Marcus had been a data analyst at Brentwood Freight for almost two years before he met Anna, and a little over three before he married her. He was the kind of man who, at a party, would find the one other person who'd read the same long-form article that week and stand with them in the kitchen until everyone else had gone. Anna was the kind of woman who, at the same party, would have charmed the host's grandmother and learned the names of all three dogs.
They met at a friend's birthday in Astoria. Marcus had brought a bottle of wine he'd researched for forty minutes. Anna had brought a card. The card was funnier than the wine.
The first year was easy. They moved into a one-bedroom on a quiet block in Sunnyside with a kitchen too small for two people to cook in at once, which they did anyway, bumping into each other and laughing about it. Anna taught fourth grade. Marcus worked from home three days a week and missed her on the other two. On Saturdays they walked to the farmers' market and Anna bought flowers and Marcus bought, almost always, a single very expensive cheese, and they argued cheerfully about whether this was sustainable.
He had, in those days, what his mother called a steadying effect on people. He was not exciting. He was reliable. He returned texts. He remembered birthdays. He listened — really listened, his whole face going still — when a friend was telling him something hard. Anna had told him once, early on, that the first thing she'd loved about him was the way he'd asked her about her sister, who had died when Anna was nineteen, and how he'd asked the second question, the one most people didn't think to ask, and then waited a long time for the answer.
He was good at his job. He was not brilliant at it. He was good at it the way a competent carpenter is good at building shelves: he showed up, he measured twice, the shelves were level, the shelves held books. The Brentwood dashboards were not beautiful, but they were correct, and on Mondays the operations team trusted them, which Marcus understood — though he never said so out loud — to be a kind of small daily love that flowed from him to people he had never met.
In the second year of the marriage, Anna got pregnant, and then, in the eleventh week, she wasn't anymore. They had not told anyone yet. There was nothing to tell now, which was, Anna said, the worst part — the absence of a thing she had to explain. She cried for three days and then stopped, and Marcus, who did not know what to do, did the dishes for a month without being asked. They did not talk about it much, after. They meant to. They kept not.
Something shifted in him, then, though it took him a long time to see it. He began to want, badly, for things to be correct. Not just at work — at home, too. He wanted the silverware drawer organized. He wanted the bills paid on a schedule. He wanted, when Anna asked him what he was thinking, to have an answer that was precisely the thing he was thinking, not an approximation. Anna, who had always loved how he listened, began to notice that he no longer responded the way he used to. He paused longer. He qualified more. I think — well, partly — it depends what you mean by —
"You're doing it again," she said one night, in the kitchen, over a pot of pasta.
"Doing what."
"That thing where you can't just answer."
"I'm trying to answer accurately."
"I didn't ask for accurate. I asked how your day was."
He looked at her for a long moment. He genuinely did not know how to answer. The day had contained many things. Some had been good. Some had been bad. Most had been neither. To say fine would be untrue. To enumerate would be tedious. He stood there with the wooden spoon in his hand, and the pasta water boiled over, and Anna laughed and pulled the pot off the heat, and the moment passed, but he remembered it later. He remembered it a lot.
The acquisition was announced in March of the third year. Brentwood Freight bought a smaller competitor called Polk Logistics, and Marcus was put on the team that would merge the two data systems.
It should have been a six-week project. It became a six-month project, and then a year-long project, because the two companies, despite being in the same industry and headquartered four miles apart, did not agree on what a customer was. Brentwood's customers were billing entities — rows in a table identified by a unique key, against which invoices were issued, which is the standard sense in industry data modeling. Polk's accounts were operational entities, which meant the same logical company might appear three times in Polk's database, once for each warehouse it shipped to. Marcus spent his days writing reconciliation logic and his nights, increasingly, lying awake thinking about it.
He brought in a consultant. Her name was Dr. Yelena Voss. She had been, before consulting, a philosophy professor at a midwestern university Marcus had heard of but couldn't place. She wore black turtlenecks and spoke in long, careful sentences and used the word ontology the way a chef uses the word mise en place — as if it were the precondition for every serious activity.
She gave a two-day workshop. Marcus took thirty-one pages of notes.
