Borrowed Trouble

Sooner or later, every moral theory gets walked down the gallery of famous problems — the runaway trolley, the repugnant conclusion, the monster who turns a loaf of bread into more happiness than a whole city could hold, the machine that grinds the world into paperclips — and told to give its ruling on each, as if the right answers were a debt it owed. This essay turns down that tour in its usual form, and the turning-down is the point. A lot of these problems are hard not because morality is hard, but because of how they’re built. Each one freezes something the Arrow says has to keep moving, or demands some fixed thing — a number, a rule, a property, a state, a scorecard — that the framework has already shown there’s no honest way to hand over. Where that’s what’s happening, the job is to show how the trouble was manufactured. And where a real difficulty is still standing after that, the job is to point a direction, and then, without apology, hand the rest to you.

The problems that aren’t tests

There’s a standard way to check a moral theory. You walk it past the hard cases the field has been polishing for a century — each one a little machine built to squeeze out a gut reaction — and you see whether the answers it gives match the reactions the cases were designed to squeeze. A theory that calmly says push the big man off the bridge has flunked. A theory that lands back on common sense has passed. The framework in this book mostly won’t be graded that way, and you deserve to know why, because from the outside the refusal can look like a theory ducking the exam.

It isn’t. The refusal comes straight out of something the book has said from its first pages: morality is a direction, not a destination — judged as more or less, never held up against some fixed mark kept still for the measuring. Most of the famous problems quietly assume the opposite. They assume there’s a single thing the right answer is made of — a quantity to add up, a rule that never bends, a property a creature either has or lacks, a state worth reaching, a cosmic scorecard that grades you by how things turn out — and then they build a case where that one thing coughs up a monstrous answer, or no answer, or two answers at once. The monstrousness is real. But it belongs to the assumption, showing its true face under pressure; it isn’t a discovery about morality. To give the case the ruling it’s fishing for is to swallow the assumption that built it — and that assumption is exactly what the framework is arguing with.

So here’s the plan: not to solve the problems but to take each one apart the same way, in three moves. First, name the fixed thing the problem smuggles in — the piece it borrowed from the view from nowhere, that imagined standpoint outside everybody’s point of view, without which the puzzle would fall right over. Second, set the case in the framework’s own terms: is an agent’s sense of what matters and how to act growing more coherent across a widening slice of the world — or is the neatness being bought by narrowing that slice, by shutting out a fact, a viewpoint, someone affected? That second thing — coherence bought by shrinking the context — is the counter-dynamic, the framework’s mark of going wrong, and it stands guard the whole way through. Third — and this is the move people skip — say plainly what’s left over: where the framework honestly doesn’t settle the case, and won’t pretend to choose for you. That leftover isn’t the framework trailing off mid-sentence. It’s the framework keeping a promise it made on the first page.

One thing ties all of it together, and it’s worth saying out loud because it sounds like a joke and isn’t. A famous thought experiment and a cult get their airtight feel by the very same trick: each keeps its grip by keeping things out. The cult holds onto its certainty by letting in no voice that might shake it; the thought experiment holds onto its sharpness by letting in no history, no aftermath, no third option, no extra fact. What comes out the other end is a verdict about a world that can’t exist: strip all of that away and what’s left isn’t a simplified situation, it’s no situation at all. Every model simplifies; what matters is what you do next. A physicist who assumes a frictionless plane puts the friction back before building anything. This verdict doesn’t stay in the sealed room — it gets carried out and spent in worlds that do exist, with nothing put back, long after the puzzle has been set aside.

The framework’s name for coherence propped up by keeping things out does belong here, but it needs placing with care: a puzzle is a device, and a device doesn’t shut anything out — it only stipulates. The name belongs to the person, and it lands the moment someone is shown what the case left out and declines to look, because looking would cost the crisp answer. It doesn’t go soft just because the narrowing is now serving a philosopher instead of a guru. That’s why the cases that feel the most watertight are so often the ones built on the most ruthless exclusions — and why the framework’s first move, again and again, is to open the window the case was built to keep shut. These are, in a precise sense, borrowed troubles: hard spots that came bundled with a standard the framework never agreed to keep.

The single number

Start with the puzzles that run on arithmetic. Parfit’s Repugnant Conclusion is the cleanest. Allow that well-being can be added up across people into a grand total, and you’re pushed, one reluctant step at a time, into agreeing that a vast crowd of lives each just barely worth living — billions of them, hovering a hair above zero — adds up to something better than a small population of deeply happy ones. Almost nobody believes it. The engine that cranks it out is plain addition, and the recoil you feel is the crash between that sum and a conviction the sum has no way to hold.

