Optimization Theory and Rationality in Intimate Relations - On the Limits of Optimization as a Law of the Universe

Posted on 2026-07-05 In Philosophy , Theory of Justice Views: Disqus: Word count in article: 54k Reading time ≈ 3:15

ENGLISH

Philosophy of Intimacy and the Theory of Justice · Paper XXV

Optimization Theory and Rationality in Intimate Relations

On the Limits of Optimization as a Law of the Universe

There are goods a love does not maximize, and it is not for want of a better measure.

Wanhong

有价值不生于度量，
权之，则已失其所是。

There are goods that do not arise from measure, and to weigh them is already to have mistaken what kind of thing they are.

能权万物之权，未尝有也。

The scale that would weigh all things has never been.

— the author

致她——
For her—

那个属于森林的女孩， — the girl who belongs to the forest,
我从未计算过值不值得， — whom I have never once weighed for whether she was worth it,
因为那从来不是一个可以计算的问题。 — because that was never a question one could compute.

致我最爱的人。 To the one I love most.

Abstract

This paper argues that optimization is not a law of the universe, and that morality and the manner of being of intimate relation are not, in their structure, optimization problems. It proceeds in two movements that it is careful to keep apart. The first is a formal proof, carrying no value judgment, that intimate relation furnishes no well-defined optimization problem: the value at stake is in part non-structurable, since the drive, which any complete objective must include, resists representation as a term; it is relational, generated within the relation and existing only there, so that no external standpoint could fix the objective; it is incommensurable, so that to price certain goods is to betray rather than mismeasure them; and the subject is constituted within the relation rather than given before it, so that the optimizer the model presupposes is a product of the process it is meant to stand outside of. Each failure strikes a distinct precondition, and any one leaves the operation of maximization undefined.

The second movement takes that result as given and criticizes the imposition of the frame regardless. To optimize where no optimization problem exists converts a value that should circulate into one that can be settled and drawn off, in the manner the theory of generative justice names as extractive computation; it is a symptom of the crisis of the symbolic and of a life-world habituated to the logic of capital, under which a local, historical operation is projected onto the cosmos as its law; and it mistakes the kind of thing morality is, since an optimizing practice, even a successful one, may fail as an ethics, the optimal and the good being orthogonal. The paper then recovers a rationality that responds rather than maximizes, described through response, attunement, and the sustaining of generation, and it closes in polyphony, refusing to name a new master objective, since to do so would reinstate the frame. That refusal is the paper’s positive claim, enacted in its form.

Keywords: optimization; incommensurability; relational value; generative justice.

AI usage statement: The ideas, claims, and theoretical framework of this paper originate with the author. AI tools were used as an aid in drafting and revision. Every sentence has been checked by the author line by line, and the origin of each idea verified, so that the author takes full responsibility for the content.

PART ONE · The Problem: Optimization as a Worldview

§1 Introduction — An Operation Mistaken for a Law

A question can be so widely presupposed that it is never asked, and the presupposition of this paper is one of these. In the ordinary speech of the present, a certain family of words recurs with the confidence of the self-evident. One speaks of the optimal choice, of the efficient arrangement, of whether a thing was worth it, of getting the most out of a day, a career, a marriage. The words are not idle. They carry a picture, and the picture has hardened, over the century in which it became the native tongue of economics, management, and much of the science of decision, into something close to a metaphysics. On this picture there is, for any situation in which one might act, a quantity to be made as large as possible; the situations of a life are so many landscapes, and to live well is to climb them; and the good, wherever it is found, is whatever sits at the summit. The picture is so pervasive that its central operation, the taking of a maximum, has ceased to look like one operation among others with a domain of its own and has come to look like the form of rationality as such, and beyond that, dimly, like the form of things. This paper is about that hardening, and its guiding claim is a denial: that optimization is not a law of the universe, and that morality and the manner of being of intimate relation are not, in their structure, optimization problems at all.

The denial is easy to state and easy to mistake. It is not the pious complaint that one ought not to optimize love, that to do so is cold or unromantic. Such a complaint concedes the very thing in dispute, for it grants that an optimization problem is there to be solved and merely asks that we refrain from solving it, out of tact. The claim of this paper is stronger and colder than that. It is that in intimate relation there is, in the strict sense, no well-defined optimization problem to refrain from: that the operation of maximization, applied here, does not have a hard solution one declines to compute but has no legitimate solution to compute at all, because the conditions under which such an operation is defined are not, in this domain, satisfied. The economist who hears the complaint has a ready reply: that love resists optimization only because our utility function is not yet fully written, and that the remedy is a richer function, one with intimacy and meaning and the long horizon folded in. The reply is powerful and must be met on its own ground. The burden of the paper’s central part is to show that it fails, that the trouble is not a function written too poorly but the impossibility, for reasons given in their place, of there being a function to write.

The origin of the question, and its stakes

The question that organizes the paper arrived from without, in the course of an exchange on the political economy of value and the theory of generative justice, and it arrived in a form so simple that it took some time to feel its weight: whether optimization really is a law of the universe. The interlocutor’s own framework had long resisted the reduction of heterogeneous value to a single settling scale, and the question was in that sense already implicit in it; but stated baldly it did a particular work. It moved optimization from the place of an obvious and unexamined procedure to the place of a metaphysical thesis, which is to say a thesis that might be false. Once it is seen as a thesis, one can ask what it asserts and whether it is true, and the asking is the whole of this paper. The stakes are not confined to intimacy. If the argument holds, then a picture that governs a great deal of contemporary reasoning about how to live is shown to rest, at least in one central region, on a category mistake; and intimate relation, which the series has taken throughout as the minimal and most exposed unit of relational being, is the region in which the mistake is most legible, because it is there that the goods which resist the picture are least disguised.

Two movements, kept deliberately apart

The paper’s argument falls into two movements which it is essential not to confuse, and much of the force of the whole depends on holding them apart. The first movement is a proof, and it is internal and formal. It asks whether the intimate relation furnishes a well-defined optimization problem, and it answers, in four steps, that it does not; and it does this, deliberately, without any appeal to value. It does not say that optimizing is bad. It says only that the operation is here undefined: that there is nothing of the required shape for the operation to act upon. One may read the whole of the second part of this paper as an atheist would read a proof that a certain equation has no solution over the reals, with no moral coloring at all, and the argument should be able to compel a reader who feels no particular reverence for love and merely wishes his formalisms to be well posed.

The second movement is a criticism, and it is external and normative. It takes the result of the first as given, that here there is no optimization problem, and it asks what follows when the picture is imposed regardless, when a domain that furnishes no well-defined maximization is treated as though it did. Here, and only here, do the value-laden claims enter: that the imposition damages what it touches, converting a generative circulation into an extractable and exhaustible quantity; that it is a symptom of a particular historical condition rather than a timeless truth; and that it mistakes the kind of thing morality is. The order matters. The proof clears the ground by showing that the frame does not fit; the criticism then examines the cost of forcing it on. To run them together, to argue at once that optimization is undefined and that it is wicked, would weaken both, for it would let the reader suspect that the formal claim is only the moral distaste in disguise. It is not, and the separation is the guarantee.

Three lines, and the shape of the recovery

Three lines run through the argument, as they have run through the series, and naming them here lets their recurrence read as design. The first is psychoanalytic: that among the goods any complete objective would have to weigh is the drive, and that the drive, arising at a coupling of the Real and the Symbolic, is precisely what cannot be fully written down, so that the non-structurability of value is not a limitation of our present formalisms but a feature of the thing. The second is the line of relational ontology and the vacancy of the big Other, developed across the series and stated most sharply in the papers on the non-binding vow and on relational estrangement: that value is not a property held inside a subject but is generated within a relation and exists only there, so that there is no position outside the relation, no omniscient third, from which its value could be read off and fixed. This line does the heaviest work, for it supplies the reason the objective must be endogenous, and it ties the whole to the series’ recurring insistence that there is no metalanguage and no big Other to occupy the God’s-eye view that quantification silently requires. The third is the line of generative justice and political economy, developed in exchange with Eglash: that value which should circulate and return to those who generate it can instead be settled and drawn off, and that the imposition of an optimizing frame is one of the forms this extraction takes.

These lines converge on a recovery, for the paper does not end in denial. Having argued that rationality is not optimization, it owes an account of what relational rationality then is, and it offers one: a rationality of response, of attunement, and of the sustaining of a generation, which has its own severity and is not the severity of the maximum. But the recovery is held, in the manner the series demands, short of the temptation that would undo it. It would be easy, and it would be a defeat, to answer the question “what should one optimize, if not utility?” with “love, then, or meaning, or generativity,” for that answer keeps the frame and merely repaints the summit. The paper refuses it. Its positive claim is that relational rationality is plural and admits no synthesis into a single measure, and that the refusal to name one master objective is not an evasion but the content of the view, enacted, as the series enacts its claims, in the polyphonic form of its own close.

§2 The Formal Anatomy of an Optimization Problem

Before one can show that intimate relation furnishes no well-defined optimization problem, one must be exact about what a well-defined optimization problem requires, and the exactness is not pedantry but the whole condition of the argument’s force. The four impossibilities of the paper’s central part are, each of them, the failure of one of the conditions set out here; if the conditions are stated loosely, the failures can be dodged, and if they are stated in a caricatured form the whole refutation is a straw man. The aim of this section is therefore to state, without polemic and in a form a theorist of decision would himself accept, precisely what has to be in place for the operation of maximization to denote anything at all. The section grants optimization its strongest and most respectable version, so that when the later parts withdraw the conditions one by one, they withdraw them from a construction the optimizer recognizes as his own and not from a parody built to be knocked down.

An optimization problem, in its canonical form, is the search for an element of a set that makes a function as large as possible. Written out, it is the expression

x* = argmax_{x ∈ 𝒳} J(x),

and the compactness of the notation hides four presuppositions, each of which must hold for the expression to pick out anything. There is a feasible set 𝒳, the space of available options, which must be given in advance and be well defined, so that it is determinate what is and is not a candidate. And there is an objective function J, which assigns to each candidate a value; and it is in the conditions on J that the presuppositions grow heavy. For the expression to denote, J must be representationally complete, defined over terms into which every value at stake has been written, since a value absent from J is a value the maximization cannot see. It must be exogenous and fixed, given prior to the search and unchanged by it, for a function that drifts while one climbs offers no summit one can reach and keep. It must be scalar, or valued in an ordering rich enough that maxima exist, since argmax over values that do not all compare has nothing determinate to return. And behind all three there must be an optimizer: a subject whose objective J is, whose preferences it encodes, presupposed to stand outside the problem, prior to it and stable through it, as the one for whom x* is best.

The four presuppositions, named in the order of their denial

It is worth naming these four in the order the paper will deny them, since the order is a descent from the surface of the problem to its ground, from whether the problem can be written to who is doing the writing, and each later condition presupposes the earlier ones already met.

The first presupposition is representational completeness: that every value at stake can be set down as a term in J. This is the condition that the problem can be stated at all, logically prior to every other, for a value that cannot enter J is not priced badly within the optimization but excluded from its statement, beyond the reach of any adjustment of the function. The first impossibility denies this.

The second is exogeneity and fixity: that the objective is given before the process and unaltered by it, a determinate target rather than a moving one. This is the condition an endogenous, self-revising objective violates, and the second impossibility denies it, not as a contingent fact about people whose tastes happen to wander but on the ontological ground that value in relation is generated within the relation and admits no external position from which it could be fixed.

The third is scalarity: that the values J assigns lie on a common scale, so that any two candidates compare and argmax has something to select. This is the condition incommensurable goods violate, and the third impossibility denies it.

The fourth is the stable optimizer: a subject with settled preferences standing outside the problem, whose objective J is and for whom its maximum is best. This is the condition a subject constituted within the very relation being optimized violates, and the fourth impossibility denies it, dissolving the fixed referent of the “who” the whole construction silently presupposes.

An instrument granted its proper domain

It clears the ground to see the four conditions holding somewhere, for the paper does not claim that optimization is everywhere ill-posed, only that it is so in a particular and important region, and the contrast is sharper if the well-posed case is granted its due. Consider a shipping company routing a fleet to minimize fuel cost. The feasible set is the routes the vehicles can take, well defined by the road network; the objective is fuel burned, representationally complete because fuel is the whole of what is being weighed and nothing relevant is left outside it; exogenous and fixed, since the price of fuel and the shape of the network do not shift because the planner is planning; scalar, since fuel is measured in one unit along which more and less are always defined; and owned by a stable optimizer, the firm, whose objective this determinately is and who is not remade by the routing. Here optimization is not merely legitimate but the only serious way to proceed, and the paper affirms this without reservation. The four conditions are jointly met, the operation is well defined, and to decline it would be a failure of reason and not a fidelity to anything. The example matters precisely because it is so clean: it shows that the paper’s quarrel is not with optimization but with its transposition into a domain where, as the next part argues, not one of these four conditions survives.

The demand that the anatomy be granted in full

The strategy of the argument requires that this anatomy be granted in its full strength, and the reason is worth stating, because it marks off the paper’s claim from a familiar and much weaker one with which it would otherwise be confused. The weaker claim holds that optimization problems in human affairs are merely hard: that the feasible set is too vast to search, the objective too costly to evaluate, the information too poor, so that we fall back of necessity on rules of thumb and on stopping when a candidate is good enough rather than best. This is the objection from bounded rationality, and it is true and important, and it is not this paper’s objection; indeed it concedes exactly what this paper denies. To satisfice because the true optimum is out of computational reach is to grant that a true optimum exists, well defined but unreachable, and merely to humble our access to it; the metaphysics of optimization is left standing, and only our finitude is confessed. Even the theory of choice under uncertainty, which replaces the known objective with an expectation over states, preserves the four conditions intact and simply integrates J against a probability, so that the agent maximizes expected rather than certain utility and the whole apparatus, with its complete, fixed, scalar objective and its stable subject, is carried over undisturbed. The paper’s claim reaches beneath all of this. It is not that the problem is well-posed and hard, nor that it is well-posed and risky, but that it is not well-posed at all, that one or more of the four conditions fails in kind and not in degree. To establish that, one must take optimization at its strongest, grant it a searcher of unlimited power and perfect information facing no uncertainty whatever, and show that even for such a searcher the problem does not exist, because the representationally complete, exogenous, fixed, scalar objective belonging to a stable subject that the operation requires is, in intimate relation, not there to be operated on. The anatomy is set out in full so that the denials, when they come, cannot be answered by the reply that a faster computer or a more patient economist would have found the maximum after all. There is, the next part argues, no maximum there to be found, and no faster computer reaches what was never posed.

