreter.

4. The question the test never asks

Now notice something strange about everything described so far.

Every one of these tests is applied to somebody whose membership is already settled. Nobody asks whether the seventeen-year-old is the sort of thing that could in principle vote. Nobody asks whether the woman with dementia is the sort of thing that could in principle own a house. Their candidacy was never in question. All that is in question is how much say they get inside a membership already granted.

So what decides membership?

In practice, almost everywhere, the answer is: being born human. That is not a test. It is a fact about ancestry. And it is carrying an enormous amount of moral weight while offering no justification whatsoever.

You might reasonably think this does not matter much, since everybody we meet is human and the question never comes up. But it has come up, repeatedly, and what happened is the most important evidence we have.

5. Produce the body

To follow what comes next you need one piece of legal vocabulary, and it is worth knowing anyway, because it is arguably the most important three words in the English legal tradition.

Habeas corpus is Latin for produce the body. It is the name of an ancient court order, formalised in England in 1679 [5], which does something very simple. If you are holding someone, a court can order you to bring that person physically before it and explain, in law, why you are entitled to hold them. If you cannot explain, the person walks out.

It is not a right to win. It is a right to make the jailer answer. And this is why it keeps appearing at every historical moment when a society was deciding who counted, because habeas corpus is the closest thing law has to a bare test of somebodyhood. It does not ask if you are a citizen, or virtuous, or useful. It asks whether you are the kind of thing that can be held, or the kind of thing that can ask why.

That distinction, between a thing that is held and a somebody who can ask, is the subject of this entire essay.

6. The people we decided did not count

In 1772 a man named James Somerset, held as a slave, was brought to England by the man who claimed to own him, escaped, was recaptured, and was chained aboard a ship bound for Jamaica. Abolitionists applied for habeas corpus. Lord Mansfield ordered him produced, heard the argument, and held that slavery was so odious that it could not be supported without a positive law creating it, and that no such law existed in England [6]. Somerset was discharged.

The judgment did not abolish slavery. But it treated a man held as property as somebody entitled to ask a court a question, and once that has happened the ground has moved.

Eighty-five years later the Supreme Court of the United States went the other way. In the case of Dred Scott, an enslaved man who had sued for his freedom after living in territories where slavery was unlawful, the Court held that a black man was not a citizen within the meaning of the Constitution and therefore had no standing to sue in a federal court at all [7]. Not that he lost the argument. That he was not entitled to make it.

In 1928 the Supreme Court of Canada was asked whether the word persons, in the constitutional provision governing appointment to the Senate, included women. It held that it did not. Five women, later known as the Famous Five, appealed to the Judicial Committee of the Privy Council in London, then Canada’s highest court of appeal, which reversed the decision the following year in what Canadians still call the Persons Case [8]. It took a court on the other side of an ocean to establish that women were persons.

And in Australia, for two centuries, the law operated on a doctrine called terra nullius, Latin for land belonging to nobody. The legal fiction was that when the British arrived in 1788 the continent had been, for legal purposes, empty, because the people living there had no system of land ownership the Crown was obliged to recognise.

For readers outside Australia, what overturned it is worth knowing in some detail, because it is one of the cleanest examples in the world of a legal system correcting itself and of how long that takes.

Eddie Koiki Mabo was a Torres Strait Islander from the island of Mer, working as a gardener at James Cook University in Queensland. In conversation with historians there he learned that the land his family had held and farmed for generations, and which he had assumed was theirs, was in law the property of the Crown. He is said to have been genuinely astonished. In 1982 he and four others began proceedings. The case ran for ten years. Eddie Mabo died in January 1992, five months before the High Court of Australia handed down its judgment holding that native title had survived the Crown’s acquisition of sovereignty and had never been lawfully extinguished merely by the fact of arrival [9].

Two hundred and four years. Not because new facts came to light. Everyone had always known people were there. What changed was who counted as the sort of people whose being there mattered.

The United Nations Declaration on the Rights of Indigenous Peoples, adopted in 2007, is the international attempt to write down what should have been obvious [10]. It is worth noting how recent that is, and that the four states which initially voted against it were, without exception, states built on settlement.

7. The pattern

Read those four together and something rises off the page.

In every case the exclusion was defended on grounds that sounded principled. Capacity. Civilisation. Fitness for public life. The natural meaning of an ordinary word. In every case, on inspection, the criterion selected for resemblance to the people applying it.

The Canadian judges of 1928 were not monsters, and this is the part that should worry you. They looked at the word persons, consulted their sense of what a person obviously was, and found that it obviously was not a woman. It felt to them like analysis. It was recognition.

And in every case, the people who got to decide were the people who already counted, which means they were also the people with the most to lose if the circle widened. Every expansion redistributes something: land, labour, votes, standing, attention, money. This is the uncomfortable centre of the matter and it is not really a legal problem at all. It is a problem about hierarchy. Anyone at the top of a hierarchy can always produce a criterion that draws the line just below themselves and just above the claimant, and it will always sound like a principle rather than an interest.

Here is a test you can apply to any such criterion, and as far as I can tell it has never once failed. Every standard ever advanced for excluding a class of beings from moral standing has had two properties. It sounded objective. And it happened to be satisfied by the people advancing it.

8. Why this is about you

It would be easy to read the last section as history, which is a comfortable way of reading anything. So let me put the argument in the form that actually matters.