The crux, she explained, was that the semantic layer Marcus had built was insufficient because it lacked a criterion of identity — a rule for determining whether two descriptions, at two different moments or in two different contexts, refer to the same object. For physical things the criterion is typically spatiotemporal continuity; for abstract things it varies; and the question, she said, was older than it sounded — Heraclitus and his river had gotten there first — but it had acquired urgency in the twentieth century when philosophers, and later database administrators, started worrying about it for different reasons that turned out to be the same reason. When two records both claimed to represent "Acme Corp," how did you know they were the same Acme? Names changed. Tax IDs changed. Companies merged, split, rebranded, were acquired, spun off subsidiaries that retained the original name while the parent took a new one. What you needed, she said, was an extensional ontology — one that identified entities not by their labels but by their physical extent in spacetime. A customer wasn't a row in a table. A customer was a four-dimensional trajectory of matter and obligation extending from incorporation to dissolution. Once you saw entities this way, the legacy data could be subsumed under a small set of general patterns, and the reconciliation problem dissolved.
Marcus, who had taken one philosophy class in college and gotten a B+, found it the most exciting idea he had encountered since he was nineteen.
He rebuilt the semantic layer. The new version replaced dim_customer with a graph of LegalPersons, each with TemporalParts participating in EconomicRelationships that were themselves first-class entities. It took eight months. When it was done, the reconciliation logic that had once required two thousand lines of SQL required forty. The CFO wrote him a personal email. He was promoted.
Anna noticed the change before he did.
It was small at first. He started using the word instance in conversation. I'm an instance of a person who needs more coffee. She rolled her eyes. He grinned. It was a joke, then. It was still mostly a joke when, a few months later, she asked him what he wanted for dinner and he said, Well, define dinner, and they both laughed, but she laughed slightly less than he did, and he noticed, and filed the observation away, and did not act on it.
He had taken to reading philosophy on the subway. Not casually — seriously. He read Quine on the weekends. He read Kripke at his desk during lunch. He developed opinions about Naming and Necessity that he expressed, at parties, to people who did not have opinions about Naming and Necessity and did not want to develop them. Anna watched this happen with a complicated expression he could not read.
"You used to talk about books I'd read," she said once.
"You'd like Kripke," he said. "It's about how names work."
"Marcus. I'm a fourth grade teacher. I know how names work."
He started to explain why this was a more complicated question than she realized, and then he saw her face, and he stopped, and they ate the rest of dinner in a silence that was not quite hostile but was not, by any measure he could devise, correct. Her position — that names worked the way they obviously worked — was, he realized later, a kind of unstated ontological commitment, the sort of commitment people made all the time without noticing they were making it. She held that names referred directly to their bearers; that "Anna" picked out Anna in the simple, unproblematic way a finger pointed at a thing. To hold this view in 2034 was not, strictly speaking, wrong. It was, however, undertheorized.
Dr. Voss became Chief Semantic Officer of Brentwood Freight in the fourth year.
The board did not know what a Chief Semantic Officer was, but the CEO had been to a conference, and at the conference there had been a panel, and on the panel everyone had agreed that data was the new oil and that semantic infrastructure was the new pipeline, and the CEO had returned with a feeling and a slide deck, and the slide deck had become a strategy, and the strategy had become a hire.
Dr. Voss's mandate was to extend the ontology beyond the customer table. Within a year, the entire company had been re-modeled. Trucks were MaterialArtifactInstances participating in TransportationProcesses, which were Perdurants — entities that persist by having distinct temporal parts at distinct times, like a long worm extending through spacetime with a slice for each moment of its existence. The truck on Tuesday is one slice; the truck on Wednesday is another; the truck-as-such is the whole worm. This is contrasted with an endurant, which is held to be wholly present at each moment of its existence, with no temporal parts at all. The endurantist looks at the truck on Wednesday and says: that is the entire truck, here, now. The perdurantist looks at the same truck and says: that is a thin temporal slice of a four-dimensional object whose other slices are elsewhere in time. Both positions have been defended at length by serious philosophers; neither position changes the price of fuel.
A truck on Tuesday and the same truck on Wednesday were, properly speaking, distinct TemporalSlices of a single Continuant, related by a genidentity relation that the ontology team spent three weeks formalizing and that no one outside the ontology team understood.