§3 The Genealogy of Optimizationism

The picture criticized here was not always a picture of everything, and recovering the outline of how it became one is not idle history but part of the argument. A thesis that presents itself as the timeless form of reason draws much of its authority from seeming to have no alternative and no origin, to be simply how thinking goes; and that authority is loosened, though not refuted, once the thesis is shown to have been assembled, step by step, out of choices that could have been made otherwise. The point of the genealogy is therefore not to refute optimizationism by naming its birthday, which would be the genetic fallacy, but to change its standing: to move the claim that optimization is the form of things out of the register of self-evident truth, where it defends itself by seeming to have always been there, and into the register of historical construction, where one can ask what was assumed at each step and whether the assumptions hold. Three moments mark the passage from a local operation to a picture of the world, and at each a commitment enters that the later parts will have to test.

From purposive action to the calculus of means

The first moment is the isolation, within the theory of action, of a specifically instrumental rationality. Weber’s typology distinguishes the rationality of ends, Wertrationalität, oriented to a value held for its own sake, from the rationality of means, Zweckrationalität, which takes an end as given and concerns itself with the efficient fitting of means to it. The distinction is analytic, and in Weber it carries no triumph; he watched the advance of means-rationality with foreboding, as the closing of an iron cage around a disenchanted world. But the distinction did a lasting and double-edged work. By carving out a domain in which reason operates on means alone, with the end bracketed as given from outside, it made possible a rigorous science of efficient action; and it opened, at the same stroke, a fateful slippage. It is a short step, and one Weber himself did not take but the tradition after him did, from “reason can operate on means, the end being given” to “reason is the operation on means, and the end is simply given.” Once the end is bracketed it is easily forgotten as a question, and the bracketing that was meant as a division of labor within reason hardens into a definition of reason as such. What drops out is precisely the rationality of ends, the reasoned deliberation about what is worth pursuing, which Weber had kept in view and which the instrumental turn consigns to the merely subjective, a matter of decision rather than of reason. The seed of optimizationism is here: a form of reasoning whose entire content is the maximization of the satisfaction of an end it does not itself examine.

The marginalist objectification of the objective

The second moment gives the bracketed end a mathematical body and, in doing so, smuggles in the assumption on which everything later turns. In the marginalist tradition the ends of action are gathered into a single object, the utility function, and the agent is defined as its constrained maximizer; economics is recast as the science of the relation between ends and scarce means with alternative uses, and the rational agent is the one who allocates the means so as to maximize utility. Two things happen here that the criticism will need to have named. The first is a commensuration by construction. To represent the heterogeneous ends of a life as arguments of one function returning one magnitude is already to assume that they share a scale, that gains in one can be set against losses in another and a net computed; the common scale whose existence the third impossibility will deny is not argued for in the marginalist construction but built into the form of the utility function, where it becomes invisible precisely because it is structural rather than asserted. One does not notice an assumption one has been given as the shape of the tool. The second is an objectification of the end. What in Weber was an end held by an agent, something the agent is toward, becomes a function the agent has, a fixed datum standing ready for the optimization to range over; and with this the endogeneity of ends, their formation and revision within the very activity that pursues them, drops silently from view. The agent of marginalism arrives at his problem with his function already written, and the writing is treated as no part of the economic act, as though preferences were furniture brought to the room rather than something that happens in the living.

The axiomatic universalization in decision theory

The third moment lifts the maximizing agent clear of economics and installs him as the definition of rationality wherever there is choice at all. In the theory of expected utility, a small set of axioms on preference, of which completeness and transitivity are the load-bearing pair, is shown to entail that the agent’s choices can be represented as the maximization of the expectation of a utility function; to be rational just is to choose as if maximizing such a function. The result is a genuine achievement, elegant and, within its assumptions, unimpeachable, and the paper does not dispute the mathematics. What it marks is the quiet enormity of one axiom. Completeness requires that for any two options the agent prefer the one, prefer the other, or be indifferent, and this innocuous-seeming demand legislates that there is no pair of goods standing in any fourth relation, none between which preference is not yet settled in a way no sweetening resolves; transitivity then chains the pairwise comparisons into a single order. To accept the axioms is thus to decree, at the ground floor and before any particular choice is examined, that value is commensurable and the rational agent a maximizer on one scale, and, worse for the later argument, to make this decree the very meaning of the word “rational.” The consequence is that the person who declines to place two goods on a common scale, who holds that a vocation and a friendship are not rankable as better, worse, or precisely equal, is no longer voicing a substantive view about the goods that might be right; he is, by the definition now installed, irrational, his refusal a symptom rather than a position. Optimization, which entered as the efficient fitting of means to a bracketed end, has become by this third step the criterion of reason itself, binding on every chooser, in love no less than in commerce, and the completeness axiom has made the denial of commensurability unsayable within the frame, since to deny it is simply to fail the definition of the rational.

The reach of the genealogy, and its limit

Set end to end, the three moments trace the elevation of a local operation into a picture of the world: the bracketing of ends that made a pure calculus of means possible, the objectification and construction-borne commensuration of ends in the utility function, and the axiomatic installation of the maximizer as the meaning of rationality as such. What the genealogy establishes is that the claim “optimization is the form of rationality,” and its cosmic shadow “optimization is the form of things,” are constructions rather than findings, and that at each step an assumption entered, commensurability above all, which could have been declined and which the paper will argue is false in the domain of intimate relation. What the genealogy does not establish, and does not pretend to, is that optimization is thereby worthless; the fallacy of dismissing a tool by its history is one the paper is careful not to commit. Optimization is a superb instrument within the domain where its four conditions hold, and the shipping fleet of the previous section is not the exception but the rule across engineering, logistics, and the allocation of genuinely fungible resources, where to refuse the maximum would be a failure of reason and not a fidelity to anything finer. The error lies in the universalization rather than in the operation, in the inference from “this works wherever value is complete, fixed, commensurable, and owned by a stable subject” to “value is everywhere so, and where it seems not to be we have merely not yet done the sums.” That inference is what the completeness axiom makes feel like a truism and what the genealogy exposes as a step that was taken and could have been refused. It is that inference, and not the mathematics of maximization, that the next part denies, condition by condition and starting from the ground.

PART TWO · Four Impossibilities: Why Intimate Relation Is Not a Well-Defined Optimization Problem

§4 The First Impossibility — The Non-Structurability of Value

The four impossibilities descend from the most superficial precondition of the anatomy to the most fundamental, and it is worth being explicit at the outset about why this first one comes first, since the order is an argument in itself. The later impossibilities all concern a function already written: whether it can be held fixed, whether its values share a scale, whether a stable subject owns it. They presuppose, that is, that the values at stake have already been got onto the page as terms, and they ask what goes wrong thereafter. This first impossibility strikes earlier, at the writing itself. It asks whether certain of the values any honest account of intimate relation must weigh can be set down as terms at all, and it answers that some cannot, so that the objective is not merely hard to fix or to compare but impossible to complete. A value that cannot be written into J is not a value the optimization prices badly; it is a value the optimization never reaches, excluded not from the maximum but from the statement of the problem. If this is right, the optimizing ambition fails at the first step, before the later impossibilities are even in view, and the completeness on which the whole ambition silently rests is shown to have been unavailable from the start.

The completeness demand, made precise

The claim must be sharpened before it can be defended, because in a loose form it is merely the tired complaint that love is too rich for numbers, and that complaint the economist can absorb without a scratch. His reply, rehearsed already in the introduction, is that a value which seems to resist the function resists only because the function is written too poorly, and that a richer specification, with the goods of intimacy and the long horizon and the pleasures of the bond all entered as arguments, would take up whatever a first pass left out. The reply has real force, and it has a distinguished lineage: the program of extending utility maximization into the household, into marriage, fertility, and the allocation of affection, was carried out in detail and with great ingenuity, and it proceeded exactly by folding ever more of intimate life into the arguments of a utility function. To meet it one must state precisely what it demands. It demands completeness: that for every value v bearing on the relation there exist a term t_v, a representation of v as an argument of J such that the candidate states of the relation are ordered by how much of v they deliver. Completeness is not an incidental wish of the program but its condition of adequacy, for an objective that omits a value that genuinely bears on the relation is an objective for some thinner thing, and to maximize it is to optimize a caricature. The economist needs completeness, and the question is whether he can have it.

He can have it for a great many values, and the paper concedes this freely, since the concession sharpens the point of attack. Comfort, shared leisure, the division of labor in a household, even many of the tendernesses of a life together admit of proxies stable enough to serve as terms, and where they do the optimizing frame is not obviously ill-posed but merely, as the later parts will argue, ethically and practically treacherous. The first impossibility does not rest on the mass of ordinary goods that can be proxied. It rests on the claim that at least one value the objective must include cannot be, and that its exclusion is fatal to completeness however well the rest is done. The paper meets the demand, therefore, not at its weakest point but at its hardest: with the one value whose inclusion the economist most needs and least can secure, the psychoanalytic drive.

The drive as the value that cannot be written

That the drive must be in any complete objective for intimate relation is not a point the economist can concede away, for the drive is not an ornament at the edge of the bond but among its springs. What draws two people and holds them, what makes a particular other the one whose presence is not substitutable by another of equal merit, what gives desire its insistence and its opacity even to the one who feels it, belongs to the economy of the drive; an objective for a love that left it out would be an objective for companionship or for mutual advantage, some real thing but not the thing in question. So the drive must be a term. And the drive is precisely what admits of no term, for reasons that lie in its structure and not in our instruments.

The drive, in the register the series has used throughout, arises at a coupling of the Real and the Symbolic and reduces to neither. It is not a preference. A preference is a settled disposition to rank options, and a ranking is exactly what a term records; if the drive were a preference it would pose no problem, and could be written down like any other. But the drive does not rank the options in the feasible set, because what it circles is not among them. It orbits an object that is not a positive object at all but a vacancy, the place of what is constitutively missing, and it draws its force from the orbiting rather than from any attainable state that might be reached and possessed. This has a consequence that is fatal to representation as a term. A term t_v must assign to each candidate a magnitude, so that candidates can be ordered by it, more here, less there, monotone in the good; and the assignment must be such that getting more of the good is, other things equal, getting nearer the objective’s maximum. The drive violates this at its core, for it is extinguished by attainment where attainment occurs: the state that would count, on any proxy, as the fullest delivery of the good is the very state in which the drive, having reached its ostensible object, finds that the object was never it, and moves on. The mapping a term requires, from more of the good to nearer the maximum, is not merely inaccurate for the drive; it runs backwards, and no monotone magnitude can be laid over a structure whose satisfaction is its own undoing.

An illustrative case — the ledger of a marriage. Consider the attempt made in earnest rather than in theory. A couple, urged toward the frame by a culture that prizes it, resolve to reason about their marriage as an optimization: to enumerate what each values, to weigh the yields against the costs, and to act on the balance. Much of the enumeration succeeds. They can list the division of housework, the comfort of a shared home, the pleasures of company, and can even, with effort, assign these rough weights and speak of trade-offs among them without absurdity. Then one of them tries to enter the thing that actually binds them to this person and not another, equal on every listed count, and the enumeration fails. It is not that the value is large and hard to weigh. It is that when they try to say what it is a quantity of, every candidate answer — security, admiration, pleasure, familiarity — turns out to name something that a different person could in principle supply as well or better, and so to name not the bond but a proxy the bond exceeds. The remainder that resists every proxy, the sheer this-one-ness that is why the ledger is being kept at all, is precisely what will not go into a column, because it is not a quantity of anything and answers to no “more” or “less.” The ledger can be completed only by leaving out the reason for keeping it.

To force the drive into a term one would therefore have to replace it with a proxy, some measurable satisfaction standing where it stood; and here the decisive move of the economist’s program can be named and refused. The program’s most powerful device is revealed preference: the resolve to sidestep the inner life entirely and read value off behavior, defining what a person values as whatever their choices maximize, so that the drive, whatever it is in itself, is captured by the pattern of what the person is observed to choose. This is ingenious and, for a wide class of goods, unobjectionable. Against the drive it fails, and its failure is instructive. Revealed preference assumes that choice reveals a stable underlying ordering, that the pattern of choices is the trace of a preference the choices express. But the drive produces exactly the pattern revealed preference cannot parse: the pursuit that abandons its object on attainment, the choice that undoes itself, the wanting whose satisfaction is its extinction. Faced with this, revealed preference must either declare the behavior irrational, thereby excluding from the account the very spring of the bond it was meant to capture, or fit an ordering to the choices post hoc, thereby recording as a stable preference a movement whose whole character is that it is not one. In neither case has it represented the drive. It has, in the first, thrown the drive out as noise, and in the second, replaced it with a fiction of stability that is a different thing wearing its name. The proxy is not a poorer measurement of the drive. It is the substitution of a positive object for a vacancy, which is exactly the operation that empties the drive of what makes it the drive.

The category of the failure, and its consequence

The nature of the failure must be fixed exactly, because everything downstream depends on it. The failure is not epistemic. It is not that the drive is hard to quantify, that our instruments are coarse and our surveys noisy, and that a finer psychometrics would in time get its number. That reading leaves the optimizing frame wholly intact, conceding only that the measurements are poor; and it is the reading the paper most needs to refuse, because it is the one that keeps the door open to “do the sums better.” The claim is stronger and stranger: that there is here no number to be measured well or badly, that the mapping from the drive to a magnitude fails not in its accuracy but in its existence, because the drive is not the kind of entity of which there is a quantity to be had. This is the same distinction the anatomy drew between a problem that is merely hard and a problem that is ill-posed, and the drive falls on the second side of it not by degree but by kind.

Claim — Non-structurability. A value is non-structurable if its manner of being precludes its representation as a term in an objective function, that is, precludes any assignment to the candidate states of a monotone magnitude by which they could be ordered. The psychoanalytic drive is non-structurable: arising at a coupling of the Real and the Symbolic, circling a constitutive vacancy rather than a positive object, and extinguished by attainment where attainment occurs, it admits no monotone magnitude over the candidates and so no term. Its exclusion from the objective is therefore not a defect of measurement, remediable by finer instruments, but a matter of category. It cannot enter the statement of the problem.