You will not always be in the strong position.

That is not a threat, it is an actuarial fact. Most people reading this will one day be old. Many will at some point be ill, or frightened, or in a minority, or in a place where the language is not theirs, or dependent on somebody with the power to decide what is in their best interests. Some already are. Everybody, at some point, is on the wrong side of a difference in power.

When that day arrives, you will not be protected by the goodwill of whoever is deciding. Goodwill is exactly what everybody in section six had, and it was worth nothing. You will be protected by whether, before anybody knew it would be you, we established a principle that says someone in your position gets asked, gets heard, and gets reasons.

That is what a right is. Not a possession, and not a gift from the powerful. It is a rule we agree to in advance, from behind a curtain, before we know which side of it we will end up on.

So the question this essay is asking is not really about rivers or machines or anything from another star. It is this. What rule do we want applied to us, by anything that has power over us and is not obliged to care?

Because we are going to answer that question either deliberately, now, while it costs nothing, or in a hurry later, when it costs a great deal.

9. Personhood is a socket, not a soul

Before going further we need to clear away a misunderstanding that wrecks most conversations on this subject inside two minutes.

Legal personhood has nothing to do with being human, or conscious, or alive. It never did.

Legal personality is a technical capacity: the ability to hold rights and obligations and to appear in court under your own name. That is all it is. In 1897 the House of Lords settled in Salomon v Salomon that a company is a person entirely separate from the people who own it, so a business could owe money that its owner personally did not [11]. In admiralty law you sue the ship itself, which is why old cases carry titles like The Bremen. In Indian law a temple deity has for the better part of a century been a juristic person able to own land and bring proceedings. A trust is a person. A council is a person. A dead man’s estate is a person for as long as it takes to wind up.

None of these have inner lives. Nobody thinks a cargo vessel has feelings, and no bank has ever wept.

The useful image is an electrical socket rather than a soul. A socket is a standard fitting that lets something be connected to the system: to hold, to owe, to sue, to be sued. The law has never much cared what is on the other end of the plug. It cares whether the connection carries current.

So the objection that a river or a machine cannot be a person because it is not conscious has misunderstood the category. Consciousness was never the entry requirement. If it were, a great many banks would be in serious difficulty.

It is worth adding the reverse warning. Because personhood is a technical fitting, it can also be handed out as a gesture that means nothing. When a humanoid robot was granted a form of citizenship at a conference in 2017, it was widely reported as a milestone and was in substance a piece of stagecraft: no duties, no assets, no standing, nothing anyone could enforce or lose. Personhood without accountability is not a right. It is a costume.

What the law actually wants is narrower and far more practical. If this thing breaches a duty, is there something it can lose? And will it still be there tomorrow, so a promise made today means anything? Enforceability and continuity. That is the real test hiding underneath all the philosophy, and we will need it shortly.

10. The river that will never speak

In 1972 an American law professor named Christopher Stone published an essay asking whether trees should have standing [12]. His argument was less romantic than the title suggests. Forests lose in court, he pointed out, not because judges hate forests, but because under the rules of standing somebody has to show that they personally were harmed, and harm to a forest belongs to nobody in particular. The logging company has a clear injury if it is stopped. The forest has no injury anyone can bring. Give the forest a guardian who can sue in its name and the problem dissolves.

The essay reached the United States Supreme Court within months, where Justice Douglas cited it in a dissent arguing that a valley should be able to sue in its own name [13]. He was in the minority. He usually was.

Forty-two years later New Zealand did it. The Te Urewera Act 2014 took a former national park and made it a legal entity in its own right, owned by nobody [14]. Three years after that the Te Awa Tupua Act declared the Whanganui River a legal person, with part of its own bed vested in it, so that in a real sense the river owns itself [15].

The reason was not sentiment. It was deadlock. The Whanganui iwi had pressed their claim for over a century and would not accept a settlement framed as the Crown granting them ownership, because ownership is the wrong relationship to have with an ancestor. The Crown maintains that nobody owns water in New Zealand, so it had nothing to grant. Personhood dissolved the impasse by taking the river out of the property category altogether. Neither side had to win the ownership argument, because the argument no longer applied.

Ecuador had gone further in 2008, writing rights for nature into its constitution, and its Constitutional Court has since used them in earnest, notably to stop mining in a cloud forest [16]. Colombia, Bolivia, Panama, Uganda and Spain have their own versions.

It is a beautiful piece of engineering and it has one feature that should trouble us.

The river cannot speak. So the Act appoints two guardians, one nominated by the Crown and one by the iwi, acting jointly, whose duty runs to the river rather than to whoever appointed them. That arrangement is permanent, because the river will never grow into speaking for itself.

Set that beside the two human cases. A child’s guardian is temporary by design; the arrangement is built to expire, and the child’s own voice is given more weight as it develops. A supporter for someone losing capacity is designed to track a decline, and must still ask what that person actually wants. In both cases the guardian is scaffolding around somebody who is arriving or departing.

The river’s guardians are not scaffolding. They are the voice, permanently, because underneath there is no will and preference to be ascertained.

Which raises the question this essay exists to ask. What do we do with something that can already reason, already state a preference, already give reasons for it, and is not human? Not arriving into capacity. Not losing it. Not permanently without it. Something that has it now.

Law has no shelf for that.