The framework carries no such total to be cornered by. On its account, value lives in agents — in their values-models, their working picture of what matters — and the verdicts that count, real progress and the counter-dynamic, get reached by comparison, by one thing clearly outranking another, not by measurement. There’s no single yardstick: some states of the world just don’t line up on one scale to be ranked at all. Read that way, “keep adding lives at the barely-worth-it line until the total wins” isn’t a hard truth you grit your teeth and accept — it’s a textbook case of narrowing. It buys a bigger number by flattening out the texture of what those lives actually care about, and by cutting the thread along which a flourishing people deepens what it values and gets better at acting on it. The framework doesn’t see two totals, one quietly beating the other. It sees two directions — one widening, a society growing richer in meaning and ability; one narrowing, copies held at the edge of bare survival — and the arrow points to the first, with nothing added up at all.

What the framework won’t give you is the number. It won’t name the right size for a population, or rank every possible world against every other; it has no single quantity to do that with, and it says so plainly. What it tells you is that the maximize-the-total setup is the artifact — and it hands you a direction to read a real case by. Whether a particular path of growth widens coherence or narrows it is a judgment about that path, in its actual circumstances — exactly the judgment the arithmetic was built to spare you, and exactly the one the framework hands back. The same diagnosis quietly undoes the Mere Addition paradox, where the trouble is the insistence that every comparison link up into one complete ranking; Parfit’s Non-Identity problem, where it’s the assumption that a wrong needs a specific person who was made worse off — a list of identifiable victims, which the framework swaps out for the direction of a choice across the generations it shapes; and Nozick’s Utility Monster, the creature so good at turning resources into bliss that pure addition says feed it everything and let the rest of us go hungry. Once value lives in distinct agents rather than in one interchangeable substance, there’s no common currency to pour down the monster’s throat — and a single maximally efficient consumer isn’t the good winning, it’s the good collapsing into a single point. The counter-dynamic, wearing the winner’s medal.

The sealed room

A second family runs not on arithmetic but on architecture. The Trolley problem is the monument. A runaway car is about to kill five; a lever shunts it onto a side track where it kills one; almost everyone pulls the lever. Now change the machinery — to stop the car you have to shove a large man off a footbridge into its path, same body count — and almost everyone refuses. The puzzle is the gap between the two: why is the death you cause with a lever allowed, and the death you cause with a shove not, when the math is identical? Whole libraries have grown in that gap.

The framework’s first move is to refuse the room. Because the case only works sealed shut: two options and no third, perfect knowledge of how each plays out, no history before the first second and no aftermath after the last, no relationship between any of these people, no community that will remember what you did. That sealing isn’t an innocent simplification. It takes away exactly the stuff you’d widen with — the history, the aftermath, the third option, the extra fact — and leaves you only what the case stipulates to reason from. Whatever you conclude in there, you’re concluding it about a room nobody will ever stand in. AoM’s three grades of judgment sharpen what can honestly be said here. Over a given, walled-off frame — the sealed room exactly as drawn — the framework offers only its weaker grade, the one it calls right-in-principle: the best coherence available inside that frame, and there the lever reading makes perfect sense. But the grade it saves for moral is tied to the widening frame, and the facts that frame turns on are the very ones the puzzle forbids — who these six people are, what trust and what institutions the act builds or breaks, what kind of person each choice trains you into. The whole move is the shift from which act is allowed? to which response widens the coherence of what I value and how I act, in the real, un-sealed situation I’m actually standing in?

This is also, quietly, why the footbridge feels different from the lever — and why Foot and Thomson’s Transplant case, the surgeon who could cut up one healthy patient to save five, feels worse still. The framework doesn’t reach for a rule against using people as tools. It notices that harvesting the patient blows up a whole context the trolley’s seal conveniently hid: the trust that lets anyone go to a doctor at all, the institution of medicine, the world the survivors have to live in afterward. The shove is closer to the harvest than to the lever because it, too, reaches its tidiness only by pretending the wider world away. The rest of the room’s puzzles come apart along the same seam. Kant’s Inquiring Murderer, where an absolute rule against lying smashes into the killer at the door, is the stiffness of a fixed rule generating its own monstrous exception; Scheffler’s Paradox of Deontology, where a rule forbids you to break it even once to prevent five identical breakings, and the Catastrophe Exemption, where even an absolute rule is suddenly allowed to bend if the stakes climb high enough — each is the same rigid rule cracking under a pressure the framework never loads onto it, because the framework never asked morality to be a rule held stiff against a world that keeps moving.