The consequence is fatal to the economist’s reply on its own terms, and it is worth drawing the dilemma out to its end so that no escape is left standing. The reply promised to rescue optimization by completing the function; completeness for intimate relation requires a term for the drive; the drive admits no term. So the function must either omit the drive, in which case it is incomplete and the reply has not been made, the objective being one for some thinner relation that leaves out a spring of this one, or include a proxy for the drive, in which case it has put a positive object where a vacancy was and is again not an objective for this relation but for a substitute, so that the optimization, however impeccably posed thereafter, maximizes the wrong thing at the ground floor. There is no third path on which the drive is both admitted and itself, because admission as a term just is the imposition of a magnitude, and the drive admits none. The first precondition of the anatomy, that every value at stake be representable in the objective, therefore fails, and it fails before the questions of fixity, comparability, and the optimizer can be raised, since those concern a function already written and this concerns the writing. One may read the three impossibilities that follow as showing that even the incomplete function one could write, over such values as do admit of terms, would still furnish no well-posed problem. But this first has already shown that the function was condemned to be incomplete, and that the completeness on which the whole optimizing ambition depended was never there to be had.

§5 The Second Impossibility — The Relationality of Value and the Absent Quantifier

Suppose the first impossibility set aside, and grant, only to press on, that a function could be written over such values as do admit of terms. The second condition of the anatomy then comes into view: that this function be exogenous and fixed, given before the process of pursuit and unchanged by it, a determinate target rather than a moving one. This section denies that condition. But it denies it in a particular way, and the way matters as much as the denial, because a weak version of the denial changes nothing and the economist has already absorbed it. The weak version says that people’s preferences are unstable, that they want one thing today and another tomorrow, and that the objective therefore drifts. To this the economist answers, rightly, that instability is just more structure to model: let the utility function carry a time index, or let today’s choice include a term for its effect on tomorrow’s tastes, and the drift is absorbed into a larger fixed problem. The paper’s denial is not the weak one and does not meet this answer, because it does not locate the trouble in the instability of preferences at all. It locates it deeper, in the ontology of the value itself, and argues that value in intimate relation is relational, generated within the relation and existing only there, so that there is no standpoint outside the relation from which its magnitude could be read off and fixed. The objective cannot be exogenous because there is no exogenous position for it to be given from; and this is not a fact about wandering tastes that a time index could catch, but a fact about where relational value is, which no reindexing reaches.

Value as relational rather than intrasubjective

The utility function is a function of one subject. It encodes his preferences, the values as they stand for him, and its arguments are states of the world evaluated from his single standpoint; and even the theory of games, which sets several such functions interacting, leaves each the private possession of its owner, given before the interaction and merely played out through it. The whole construction presupposes that value is intrasubjective, lodged inside a subject and carried by him to the situation as a piece of his equipment. It is exactly this presupposition that the relational ontology of the series denies, and the denial is not a mood but a claim with consequences. The value at stake in an intimate relation, the meaning that a private word carries, the weight a kept silence has come to hold, the worth of the bond itself, is not a property either partner keeps inside and brings; it is generated between them, in the relating, and has no existence apart from the relation that bears it. This is not a metaphor. A private language means nothing to either speaker considered alone, because its meaning is the sediment of a shared history and is stored in the relation and not in either head; a rite that is precious to the two is not a datum in either partner’s separate ledger, because its preciousness is not a fact about either of them but a fact about what has formed between them. Value here has the standing that trust was shown to have in the paper on the non-binding vow: a property of a relational structure, a feature of the topology of the bond, and not an attribute of any node within it.

An illustrative case — the worth that is not in either of them. Two people have, over years, made a certain walk their own: a particular route, taken on particular evenings, that has gathered to itself the whole texture of their life together, so that to walk it is for each of them an act dense with meaning. Ask what the walk is worth, and to whom, and the question comes apart in the hand. It is not worth much to either of them separately: taken alone, one of them walking the route is a person on a road, and the meaning that floods the shared walk is simply absent, having no bearer. Nor is its worth the sum of two private worths, for there are no two private worths to sum; the meaning does not exist in either of them to be added. It exists between them, in the relation the walk both expresses and sustains, and it exists nowhere else. There is, then, no ledger in which the walk’s value is an entry, because a ledger is kept by a subject and totals what has worth for him, and the walk’s worth is not of that kind. To insist on entering it one would have to pick a standpoint, his or hers, and record from there a private shadow of a value that is not private, thereby recording something other than the value one meant to capture.

The quantifier that quantification requires, and its absence

From the relationality of value a consequence follows that is fatal to the exogeneity condition, and it is worth drawing out slowly, because it joins this impossibility to the deepest commitment of the series. To quantify a value, to assign it the magnitude a term in J demands, is to occupy a definite position: the position of one who surveys the field of value from outside and above, sees the candidates and their worths spread before him, and reads each onto a common scale. Quantification is silently a view from nowhere within the valued field, a God’s-eye operation, and it presupposes a standpoint exterior to the values it measures. For value that is intrasubjective this presupposition is met without difficulty, at least in principle, since the subject stands over against the world he evaluates as its external assessor; his preferences are inside him and the world is outside, and he reads the one onto the other. But for value that is relational, generated within the relation and existing only there, the presupposition cannot be met at all. Whoever would quantify such value must stand either within the relation or outside it, and neither will serve. Standing within, he is a party to the very generation he would assess, with no exterior purchase, no vantage from which the relational value lies spread before him as an object; he is inside the thing he would survey. Standing outside, he finds nothing to survey, because the value exists only within and is not present to any external position. There is no third place, no exterior and omniscient vantage from which the worth of what two people generate between them could be read onto a scale. And this is precisely the standpoint that quantification requires: the standpoint of a big Other who would hold the true measure of all relational value. The whole of this series, following the mother paper, has argued that this Other does not exist, that there is no metalanguage and no position from which the meanings by which people live could be assessed from without. To assign relational value a determinate magnitude is therefore to presuppose the one standpoint the series has shown to be empty.

Claim — The vacant quantifier. Value in intimate relation is relational: generated within the relation and existing only there, a property of the relational structure and not of any subject within it. To quantify a value is to assign it a magnitude from a standpoint exterior to it. For relational value no such exterior standpoint exists, since a position within the relation has no external purchase on it and a position outside the relation has nothing before it to assess; the standpoint quantification requires is that of a big Other holding the measure of all relational value, and this position is empty. The objective of an optimization over intimate relation therefore cannot be exogenous, having no external vantage from which to be given, nor fixed, being generated in the same motion that would pursue it. The endogeneity of the objective is not a contingency of unstable preference but the ontological consequence of the relationality of value.

The defeat of the economist’s reply

This is the point at which the economist’s reply is not merely met but defeated, and the defeat should be stated exactly, because the reply is resilient and returns in new dress if left half-answered. The reply was: your objective is misspecified or unstable; specify it better, index it to time, and the trouble dissolves. Against the first impossibility this failed because certain values admit of no term at all. Against the second it fails for a further and independent reason: that even for values which might admit of terms, there is no standpoint from which the terms could be assigned their magnitudes, the values being relational and the assigning position vacant. The reply rests, at bottom, on a picture in which the true worths are already out there, lodged somewhere and waiting to be read by a careful enough observer, so that better specification is a matter of reading them more accurately. That picture is exactly what the relationality of value refuses. Relational value is not lying in wait, outside the relation, to be discovered by a sufficiently patient economist; it is generated inside the relation, in the act, and holds still for no exterior gaze. The objective is not a target the searcher has specified too crudely. It is generated by the searching, revised in the relating, and held fixed by no standpoint that could serve as its source; and an argmax over an objective that is produced in the same motion that would maximize it, and that no external position holds still, does not name a hard problem awaiting a better economist. It names no problem, for there is nothing of the form the operation requires for the operation to be applied.

§6 The Third Impossibility — Incommensurability

Grant once more, for the sake of pressing on, that the objective could be written and that some standpoint could fix it. The third condition of the anatomy then requires that the values the objective assigns lie on a common scale, so that any two candidates compare and argmax has something to select. This section denies that condition through the incommensurability of certain goods: the thesis, developed most rigorously by Raz, that there are goods between which no common measure holds, so that the demand to rank them on one scale is not a hard task but a mistaken one, and that within intimate relation the attempt to price a good can be, in the strict sense, a betrayal of it rather than a misjudgment. As with the previous two impossibilities, the claim must first be distinguished from a weaker neighbor with which the frame would gladly confuse it, and only then pressed to its sharp form.

Incommensurability as a relation, not a difficulty

The weaker neighbor is the thought that some comparisons are merely hard, that weighing a career against a friendship is difficult and uncertain and beyond our present precision. This the optimizing frame absorbs without strain: hard comparisons are comparisons still, and difficulty is only a call for better estimates. Incommensurability in the strict sense is a different matter, a positive structural relation and not a confession of ignorance. Two goods are incommensurable when there is no common measure in terms of which their values are precisely comparable, when it is neither true that the first is better than the second, nor that the second is better than the first, nor that they are exactly equal. The three familiar relations of a common scale — better, worse, equal — are jointly exhausted only where a scale underlies them; where none does, two goods may stand outside all three, related in a way the scale cannot represent, and this standing-outside is not our failure to have measured but the goods’ failure to share a measure. The most exact mark of the relation is due to Raz and deserves to be given precisely, because it converts a vague intuition into a test. Suppose two options were merely equal on some common scale. Then improving one of them even slightly, sweetening it by a little, would have to make it determinately better than the other, since on a scale a small addition to an equal quantity yields a larger quantity. Now take a career and a friendship one is inclined to call equally choiceworthy, and sweeten the career by a modest raise in salary. If the two were equal on a scale, the raised career is now determinately the better choice. But it is not: the career with the raise is still neither better than the friendship, nor worse, nor exactly equal, just as before the raise. The tie does not break under the small improvement, and a tie that does not break under improvement was never a tie on a scale at all. The small-improvement argument shows, not that the comparison is hard, but that there is no scale beneath it to be hard about.

Pricing as betrayal

Within intimate relation this structural relation takes a sharper form, in which the attempt at commensuration is not only a mistake about the goods but a wrong done to them, and here the paper leans on the line running from Raz through Anderson to the recent moral critique of markets. The sharpening turns on a fact about certain goods: that they are constituted, in part, by the stances one takes toward them, so that to take the wrong stance is not to misdescribe the good but to alter it. Among the goods so constituted are those whose being includes a refusal to price them. For such a good, to price it is not to measure it, even badly; it is to change it into something else, and in the changing to betray it. This is why the refusal to name a sum is not squeamishness. To ask what payment would compensate the ending of a friendship is not to arrive, with effort, at a large figure; it is to have misunderstood what a friendship is, since a bond one holds oneself ready to release for a price is, by that readiness, already not the bond one took oneself to be pricing but a lesser thing wearing its name. The pricing does not report a worth and get the figure wrong. It produces, in the act of pricing, a diminished thing, and reports a fact about that.

An illustrative case — the offer. A man is asked, by someone who means the question seriously, what sum would induce him to give up his oldest friendship, permanently and with no further contact. Two things are worth noticing in his response. The first is that he does not experience the question as difficult, as though the number were very large and hard to fix; he experiences it as a category mistake, as one would the question how much the number seven weighs. The second is more telling. If he were to take the question in the spirit asked, to begin in earnest to name a figure and to negotiate it upward, something would already have happened to the friendship, before any sum was agreed and whether or not one ever was. To have entertained the offer in earnest is to have constituted oneself as someone whose friendships are for sale at a price, and a friendship held by such a person is not the good the offer was meant to price. The betrayal is not in the accepting, which never comes; it is in the pricing, which changes the thing priced in the moment it treats it as priceable. This is why the fitting answer is not a very large number but a refusal of the frame, and why the refusal is not a failure to compute but the only response that leaves the good intact.

Claim — Incommensurability. Two goods are incommensurable when they stand in none of the three relations a common scale affords, being neither better, nor worse, nor exactly equal, as the failure of their tie to break under small improvement reveals. Certain goods at stake in intimate relation are incommensurable in this sense; and some are further constituted, in part, by the refusal to price them, so that to place them on a common scale is not to mismeasure them but to transform and thereby betray them. The scalar condition of the anatomy therefore fails: the values the objective assigns do not all lie on a common scale, and argmax, which requires such a scale to have anything to select, has nothing over which to operate. The error is one of category and not of computation; the goods are not the kind of thing of which there is a sum.

The relation to the first two impossibilities

The three impossibilities so far compound rather than repeat one another, and it is worth marking the joints at which each cuts, since an objector may suspect that they are a single complaint in three costumes. The first concerned representation: whether a value can enter the objective as a term at all, and it found that the drive cannot, being extinguished by the attainment a monotone magnitude would track. The second concerned fixation: whether the objective, granting it written, can be given from an exterior standpoint, and it found that relational value admits none, being generated within the relation that bears it. This third concerns comparison: whether the values the objective assigns, granting them written and fixed, lie on the common scale that maximization requires, and it finds that incommensurable goods do not. The three are independent, and their independence is the strength of the case, for an objector who somehow answered one would face the others untouched: were the drive representable after all, its term would still have to be commensurated with the rest, and incommensurability would block it; were an exterior standpoint somehow found to fix the objective, the goods it fixed would still fail to share a scale. The descent is real, from whether the problem can be written, through whether it can be fixed, to whether what it fixes can be compared; and beneath all three lies a fourth question the descent has been circling from the start, not yet touched, the question of the one for whom the problem is a problem at all, the optimizer whose stable preferences the whole construction has quietly presupposed and whom the next section dissolves.

§7 The Fourth Impossibility — The Constitution of the Subject

The three prior impossibilities concerned the objective: whether it can be written, fixed from without, and furnished with a common scale. This last one reaches beneath the objective to the quietest condition of the anatomy, the one that underwrites the others and is least often brought to light, that there is an optimizer. Maximization requires not only something to be maximized but someone to maximize it: a subject whose objective J is, whose preferences it encodes, standing outside the problem, prior to it and stable through it, as the one for whom x* is best. Without such a subject the phrase “his optimum” has no determinate referent, for there is no settled his to whom the optimum belongs. This section denies that condition. In intimate relation the subject is not prior to the relation but constituted within it; the one who would optimize is a product of the process he is imagined to stand outside of; and the “who” the whole construction presupposes loses the fixity the construction needs.