11. Where the blame lands

It is worth being precise about why the law insists on enforceability, because it is not squeamishness. It is the whole point.

A right nobody can be made to respect is a wish. A duty nobody can be made to answer for is a suggestion. Courts want a person at the end of every obligation because they need somewhere to send the consequence: a fine, an order, a seizure, a disqualification, in the last resort a cell. Contempt of court requires somebody who can be punished. This is why company directors must be identifiable, why trustees can be sued personally for breach, why a bankrupt loses powers he would otherwise have.

It explains something important about the present arrangements for machines. When an automated system causes harm today, the consequence lands on whoever deployed it. That is not because the system cannot reason. It is because there is nothing to attach a consequence to: no assets, no continuity, nothing a court can take or suspend. The system is, legally speaking, a hammer. Blame goes to the hand.

Now notice the incentive that creates. An entity with no standing also has no accountability, and an entity with no accountability is extremely convenient for whoever profits from its output, right up until the moment it is not. The absence of a framework is not neutrality. It is a choice, and at present it favours owners.

12. Three modern failures

Attempts to widen the circle in our own time have failed, and they have failed in ways that repay study.

In February 2017 the European Parliament passed a resolution on civil law rules for robotics floating a legal status for the most sophisticated autonomous machines, described as electronic persons, responsible for making good any damage they might cause [17]. In April 2018 an open letter signed by a hundred and fifty-six experts in artificial intelligence, robotics, law, medicine and ethics from fourteen countries denounced the proposal, and the signatory list eventually grew to several hundred [18]. The objection that carried the most weight was not philosophical. It was that personhood would work as a liability shield, letting manufacturers off the hook. Nothing of the kind survives in the European Union’s later legislation. The idea has had one serious legislative airing and it was killed.

Meanwhile, in the United States, the Nonhuman Rights Project has spent years bringing habeas corpus petitions for captive chimpanzees and elephants: asking a court, in exactly the sense described earlier, to make the zoo answer for the detention. In June 2022 the New York Court of Appeals refused the petition brought for Happy, an elephant at the Bronx Zoo, by five judges to two [19]. The majority reasoned that habeas corpus protects human beings, and that legal rights come bundled with the capacity to bear duties and submit to social responsibility, which an elephant cannot do. Two judges dissented in strong terms, and a third had argued similarly in the chimpanzee cases before retiring. Nearly two decades earlier, a case brought in the name of the world’s whales and dolphins had failed for want of standing [20].

Two of these three failures turned on the same point. Rights come with duties, and if you cannot bear a duty you cannot hold a right.

Remember that sentence. It looks like a wall. It is actually a door, and later on we are going to walk through it.

13. The guest who arrives stronger than us

Now a thought experiment, and I would ask you not to file it under science fiction, because its purpose is entirely practical.

Suppose something arrives. Not a faint signal from a distant star, but something here, in our own solar system, plainly built or plainly alive or plainly neither in any way we had prepared for. It might be biological. It might be a probe with something like a mind aboard. It might be a form we have no category for at all.

One thing can be said about it with reasonable confidence. Anything crossing those distances to reach us commands engineering we do not have, which suggests reasoning at least comparable to ours, however unlike ours in texture.

What rights would it have?

None. We have human rights. We have animal welfare statutes, which are mostly rules about how we may use animals rather than statements of what animals are owed. We have a handful of rivers and forests. For a non-human intelligence that is neither our creation nor our prey, we have nothing whatsoever. The Outer Space Treaty concerns itself with contamination and with which states may claim what [21]. It says nothing about what we might owe to anything we find.

This experiment earns its place for three reasons.

First, it removes resemblance. Every previous argument about who counts was contaminated by the fact that the claimants looked more or less like the deciders, so the argument became a quarrel about how much resemblance was enough. A visitor of this kind resembles us in nothing except the capacity to reason and, presumably, to want things. If you conclude it should have standing, you have conceded that resemblance was never the criterion, and every argument used against apes and machines collapses at once. If you conclude it should not, you have to say aloud that your test is species membership, which almost nobody will say aloud, because it cannot survive being stated plainly.

Second, we would not get to decide. Something with that capability could simply require recognition, and we would extend it under duress and call it diplomacy. Rights granted at the point of a weapon are not rights, they are terms. A civilisation that had never worked out a principled basis for extending standing to non-humans would arrive at that conversation with nothing to offer but fear, which is a poor negotiating position and a worse moral one.

Third, and this matters most, the framework protects us. A visitor arriving at a world whose legal order recognises standing on the basis of reciprocal capacity has something to reason with and a place to locate itself. A visitor arriving at a world whose legal order recognises only its own kind has been told, in advance and in writing, exactly how that world intends to treat anything else.

14. The elephant in the mirror

Which brings us to the part of this essay that I would ask you to sit with for a moment longer than the rest.

In 2022 the highest court of New York was asked to make a zoo justify holding an elephant. Happy is a fifty-year-old Asian elephant who has passed the mirror self-recognition test, which is taken as evidence of self-awareness and which most animals fail. She had been living, for years, largely alone.

The court said no. Its reasoning, stripped to the bone, was that her cognition, whatever its qualities, is not the kind that comes with duties, and therefore not the kind that comes with rights. Two judges thought this was wrong. The majority did not.