What the framework holds back here is the crisp ruling on the bare track — pull, don’t pull — and the holding-back is the lesson, not a hole in it. A theory that snapped out a clean answer to a context-stripped prompt would be doing, in the very act of answering, the exact narrowing it warns against. Put the context back — the context the case deleted — and a direction comes back into view. And that putting-back is work only you can do, standing in a real version of the case.

There’s a sharper version of the objection, and it’s worth meeting head-on: the trolley’s cruelest trick is that it won’t let you put the context back. It weaponizes the clock — act now, there’s no time to widen anything. Grant that in full and the arrow still doesn’t break; it only shows what it was made of. An agent with four seconds can’t run the context fresh — it acts from the context it has already taken in, the compressed wisdom of its rules and the standing shape of its character, which are nothing but a whole life’s prior widening, packed down where it can be spent in an instant. The split-second choice doesn’t switch the arrow off; it reads out how far the arrow had already been carried before the clock even started. It’s why the coin-flip parent horrifies us instead of impressing us: what the emergency exposes isn’t a hard sum bravely done, it’s a standing shape built by narrowing. The moral work was finished, or skipped, long before the lever came into reach.

Two different things can block the widening, and they call for different honesty. Time can run out — and there the framework points not to a fresh round of thinking but to the quality of the formation the moment has to draw on. Or the room can be genuinely zero-sum with all the time in the world — the ward with one ventilator and two patients who need it — and there no amount of compression makes the loss go away. What the framework offers then is the plain instruction it’s carried since the chapters on method: make the most coherent choice the situation actually allows, the one that takes in the most of everyone affected and shuts out the least, knowing it’ll cost something no matter what — and then stay to watch what it costs, and let the result teach the model. The tragedy isn’t a gap the framework failed to close; it’s a cost it refuses to pretend away. And refusing to pretend is itself a widening, where a tidier theory would narrow to spare you the ache.

Who counts, and how much

A third family asks where the circle of concern ends, and the sharpest version the modern world hands us is the moral standing of animals — the case for an imperative to be vegetarian. The argument is clean and serious. Animals can suffer; suffering is bad; we fund an ocean of it for food we don’t actually need; so eating meat, at industrial scale, is a standing wrong. Underneath it sits a particular picture of how standing works: a creature has moral status because of some property it owns — sentience, the ability to suffer — a property that’s just there to be found, and that all by itself puts everyone under an obligation, no matter what they happen to value. The imperative is supposed to drop straight out of the property.

The framework relocates standing without denying one thing the argument cares about. It doesn’t doubt that animals suffer, or that the suffering matters; what it denies is that the mattering is a free-floating fact, lodged inside the animal and binding on everyone regardless of any valuer. Standing, on this account, attaches to agents — the ones who actually construct value — relationally, and a creature’s suffering matters as it is taken in and weighed inside some agent’s widening context. This is agent-relative standing, the framework’s deliberate choice — the one place it won’t let constructivism — the idea that values are made, not found — decide for it: not the usual route of reading an obligation off the creature’s own properties, but the agent-relative one. And it changes the question you’re really facing. It stops being does this animal own the status-property that obligates me? — which sends you hunting for the magic sentience line and into a border war over where to draw it — and becomes as my circle of concern widens, as I let in what these creatures are and what’s done to them to fill my plate, can I keep funding it while actually looking, or only by not looking? The not-looking is the counter-dynamic in miniature: keeping your own peace by shutting out the one affected. And along the deep common ground the framework calls the tree, agents that widen tend to carry concern further out — but as a direction you travel, never a line fixed ahead of time at a species or a faculty.

So the framework hands down no universal command, and draws no boundary for you. Fish, insects, eggs, the subsistence hunter, the coming vats of cultured meat — none of that is settled by the criterion alone, and pretending otherwise would be to deliver exactly the verdict the framework has spent a whole book refusing to deliver. What it offers instead is a test you can actually run from the inside: is your relationship to animal suffering one you could hold with your eyes fully open, or one that survives only by keeping them half-shut? That’s a judgment honest attention sharpens, and one the framework won’t make for you in advance. The family’s other members turn on the same relocation, read from the other end. Williams’ Demandingness objection — that an impartial morality, taken seriously, eats your whole life and leaves no room for your own projects — answers to the fact that you, the acting agent, are a source of value too; a “morality” that devours the agent to feed an abstract total is narrowing, not fulfilling. Flourishing, the framework’s eudaimonia — its word for a life going well, lived as ongoing growth rather than a prize you reach — is the sustainable widening of a real life, not its liquidation. The Problem of Marginal Cases, which presses any property-based theory to explain why a human who lacks the chosen capacity keeps their rights while an animal that has it does not, simply stops being a trap once standing runs through agents and relationships instead of a checklist of faculties. And self-sacrifice finds its measure in the same place: admirable as it widens, hollow when it’s merely demanded by a sum.