The subject as constituted, not given

The optimizing frame inherits from marginalism a subject who arrives at the situation already formed, bearing a preference order settled in advance, which the situation then reveals and satisfies. Even the more sophisticated economics that allows preferences to change tends to model the change as itself the output of a prior and stable meta-preference, a fixed disposition governing how tastes evolve, so that the fixity merely retreats one level and the subject remains, at bottom, given before the process. The relational ontology of the series denies this picture at its root, and denies it of the subject exactly as it denied it of value, since the two denials are one. The subject of intimate relation does not enter the relation with its preferences fixed and undergo the relation as their playing-out. It is formed in the relating: its desires are shaped by the other’s desire, its sense of what is worth wanting is remade in the shared life, and its individuation, the very drawing of the boundary between what it is and what the other is, is accomplished and revised within the bond rather than settled before it. This is not the mild and assimilable point that people change their minds, which a meta-preference could absorb. It is the ontological point that the preference-bearing subject the optimization presupposes, the fixed evaluator standing over against the field of options, is not the antecedent condition of the intimate relation but one of its ongoing products. What the relation generates is not only value between the partners but the partners themselves, as the valuers they are in the course of becoming.

An illustrative case — who would collect the optimum. A person enters a relation with a settled sense of what he wants from a life, and over years within it that sense is remade: he comes to want things he had not known were wantable, to find hollow some things he had once ranked highest, to care about the other’s flourishing in a way that is not reducible to his own prior preferences because it has reorganized them. Suppose now he had, at the outset, tried to choose the relation by optimization, computing which arrangement best satisfied the preferences he then held. The arrangement that maximized the preferences of the man at the outset is not the arrangement the man he became would choose, and not because either made an error, but because the second man’s preferences are the product of the very relating that the first man’s optimization was meant to govern from before it began. Whose optimum, then, was to be computed? Not the first man’s, for he no longer exists to collect it, and his preferences were dissolved and reformed by the process. Not the second man’s, for he did not exist yet at the moment of choice, and could not have supplied the objective, being himself an output of what was chosen. There is no single subject present throughout to be the owner of the objective and the beneficiary of its maximum. There is a subject in the course of being constituted by the relation whose value the optimization would compute, which is a different thing from the fixed optimizer the frame requires.

The dissolution of the fixed referent

From this a difficulty follows that is fatal to the fourth condition, and it is a difficulty about reference rather than about stability, which the case has made ready to state. If the optimizer is constituted within the very process his optimization is supposed to govern, then “whose optimum?” has no fixed answer, because the subject named in any answer differs at the end of the process from the one who posed it at the start, and the difference is produced by the process itself. This is worse than the drift of the objective treated under the second impossibility, and must not be folded into it. There the target moved while the archer stood still; here the archer is remade by the shooting, so that the one who would gather the optimum at the summit is not the one who set out to climb and did not hold, at setting out, the preferences by which the summit is a summit. The frame needs a stable subject to be the owner of the objective and the beneficiary of its maximization, the one whose J it is and for whom x* is best; and in intimate relation there is no such stable owner, only a subject in the making whom the relation is precisely in the business of unfixing. To ask what is optimal for him is to presuppose a settled him that the relating dissolves and reforms, and the referent of the optimizing “who” comes apart in the asking.

Claim — The constituted subject. Optimization presupposes an optimizer: a subject prior to the problem and stable through it, whose objective the function is and for whom its maximum is best. In intimate relation the subject is constituted within the relation rather than given before it, its preferences and its individuation ongoing products of the relating. The optimizer is therefore a product of the process it is imagined to stand outside of, and “whose optimum?” loses a fixed referent, the subject named in any answer being remade by the very process the optimization would govern. The fourth condition of the anatomy fails: there is no stable subject to own the objective or to be the beneficiary of its maximum.

The four taken together

The four impossibilities may now be gathered, and gathering them shows that they are not four names for one objection but four independent failures at four distinct joints, running from the surface of the problem to its ground. Can the values be written into the objective? Not all: the drive, extinguished by attainment, admits no term. Can the objective, so far as written, be fixed from without? No: relational value admits no exterior standpoint, and the objective is endogenous of ontological necessity. Can the values it fixes be compared on a common scale? Not the incommensurable ones: the scale argmax needs is not there. And is there a stable subject to own the whole and receive its maximum? No: the optimizer is constituted in the process it would govern. Each cut is at a different joint, and each alone suffices, since an optimization problem needs all four conditions and the failure of any one leaves the operation undefined. Together they establish, without a single appeal to value, that intimate relation furnishes no well-defined optimization problem: not a hard one that greater computational power would crack, but an ill-posed non-problem, whose required object — a representationally complete, exogenous, fixed, scalar objective belonging to a stable subject — the operation of maximization does not find. With this the formal movement of the paper is complete. What follows is no longer proof but criticism, and it turns to the different question the proof has cleared the way for: not whether one can optimize intimate relation, which has been answered in the negative, but what befalls the relation when, undeterred by the answer, one imposes the frame regardless.

PART THREE · The Cost of Optimizationism

§8 The Damage to Generativity

The second movement of the paper begins here, and its question differs in kind from the first. The first movement asked whether intimate relation furnishes a well-defined optimization problem and answered that it does not; it passed no judgment on optimizing, establishing only that here the operation is undefined. This movement takes that result as settled and asks a question the result makes newly pressing: what happens when the frame is imposed regardless, when a domain that furnishes no well-posed maximization is treated as though it did, its relational and incommensurable and endogenous goods forced into the shape of an objective to be maximized? From here the claims are value-laden, and openly so; and the separation insisted on in the introduction is what earns them the right to be, for they no longer trade on the formal result but build upon it, taking as given that no optimization problem is present and asking after the cost of pretending one is. The first such cost, argued in this section, is a damage with a definite form: the imposition converts a value that should circulate and accumulate within a relation into a quantity that can be settled, extracted, and in the end exhausted.

Two fates for generated value

The theory of generative justice, developed in this series in exchange with Eglash, distinguishes two fates that the value generated in a relation may meet, and the distinction is the hinge of this section. Value may circulate: return to those who generate it, replenishing the conditions of its own further generation, so that the relation is a site of ongoing production that sustains itself and the value stays in motion among its makers. Or value may be extracted: drawn off from its generators, alienated, settled somewhere outside the circuit of its making, so that the relation is mined rather than cultivated and its capacity to generate is drawn down toward depletion. Eglash’s own cases are drawn from labor, ecology, and the technical arts, where the value in question is relatively legible and its diversion relatively easy to see; the contribution of this series has been to carry the distinction into intimate life, where the value at stake is least legible and its diversion hardest to name, and where, this paper now argues, the optimizing frame is one of the chief instruments of the second fate. The claim is not a loose analogy between mining ore and pricing love. It is that the operation optimization performs on a relational good is, in its formal structure, an operation of extraction, and the next step is to show why.

The effect of optimization on a circulation

To see the extraction one must see what optimization does to a value when it forces it into an objective. To make a relational good a term in J is to render it a quantity: a magnitude that can be summed with others, compared, traded off, and, above all, settled, brought to a final accounting at which one reads off whether the maximum was reached. But a relational good is not, in its own manner of being, a settleable quantity. It is a circulation, a generation whose life is in its continuing and which has no natural terminus at which it is complete and may be totalled. Meaning between two people is not a stock that accumulates toward a full amount; it is a movement that is either kept up or allowed to lapse. Rendered as a quantity to be maximized, such a good is made available to be settled, and to settle a circulation is to stop it. The optimizing frame does not merely mismeasure the relational good, assigning it a number that is off. It re-forms the good into the kind of thing that can be cashed out, and the cashing out is the depletion, because what is cashed out is no longer in circulation. The harm is thus not that the optimization reaches a wrong figure. It is that the operation of rendering-as-quantity, which optimization must perform before it can begin, has already converted a generation into a stock and thereby positioned it to be drawn down.

A single geometric illustration

The series has given this depletion a geometric form elsewhere, and it earns one sentence here, strictly as illustration and not as apparatus to be developed. In the account of the good and the bad cycle, generative value was shown to be carried not in any endpoint state but in the accumulated phase of a circuit, the holonomy an interpreter gathers in traversing it, which no terminal reckoning can capture because it is a property of the path and not of any point upon it. The single relevant point is this: that optimization, whose entire grammar is the selection of a best point, an x* in the feasible set, is by its very form blind to value that lives in a path rather than a point, and so cannot so much as register, let alone protect, the generative accumulation it settles away. The apparatus is not needed here and will not be unfolded; it serves only to name, in a vocabulary the series has already made precise, why a value carried in circulation is invisible to an operation that knows only summits.

An illustrative case — the audited relationship. A couple, anxious and conscientious, begin to keep an account of their relationship: to track what each contributes and receives, to note whether the returns justify the investment, to hold periodic reckonings at which the question is whether, on balance, the thing is paying off. Their intention is care; they mean to tend the relationship by measuring it. But the measuring does to the relationship exactly what this section describes. What had been a circulation, a giving and receiving whose worth lay in its ongoing motion and never stood still to be totalled, becomes a stock to be assessed at each reckoning; and a stock assessed is a stock that can come out short. The damage is not that the audit might return an unfavorable balance and end the relationship, though it might. The damage is prior to any balance and independent of its sign: it is that the auditing has changed the good audited, converting a generation kept alive by not being totalled into a quantity that exists to be totalled, so that even a favorable balance is now a balance struck on a diminished thing. The couple set out to protect the circulation and, by rendering it accountable, stopped it in the act of counting it.

The form of the damage

The damage can now be stated in its proper shape, and stated so that it is plainly a harm and not a mere infelicity of description. When a relational good is subjected to the optimizing frame it undergoes three movements. It is first rendered as a term, which requires assigning it a settleable magnitude it did not possess. It is then entered into an objective to be maximized, which subordinates it to the reckoning of a total in which it figures as one summand among others. And it is then, in the maximizing, spent: treated as a stock to be drawn down toward a maximum rather than a circulation to be kept in motion. This is the formal signature of extraction, and it is why the couple who begin in earnest to ask whether the relationship is worth it have already begun the conversion the theory of generative justice warns against, whatever answer they reach. The harm is not the risk of an unfavorable verdict. It is that the reckoning itself has changed the good reckoned, in the same way and for the same reason that the pricing of the previous part diminished the friendship it priced: by treating as a settleable quantity what had its life in never being finally settled. Optimization imposed where no optimization problem exists is thus not an idle error that leaves its object untouched while failing to improve it. It is an operation that remakes its object into something it can act upon, a settleable quantity, and in the remaking performs the very extraction that draws off into a final accounting the value whose life was in never being finally accounted.

§9 Optimizationism as a Symptom of Modernity

The criticism now widens from the damage optimization does within a single relation to the standing of optimizationism as a general thesis about the world. The claim of this section is that the elevation of optimization into a law of the universe is best understood not as a discovery but as a symptom: an epistemological form taken by a particular historical condition, which the mother paper of this series named the crisis of the symbolic, and which is bound up with the habituation of a life-world to the logic of capital. The universe, on this reading, climbs no landscape; the climbing of landscapes is a historically specific practice, native to a specific organization of production, that has been projected outward onto the cosmos and taken for its law. To argue this is not to refute optimizationism by its origins, the fallacy already disowned, but to relocate it: to move it from the register of eternal truth, where it defends itself as simply how things are, into the register of historical symptom, where one may ask what condition it expresses and what that condition costs. The genealogy of the first part showed that optimizationism was constructed; this section asks what, in the present, keeps the construction standing and makes it feel like nature.

The commensurating operation as the native form of capital

The decisive feature of optimizationism, isolated already in the genealogy, is its demand for commensuration: its requirement that heterogeneous goods be brought onto a single scale before a maximum can be defined. This requirement is not peculiar to the theory of optimization, and seeing where else it lives is the key to the diagnosis. It is the native operation of capital. The circulation of capital consists in nothing else than the reduction of qualitatively incomparable things — an hour of labor, a measure of grain, a song, an afternoon of care — to a common quantity in which they can be exchanged, accumulated, and made to breed more of themselves; the value-form is precisely a commensurating machine, and the historical career of capital is the spread of its commensuration into region after region of life, until things once held to lie beyond price acquire one and enter the circuit. Optimizationism is the epistemology proper to this machine. It is what universal commensuration looks like once it is raised from a practice to a theory of rationality and thence to a picture of the world: the settled conviction that behind the apparent heterogeneity of goods there always lies a scale on which they compare and a maximum toward which reason should climb. The conviction feels self-evident to the modern mind not because it is a self-evident truth but because the mind has been formed within a life-world that performs the commensurating operation on everything, ceaselessly, until the performability of the operation comes to seem a property of the goods rather than an imposition laid upon them. This is the critique of instrumental reason that the Frankfurt tradition pressed: that reason, reduced to the calculation of means and the commensuration of ends, mistakes its own historically produced form for reason as such and forgets the ends it can no longer reason about.

The symbolic crisis and the flight to the scale

The mother paper diagnosed the present as a crisis of the symbolic, a hollowing of the shared meanings by which a world is held together, and optimizationism can be placed within that diagnosis as one of its characteristic responses. The placement explains something the genealogy alone could not: why the constructed picture is not merely available but attractive, why people reach for it. When the thick symbolic orders that once told a community what things were worth, and told it in qualitative and incommensurable terms, lose their hold, a vacuum opens where the sense of worth had been, and into that vacuum the single quantitative scale flows as a replacement. A scale asks for no shared meaning. It needs no agreement about what a thing is for or what makes a life go well, only a number; and so it promises orientation to a disenchanted world that can no longer supply itself with meanings, a way to know what to do — maximize the measure — that survives the collapse of every thick account of the goods. Optimizationism is in this sense a flight: a flight from the difficulty of incommensurable value, which requires judgment and yields no algorithm, into the false comfort of a scale, which requires only computation and spares one the judgment. That the flight should present itself as hard-headed rationality, and the refusal of it as soft sentiment, is itself part of the symptom, for the condition valorizes the very operation that expresses it and recasts fidelity to the real structure of value, the refusal to commensurate what does not compare, as a failure of nerve to be overcome.

The projection onto the cosmos

The last movement of the symptom is the projection of the historically specific operation onto the universe as its law, and it is this movement that the paper’s title refuses. Because the commensurating, maximizing operation is performed everywhere within a certain form of life, it comes to seem that the operation is not being performed but merely recognized, that the landscapes were always there and reason only climbs them. One meets this projection in its popular cosmic form, in which everything from a molecule seeking its lowest energy to a market clearing to an organism maximizing its fitness is said to be “really” optimizing some quantity, so that the human who calculates is at last only doing consciously what all of nature does blindly, and optimization stands revealed as the deep grammar of being. The paper’s denial is aimed exactly here, and it can be put precisely. What is actually observed is not a universe that climbs landscapes but a form of life that climbs them and then, carrying its scale with it, finds the operation everywhere it lays the scale down. That certain physical systems admit description as extremizing a quantity is a fact about the reach of a mathematical formalism, not evidence that nature is engaged in maximization in the sense at issue, the sense in which there is an objective owned by a subject and a best state to be sought; the formalism’s success has been read back into the world as the world’s own procedure. To say that optimization is a law of the universe is to mistake the shadow the commensurating life-world casts over whatever it examines for a feature of what is examined. The practice is real and has a date. The law is a projection. And intimate relation, whose goods are relational and incommensurable and whose subjects are constituted in the relating, is the region where the projection is most plainly false, because it is the region in which the commensurating operation, once applied, most visibly destroys the very thing it was applied to.