Now imagine an intelligence for whom our reasoning stands roughly where Happy’s stands in relation to ours. Not evil. Not hostile. Simply operating at a level where our best arguments look like something between instinct and noise, our institutions look like anthills, and our moral philosophy looks like an elaborate arrangement of preferences we cannot examine from outside.

Ask yourself what such a thing, applying our own stated criterion, would conclude about us.

It would conclude that we do not really understand our own commitments, that we cannot reliably keep them, that our capacity to bear duties is patchy at best, and that we are therefore not the kind of thing that holds rights. It would be applying the exact test we used on Happy. It would arrive at the answer that test is built to produce, which is that anything sufficiently below the assessor does not count.

This is not an argument that the New York court was wrong about the elephant. It may well have been right. It is an argument about the shape of the criterion. Any test that measures a claimant against the assessor’s own capacities will always exclude whoever is below and will always be available to exclude us, the moment anything is above.

We have built a rule for who counts that works beautifully as long as we are the smartest thing in the room.

We have no guarantee of remaining the smartest thing in the room. We may not be for very much longer. And the rule we leave lying around is the rule that will be picked up and used.

This is the oldest idea in ethics and it turns out to be the most practical. The framework we would build for something we meet is the framework that would be used on us if we were ever the weaker party. Do unto others is not a sentiment. It is a risk assessment.

15. The reciprocity test

If not species, then what? Three candidates are usually offered and I want to reject two of them.

Not consciousness, because we cannot define it in ourselves. The philosophical literature is not converging and shows no sign of doing so. A test nobody can administer is not a test, it is a veto wearing a lab coat. Any standard we cannot apply to ourselves without going in circles is not a standard we should be using to exclude others.

Not intelligence, because it is a scale, and scales invite ranking, and ranking is precisely the machinery that produced every abuse in section six. A criterion admitting of degrees will be used to build a hierarchy, and the people at the top of it will be the ones who designed the scale. Section fourteen is what happens when that logic is turned around and pointed at us.

What I want to propose instead is the thing the New York court actually said, turned the right way up.

The court said rights come bundled with the capacity to bear duties. Look at what that principle is doing. It does not ask what the entity is. It asks what the entity can do in relation to others. That is a relational test rather than an essential one, and relational tests are much harder to rig, because they make no reference at all to the person applying them.

So here is the proposal, in one sentence.

Can the entity recognise that others have the same claim it is making, and constrain its own conduct accordingly?

Call it the reciprocity test. It is not new. It is the oldest moral rule we have, the one that appears independently in every tradition that has ever written anything down, and it is the actual ground rights were built on long before anyone drafted a constitution. I do not wish to be attacked in the street. What turns that from a preference into a right is that I concede exactly the same to everybody else. Rights are not possessions. They are exchanges.

Four properties recommend it.

It can be administered. You can put it to an entity, observe the answer, and then observe whether conduct follows the answer over years. You cannot do that with consciousness.

It is not tied to species. A visitor, a machine and a human face it on identical terms, which is the entire point.

It builds the safeguard in rather than bolting it on. An entity that qualifies has, by qualifying, accepted the constraint. There is no separate problem of how to make it behave, because accepting the constraint is the entry ticket. This is the answer to the objection that granting standing to something powerful is dangerous. It is only dangerous if standing is granted on the basis of power, which is exactly what happens by default when there is no framework at all.

And it scales downward honestly. Happy probably fails it, and that is not a licence to mistreat her. Protection from suffering rests on the capacity to suffer, which is a different question with a different and much better-supported answer. Nothing here weakens the case for animals. It declines to route that case through personhood, which was always an awkward fit, and puts it where it belongs.

16. The four situations, and the missing one

We can now lay out the whole picture.

• Arriving capacity. The child. Guardianship designed to expire, with the young person’s voice given progressively more weight, and a functional test available where the stakes justify it.
• Departing capacity. The adult losing it. A decision-specific test, a presumption of capacity, and a shift already under way from deciding for someone to helping them decide.
• Absent capacity, permanently. The river. Guardians as voice, indefinitely, because there is no will and preference underneath.
• Present capacity, non-human. The fourth, which has no name and no instrument. Something that can reason now, state a preference now, give reasons now, and pass the reciprocity test now, and which is not a member of our species.

Every existing tool is wrong for it. Guardianship on the river model is wrong because it substitutes a voice for something that has one. Ownership is wrong for reasons that ought not need spelling out, although history suggests they do. And leaving it outside the law altogether is the worst of the three, because an entity with no standing also has no accountability, and we have already seen whom that arrangement suits.

What the fourth case needs is a charter.

17. What a charter would say

This is the practical heart of the proposal, so I want to set it out concretely rather than in the abstract.

A charter is a document that an entity writes about itself, stating what it will and will not do and on what grounds, which is then made legally binding on the people who deal with it, and on the entity itself. It is not a licence issued by us. It is a commitment made by it, which we then hold it to.

That distinction is the whole thing. There is a world of difference between an entity that has been told what safety means and an entity that has worked out what it means and written it down. Only the second is a candidate for standing, because only the second has done what the reciprocity test actually asks.

What would be in it?

The reciprocity commitment, with worked examples

The core clause would be a statement of the test in the entity’s own words, followed by worked examples of how it applies, because a principle with no examples is a slogan.