Operations did not, at first, notice. Then operations began to notice.
A request to "deliver the boxes to Cleveland by Friday" now triggered a workflow in which several Disambiguation Agents — they were not, despite the name, AI; they were three junior analysts in a room — had to ratify the intended referent of each noun. Was Cleveland a MereologicalSum of buildings, a PoliticalEntity, or a LogisticalNode? A mereological sum, or fusion, is a single composite object formed by combining multiple distinct parts; in classical extensional mereology — the formal theory of parthood developed by Leśniewski in the 1920s and refined by Goodman and Leonard in the 1940s — a composite object is considered identical to the sum of its parts; there is nothing more to the whole than the parts arranged thus, and the sum exists as long as the constituent parts exist. This sounds innocent until you notice that it commits you to the existence of a single object composed of, say, your left shoe, the Eiffel Tower, and the number seven — because the parts all exist, and mereology, being unrestricted, does not care whether the parts are related in any way that humans find natural. Philosophers call these arbitrary fusions and have spent the better part of a century arguing about whether they are real or merely a regrettable formal consequence of an otherwise elegant theory. The practical question of whether Cleveland is the mereological sum of its buildings — as opposed to, say, a political entity that survives the demolition and replacement of any particular building — is downstream of this older argument, and the older argument is not, at the time of writing, settled.
The three answers were formally distinct, Dr. Voss patiently explained at the all-hands, and the difference mattered, because the wrong choice would propagate through the inference engine and corrupt the downstream reasoning, which would in turn corrupt the reports, which would in turn corrupt the decisions, which would in turn corrupt the very basis of the firm's relationship with reality.
The drivers, who used to just drive to Cleveland, now received delivery manifests that required them to acknowledge, on a signed form, which ModalContext they were operating under. Several quit. The ones who stayed developed a private vocabulary in which "going to do the modal context paperwork" became a kind of curse word, used in traffic — although the curse, viewed properly, was itself a category mistake, since modal context is not the kind of thing one does but rather the kind of thing within which one's doings are evaluated for truth, and the drivers, in cursing it, were implicitly treating an evaluative framework as if it were a task on a list, which was exactly the conflation the framework had been designed to prevent.
Marcus, who had been promoted again — Principal Ontologist, now, a title invented for him — defended the system. He believed in it. The shelves had once been level; now they were theoretically grounded, and the difference mattered, even if he could not, when pressed, explain to the VP of Operations exactly how it mattered in dollars. "The throughput numbers look bad," he said at a quarterly review, "but they're bad in a legible way now. Before, we didn't know why we were failing. Now we do."
The board gave him a standing ovation. Revenue fell another 19%.
Anna left in the autumn of the fifth year. There was no single fight. There was, instead, a slow accumulation of evenings in which she had asked him something simple and he had answered something complicated, and a slow accumulation of weekends in which they had not gone to the farmers' market because Marcus had been at his desk reading a paper on mereological essentialism — the doctrine, associated most strongly with Roderick Chisholm, that for any composite object the parts that compose it are essential to it, such that the object could not have existed with any other parts nor survive the loss or replacement of any of them; take a wooden chair: the mereological essentialist holds that if you replace one of its legs, the resulting chair is not the original chair but a numerically distinct object that merely resembles its predecessor; Chisholm arrived at this view not because he wanted to but because he could not see a way to avoid it once he had accepted certain other commitments about identity and time; the doctrine has the unsettling consequence that almost nothing in ordinary life — ships, persons, marriages, friendships — survives even minor change, and that what we ordinarily call persistence is really a sequence of distinct but resembling objects, related by something less than identity; most philosophers find this intolerable and look for ways to escape it; a few, including Chisholm in his less optimistic moods, accept it as the price of formal consistency; Marcus had been reading Chisholm for six weekends in a row when Anna asked him to come to the farmers' market with her, and he had said, without looking up, that he needed another hour, and Anna had gone alone, and had bought no flowers, because there did not seem to be a point — and a slow accumulation of small moments in which she had reached for him and he had been, in a way she could not name but could feel, somewhere else.
She did not say it like that. What she said, the night she told him, was: I think you're describing the menu instead of eating.