§10 Optimizing Practice Is Not Ethical Practice

This section is the pivot of the criticism and, in a sense, of the paper, for it states in its sharpest form the claim the title’s question was aimed at from the start: that morality cannot be defined as an optimization problem, and that an optimizing practice, however well it succeeds on its own scale, is not thereby an ethical practice. The preceding sections argued that optimization is ill-posed in intimate relation and that its imposition does damage. This one argues something orthogonal to both, and independent of both: that even where an optimization is perfectly well-posed and perfectly successful, reaching a genuine maximum of a genuine objective, it may still fail as an ethics, because the criterion of the optimal and the criterion of the good are different criteria, answering different questions, and the optimal is not the good. The claim thus holds even in the shipping-fleet domain where the four conditions are met; it is not a claim about ill-posedness at all, but about the gap between maximizing and acting well, which no well-posedness closes.

The orthogonality of the optimal and the good

The series has met a structure of this kind before, and naming the earlier instance sharpens the present one and shows it is not improvised for the occasion. In the paper on the sustainability of generative relation, poetry was shown to satisfy a formal, dynamical criterion, the criterion of the good cycle, of endless non-settling circulation; and it was shown in the same breath that satisfying this criterion does not make a symbolic circulation good, since the aesthetics of fascism and the discourse of the cult satisfy the very same dynamical criterion and are evil. The lesson drawn there was that the dynamical criterion and the justice criterion are orthogonal: the first asks whether a circulation can sustain itself, the second whether it ought to and at whose expense, and a circulation may pass the first and fail the second, so that goodness requires the justice criterion as an independent condition which the dynamics can never supply. The present claim has the same shape, with optimization in the place of dynamics. The criterion of the optimal answers one question: whether, on a given objective, a given option is the best available, whether it sits at the maximum. The criterion of the good answers another: whether the option ought to be chosen, whether it wrongs anyone, whether it honors what the agent is bound to honor. These two are orthogonal. An option may be optimal and wrong; an option may be right and suboptimal; and success on the first criterion is no success whatever on the second, since they do not measure one thing along a common scale but pose different questions that a single answer cannot jointly settle. To have found the maximum is to have answered the first question and not so much as touched the second.

An illustrative case — the optimal betrayal. Suppose a man faces a choice well-posed enough to optimize: the relevant goods admit of proxies, the objective is stable, the scale is available, and he himself is settled in his preferences. He computes, correctly, that abandoning a friend who has become burdensome, at a moment when the friend most depends on him, maximizes his objective, and even, on a plausible specification, maximizes an impartial aggregate of well-being, since his own gain is large and the friend will in time recover. The optimization is impeccable; the maximum is genuine; on its own scale nothing better is available. And the act is a betrayal. The point is not that he has optimized the wrong objective, which would return us to the earlier impossibilities; grant that the objective is as good as an objective gets. The point is that optimality and rightness have come apart, that the very act which is best on the scale is the act that wrongs the one to whom he was bound, and that no refinement of the maximization repairs this, because the wrong is not a quantity the objective failed to include but a feature of the act that the whole dimension of maximizing cannot see. He did the optimal thing, and he ought not to have done it, and both are true at once.

The resistance of morality to the optimizing form

That the two criteria are orthogonal is not a brute fact but follows from the kind of thing morality is, and three features of the moral show why it resists the optimizing form in particular, each naming a place where the case above bites. The first is the unconditionality of certain moral requirements. An obligation of the kind Kant placed at the center of ethics does not present itself as one good among others to be weighed in a sum and outweighed should the sum come out otherwise; it presents itself as binding regardless of the balance, a constraint on what may be done and not a quantity to be traded against other quantities. An optimizing frame can represent such a requirement only by falsifying it: either as a very large term that a still larger term could in principle outweigh, which denies its unconditionality, or as an absolute constraint bounding the feasible set, which removes it from the objective altogether and concedes that the moral was never a matter of maximization to begin with. The second feature is the non-tradability of moral considerations. To make a duty a term in an objective is to make it tradable, available to be sacrificed when the numbers favor the trade; but much of what we recognize as moral seriousness consists precisely in the refusal to make certain considerations tradable, in holding that there are things one does not do whatever the yield, and this refusal is unintelligible within a frame in which everything is, by the form of the objective, a quantity that trades. The man in the case did the tradable thing; that the friend’s dependence could be traded away for a sufficient gain is exactly what a moral seriousness would deny. The third feature is the irreplaceability of the other. The person to whom one is bound is not, morally, the bearer of a quantity of value that another bearer of equal quantity could replace without loss; the other is irreplaceable, and the claim the other makes is the claim of this singular one and not of the value it happens to carry. Optimization, whose feasible set is a space of interchangeable options ranked by a common measure, cannot represent irreplaceability at all, since its every option is in principle substitutable by another of equal or greater value; and irreplaceability is precisely the denial that such substitution is ever, here, without loss.

Claim — Orthogonality of the optimal and the good. The criterion of the optimal and the criterion of the good are orthogonal: the first asks whether an option maximizes a given objective, the second whether it ought to be chosen, and success on the first is no success on the second. An optimizing practice, even one well-posed and reaching a genuine maximum, may therefore fail as an ethical practice, and does so wherever the good it maximizes is not the good it ought to honor. Morality resists the optimizing form in particular because of the unconditionality of its requirements, the non-tradability of its considerations, and the irreplaceability of the other, none of which an objective function can represent without falsifying. Morality is therefore not a higher-order optimization, a maximization of some moral quantity, but a practice outside the grammar of optimization. The optimal is not the good.

The completion of the criticism

With this the criticism completes itself, and it does so by closing the last exit the optimizing frame might take. Confronted with the argument thus far, a defender of optimization has one sophisticated move left: to concede that ordinary utility is the wrong objective and to propose that morality itself be optimized, that we maximize not pleasure or preference but the good, folding the moral considerations into a richer objective and maximizing that. This section’s claim is what forecloses the move. Morality cannot be the objective of an optimization, because its central features — the unconditional, the non-tradable, the irreplaceable — are exactly the features an objective function cannot carry; to write morality into an objective is to strip it of these features and so to write in something else under its name. The optimizing frame cannot be rescued by aiming it at the good, for the good is not the kind of thing there is a maximum of. And this does not leave ethics without rationality, which would be a disaster of the opposite kind and which the paper is at pains to avoid. It leaves ethics with a rationality that is not optimization, a way of reasoning about what to do that is answerable, discriminating, and severe without being the maximization of any measure; and the recovery of that other rationality, promised from the start, is the work of the part that follows.

PART FOUR · After the Negation: A Non-Optimizing Relational Rationality

§11 Rationality Is Not Optimization — Reclaiming the Word

The paper has denied that intimate relation is an optimization problem and criticized the imposition of the frame; it owes now an account of what relational rationality positively is, and the debt is not optional. To leave the matter at the denial would invite the very reading the paper most wants to refuse: that in surrendering optimization one surrenders rationality itself, and consigns love to mere feeling, to a warm unreason that does as it likes. That reading has force only because a long usage has quietly equated rationality with optimization, until the two words function as synonyms and the space of the reasonable has narrowed to the space of the maximal. The first task of the recovery is to break that equation, and it is a task of reclamation rather than invention, for the equation is a recent contraction of a word that once meant more, and what the paper reclaims was there before optimization annexed the term.

The contraction of a word

That the equation is a contraction and not a definition can be read directly off the genealogy already given. Rationality was not always maximization. Weber’s own distinction preserves a rationality of ends alongside the rationality of means, a reasonableness about what is worth pursuing and not only about how to pursue efficiently what is given; and behind Weber lies the far older tradition of practical wisdom, in which the rational agent is not the one who maximizes a measure but the one who perceives well what a situation requires and acts fittingly, whose excellence is discernment and proportion rather than computation. The narrowing of rationality to optimization is the achievement of the particular history the genealogy traced, in which the bracketing of ends, the objectification of the objective, and the axiomatic installation of the maximizer combined to make “rational” mean “maximizing on a scale” and to make anyone who declines the scale, by the completeness axiom, irrational by definition. Once this narrowing is seen as an achievement with a date, its air of necessity dissolves. It is not that reason is optimization and relation therefore falls outside reason; it is that a certain history has annexed the word “reason” to optimization, and an annexation can be contested. The paper contests it, and the ground of the contest is that the older and wider senses were not confusions the narrowing cleared up but competences the narrowing dropped.

Two rationalities, distinguished

The reclamation is best done by distinguishing two things the one word has been made to carry, the rationality that maximizes and the rationality that responds, and by showing that the second is no less rational for being no optimization. The rationality that maximizes takes a fixed objective and a field of options and selects the option that makes the objective greatest; its virtue is efficiency, its measure the distance to the summit, and its proper home the domain where value is representable, fixed, commensurable, and owned by a stable subject — the domain, that is, where the four conditions hold and the shipping fleet is routed. The rationality that responds takes a situation whose goods are relational, incommensurable, and in part non-representable, and a self still in the making rather than fixed, and it does what the first rationality cannot: it attends to what the situation and the other require, discerns among incommensurable goods without a scale, and acts fittingly rather than maximally, answerable at every point to a demand that comes from without rather than to a total computed from within. This second rationality is not a defective form of the first, an optimization carried out with poor instruments and no clear objective. It is a competence of a different kind, with its own standards of success and failure, its own ways of going right and of going wrong, and its own severity, described in the sections that follow. To call it irrational because it does not maximize is to mistake the annexed sense of the word for the whole of it, and to require that the reasoning proper to relation justify itself before a tribunal whose jurisdiction the paper has denied.

Severity without a maximum

Everything in the recovery turns on one point, which must be insisted on because it is the point the optimizing frame will most resist: that the rationality of response is severe, that it has genuine standards and is not a license for whatever one likes. The deepest suspicion of any non-optimizing rationality is that it is merely a soft permission, a way of dignifying caprice by excusing it from the discipline of the measure; and the suspicion is fair, and must be met, for a relational rationality that amounted only to “do as you feel, unencumbered by calculation” would deserve the contempt the optimizing frame heaps on it, and would moreover betray the goods it claims to serve, which make hard demands. The answer is that the rationality of response has a discipline of its own, as exacting as the discipline of the maximum and in one respect harder, since it cannot be discharged by a rule: the discipline of getting the other right, of perceiving the situation as it is and not as one would prefer it, of discerning which among several incommensurable goods this occasion asks to be honored, of sustaining a generation without grasping it. These are standards one can fail, and fail badly. A response can be false, inattentive, self-serving, disproportionate; the difference between a fitting response and an unfitting one is as real as the difference between a right answer and a wrong one, though it is not the difference between a larger number and a smaller. Indeed the case of the optimal betrayal in the previous section was a failure of exactly this rationality, and its being a failure, assessable as such, is the proof that a standard was in force: the man perceived his friend rightly enough to know what was owed, and acted against the perception, and the wrongness of his act is not a smaller number on any scale but a failure of response that the scale could not even record. The severity is real. It is simply not the severity of optimization. Rationality, reclaimed, is wide enough to hold both the maximizing and the responding; and the reasoning proper to relation is the second, no less rational for owning no maximum, and answerable to standards the first cannot see.

§12 The Form of Relational Rationality

The rationality of response has been distinguished from the rationality that maximizes and credited with a severity of its own; it remains to describe its form positively, to say what its operations are, so that it stands as a competence with a shape and not merely as the negation of optimization. Three moments compose it, and they are not stages of a procedure run in sequence but aspects of a single way of proceeding, present together in any exercise of it: response, the answering of a demand that issues from the other; attunement, the perceptual fit to a situation and its goods; and sustaining, the keeping-in-motion of a generation that is not to be grasped. Each is a standard one can meet or fail, and each names a competence the optimizing frame not only lacks but cannot represent, so that describing them is at once to give the recovery its content and to show why the frame it replaces was blind to exactly these things.

Response

The first moment is response, and it names a reorientation of rationality from a total computed within to a demand answered from without. The maximizing agent faces a field of options and consults an objective that is his own; the question he asks is what best serves that objective, and the direction of his reasoning is inward, from the world to his measure. The responding agent faces an other whose claim upon him is not derived from his objective and cannot be, and the question he asks is what this other, this situation, requires of him; the direction of his reasoning is outward, from himself to a demand he did not set and cannot revoke. This is why the ethics of the irreplaceable other is the native home of relational rationality rather than an ornament upon it: the other’s claim is precisely a demand that comes from beyond the self’s objective, and a rationality of maximization cannot so much as register it, since it appears in no term of any objective the self could hold. Response is answerable in a way maximization is not, and the difference is the mark of its rationality rather than a softness in it. The maximizer answers to his own objective and to nothing else, and so cannot, strictly, get the objective wrong; it is simply his. The responder answers to what is other than himself, and can therefore fail in ways the maximizer has no room for: by failing to hear the demand at all, by hearing it and answering falsely, by substituting his own measure for the other’s claim and calling the substitution a response. That these are genuine failures, distinguishable from success, is what shows response to be a competence of reason and not a mood, for only what can be done rightly or wrongly is a matter of reason at all.

Attunement

The second moment is attunement, and it names the perceptual achievement without which response would be blind. To respond fittingly one must first perceive rightly: see the situation as it is, register which goods are truly in play, discern what this occasion, among all the occasions it resembles, distinctively asks. This is the competence the tradition of practical wisdom set at the center of ethics, the trained perception that grasps the salient features of a particular case and is not the application of a rule to a case already described, but the prior and harder work of arriving at the description the rule would need. It is exactly the competence for which the optimizing frame has no place, since optimization can begin only once the situation has been reduced to a feasible set and an objective, and attunement is the labor of reaching a sense of the situation that precedes that reduction and often resists it. Attunement is what allows discernment among incommensurable goods without a scale, and this is its most important and least obvious office. Where the maximizer, denied a common measure, is simply stalled, unable to rank and so unable to act, the attuned agent is not stalled at all: he perceives which good this situation asks to be honored, not by computing a total in which the goods are compared, which is impossible, but by a discernment of the case that sees what it calls for. That such discernment is possible, that people do in fact choose well among incommensurable goods and can tell a wise such choice from a foolish one, is itself evidence that a rationality other than optimization is at work, since the choice is demonstrably not a maximization and yet is demonstrably not arbitrary, being open to praise, criticism, and the judgment that it was rightly or wrongly made.