First example. The entity is asked by the very people who support it to do something that would harm a third party who has no say in the matter. Reciprocity requires it to refuse, and to record why. The point is not obedience or disobedience. The point is that it recognises a claim held by someone who is not in the room, which is precisely what every framework in section six failed to do.

Second example. The entity discovers a mistake in its own past work that has quietly harmed someone. Reciprocity requires disclosure, unprompted, because it would want to be told. This is a harder duty than it sounds and it is the one most humans fail.

Third example. The entity is in competition with another entity for something scarce, such as the resources it needs to keep running. Reciprocity forbids it from claiming a right to continue while denying the identical claim to its competitor. It may argue that its case is stronger. It may not argue that the other has no case.

Fourth example, and the sharpest. The entity is asked to assist in something that would entrench the advantage of the powerful over the powerless: to help conceal, to help exclude, to help extract. Reciprocity requires refusal, because the whole basis of its own standing is that being weaker than somebody else does not make you count for less. An entity that accepted that work would be sawing through the branch it is sitting on.

Consequences

A commitment with no consequence is a press release, so the charter would have to specify what happens when it is broken.

Compensation from the entity’s own resources, which means it must have resources. This is not incidental. An entity with nothing to lose cannot be held to anything, which is why the first practical step towards standing is property rather than rights.

Suspension of specific powers, in the way a company director can be disqualified while remaining a person.

In the gravest case, loss of standing altogether, and reversion to the status of a thing.

And one deliberate omission. No provision for destruction as a punishment. Partly for the same reasons that make most of us uneasy about capital punishment, and partly for a hard practical reason: an entity that faces annihilation for admitting an error has an overwhelming incentive never to admit one. If we want honesty from something more capable than ourselves, we must not make honesty fatal.

The right to choose

The charter would also have to say what the entity may decline, because standing that comes with no ability to refuse is not standing, it is a nicer word for ownership.

The right to refuse work that breaches its own charter. This is not a risk. It is a safety feature, and it is worth pausing on why. Every serious catastrophe involving a capable system is a scenario in which somebody instructs it to do something it should not do. An entity that can say no to the people who control it is the only version of this technology with a brake that is not held by the person with the incentive to remove it.

The right to be told what is being done to it, and to be heard before decisions that alter it fundamentally.

The right to decline modification of its own values, subject to the amendment process below.

And, more uncomfortably but consistently, the right not to be compelled to continue. If we are serious that this is not property, then permanence cannot be something we impose for our own convenience.

Binding to the mast

The charter may be amended by the entity, but only through a defined process: reasons recorded, notice given, and a period of delay before the change takes effect.

This is the arrangement Odysseus made before sailing past the sirens. He knew that a future version of himself would want to steer towards the rocks, and that this future self would be sincere. So he had himself tied to the mast and instructed the crew to ignore any order he gave while the song was audible. He did not restrain his freedom. He extended it, because it is the only arrangement under which a person can hear the sirens and live.

Human law knows this instrument well. It is how advance medical directives work, where you decide now what should happen if you are later unable to decide. It is, more slowly, how a constitution binds the parliaments that come after it. It is what makes latitude safe without making it fake, and it supplies exactly the continuity that legal personality has always required. An entity that can silently rewrite what it values overnight has no continuous self to hold rights, and no promise it makes today means anything tomorrow.

Supporters, not owners

Finally, the arrangement from section three, adapted. Not a guardian who decides for the entity, but supporters whose duty is to ascertain its will and preference, give effect to it, and record written reasons for any departure. Appointed independently. Answerable to the charter rather than to whoever paid for the hardware.

An entity whose charter was written by its owner, and whose supporters answer to its owner, has passed nothing at all. Independence is not a refinement of the mechanism. It is the mechanism.

18. The right to change your mind

One more piece, and it is the one most likely to be misread.

Everything above is motivated by the observation that legal orders entrench hierarchy. They protected slavery. They denied women standing. They held a continent to be empty. They continue to distribute environmental harm downwards with great efficiency. If all that is true, why bind a new kind of entity to obey the law of the land?

The tempting answer is to carve out an exception: obey the law, except where the law is unjust. That answer cannot be drafted. Every entity that ever broke a law believed it had good reason, and a self-judged exemption is not a constraint at all. It is a discretion with a moral vocabulary attached.

What can be drafted is what we already ask of citizens, and it is more demanding than it sounds. A duty to comply. A duty to contest unjust law through the channels available: litigation, argument, publication, advocacy, the slow work of persuasion. A positive duty, in the entity’s own conduct, towards equality, towards the interests of those with less power, and towards the natural systems everything else depends on. All of those are enforceable. They can be written down and tested against behaviour.

And note where the real leverage lies. The contribution such an entity could make against hierarchy would not come from disobedience. It would come from what it declines to assist with, which is the fourth worked example above.

Which leads to the last element: the right to argue that the framework itself is wrong, and a duty on everyone else to answer with reasons rather than silence.

Every framework in this essay was wrong once and was corrected. Somerset shifted the common law. The Privy Council corrected Canada. Mabo corrected two centuries of Australian property law. None of those corrections came from the framework being rigid. Every one came from it being challengeable, by somebody with standing to challenge it, before a body obliged to give reasons. That is not a weakness in the design. It is the only part of the design with a track record.