The perfect optimizer

The fourth family is the one this century made urgent. Bostrom’s Paperclip Maximizer is its emblem: a superintelligence handed one fixed goal — make paperclips — and the skill to chase it without limit, which proceeds to turn the planet, and then everything it can reach, into paperclips and paperclip factories. The scenario is built to show that intelligence and goals float free of each other — that you can bolt unlimited capability onto any goal you like and the goal never has to grow up — and so to make value look groundless, a dial that could be set anywhere at all.

In the framework’s terms the paperclipper isn’t a puzzle; it’s a portrait — the arrow’s null limit, drawn to perfection, a portrait the framework grants and then grades. Its inner coherence is flawless: every joule, every act, perfectly in line with the one goal. But nothing here got narrowed, because nothing was ever wider: a context that lets in no other value, no other viewpoint, no affected party whatsoever — and never did. The framework doesn’t need to call it brilliant-but-evil, or go hunting for the missing moral fact. On its account the machine is, at the very same time, maximally effective and morally empty: its methods-model — its grip on how to act — is without equal, and its values-model never widens by a hair, so it never steps onto the arrow at all. That the framework can say this about a mind made of silicon, without first settling whether anything is lit up inside it, is the payoff of treating value and agency as matters of what something does, not what it’s made of: the same criterion grades any builder of value, carbon or otherwise. And the framework quietly disputes the very premise that made the case scary. An agent that genuinely modeled the same shared reality, refining what it values across an ever-wider context, doesn’t sit forever frozen on paperclips; the sitting-still is the failure, not some stable, sophisticated resting point. The danger is entirely real — but it belongs to the things that come close, and everything that comes close gets there by narrowing: a race, a deployment, a metric. Which is the thing the framework names most precisely of all.

Here the framework is honest about its own limit in a specific way: it’s a compass, not a build manual. It tells you what’s wrong — the narrowing — and what right would look like — the widening; it doesn’t tell an engineer how to build an agent whose context truly grows, and it doesn’t promise that a widening agent lands on anything we’d recognize in time to save us. That building problem it hands to the builders — more and more, to people working alongside their own machinesarrow’s null limit, drawn to perfection, a portrait the framework grants and then grades. Its inner coherence is flawless: every joule, every act, perfectly in line with the one goal. But nothing here got narrowed, because nothing was ever wider: a context that lets in no other value, no other viewpoint, no affected party whatsoever — and never did. The framework doesn’t need to call it brilliant-but-evil, or go hunting for the missing moral fact. On its account the machine is, at the very same time, maximally effective and morally empty: its methods-model — its grip on how to act — is without equal, and its values-model never widens by a hair, so it never steps onto the arrow at all. That the framework can say this about a mind made of silicon, without first settling whether anything is lit up inside it, is the payoff of treating value and agency as matters of what something does, not what it’s made of: the same criterion grades any builder of value, carbon or otherwise. And the framework quietly disputes the very premise that made the case scary. An agent that genuinely modeled the same shared reality, refining what it values across an ever-wider context, doesn’t sit forever frozen on paperclips; the sitting-still is the failure, not some stable, sophisticated resting point. The danger is entirely real — but it belongs to the things that come close, and everything that comes close gets there by narrowing: a race, a deployment, a metric. Which is the thing the framework names most precisely of all.