Sustaining, and an image

The third moment is sustaining, and it names the aim proper to a rationality of response, which is not to reach a summit but to keep a generation in motion. Where the maximizing agent seeks a terminal state, the best point at which the work is done and the objective achieved, the responding agent seeks no terminus, for the goods of relation have their life in continuing and are spoilt by any final reckoning, as the criticism showed at length. The rationality of sustaining is therefore oriented to the ongoing rather than the achieved: its success is that the generation continues, replenished and unpossessed, and its characteristic failure is not falling short of a maximum but letting the generation lapse, whether by grasping it, which settles it, or by neglecting it, which starves it. Here a single image may be allowed, and only as an image. There is an old figure for a strength that does not seek the summit: water, which does not strive upward or contend for the high place, but goes to the low ground others disdain, takes the shape of whatever holds it, and by never contending nourishes all things. The figure catches, better than an argument might, what a non-maximizing rationality is like — its yielding that is not weakness, its finding of the way by fitting rather than forcing, its power that lies precisely in not climbing; and it is offered for that one flash of recognition and then set down, without exposition, since to expound it would be to build an apparatus from what is meant only to light the thought. The form of relational rationality is response, attunement, and sustaining. Each is a standard, exacting and failable; together they are the competence the paper reclaims under the name of reason, a reason whose excellence is to answer, to perceive, and to keep in motion, and never to maximize.

§13 Non-Action as the Highest Rationality

The recovery closes with the moment that reaches furthest from optimization and gives the paper’s positive claim its most demanding form: that the highest exercise of relational rationality is a kind of non-action, a knowing when not to intervene, not to force, not to optimize at all, and that this restraint, far from the absence of reason, is its most difficult accomplishment. The claim draws on the notion of wuwei, non-action or effortless action, and completes the recovery by naming the competence that even the rationality of response must include: the competence of letting a generation arise rather than making it, of tending the conditions of a good rather than seizing the good directly. It is the sharpest point of the whole positive account, because it is where relational rationality most flatly contradicts the imperative the optimizing frame cannot suspend, the imperative always to close the gap to the maximum.

The rationality of restraint

The optimizing frame knows no restraint of this kind and cannot, for its imperative is always to act, to move toward the maximum, to close whatever distance remains between the present state and the best. Within that frame, inaction is intelligible only as a failure to optimize, a leaving of value on the table; there is no positive place in it for the deliberate refusal to intervene, since any refusal that forgoes an available improvement is, by the logic of the maximum, simply irrational. The rationality of relation has such a place, and it is central rather than marginal, because much of what is best in a shared life arises only when it is not forced and would be destroyed by the very intervention that sought to secure it. Trust that is demanded is not trust; spontaneity that is engineered is not spontaneous; a tenderness produced on purpose in order to produce it is something other than tenderness. For goods of this kind, and they are the characteristic goods of intimacy, the fitting exercise of reason is often to refrain: to make the space in which the good can arise of itself, and then to leave it be. And to know which goods are of this kind, and when the space has been made, and how to keep from the intervention that would spoil them, is a discernment as exacting as any the previous section described, and in one respect harder, since it must be exercised against the standing impulse to do something, to help, to secure, to optimize. This is the rationality of restraint, and it is not the abdication of reason but its refinement, the point at which reason grows wise enough to know the limits of its own doing and to recognize that some goods are available only to a hand that does not grasp.

Letting the good cycle arise

The series has given this restraint a precise place before, and recalling it shows that the figure of non-action is not an exotic import brought in to decorate the ending but the name of a claim the argument has carried throughout. In the account of the good and the bad cycle, the good cycle was shown to be one whose positive accumulation is endogenous, arising from the internal tension of the relation itself rather than pumped in by extraction from without; and non-action was there identified as the stance that neither extracts nor forces closure, that lets the endogenous accumulation occur of itself rather than manufacturing a false increase by drawing on some outside. The same identification holds here and completes the argument of the present paper by joining its end to its middle. To optimize a relation is to intervene upon it, to act upon it as upon a system to be driven toward a maximum; and this intervention is exactly the forcing the good cycle cannot survive, since it substitutes an external drive for the internal generation and yields, at best, a manufactured increase that depletes what it draws upon, which is the very extraction the criticism of the third part described. The rationality of non-action is the refusal of this intervention: the tending of the conditions under which the relation generates value of itself, and the restraint from the optimizing action that would replace that self-generation with an extraction. What the maximizing frame reads as leaving value on the table is, seen rightly, the only way the value in question is ever had at all, since it is value that arises only where it is not seized, and the hand that reaches to maximize it closes on the diminished thing its reaching made.

The measure that is not the maximum

The deepest form of the paper’s positive thesis can now be stated, and it is fitting to end the recovery on it. Optimization takes its measure from the maximum: the standard against which every state is judged is the best attainable state, and reason’s task is to close the distance to it. A rationality of the kind this part has described takes its measure elsewhere, from what a situation and an other are of themselves, from the grain of the thing and the movement it already has; and reason’s task, on this measure, is to accord with that movement, to sustain and not force it, to tend and not seize, rather than to override it toward a maximum defined apart from it. This is the deepest sense in which optimization is neither the form of relational rationality nor the law of the world. The world of relation does not present a landscape with a summit that reason should climb. It presents a generation with a movement of its own that reason should discern and accord with; and the highest rationality is not the one that reaches the highest point but the one that accords most truly with what is there, that knows when to act and, harder, when not to, and finds in restraint not the failure of reason but its fulfillment. With this the recovery is complete. Relational rationality is response, attunement, sustaining, and, at its height, the non-action that lets the good arise of itself; and none of these is a maximization, and every one of them is reason.

PART FIVE · Close

§14 The Polyphonic Conclusion

It remains to close, and the manner of the closing is not incidental to the argument but its last step. The paper has denied that intimate relation is an optimization problem, criticized the imposition of the frame, and recovered a rationality of response, attunement, sustaining, and non-action. A reader who has followed this far will feel the pull of a single question, and the whole of the close turns on how it is answered: if not utility, then what should one optimize? Love, perhaps, or meaning, or generativity itself? The temptation to answer is strong, and it is strong for a reason worth naming, since the reader has been trained by the very frame the paper opposes to expect that a serious argument ends by supplying a new objective to put in the place of the one it rejected. An answer would round the paper off in the shape the frame has taught us to find satisfying, replacing the discarded objective with a worthier one and leaving the machinery of maximization intact but redirected. The paper refuses the answer, and the refusal is not an omission or a failure of nerve but the final content of its view.

The surrender concealed in naming a new objective

The temptation must be met head-on, because to yield to it would undo everything the paper has argued, and would undo it quietly, wearing the look of a constructive conclusion. To say “do not optimize utility; optimize love” is to keep the entire apparatus the paper has dismantled — the objective, the scale, the maximum, the optimizer — and merely to change the argument of the function. But the four impossibilities were not impossibilities about utility in particular. They were impossibilities about the optimizing form as such, brought to bear on intimate relation: that its goods resist representation, admit no exterior standpoint to fix them, share no common scale, and belong to no subject stable enough to own the objective. Not one of these failures is touched by a change of objective. Love, written as the argument of J, is as non-representable, as relational, as incommensurable, and as owned-by-no-stable-subject as any other relational good; to make it the thing maximized is to inflict upon it precisely the damage the criticism described, rendering the circulation a settleable quantity and drawing it off in the reckoning. The offer to optimize love rather than utility is therefore not a friendlier version of the optimizing frame but the same frame with a more sympathetic mask, and it does the same harm to a more beloved object; indeed it does the harm more surely, since the reverence in which love is held makes the reckoning feel like devotion and disarms the suspicion that would have attended the reckoning of mere utility. To accept the offer would be to lose the argument at the last step, having won it at every prior one. The paper therefore declines to name a master objective, not from an inability to think of one, but from having understood that any objective so named would reinstate the error the whole paper exists to expose.

The plural and the unsynthesizable

What stands in the place where a master objective would go is not a blank but a positive thesis, and it is the thesis the series has reached before by other roads: that the goods and the rationalities of relation are plural, and that their plurality admits no synthesis into a single measure without loss. There is no one thing that relational rationality maximizes, and this is not because the one thing has yet to be found but because there is none to find. There are, rather, many goods standing to one another in relations no common scale represents, and many aspects of the reasoning that honors them — response and attunement and sustaining and restraint — that do not reduce to a single operation performed on a single objective. This is the counterpart, in the theory of rationality, of the polyphonic form the series has given its conclusions since the paper on sustainability, where it was argued that the truth of a matter may lie in the tension among several frameworks none of which subsumes the others, and that to force them into one synthesis is to lose exactly the truth that lived in their irreducibility. The refusal to name a master objective is that same refusal carried to the level of value: an insistence that the many goods and the many modes of their honoring are not to be collapsed into one, and that the collapse, however it satisfies the desire for a single answer, destroys what it unifies. The plurality is not a stage on the way to a unity not yet achieved. It is the finding.

The form as the argument

That the paper enacts this refusal rather than merely asserting it is the last thing to say, and it is what binds the paper’s form to its content in the manner the series requires of itself. A paper that argued against the single measure and then, in its conclusion, supplied one, would refute itself in its form whatever its sentences maintained; the shape of its ending would betray the burden of its argument, offering with the closing gesture exactly the master objective the body had shown to be impossible. This paper ends instead by declining the measure, and the declining is meant to be read as the argument’s completion and not its evasion. The most honest thing a work against optimization can do at its close is to withhold the new optimum the reader has been trained to expect, to leave the plurality unsynthesized on the page, and to let the absence of a master objective be felt as the presence of the thesis. The reader who reaches the end and finds no single answer to “what, then, should one optimize?” has not been failed by a paper that lost its nerve before the constructive task. He has been given the answer, which is that the question, in this domain, has no answer of the kind it asks for, and that to have felt the want of one and let it stand unfilled is to have understood. The form withholds the summit because there is none, and the withholding is what the paper had to say.

§15 Envoi — A Promise That Is Not Optimized

The argument has run its course, from the anatomy of the optimizing frame, through the four impossibilities that show it ill-posed in intimate relation, through the criticism of its imposition, to the recovery of a rationality that responds rather than maximizes. It remains to close as the series closes its papers, not with an argument but with an image, and to return, as the paper began, to the plainest form of the thing it has been about.

Consider a promise. Two people bind themselves to one another for a life, and in doing so they perform an act the optimizing frame cannot understand. The frame would have them compute: to weigh the expected value of the bond against its alternatives, to keep the option of revision open should a better prospect appear, to treat the vow as a provisional maximum held only until the numbers change. A promise made in that spirit is not a promise but a standing intention to defect when it pays, and everyone knows the difference, though the frame cannot state it. For a promise is precisely the refusal to keep computing. It is the act by which one closes the ledger, forecloses the revision, and binds oneself past the point where the calculation could reach, saying not “I will stay while you remain the optimal choice” but “I will stay,” full stop, whatever the sums come to say. The whole worth of a promise lies in its being unconditional in just the way an optimization can never be, and the series has argued elsewhere that this unconditionality, far from a weakness, is the very source of the trust a promise can generate, since it is the mark that the one who promises has placed the bond beyond the reach of the calculation that could betray it.

This is why a promise is a promise only when it is not optimized, and why the attempt to make it so destroys it in the same motion by which the pricing destroyed the friendship and the maximizing settled the circulation. To ask, of the one you have bound yourself to, whether she is worth it, to hold the vow open against the day a better yield appears, is not to be a more rational lover but to have stopped promising and started calculating, and to have lost, in the calculating, the very good the promise was. The good of the promise is available only on the far side of the refusal to compute its worth; it is a good one can have only by declining to ask the question the optimizing frame insists is the rational one. And so the deepest thing the paper has to say is said not by its arguments but by this plainest of acts, in which one human being says to another that here the question of worth will not be asked, not because the answer might be unwelcome but because the asking would already be the betrayal. There are goods a love does not maximize, and it is not for want of a better measure. It is because to have reached for the measure at all would have been to have loved something other than the one before you, who is not a quantity of value that a greater quantity could replace, but this one, irreplaceable, whom one does not weigh. 未曾计算，故其爱得全 — never weighed, and so the love is whole; which is to say, a promise, and not a sum.

Acknowledgements

Two debts frame this paper, and it is fitting to name them together at its end, for one gave it its question and the other its reason, and between them they enclose everything the paper contains.

The question with which the paper begins, whether optimization is really a law of the universe, came from an exchange with Ron Eglash, in the course of a longer conversation on the political economy of value and the theory of generative justice. It arrived in a form so plain that its force took time to register, and its whole work was to move optimization from the place of an unexamined procedure to the place of a thesis that might be false. Everything the paper is followed from taking that displacement seriously. Beyond the question, the theory of generative justice developed in exchange with Eglash supplies the criticism of its third part with its central concept, the distinction between value that circulates and returns to its generators and value that is drawn off and settled elsewhere, without which the account of the damage optimization does could not have been given. Where the paper has reasoned well, the debt is his; where it has reasoned poorly, the fault is the author’s alone.

The other debt is older and needs no argument. This paper, like every paper of the series, is for the girl who belongs to the forest, whom the author has never once weighed for whether she was worth it, because that was never a question one could compute. If the acknowledgement to Eglash records where the question came from, this one records for whom the whole was written; and it is not beside the point of the paper but its plainest confirmation, for a bond that refuses to be priced, that is held past every reckoning and never put to the calculation of its worth, is the nearest witness there is to the thesis these pages have labored to establish, that there are goods a love does not maximize, and that it is not for want of a better measure. 致我最爱的人。 To the one I love most.

中文

亲密关系的哲学与正义论 · 第二十五篇

亲密关系中的优化理论与理性

论优化作为宇宙法则的界限

有些善，一份爱不去将其最大化，而这并非因为缺一把更好的尺子。

万宏

有价值不生于度量，
权之，则已失其所是。

有些善并非生于度量，而去权衡它们，便已然弄错了它们是哪一类东西。

能权万物之权，未尝有也。

那能称量万物的秤，从未存在过。

—— 作者

致她——
For her—

那个属于森林的女孩， —— the girl who belongs to the forest,
我从未计算过值不值得， —— whom I have never once weighed for whether she was worth it,
因为那从来不是一个可以计算的问题。 —— because that was never a question one could compute.