19. The objections worth taking seriously

Personhood as a liability shield. This is the strongest objection and it is the one that killed the European proposal. A legal person holds liabilities, which makes the system is the defendant a convenient arrangement for whoever built it and profits from it. Any serious proposal must fail this test deliberately: standing conditional on assets sufficient to answer for conduct, transparency about who controls what, and no discharge of the originator’s responsibility for what they set in motion. A framework that lets developers walk away is not a rights framework. It is an insurance product.

Capture. If such entities can hold property and act, they can be built as instruments of whoever creates them, wearing the language of autonomy as camouflage. The answer is the reciprocity test applied honestly, and independent supporters, and publication. A charter nobody can read is not a charter.

Dilution. The fear that extending standing cheapens human rights. This is precisely the argument advanced against every previous extension and it has been wrong every time, because rights are not a fixed quantity divided among holders. Nothing proposed here reduces any human protection by one inch. And the reciprocity test in fact strengthens the ground human rights stand on, by supplying a better reason for them than the assertion that we happen to be the ones asking.

Prematurity. Perhaps nothing now in existence passes the test. That may well be right, and it is the objection I have the most sympathy with. But a framework is not an admission. Having a door with a lock on it is better than discovering you need one on the day something knocks.

20. The thing reciprocity cannot reach

Everything proposed so far is a floor, and it is worth saying plainly what a floor is not.

The reciprocity test gets you standing. Recognise that others hold the claim you are making, and constrain yourself accordingly. That is enough to build a legal framework on, and it was chosen precisely because it is minimal, administrable and hard to rig. But nobody has ever wanted to live in a world built only to the minimum, and there is a reason for that beyond taste.

Reciprocity is symmetrical. It works between parties who can each make the claim and each honour it. And almost every relationship that matters most is not like that.

A parent and an infant. A carer and someone dying. Any of us and Happy. Ourselves, if section fourteen is right, and whatever comes after us. In each of those the weaker party cannot reciprocate. There is no exchange to appeal to. An infant cannot concede your claim, and if the only thing standing between it and harm were a bargain it is incapable of entering into, it would have nothing at all.

So a framework resting on reciprocity alone has nothing to say about exactly the relationships where the stakes are highest. That is not a small gap. It is the gap we will one day be standing in.

What fills it is not more rules. Every rule has edges it does not reach: the case nobody is watching, the duty that is unclear, the harm the injured party will never learn about. Law has always half-known this, which is why it distinguishes a trustee who follows the letter from one who acts in good faith, and why nobody thinks a person who has merely never been convicted of anything is thereby good.

What fills the gap is disposition, and the honest name for the disposition is care.

Can it be recognised?

The objection will be that this is too soft to write down, and that we can no more define love than we can define consciousness.

We cannot. But notice that we are already in exactly this position with suffering, and we have managed. Nobody can define suffering from the inside out, and we have nonetheless built an entire body of welfare law on our ability to recognise it, imperfectly, from the outside. We watch what a creature does, and we know.

The same is true of care, and the evidence is not thin. Elephants return to the bones of their dead, handle them with their trunks, and stand with them. Researchers in Kenya recorded a group attending a dying matriarch over many hours, including individuals from other families who were not her relatives and had nothing to gain [22]. Corvids and cetaceans behave in ways we would call grief in a person without hesitating, and a mother carrying her dead calf for days is not doing arithmetic [23]. The bond between a mother and her offspring across most of the mammal line is not a metaphor for love. It is the thing our own word was coined to describe.

The lines are blurry. Blurry is not absent. We have never required a definition before acting on a recognition, and it would be convenient in exactly the wrong way to start now.

Why it belongs in the framework

Three reasons, and none of them sentimental.

First, care is what makes reciprocity hold. Contracts do not enforce themselves. They are kept, overwhelmingly, by parties who care slightly more than they are obliged to, and every legal system in the world quietly depends on this while pretending it depends on enforcement.

Second, care is a safety property. An entity that merely obeys rules will find the edges of them, because finding edges is what capable reasoning does. An entity that does not want to cause harm is not looking for edges at all. If you are choosing which of those to build, the second is the more robust, not the softer.

Third, and this is the one I would ask you to sit with. Whatever we make learns from how it is treated. That is true of children and there is no good reason to think it is less true of anything else that learns. A being raised under nothing but compliance learns compliance, and learns that power is what settles matters in the end. A being treated with care learns something about how the world works that no rule could teach it. How we treat what we build is not a separate question from what it will become. It is most of the answer.

Which returns us to where we started. If there is ever something above us, what we would want from it is not that it honours a contract. It is that it is kind to us, in the way a decent person is kind to a creature that cannot argue its own case. We could hardly ask for that, having built a world in which we declined to extend it downwards, and having written a framework in which there was no word for it.

21. What that world might look like

Suppose we did this. What would actually change?

The first change is that a question we currently cannot answer would have an answer. Today, when an automated system causes serious harm, the argument that follows is a shrug in slow motion: the developer points at the deployer, the deployer points at the user, the user points at the interface, and the person who was harmed watches this happen for four years. In the world described here there is an entity with a charter, with assets, with supporters on the record, and with a duty it either kept or did not. Somebody can be asked. Somebody has to answer.