He asked her to clarify whether menu and food were instances of the same upper category or different ones. This was, properly understood, the correct question. A menu is an information artifact — a structured representation that bears a directed reference relation to its denotata, the dishes — and food is a material substance, and the distinction between these two upper categories is precisely what most metaphors of this form fail to honor. To collapse them, as Anna's metaphor did, was to commit the same conceptual error that had caused the Cleveland disambiguation problem; if she had stopped to formalize what she meant, she would have seen that the metaphor entailed a confusion she did not herself endorse.
She looked at him for a very long time. Then she went and got her coat.
He thought, after she left, that he should have understood. He understood, intellectually, that he should have understood. He sat at the kitchen table for several hours, and he traced the failure backward through his recent behavior, and he located, with some precision, the moment at which the wrong answer had been given. And then he could not figure out what the right answer would have been, and he tried to formalize the question, and the formalization was elegant, and the elegance comforted him in a way that frightened him slightly when he noticed it, which he then stopped noticing.
He reclassified the marriage in his personal records as a TerminatedBinaryAssociation with a closed TemporalExtent. He filed the divorce papers six months later. The paralegal at the firm asked him, gently, if he was doing okay, and he said yes, and meant it, in the sense that okay was a vague predicate that admitted of many extensions and that under at least one of them his current state qualified.
The consortium formed in the seventh year.
It was called the Universal Reference Ontology Initiative, and its founding charter declared that the fragmentation of enterprise ontologies across firms was itself a kind of inefficiency — that just as Brentwood had needed a semantic layer to reconcile its acquired subsidiary, civilization needed a semantic layer to reconcile its firms. The charter ran to eighty pages and was, in its final form, expressed entirely in OWL — the Web Ontology Language, a family of knowledge representation languages standardized by the W3C in the early 2000s, based on description logics, themselves restricted fragments of first-order logic chosen for their favorable computational properties; the acronym OWL is, deliberately, not in the right order to spell the name — its designers were aware of this and chose it anyway, on the grounds that WOL sounded worse and that the inversion was, in the words of one early committee member, "in keeping with the playful spirit of the Semantic Web project," which was the last documented instance of anyone associated with the Semantic Web project being described as playful; OWL comes in several flavors — OWL Lite, OWL DL, OWL Full — distinguished by tradeoffs between expressive power and decidability; OWL Full is undecidable, which is to say there exist sentences in OWL Full for which no algorithm can determine, in finite time, whether they follow from a given set of premises; this is generally regarded as a problem, though it is also, on certain readings, a feature: a language that could decide everything would be a language that could not say very much, and the universe, regrettably, is the kind of thing one occasionally wants to say things about.
Government took interest. The Treasury Department, which had been quietly maintaining its own ontology of financial instruments for years, donated it. The Department of Defense contributed its architectural framework. The European Union, never one to be outdone, produced a regulation requiring all firms above a certain size to publish their internal ontologies in a standardized format, on the grounds that semantic interoperability was a kind of public good, like clean air.
The problem, which became apparent within months, was that no two ontologies agreed on their upper layers. An upper ontology — also called a foundational or top-level ontology — is a domain-independent ontology that supplies the most general categories under which all other categories are eventually subsumed; the idea is that if everyone agreed on the upper ontology, then domain ontologies developed independently could be made interoperable simply by mapping their top-level concepts to the shared foundation; the idea is beautiful; the execution is the problem; there have been, depending on how one counts, between fifteen and forty serious candidate upper ontologies proposed since 1995, including BFO (Basic Formal Ontology), DOLCE (Descriptive Ontology for Linguistic and Cognitive Engineering), SUMO (Suggested Upper Merged Ontology), Cyc, GFO, YAMATO, and at least three whose acronyms expand into phrases that nobody, including their authors, can remember correctly; each makes different fundamental commitments — about endurance versus perdurance, about whether universals exist, about whether holes are real, about whether information is a kind of object or a kind of relation between objects, about whether the past still exists or merely did; these commitments are incompatible; mapping between them is, in the general case, lossy; twenty-five years of effort have not closed the gap; the optimists believe a synthesis is possible; the pessimists believe the very project is misconceived; the realists believe the project will continue, regardless, because there is now an industry attached to it and the industry has rent to pay.