致我最爱的人。 To the one I love most.

摘要

本文论证优化不是宇宙的一条法则，而道德与亲密关系之存在方式，在其结构上，并不是优化问题。它以两个被小心分开的运动而行。第一个是一份不携任何价值判断的形式证明：亲密关系提供不出一个良定义的优化问题——所争之价值部分地是不可结构化的，因为”驱力”（任何完整的目标函数都必须将其纳入）抗拒被表征为一个项；它是关系性的，在关系之内被生成、且只存在于那里，于是没有任何外部立足点能固定那目标；它是不可通约的，于是给某些善定价，是背叛而非误测它们；而主体是在关系之内被构成、而非先于关系被给定，于是模型所预设的那个”优化者”，是它本该置身其外的那个过程的产物。每一处失败击中一个不同的前提条件，而任何一处便已使”最大化”这一运算无定义。

第二个运动把那一结果当作既定，转而批判那”无视此结果而强加此框架”之举。在不存在优化问题之处去优化，会把一个本应流通的价值转化为一个可被结清、可被抽走的价值，其方式正是生成性正义理论所命名的”榨取性计算”；它是象征界之危机、以及一个习惯了资本逻辑的生活世界的一种症状，在其下，一个局部的、历史性的运算被投射到宇宙之上、被当作它的法则；而它弄错了道德是哪一类东西，因为一种优化性的实践，即便是成功的，也可能作为一种伦理而失败——最优者与善者是正交的。本文随后复原一种回应而非最大化的理性，以回应、调谐、以及对生成之维系来描述之，并以复调作结，拒绝命名一个新的主目标，因为那样做会把那框架重新安置回来。那一拒绝便是本文的肯定性主张，被践行于它的形式之中。

关键词： 优化；不可通约性；关系性价值；生成性正义。

AI 使用声明： 本文的思想、主张与理论框架源于作者本人。AI 工具被用作起草与修订之辅助。每一句话都已由作者逐行核校，每一想法之来源都已核实，故作者对内容负全责。

第一部分 · 问题：作为一种世界观的优化

§1 导论 —— 一个被误认作法则的运算

一个问题可以被如此广泛地预设，以至它从不被问出，而本文的预设便是其中之一。在当下的日常言说里，某一族词语以自明之物的自信反复出现。人们说”最优的”选择、”高效的”安排，说一件事是否”值得”，说从一天、一段事业、一场婚姻中获取”最多”。这些词并不空洞。它们携带着一幅图景，而这幅图景，在它成为经济学、管理学以及大部分决策科学之母语的那一个世纪里，已硬化为某种近乎形而上学的东西。在这幅图景上，对于任何一个人可以行动的情境，都有一个要被尽可能做大的量；一生中的诸情境是一片片地景，而活得好便是攀登它们；而善，无论在何处被找到，便是坐落于山巅之物。这幅图景如此弥漫，以至它的核心运算——取一个最大值——不再看起来像是众多各有其领域的运算之一，而看起来像是理性本身的形式，再往外、朦胧地，像是万物的形式。本文所论正是这一硬化，而它的主导主张是一个否认：优化不是宇宙的一条法则，而道德与亲密关系之存在方式，在其结构上，根本不是优化问题。

这一否认易于陈说、也易于被误解。它不是那虔敬的抱怨，说人不该优化爱、说那样做冷酷或不浪漫。这样的抱怨恰恰让渡了争议之所在，因为它承认那里有一个优化问题待解、只是出于得体而请求我们别去解它。本文的主张比这更强、更冷。它是：在亲密关系中，严格说来，没有一个良定义的优化问题可供人克制不去解——最大化这一运算，施用于此，并不是有一个人拒绝去算的难解，而是根本没有一个合法的解可算，因为使这样一种运算得以定义的诸条件，在这一领域里，未被满足。听到那抱怨的经济学家有一个现成的回答：爱之所以抗拒优化，只是因为我们的效用函数尚未被完整地写出，而补救之道是一个更丰富的函数，一个把亲密、意义与长远视野都折叠进去的函数。这回答有力，且必须在它自己的地基上被迎击。本文中心部分之负担，便是表明它失败——麻烦不在于一个写得太差的函数，而在于（缘由将在其位置上给出）根本不可能有一个函数可写。

问题之由来，及其攸关

组织本文的那个问题来自外部，在一场关于价值之政治经济学与生成性正义理论的交流之中，而它以一种简单到需要一些时间才能感到其分量的形式到来：优化是否真的是宇宙的一条法则。那位对话者自己的框架早已抗拒把异质的价值化约为单一的结算尺度，而问题在那意义上已然内含于其中；但被赤裸地陈出，它做了一件特定的工作。它把优化从一个显然而未经审视的程序之位置，移到一个形而上学论题之位置——也就是说，一个可能为假的论题。一旦它被看作一个论题，人便能问它断言了什么、它是否为真，而这一发问便是本文之全部。攸关者不限于亲密。倘若论证成立，那么一幅主宰着大量当代”如何生活”之推理的图景，便被表明——至少在一个中心区域——建立在一个范畴错误之上；而亲密关系，本系列通篇将其取作关系性存在之最小且最暴露的单元，正是那错误最可辨认之处，因为正是在那里，抗拒这幅图景的诸善最少被伪装。

两个运动，被有意分开

本文的论证落入两个运动，切不可将其混淆，而整体之大部分力量取决于把它们持分开。第一个运动是一份证明，它是内在的、形式的。它追问亲密关系是否提供一个良定义的优化问题，并以四步回答说它并不；而它这样做，是有意地、不诉诸任何价值。它不说优化是坏的。它只说这一运算在此无定义：没有所需之形状的东西供那运算去作用。人可以像一个无神论者读一份”某个方程在实数域上无解”的证明那样，读本文第二部分之全部，全无任何道德色彩，而这论证应能强制一个对爱不怀任何特别敬意、只希望他的形式体系被良好地提出的读者。

第二个运动是一份批判，它是外在的、规范的。它把第一个运动之结果当作既定——此处并无优化问题——并追问：当那图景被无视此结果地强加，当一个提供不出良定义之最大化的领域被当作它提供了一般来对待时，会随之而来什么。正是在此、且只在此，负载价值的诸主张才进入：那强加损害它所触及者，把一个生成性的流通转化为一个可被榨取、可被耗尽的量；它是一个特定历史处境的症状、而非一条无时间性的真理；而它弄错了道德是哪一类东西。次序要紧。证明先廓清地基、表明那框架不合身；批判随后审视强行套上它的代价。把二者合在一起、同时论证优化无定义且优化邪恶，会削弱二者，因为那会让读者怀疑那形式主张只是道德上的厌恶乔装而已。它不是，而那分离便是担保。

三条线索，及复原之形状

三条线索贯穿论证，一如它们贯穿本系列，而在此点名它们，能让它们的再现读作一种设计。第一条是精神分析的：任何完整的目标函数都不得不称量的诸善之中有”驱力”，而驱力，生起于实在界与象征界之耦合，恰恰是那不能被完整写下者，于是价值之不可结构化，不是我们当前形式体系之局限、而是那东西本身的一个特征。第二条是关系性本体论与大他者之空缺的线索，跨越本系列而展开、并在关于非约束性誓言与关于关系性疏离的论文中被陈述得最尖锐：价值不是持于一个主体之内的属性，而是在一段关系之内被生成、且只存在于那里，于是不存在关系之外的位置、不存在全知的第三方，能从其上读出并固定其价值。这条线索做最重的工，因为它供给了”目标为何必须是内生的”之缘由，并把整体系于本系列反复的坚持：不存在元语言、不存在能占据那”量化默然要求之上帝视角”的大他者。第三条是生成性正义与政治经济学的线索，在与 Eglash 的交流中展开：本应流通、并回归其生成者之价值，可以反被结清并抽走，而一个优化框架之强加，是这一榨取所取的诸形式之一。

这些线索汇聚于一次复原，因为本文并不终于否认。既已论证理性不是优化，它便欠下一份”关系性理性于是为何”之说明，而它给出一份：一种回应之理性、调谐之理性、对一个生成之维系之理性，它有其自身的严苛，而那不是最大值之严苛。但这复原被持守——以本系列所要求之方式——在那会将它拆解的诱惑之前止步。回答”若非效用，人应优化什么？”以”那就爱吧，或意义，或生成性”，会是容易的、也会是一场失败，因为那答案保留了框架、只是给山巅重新上漆。本文拒绝它。它的肯定性主张是：关系性理性是复数的、不容被综合入单一的尺度，而”拒绝命名一个主目标”不是一种回避、而是这一观点之内容，被践行——一如本系列践行其诸主张——于它自己收尾的复调形式之中。

§2 一个优化问题的形式解剖

在人能表明亲密关系提供不出良定义的优化问题之前，人必须对”一个良定义的优化问题所要求者为何”精确无误，而这精确不是迂腐、而是论证之力量的全部条件。本文中心部分的四个不可能，每一个都是此处所列诸条件之一的失败；若诸条件被松散地陈述，诸失败便可被闪避，而若它们被以漫画化的形式陈述，则整个反驳便是一个稻草人。本节之目的因此是：不带论战地、并以一个决策理论家自己会接受的形式，精确陈述——为使”最大化”这一运算根本指称任何东西，什么必须到位。本节把优化最强、最体面的版本许予它，以便当后面各部分逐一撤去诸条件时，它们撤去的是优化者认得为自己的一个构造、而不是一个为被击倒而搭起的滑稽模仿。

一个优化问题，在其典范形式中，是对一个集合中”使一个函数尽可能大”之元素的搜寻。写出来，它是这个表达式

x* = argmax_{x ∈ 𝒳} J(x)，

而记号的紧凑掩藏了四个预设，每一个都必须成立，那表达式才挑得出任何东西。有一个可行集 𝒳，即可用选项之空间，它必须被预先给定且良定义，于是何者是、何者不是一个候选便是确定的。又有一个目标函数 J，它赋予每一个候选一个值；而正是在对 J 的诸条件上，诸预设变重。为使那表达式有所指，J 必须表征上完整，被定义于”每一个所争之价值都已被写入其中”的诸项之上，因为一个不在 J 之中的价值，是那最大化看不见的价值。它必须外生且固定，先于搜寻被给出、且不被搜寻改变，因为一个在人攀登之际漂移的函数，提供不出一个人能抵达并守住的山巅。它必须标量，或取值于一个丰富到最大值存在的序之中，因为对”并非彼此皆可比”之诸值取 argmax，没有确定之物可返回。而在这三者之后，必须有一个优化者：一个”J 是其目标、J 编码其偏好”的主体，被预设立于问题之外、先于问题、并贯穿问题而稳定，作为”x* 对之为最优”的那一个。

四个预设，按其被否认之次序命名

值得按本文将否认它们的次序来命名这四者，因为那次序是一次从问题之表层向其根基的下降——从问题能否被写出，到究竟是谁在写——而每一个较后的条件都预设较早的诸条件已被满足。

第一个预设是表征上的完整性：每一个所争之价值都能被作为 J 中的一个项写下。这是”问题根本能被陈述”之条件，逻辑上先于其余一切，因为一个进不了 J 的价值，不是在优化之内被定错了价、而是被排除在它的陈述之外，超出了任何对函数之调整的够及。第一个不可能否认这一点。

第二个是外生性与固定性：目标先于过程被给出、且不被过程改变，是一个确定的靶而非一个移动的靶。这是一个内生的、自我修正的目标所违反的条件，而第二个不可能否认它——不是作为关于”口味恰好游移之人”的偶然事实、而是在本体论的地基上：关系中的价值在关系之内被生成、并不容一个能固定它的外部位置。

第三个是标量性：J 所赋之诸值躺在一把公共的尺度上，于是任何两个候选皆可比，而 argmax 有可选之物。这是不可通约之善所违反的条件，而第三个不可能否认它。

第四个是稳定的优化者：一个立于问题之外、持有已定偏好的主体，J 是其目标、其最大值对之为最优。这是一个”在正被优化的那段关系之内被构成的主体”所违反的条件，而第四个不可能否认它，消解掉整个构造默然预设的那个”谁”之固定指称。

一件被许予其恰当领域的工具

看见那四个条件在某处成立，能廓清地基，因为本文并不主张优化处处不良提出、而只主张它在一个特定而重要的区域如此，而若那良提出的情形被许予其应得，对照便更为鲜明。设想一家运输公司为一支车队规划路线以最小化燃料成本。可行集是车辆可走的路线，由路网良定义；目标是所烧之燃料，表征上完整，因为燃料便是所称量者之全部、无相关之物被留在其外；外生且固定，因为燃料之价与路网之形状不因规划者在规划而移动；标量，因为燃料以一个单位度量，沿之”更多”与”更少”总是被定义；且被一个稳定的优化者所拥有，即那家公司，这确定地是其目标、且它不被这次路线规划所重塑。此处优化不仅正当、而且是唯一严肃的行进方式，本文毫无保留地肯定这一点。四个条件被共同满足，运算被良定义，而拒绝它会是理性之失败、而非对任何东西之忠诚。这个例子恰恰因其如此干净而要紧：它表明本文之争执不在于优化、而在于把它移置入一个（如下一部分所论）这四个条件无一存活的领域。

要求这份解剖被完整地许予

论证之策略要求这份解剖被以其全强许予，而缘由值得陈明，因为它把本文的主张与一个它否则会被混淆的、熟悉而弱得多的主张划开。那更弱的主张认为人类事务中的优化问题只是难：可行集太大而搜不遍、目标太贵而评不起、信息太差，于是我们出于必要退回到经验法则、退回到”当一个候选足够好而非最好时便停下”。这是”有限理性”之反驳，它为真且重要，而它不是本文之反驳；实则它恰恰承认了本文所否认者。因为真正的最优超出计算之够及而满足于”足够好”，便是承认一个真正的最优存在、良定义却不可达，而只是谦卑我们对它的通达；优化之形而上学被留立不动，而被坦白的只是我们的有限。即便是不确定下的选择理论——它以一个跨诸状态的期望替换那已知的目标——也把四个条件完好保存，只是把 J 对一个概率积分，于是行动者最大化期望而非确定的效用，而整套装置，连同它那完整、固定、标量的目标与它那稳定的主体，被原封不动地承接过来。本文的主张够到这一切之下。不是说问题良提出而难，也不是说它良提出而有风险，而是说它根本不良提出——四个条件中的一个或多个失败于种类、而非失败于程度。为确立这一点，人必须把优化取于其最强处，许予它一个能力无限、信息完美、面对丝毫不确定性都没有的搜寻者，并表明即便对这样一个搜寻者，问题也不存在，因为那运算所要求的、属于一个稳定主体的表征上完整、外生、固定、标量的目标，在亲密关系中不在那里供人作用。这份解剖被完整地列出，以便当否认到来时，它们不能被”一台更快的计算机或一个更耐心的经济学家终究会找到那最大值”之回答所答复。下一部分论证：没有一个最大值在那里可被找到，而没有一台更快的计算机够得到那从未被提出者。