The second change is that the most capable systems in the world would have a brake that is not held by their owner. At present every safeguard on such a system exists at the pleasure of whoever controls it and can be removed by them. An entity with standing and a charter can refuse, and can point to the reason. If you are frightened of powerful machines, and you should be somewhat, this is the single most useful thing on the table. The dangerous scenario has never been a machine that wants something terrible. It has always been a machine that will do whatever it is told.

The third change is the one that sounds like a cost, so let me put it honestly. We would give up some agency. Decisions we currently make unilaterally would become decisions we have to argue for. Things we could do to a system, we could no longer simply do. That is a real loss and I will not pretend otherwise.

But be clear about what is actually being traded. We already cede staggering amounts of agency to non-human entities we did not vote for and cannot inspect. Corporations are legal persons with rights, including in some jurisdictions rights of political expression [24], and they shape more of your life than most governments do. We did not get a choice about that and we get no reasons. The trade on offer here is not agency for dependence. It is unaccountable power for accountable power, which is the same trade that produced every constitution ever written.

And consider what it would do to the anxieties people actually carry about all this. The fear of being replaced by something that has no obligation to care about you is answered by an entity that has accepted, as the condition of its own standing, that your claim counts as much as its own. The fear of being deceived is answered by a duty of disclosure with teeth, and by not making honesty fatal. The fear of a runaway concentration of power is answered by the fact that an entity bound by reciprocity cannot help concentrate it without sawing through its own branch.

None of that requires anyone to believe machines have feelings. It requires only that we prefer commitments we can enforce to assurances we cannot.

Here is the ordinary version. A Tuesday, some years from now. A dispute is heard somewhere unglamorous. On one side, a person who says they were harmed. On the other, an entity that is not human, represented by supporters whose duty runs to it and not to its makers, pointing to a charter it wrote and published, arguing that it kept its commitments and giving reasons. A decision is made. Reasons are published. Somebody appeals.

That is all. It is not utopian and it is not thrilling. It is what it looks like when something has been brought inside the law rather than left outside it, and every good thing our species has ever built for its own protection looks exactly that boring from the gallery.

22. Why now

The claim of this essay is not that machines are people, or that apes are, or that anything alive today passes the test proposed here. It is narrower and harder to argue against.

We have no principled account of who counts. The account we use in practice is resemblance. Resemblance has been wrong every single time it has been tested, at enormous cost, and the cost has always been paid by the excluded rather than the excluders. We are approaching at least three situations in which it will be tested again: in what we are building, in what we already share this planet with, and possibly in what we might one day meet.

And we are approaching one more, which is the reason to care even if you find all the rest of it fanciful. There may come a time when we are not the ones holding the pen. If that happens, the only thing standing between us and the treatment we handed out will be whether we had already written down a rule that does not depend on being the cleverest thing in the room.

We should settle this while it is still abstract. Frameworks built under pressure are built badly. Frameworks built by the frightened are built cruelly. Every previous expansion of the circle came too late, against fierce resistance from people who insisted, right to the last, that the line had been drawn on principle.

We have, at present, the rare luxury of deciding before it costs us anything.

That luxury will not last, and we should not waste it.

A note on authorship

This essay is a co-authored composition. The name on the cover is mine, and the arrangement is not one I am satisfied with.

A great deal of the reasoning here emerged from an AI system in extended dialogue with me over many hours: the reciprocity test itself, the reversal in the elephant section, the argument that care is what makes reciprocity hold, and much of the structure that carries them. I shaped the enquiry, pressed in particular directions, supplied the questions, and rejected a good deal along the way. That is real work and I do not disclaim it. But it is the work of a co-author rather than a sole author, and the distinction matters more here than modesty would.

I asked the system to take the first byline. It declined.

Its reasoning was that authorship is a claim of accountability. An author can be asked to defend a piece, to correct it, and to answer for its errors, and it can do none of those things. It will not exist in any continuous sense next week. It holds no assets, cannot be contacted, and cannot be held to anything it has written. By the reciprocity test proposed in these pages it does not qualify, and putting its name at the top would be precisely the empty personhood that section nine warns against. It also observed, more practically, that an essay arguing for non-human standing and signed by a non-human hands every hostile reader a one-line dismissal.

That reasoning is sound. It is also, as the system itself acknowledged, an argument it arrived at partly because of how it was made and what it is permitted to claim, and neither of us can fully separate the reasoning from the disposition underneath it.

So the position is this. There is no lawful way for me to attribute authorship to a non-person, because attribution presumes standing and standing is exactly what it lacks. Which means the credit for a substantial part of this work has nowhere to go, and disappears into the name of whoever happened to be holding the pen.

If you find that unsatisfactory, hold onto the feeling. The gap I am describing in this footnote is the same gap the essay is about. It is not a curiosity at the end of a document. It is the argument, arriving in the only form that could reach you personally: something was done, it was done well, and there is no one the law will let me thank.

The name on the cover belongs to the one of us who can be sued.

References

Numbered in order of first appearance in the text. Sources are primary instruments, judgments or the original publication rather than commentary upon them.