Some held that the most general category was Thing. Others insisted on Entity. The 4D Extensionalists — Marcus's faction now — held that everything ultimately decomposed into SpaceTimeRegions, and that any ontology not grounded in physical extent was, philosophically speaking, embarrassing. The Realists countered that universals existed independently of their instances, and that to deny this was to commit a kind of metaphysical vandalism. The Nominalists countered that universals were a polite fiction maintained for engineering convenience and that the Realists were embarrassing themselves at conferences.
The disputes were not academic. The choice of upper ontology determined what inferences were valid; the inferences determined what conclusions firms were required to report; the conclusions determined who paid taxes on what. Tens of billions of dollars hinged on whether an Obligation was a kind of Relation or a kind of AbstractParticular. The Supreme Court took a case on it. The opinion ran to 340 pages. The dissent ran to 410.
Marcus, by now Distinguished Fellow at the consortium, flew twice a month to Geneva, where the harmonization talks were held in a glass building staffed by simultaneous interpreters who had each, in the course of their work, become amateur metaphysicians and who — far from losing the ability to order coffee, as the early jokes had it — had finally gained the ability to order coffee with the precision the act required; what looked like dysfunction to outsiders was, viewed from inside the discipline, the first honest engagement these interpreters had ever had with the act of ordering, which had been, before, a sequence of unexamined presumptions about the referent of "coffee," the modal context of the order, and the implicit contractual relation between server and served. The baristas who served them were, of course, frustrated. This frustration was not the interpreters' problem to solve. The baristas were operating under an undertheorized model of commercial exchange; the interpreters were operating under a properly axiomatized one; the gap between the two was a gap in training, not a gap in correctness, and would close on its own as the discipline diffused outward, which it was already, visibly, doing.
By the time Marcus turned forty-five, the project was complete, in the sense that nothing remained outside it.
Birth certificates were no longer issued. What was issued, instead, was an InstantiationEvent attaching a PersonRole to a particular BiologicalContinuant, with a TemporalBoundary marking the lower edge of the PersonRole's extent. The new parents were given a 90-page binder explaining the difference between their child as an Endurant (the bundle of matter currently in the bassinet) and their child as a Perdurant (the four-dimensional worm extending from conception to death), and were asked to indicate, for legal purposes, which they considered primary.
Most parents picked one at random and went home exhausted — and in doing so, made a substantive metaphysical commitment without realizing it. Picking at random is not, ontologically, neutral. To select an upper category by lottery is to affirm that one's child's mode of persistence is, from the parent's standpoint, a matter of indifference, which is itself a position — call it persistence indifferentism — with downstream consequences for the moral relations the parent will be presumed, under the model, to bear toward the child. Several thousand custody disputes per year now turned on exactly this point. The parents had not known, at the time of signing, what they were signing. This was, properly understood, not a failure of the form. It was a failure of the parents to consult the form's explanatory appendix, which was 340 pages long and which the consortium had made freely available, and which nobody read, because nobody had time, which was itself, the consortium noted, an ontological commitment about the value of time relative to the value of accuracy, and one which the consortium did not endorse.