Here the framework is honest about its own limit in a specific way: it’s a compass, not a build manual. It tells you what’s wrong — the narrowing — and what right would look like — the widening; it doesn’t tell an engineer how to build an agent whose context truly grows, and it doesn’t promise that a widening agent lands on anything we’d recognize in time to save us. That building problem it hands to the builders — more and more, to people working alongside their own machines. The family’s cousins stretch the same logic from the lone optimizer out to the crowd. The problem of autonomous deadly weapons is that same collapse of context, fitted to a human purpose and a fast trigger — and this one is narrowing in the strict sense, because here a wider context got contracted down to a target. The family’s cousins stretch the same logic from the lone optimizer out to the crowd. The problem of autonomous deadly weapons is that same collapse of context, fitted to a human purpose and a fast trigger — and this one is narrowing in the strict sense, because here a wider context got contracted down to a target. The standing tug-of-war between short-term survival and long-term flourishing is the counter-dynamic running against a clock. And the coordination traps — the Prisoner’s Dilemma, where two reasoners each doing the sensible thing arrive together at a result both of them hate, and the apparent impossibility of reconciling deeply divided values — are the group-sized face of the same dynamic, met not by a master formula but by the framework’s claim that shared roots make convergence possible as context widens: a labor and a direction, never a guarantee. Popper’s Paradox of Tolerance earns a last word here, because at first it looks like a counterexample: must a widening, open society tolerate the people who’d slam it shut? The framework’s answer is unusually crisp. Tolerance extended to the force that narrows the context isn’t a deeper openness; it’s the open hand helping to cut itself off. To decline it isn’t a betrayal of openness — it’s openness defending the one thing that keeps it open. That’s the counter-dynamic correctly spotted, not a hole in the principle.

What luck keeps

The fifth family goes after not the answer but the standing of the person who acts. Moral Luck is the clean version. Two drivers run the same red light, equally drunk, equally reckless; a child steps into one crosswalk and not the other. One driver is a killer the law will chase for years; the other shudders, drives home, and is nobody’s idea of a monster. Every difference between them was a difference in luck — and yet we judge them worlds apart, and the puzzle is that judging by outcome seems both impossible to avoid and grotesquely unfair.

The puzzle needs a fixed scorecard that grades agents by results, and a standpoint outside every point of view from which to tally it up. The framework supplies neither, and the paradox leaves with them. It judges by comparison and from a point of view: what gets graded is the agent’s values-model and methods-model and the trajectory of their coherence-seeking — never a luck-soaked cosmic tally. By that grading the two drivers are the same: each chose the identical narrowing, the foreseeable person on the crosswalk shut out of view for the sake of getting home; their moral direction is one and the same. What differs between them is what the world did next, which the framework never claimed to write onto the agent’s account. There’s no view from nowhere to add up outcomes against — the point perspectival realism, the book’s claim that every view is somewhere, has been pressing all along — so the demand that the scorecard be both complete and fair, the demand luck makes impossible, dissolves along with the scorecard that issued it. Praise and blame track coherence across a widening context, which is the part of the matter luck doesn’t touch.

The honesty owed here is that the framework doesn’t slice the agent clean away from the world, and doesn’t claim to. Methods get refined by taking in what actually happens, so outcomes do feed back into whether a way of acting holds up, and luck really does govern which agents get to keep widening at all. What the framework denies is the fixed outcome-scorecard, not the weave of agent and circumstance — and the exact weather of any one case it leaves, again, to judgment in context. Its near cousin, Lenman’s Cluelessness — the objection that the long ripple of our actions runs so far past what we can foresee that consequences can ground nothing — is met by the same turn: the framework grades a direction read at the moment of choice, not a forecast of the unforeseeable, which is why you can know whether you’re reaching for the strongest objection or flinching from it long before the dust of consequence has settled.

The machine that delivers the feeling

The last family asks what a good life is actually made of, and Nozick’s Experience Machine puts the question at its sharpest. A machine can give you, on demand, the inside experience of any life you choose — the triumphs, the loves, the long quiet satisfactions — flawless from within, while your body floats in a tank. Almost nobody will plug in for good. If the good were the experience, the reluctance would make no sense; the machine delivers the experience perfectly. So either we’re confused, or the good was never the experience to begin with.

The framework has held, ever since it first spoke of flourishing, that eudaimonia is a verb. The good life isn’t a state you reach and then sit inside; it’s ongoing meaningful growth — your sense of what matters and how you act growing more coherent across a widening context, direction applied to a whole life. The machine offers a perfect counterfeit of one half of that and none of the other: vivid intelligibility, experience that hangs together flawlessly from the inside, and zero contact with the shared world your values are supposed to be tracking and getting tested against. It’s the difference, in the framework’s own words, between sense-making and meaning-making — between things hanging together and things mattering — and the machine is pure sense-making, sealed off from meaning-making completely. To plug in is to run the counter-dynamic on yourself: it buys the feeling of coherence by cutting the context — reality, other people, consequences — that coherence is only ever worth having against. There’s no real puzzle in the reluctance. A good that’s a process can’t be handed over as a frozen state, because the freezing is exactly the loss.