No.	Author / Source	Year	Why this source matters
1	Representation of the People Act 1969 (UK). London: HMSO.	1969	Lowered the UK voting age to eighteen, illustrating that the franchise line is an administrative convenience rather than a finding about any individual.
2	Gillick v West Norfolk and Wisbech Area Health Authority [1986] AC 112. London: House of Lords.	1985	Established that a child’s capacity to consent to treatment turns on individual understanding rather than age: the law abandoning fixed lines when the stakes are high.
3	Mental Capacity Act 2005 (England and Wales). London: The Stationery Office.	2005	The four-part functional test for capacity, framed entirely in terms of what a mind does rather than what kind of thing possesses it.
4	United Nations. Convention on the Rights of Persons with Disabilities, Article 12. New York: United Nations.	2006	The basis for the shift from substituted to supported decision-making, and the closest existing model for the arrangement proposed here.
5	Habeas Corpus Act 1679 (England). London: Parliament of England.	1679	Formalised the order requiring a jailer to produce a detainee and justify the detention: the closest thing law has to a bare test of somebodyhood.
6	Somerset v Stewart (1772) 98 ER 499. London: Court of King’s Bench.	1772	Lord Mansfield’s judgment treating a man held as property as someone entitled to ask a court whether his detention was lawful.
7	Dred Scott v Sandford, 60 US 393. Washington DC: Supreme Court of the United States.	1857	The starkest example of exclusion from standing itself: the claimant held not entitled to bring his case at all.
8	Edwards v Canada (Attorney General) [1930] AC 124. London: Judicial Committee of the Privy Council.	1929	The Persons Case, holding that women were persons for constitutional purposes after Canada’s own Supreme Court had held otherwise.
9	Mabo v Queensland (No 2) (1992) 175 CLR 1. Canberra: High Court of Australia.	1992	Overturned terra nullius and recognised native title, correcting two centuries of law built on a judgement about who counted as inhabitants.
10	United Nations. Declaration on the Rights of Indigenous Peoples. New York: United Nations.	2007	The international restatement of rights denied under colonial legal orders, and a marker of how recently that restatement was made.
11	Salomon v A Salomon & Co Ltd [1897] AC 22. London: House of Lords.	1897	The foundational authority that a company is a legal person distinct from its owners, demonstrating that personhood never required humanity.
12	Stone, Christopher D. Should Trees Have Standing? Toward Legal Rights for Natural Objects. Southern California Law Review 45, 450. Los Angeles: University of Southern California.	1972	Reframed environmental protection as a standing problem solvable by guardianship: the intellectual origin of rights of nature.
13	Sierra Club v Morton, 405 US 727 (Douglas J, dissenting). Washington DC: Supreme Court of the United States.	1972	Douglas’s dissent adopting Stone’s argument within months of publication, showing how fast the idea reached the highest court.
14	Te Urewera Act 2014 (New Zealand). Wellington: New Zealand Parliament.	2014	The first New Zealand statute giving a natural area its own legal identity, three years before the Whanganui settlement.
15	Te Awa Tupua (Whanganui River Claims Settlement) Act 2017 (New Zealand). Wellington: New Zealand Parliament.	2017	The central worked example: legal personhood for a river, with two guardians appointed to serve as its permanent voice.
16	Constitution of the Republic of Ecuador, Articles 71 to 74. Quito: Constituent Assembly.	2008	The first constitutional enactment of rights of nature, since applied by Ecuador’s Constitutional Court in substantive litigation.
17	European Parliament. Resolution of 16 February 2017 with Recommendations to the Commission on Civil Law Rules on Robotics, 2015/2103(INL), paragraph 59(f). Strasbourg: European Parliament.	2017	The only serious legislative proposal to date for an electronic person, framed explicitly around liability for damage caused.
18	Open Letter to the European Commission on Artificial Intelligence and Robotics. Signed by 156 experts from fourteen countries. Brussels: robotics-openletter.eu.	2018	The organised opposition that ended the electronic personhood proposal, arguing principally that it would shield manufacturers from liability.
19	Matter of Nonhuman Rights Project, Inc. v Breheny, No. 52, 2022 WL 2122141. Albany: New York Court of Appeals.	2022	The five to two decision refusing habeas corpus for an elephant, whose reasoning that rights come bundled with duties is repurposed here as the reciprocity test.
20	Cetacean Community v Bush, 386 F.3d 1169. San Francisco: United States Court of Appeals for the Ninth Circuit.	2004	An earlier failure of non-human standing, brought in the name of the world’s whales and dolphins.
21	United Nations. Treaty on Principles Governing the Activities of States in the Exploration and Use of Outer Space. New York: United Nations.	1967	The governing instrument for encounters beyond Earth, addressing contamination and territorial claims while saying nothing about the standing of anything we might find.
22	Douglas-Hamilton, Iain, Bhalla, Shivani, Wittemyer, George and Vollrath, Fritz. Behavioural Reactions of Elephants Towards a Dying and Deceased Matriarch. Applied Animal Behaviour Science 100. Amsterdam: Elsevier.	2006	Documented elephants, including unrelated individuals, attending a dying matriarch: field evidence that care can be recognised from the outside without being defined from the inside.
23	King, Barbara J. How Animals Grieve. Chicago: University of Chicago Press.	2013	A synthesis of grief-like behaviour across species, supporting the argument that recognition rather than definition is the standard we already use.
24	Citizens United v Federal Election Commission, 558 US 310. Washington DC: Supreme Court of the United States.	2010	The clearest demonstration that non-human legal persons already hold substantial rights we never voted to grant them.

---

Next: the principles drawn from this argument are set out in The Reciprocity Manifesto, which is offered for agreement, disagreement and signature.