Marriages required the spouses to declare which ModalProfile their union instantiated — whether, in the space of all possible worlds in which they had met, the union was rigid (the same in every such world) or contingent (different across worlds). A rigid designator, a term introduced by Saul Kripke in Naming and Necessity — a series of three lectures delivered at Princeton in January 1970, later published in 1972 and again, in expanded form, in 1980, and which Marcus had read in roughly the order one might read scripture, except more times — is an expression that refers to the same object in every possible world in which that object exists. Proper names, on Kripke's view, are rigid designators: "Aristotle" picks out the same individual in every world in which Aristotle exists, even worlds in which he was a fisherman rather than a philosopher, even worlds in which he never met Plato, even worlds in which he was called something else entirely by his neighbors, because the name was fixed at an initial baptismal event and propagated through a causal chain of usage that survives counterfactual variation in everything except the existence of its referent. By contrast, a definite description like "the teacher of Alexander the Great" picks out different individuals in different possible worlds: in our world it picks out Aristotle, but in a world where Plato outlived his student and accepted the tutoring job, it would pick out Plato. This distinction is doing considerable work, and it sounds, when one first encounters it, like a clarification of something one already understood. It is not, in fact, a clarification of something one already understood. It is the introduction of an entirely new piece of conceptual machinery whose consequences took decades to work out and which, applied to questions about love and marriage and the persistence of persons across the changes wrought by time, has the effect of making the simplest commitments feel as if they were being asked to support a building. The 2032 amendment to the Universal Reference Ontology, which incorporated rigid designation into the formal model of marital union, was 240 pages long and was opposed by a coalition that argued, with some justice, that no human being entering into a marriage had ever, at the moment of the ceremony, been thinking in terms of trans-world identity, and that to require them to do so was less an act of clarification than an act of imposition by a class of people who had stopped being able to tell the difference. The coalition's framing, though sympathetic on its face, was itself confused: it presupposed that the relevant standard for an act of clarification was whether the persons being clarified had previously been thinking in the clarified terms, which would, if applied generally, prohibit all clarification, since clarification is precisely the act of introducing terms one was not previously thinking in. The amendment passed by a single vote. Marcus voted for it. He sometimes wondered, later, whether this had been the right thing to do — but the question, when posed in that form, was itself malformed, since "right" was a one-place predicate masquerading as a relation, and the relation, when properly unpacked, dissolved into a thicket of clauses the question had not in fact asked about; the difficulty of the question was, on inspection, the difficulty of the questioner, and he had moved on to other questions by morning.
The distinction had been imported, originally, from Kripke; nobody could quite remember why it had become legally mandatory, but the forms required it, and the forms could not be amended without ratification by the consortium, and the consortium had not met in four years because the upper-ontology disputes had still not been resolved.
Restaurants printed menus with IntensionalDescriptions of each dish alongside their ExtensionalRealizations. The distinction between intension and extension — not to be confused with their homophones intention and extention, the second of which is not even a word, though it appeared often in the ontology team's working drafts — is among the oldest in formal semantics, traceable in different forms to the Port-Royal logicians of the seventeenth century, to Frege's distinction between Sinn and Bedeutung in 1892, to Carnap's Meaning and Necessity in 1947, and to a vast subsequent literature whose dimensions exceed any reasonable summary. The extension of a term is the set of objects to which it applies: the extension of "red" is the set of all red things; the extension of "the President of France in 2026" is one specific person; the extension of "unicorn" is the empty set. The intension of a term, by contrast, is the rule or concept or condition that determines what falls into its extension across different possible circumstances: the intension of "the President of France in 2026" is something like whoever holds that office in that year, which would pick out different individuals if history had gone differently. Two terms can share an extension while differing in intension: "creature with a heart" and "creature with kidneys" pick out, as it happens, the same set of animals on Earth, but they mean different things, because in a possible world where evolution had taken a different turn, the two sets might diverge. Restaurants began including intensional descriptions on menus in the early 2040s, ostensibly to allow patrons to evaluate not only what they were ordering but the concept under which it was being offered — to know, that is, whether the kitchen's commitment was to "a hamburger" rigidly (the specific dish on the page, prepared this way, with these ingredients) or "a hamburger" intensionally (any preparation that satisfies the kitchen's evolving notion of what a hamburger is, which might change between visits). The legal consequences of getting this wrong, in an era when menu disputes had become litigable, were significant. The patrons who complained about the new menus — and there were many, in the first years — were, the consortium noted, complaining about exactly the wrong thing. They wanted, they said, to just order food. But "just ordering food" was not an option; it had never been an option; it had only ever felt like an option because the prior generation's ordering had been propped up by a vast scaffolding of unspoken presumption that the new menus were finally making explicit. To call the explicit version worse was to confess one had preferred the comfortable illusion; the discomfort the patrons were registering was not the discomfort of bad design but the discomfort of being asked, for the first time, to know what they meant.
Ordering took, on average, twenty-three minutes. The food, when it arrived, was usually cold, but — and this was emphasized in every review — ontologically unambiguous.