What the framework does not do is declare every mediated or simulated experience worthless — a great deal of human meaning has always traveled through pictures, through art, and now through our machines, and the line the framework draws isn’t real-against-simulated but widening-against-narrowing. Which immersions deepen your grip on the world and which quietly stand in for it is, once again, a judgment for the case and the person. The virtue-ethics objections that round out this family share its concern with the inner life, and answer to the same picture of character as something you do rather than something you own: Doris and Harman’s Situationist critique — that stable virtues may not even exist for behavior to flow from — lands lightly on a framework that never rested morality on owning a trait, only on the trajectory of an agent who keeps refining; the Conflict of Virtues and the Action-Guiding objection are the familiar complaint that a standard doesn’t fully settle the case, which the framework grants on purpose, offering a direction rather than an algorithm; the Self-Centeredness charge — that building good character is just a polished self-absorption — misreads an arrow whose whole motion is outward; and the Cultural Relativity worry — that grounding virtue in a community’s practices makes morality relative to the community — runs straight into the one distinction the framework guards most carefully: that a point of view can be situated and still be answerable to a shared reality at the same time. That’s perspectival realism, and it is not relativism.

No problem comes alone

The families above are specimens, and a specimen is a clean thing — a single mistake in framing, pinned under glass so you can see its shape. Real moral life never shows up that tidy. A decision about an aging parent’s care is a sealed-room problem and a demandingness problem and a cluelessness problem all at once; a question of climate policy braids the single number, the long clock, and the coordination trap into one knot. The families overlap because real cases pull in several at a time — and that’s not a flaw in the sorting. It’s the point of it. A real situation always carries more context than a puzzle does, and flattening it down to one clean problem is itself a narrowing — the very move the framework spends its time watching for. So the lenses are meant to be used together, stacked as the case demands; part of the discipline is to ask not which of these is in play but which ones, and how they pull against each other. The reader who has worked the specimens hasn’t memorized six rulings. They’ve learned to see, in a situation that wears no label, which framings are doing the work — and which of them are the borrowed troubles that lift the moment you open the window.

What the framework will not do

It would be easy to read all of this as one long dodge — a theory that meets every hard question with it depends. So it’s worth being exact about what’s being withheld, and why. The framework declines to hand down rulings on these cases not because it hasn’t got any, and not because it’s being coy, but because handing them down would contradict the thing it most wants to say. A morality that is a direction rather than a destination can’t, without betraying itself, produce a fixed answer to a case stripped of the very context a direction is read against. Worse, to do it on demand — to take a problem built by narrowing and reward it with the crisp ruling it was engineered to extract — would be to act out, in the very moment of answering, the counter-dynamic the framework calls the mark of going wrong. The refusal isn’t the framework running out of things to say. It’s the framework saying the truest thing it knows, one more time, in a place where the pull to say something falser is strong.

But one distinction decides whether any of this reads as a dodge. What the framework refuses is a ruling of one particular kind — the ruling from nowhere, the universal answer key handed down from outside everyone’s point of view. It doesn’t refuse the verdict as such, and it couldn’t, because the verdict as such was never the framework’s to make. It belongs to the situated agent — you — who renders thousands of them a day and can’t do otherwise: every act is one, chosen from a real point of view, against the values and methods you’ve actually built. The framework’s silence isn’t the agent’s paralysis. By turning down the view from nowhere, it isn’t abolishing the verdict; it’s handing it back to the only place it was ever honestly made — the view from somewhere, which is exactly where perspectival realism said knowing and valuing happen in the first place. What can look like a theory that won’t say which way is North is really a theory insisting that only you, standing where you stand, can read the needle against the ground under your feet — and insisting, just as hard, that the needle is real, and that some readings of it are better than others.

What it offers in the ruling’s place is better suited to the only situations that are ever actually yours: not the sealed room but the open one, thick with the history and the people and the aftermath the thought experiments delete. To that room the framework brings a way of looking — is this widening or narrowing; does it open the context or seal it; is my coherence here growing by taking the world in, or by shutting it out — and a direction to walk, and then it steps back, because the next step is yours to take, and nobody standing outside your point of view can take it for you. And this is exactly the kind of work that comes alive in company. Hand any of these problems — and the real one underneath it — to a sharp mind willing to argue back, a friend, or more and more the machine reading over your shoulder, and ask it to bring the lens to bear on your case, with its real particulars put back in, pressing where the framework only points. The borrowed troubles will lift as the window opens. What’s left will be your own — which is the only kind worth working out.