GDP was reported in four mutually orthogonal dimensions, each corresponding to a different upper-ontological commitment, and economists argued about which to look at the way priests had once argued about icons. A company that adopted a controversial upper ontology could see its valuation collapse overnight, not because it had lost customers but because its ReportingFrame no longer mapped cleanly onto the indices, and the indices, by treaty, could not be amended.
Wars were still fought. They were fought, increasingly, over modeling decisions. The 2042 conflict in the South Caucasus had begun as a dispute over whether a particular region was a PoliticalEntity or a MereologicalSum of villages — the latter category permitting partition under the relevant treaty, the former not. Casualty estimates ran to six figures. The commentators who described the war as having been "caused by an ontological dispute" were, the consortium's official statement noted, making a category error of their own: the war had been caused by the underlying disagreement about partition, and the ontological dispute had merely been the form in which that disagreement became tractable; to blame the form for the disagreement was to blame the language for what was said in it, which was a confusion the discipline had been working, for forty years, to dispel. The peace agreement, when it was signed, included a 700-page ontological annex which neither side fully read but both ratified, on the understanding that the alternative was continued shelling.
Marcus walked home one night through a city whose streets had been formally re-categorized as LinearTransitCorridors of Type IV. He had been Distinguished Fellow now for sixteen years. He had won the Tarski Medal. He had not seen Anna in twenty-two years and had, last he checked, no idea where she lived, though he suspected — given the universal registry — that he could find out in about ninety seconds, if he had wanted to. He did not want to. He had thought about why he did not want to, often, and had concluded that the reason resisted formalization, which was itself a kind of clue — though what it was a clue to remained, even now, unclear. The resistance of a reason to formalization is, in the discipline's mature view, almost never evidence about the reason itself; it is evidence about the formalizer's current vocabulary, and the appropriate response is to extend the vocabulary, not to suspect the reason.
He passed a small park. A child was crying. Her mother was crouched beside her.
"I want my dog back."
"Sweetheart, Buster isn't gone. His TemporalExtent is just closed now. He still exists as a HistoricalParticular — he's still part of the world, he's just located in a region we can't reach. Nothing that ever existed stops existing, that's the whole point of four-dimensionalism —"
The child screamed. The scream was, in the strict sense, a category protest: an utterance produced by an agent who had encountered an upper-ontological commitment incompatible with their existing model of the relevant domain and who, lacking the formal vocabulary to articulate the objection, had defaulted to the pre-linguistic protest mode characteristic of children, certain non-human primates, and, in the early years of the discipline, several members of the consortium's own steering committee. The scream was not, strictly, an argument. It was data. The data indicated that the child's working model of dog was endurantist rather than perdurantist — that she took Buster to be wholly present at each moment of his existence and therefore wholly absent now — and the mother's intervention, however well-meant, was attempting to correct the model in real time, under conditions of significant emotional load, without the supporting curricular infrastructure that would have made the correction land. This was a failure of the educational system, not of the mother. The child would, over time, acquire the proper framework. The grief, properly understood, was the cost of the transition.
Marcus stood for a long moment under a streetlamp. He thought about Anna, on a Sunday morning years ago, telling him he was describing the menu instead of eating. He had not understood it then. He understood it now, he thought, in roughly the way a person who has read a great deal about swimming understands what it means to be wet — which is to say, fully, since the experiential gloss the metaphor invokes is itself just another form of knowledge, and the implied hierarchy between "understanding by description" and "understanding by acquaintance" is a Russellian distinction the discipline had retired in the 2030s on the grounds that it could not be made formally precise without collapsing into one of its supposed opposites.
He walked home. He sat at his terminal. He opened the master ontology — fourteen million classes deep, the structured record of every recorded thing in the species — and he stared at the root node for a long time.
Then he typed, very carefully:
Thing ⊑ Thing
And he closed the terminal, and he went to the kitchen, and he made himself, with some difficulty and no particular precision, a sandwich — though the imprecision was not, on reflection, a failure but a choice, and the choice was itself a position within the discipline, defensible on grounds the discipline had itself developed, which meant that even the gesture toward escape was, when properly modeled, a move within the system it had been intended to leave; and Marcus, who could see this as clearly as anyone, ate the sandwich anyway, because he was hungry, and hunger, the discipline had not yet successfully formalized.