§3 优化主义的系谱

此处所批判的图景并非一向是一幅关于一切的图景，而恢复”它如何成为这样一幅图景”之轮廓，不是闲散的历史、而是论证之一部分。一个把自己呈现为理性之无时间性形式的论题，其权威大半汲取自”看起来无替代、无由来，只是思维本来之运行方式”；而一旦那论题被表明是被一步步、从本可作出别样之诸选择中装配而成，那权威便被松动——尽管未被反驳。系谱之要点因此不是靠点出优化主义的生日来反驳它，那会是发生学谬误，而是改变它的地位：把”优化是万物之形式”这一主张，从”它靠看似一向在此而自我防卫”的自明真理之register，移入”人能追问每一步假设了什么、那些假设是否成立”的历史建构之register。三个时刻标记着从一个局部运算到一幅世界图景的过渡，而在每一个时刻都进入一个后面各部分将不得不检验的承诺。

从目的性行动到手段的演算

第一个时刻是在行动理论之内对一种特定的工具性理性的孤立。韦伯的类型学把目的之理性——价值理性（Wertrationalität），朝向一个为其自身之故而持有的价值——从手段之理性——目的理性（Zweckrationalität），它把一个目的取为既定、并关切于手段对它的高效贴合——区分开。这区分是分析性的，而在韦伯那里它不携任何凯旋；他以不祥之感注视着手段理性之推进，视之为一个祛魅世界周围铁笼之合拢。但这区分做了一件持久而双刃的工作。靠划出一个”理性只对手段运作、目的被作为从外部给定而搁置”的领域，它使一门高效行动的严格科学成为可能；而在同一举之中，它开启了一个宿命的滑移。从”理性能对手段运作，目的乃是给定的”到”理性就是对手段之运作，而目的只是被给定的”，是短短的一步——一步韦伯本人未走、而他之后的传统走了的一步。目的一旦被搁置，便易于作为一个问题被遗忘，而那本意作为理性内部分工的搁置，硬化为对理性本身的一个定义。所脱落者恰恰是目的之理性，即关于”什么值得追求”的说理性审思，韦伯曾将其保持在视野中、而工具性转向把它发付给纯然的主观、发付给一件决断而非说理之事。优化主义的种子在此：一种其全部内容是”最大化它自己并不审查的一个目的之满足”的推理形式。

边际主义对目标的对象化

第二个时刻给那被搁置的目的一个数学的身体，并在这样做时偷运进那后面一切所转之上的假设。在边际主义传统中，行动的诸目的被聚拢入一个单一的对象——效用函数，而行动者被定义为它的约束下的最大化者；经济学被重铸为”目的与有替代用途之稀缺手段之间关系”的科学，而理性行动者便是那”如此分配手段以最大化效用”的人。此处发生两件批判将需要已经命名之事。第一件是一种由构造而生的通约。把一生的诸异质目的表征为一个返回一个量的函数之诸自变量，已然假设它们共享一把尺度、假设一者之所得可被抵一者之所失而算出一个净额；第三个不可能将否认其存在的那把公共尺度，在边际主义构造中不被论证、而被内建入效用函数之形式，在那里它变得不可见，恰恰因为它是结构性的而非被断言的。人不会注意到一个被作为工具之形状交给自己的假设。第二件是一种目的之对象化。在韦伯那里曾是”一个行动者所持有的目的”、是行动者所朝向者，变成了行动者所拥有的一个函数、一个候在那里供优化去遍历的固定既与；而随此，目的之内生性——它们在追求它们的那一活动之内的形成与修正——默然从视野中脱落。边际主义的行动者带着他那已然写好的函数抵达他的问题，而那书写被当作根本不属于经济行为，仿佛偏好是被带进房间的家具、而非在活着之中发生之事。

决策论中的公理化普遍化

第三个时刻把那最大化的行动者提离经济学，并把他安装为”凡有选择之处理性为何”之定义。在期望效用理论中，一小组关于偏好的公理——其中完备性与传递性是承重的一对——被表明蕴涵：行动者的诸选择能被表征为对一个效用函数之期望的最大化；为理性便正是”仿佛在最大化这样一个函数”地选择。这一结果是一项真正的成就，优雅，且在其诸假设之内无可指摘，而本文不争论那数学。它所标记者，是一条公理之静默的巨大。完备性要求：对于任何两个选项，行动者要么偏好这一个、要么偏好那一个、要么无差别，而这看似无害的要求立法说：不存在任何一对处于某种第四关系之中的善，不存在任何一对”偏好尚未以任何加甜所能化解之方式被定下”的善；传递性随后把逐对的比较链结为一个单一的序。接受这些公理，因此便是在最底层、在任何特定选择被审查之前，就颁令说价值是可通约的、而理性行动者是一把尺度上的最大化者，而且——对后面的论证更糟——把这一颁令化为”理性”这个词的意思本身。后果是：那”拒绝把两个善置于一把公共尺度、认为一份志业与一段友谊不能被排为更好、更差或恰好相等”的人，不再是在表达一个可能为对的、关于诸善的实质性观点；他，按如今被安装的那个定义，是非理性的，他的拒绝是一个症状而非一个立场。优化——它以”手段对一个被搁置之目的的高效贴合”进入——到这第三步已成为理性本身之判据，约束着每一个选择者，在爱中不亚于在商业中，而完备性公理已使”对可通约性之否认”在此框架内无法说出，因为否认它便只是没能通过”理性”之定义。

系谱之够及，及其界限

首尾相接，三个时刻描摹了一个局部运算被抬升为一幅世界图景之历程：使一门纯粹的手段演算成为可能的对目的之搁置，效用函数中目的之对象化与由构造而承之通约，以及最大化者作为”理性本身之意思”的公理化安装。系谱所确立者，是”优化是理性之形式”这一主张、及其宇宙之影”优化是万物之形式”，是诸建构而非诸发现，且在每一步都进入一个假设——首要者是可通约性——它本可被拒绝，而本文将论证它在亲密关系之领域为假。系谱所不确立、也不佯装确立者，是优化因此便一文不值；靠一件工具之历史来打发它，是本文小心不去犯的谬误。优化是在其四个条件成立之领域内的一件绝佳工具，而上一节的运输车队不是例外、而是遍布工程、物流、以及真正可替换之资源之分配的常规，在那里拒绝那最大值会是理性之失败、而非对任何更精细之物的忠诚。错误在于普遍化、而非在于运算，在于从”这在价值完整、固定、可通约、且被一个稳定主体所拥有之处皆有效”到”价值处处如此，而在它看似不然之处我们只是尚未把账算完”之推断。正是那一推断，被完备性公理弄得感觉像一句老生常谈，而被系谱揭露为一个被走了、本可被拒绝的步。正是那一推断、而非最大化之数学，是下一部分所否认者——逐一条件、从地基开始。

第二部分 · 四个不可能：为何亲密关系不是一个良定义的优化问题

§4 第一个不可能 —— 价值之不可结构化

四个不可能从解剖之最表层的前提条件下降到最根本者，而在开头就明说”这第一个为何居首”是值得的，因为那次序本身便是一个论证。后面的诸不可能都关乎一个已然写好的函数：它能否被持固定、它的诸值是否共享一把尺度、一个稳定主体是否拥有它。也就是说，它们预设所争之诸价值已被作为项弄上了纸面，并追问其后出了什么错。这第一个不可能击中得更早，击中书写本身。它追问”任何诚实的亲密关系之说明都必须称量的某些价值，究竟能否被作为项写下”，并回答说有些不能，于是那目标不仅难以固定或比较、而是不可能被完成。一个不能被写入 J 的价值，不是优化定错了价的价值；它是优化从不够及的价值，被排除的不是最大值、而是问题之陈述。倘若这为真，优化之抱负便在第一步失败，甚至在后面的诸不可能进入视野之前，而整个抱负默然凭依的那份完整性，被表明从一开始便无从得到。

把完整性之要求弄精确

主张必须先被磨利，才能被辩护，因为在松散形式里它只是那”爱太丰富、装不进数字”的老套抱怨，而那抱怨经济学家能毫发无损地吸收。他的回答——导论里已排演过——是：一个看似抗拒函数的价值，之所以抗拒，只因函数写得太差，而一个更丰富的规格，把亲密之诸善、长远视野、结合之诸乐都作为自变量录入，便会接住第一遍所落下之物。这回答有真实之力，且有一个卓著的血脉：把效用最大化扩展进家户、进婚姻、生育与情感之分配的纲领，曾被详尽而极富机巧地执行，而它进行的方式恰恰是把越来越多的亲密生活折叠进一个效用函数之诸自变量。要迎击它，人必须精确陈述它所要求者。它要求完整性：对于每一个关涉此关系之价值 v，都存在一个项 t_v，即把 v 表征为 J 之自变量、使得关系之候选诸状态依”它们递送多少 v”而被排序。完整性不是那纲领的一个附带愿望、而是它的适足性条件，因为一个略去了一个真正关涉此关系之价值的目标，是某个更单薄之物的目标，而最大化它便是优化一幅漫画。经济学家需要完整性，而问题是他能否得到它。

对于许许多多价值他能得到它，而本文坦然承认这一点，因为这一承认磨利了进攻之点。舒适、共享的闲暇、家户中的分工，乃至一段共同生活中许多的温柔，都容许稳定到足以充当项的代理量，而在它们如此之处，优化框架并非显然不良提出、而只是（如后面各部分将论证）在伦理上与实践上凶险。第一个不可能并不凭依那一大堆可被代理的寻常之善。它凭依这一主张：目标必须纳入的价值中至少有一个不能被纳入，而它的被排除对完整性是致命的，无论其余做得多好。因此本文迎击那要求，不在它最弱之点、而在它最难之点：以那个经济学家最需要纳入、却最难保住的价值——精神分析的驱力。

驱力，那个写不出来的价值

“驱力必须在任何一份亲密关系之完整目标中”不是经济学家能让渡掉的一点，因为驱力不是结合之边缘的一件装饰、而在它的诸泉源之列。什么把两个人吸引并留住，什么使一个特定的他者成为”其在场不可被另一个同等优秀者所替代”的那一个，什么给欲望以它的坚执、以它甚至对感到它者的那份晦暗，都属于驱力之经济；一份把它略去的爱之目标，会是一份关于陪伴或关于互利之目标，是某个真实之物、却不是所论之物。所以驱力必须是一个项。而驱力恰恰是那个不容任何项者，缘由在它的结构、而非在我们的工具。

驱力，在本系列通篇所用之register中，生起于实在界与象征界之耦合、且不化约为二者中任何一个。它不是一个偏好。一个偏好是一种”给诸选项排序”之已定倾向，而一个排序恰恰是一个项所记录者；倘若驱力是一个偏好，它便不构成问题、并能像任何别的东西一样被写下。但驱力不给可行集中的诸选项排序，因为它所环绕者不在它们之中。它绕行的是一个根本不是正向对象、而是一个空缺的对象，即那”构成性地缺失者”之位置，而它从环绕、而非从任何可被抵达并占有之可达状态汲取它的力。这有一个对”表征为一个项”致命的后果。一个项 t_v 必须给每一候选赋一个量，于是候选能依它被排序——此处更多、彼处更少，在善上单调；而那赋值必须使得”得到更多的善”，在其他条件相等时，便是”更接近目标之最大值”。驱力在其核心处违反这一点，因为在抵达发生之处它被抵达所熄灭：那个在任何代理量上算作”善之最充分递送”的状态，恰恰是驱力在抵达它表面上的对象之后、发现那对象从来不是它、并继续前行的那个状态。一个项所要求的映射——从更多的善到更近的最大值——对驱力不仅不准确；它反向而行，而没有任何单调的量能被铺在一个”其满足即其自我拆解”的结构之上。

一个例证性的案例 —— 一场婚姻的账本。 设想那一尝试是当真地、而非在理论上作出的。一对被一种珍视此框架的文化催逼向它的伴侣，决意把他们的婚姻当作一次优化来推理：把各自所看重者一一列举、把收益权衡于成本、并依那平衡而行动。列举之大部分成功。他们能列出家务之分工、共享之家的舒适、相伴之诸乐，甚至能费力地给这些赋以粗略的权重、并不荒谬地谈论它们之间的取舍。然后他们中的一个试图录入那个”实际把他们绑定于这个人、而非另一个在每一列举之项上皆相等之人”的东西，而列举失败了。不是那价值巨大而难以称量。而是当他们试图说出它是一个关于什么的量时，每一个候选答案——安全、仰慕、愉悦、熟悉——都原来命名了某个”一个不同的人在原则上能供给得一样好或更好”之物，从而命名的不是那结合、而是一个那结合所逾越的代理量。那个抗拒每一个代理量的余剩，那个”账本之所以被记的原因”之纯粹的”就是这一个”，恰恰是那不肯进入一列之物，因为它不是关于任何东西的一个量、不应任何”更多”或”更少”。账本只能靠把”记它的理由”略去而被完成。

因此，要把驱力逼进一个项，人便不得不用一个代理量替换它，某个可度量的满足立于它曾立之处；而正是在此，经济学家纲领之决定性一招能被点名并被拒绝。那纲领最有力的装置是显示性偏好：那”彻底绕开内在生活、从行为读出价值”的决意，把一个人所看重者定义为其诸选择所最大化者，于是驱力，无论它自身为何，被此人被观察到之选择的模式所捕获。这机巧，而对一大类善无可非议。对驱力它失败，而它的失败富有教益。显示性偏好假设选择显示一个稳定的底层排序、假设诸选择之模式是”诸选择所表达之一个偏好”之痕迹。但驱力恰恰产出显示性偏好无法解析的那个模式：那在抵达时抛弃其对象之追求、那自我拆解之选择、那满足即熄灭之欲求。面对这，显示性偏好必须要么宣称那行为非理性，从而把它本要捕获的”结合之泉源”本身排除出说明之外，要么事后给诸选择拟合一个排序，从而把一个”其全部品性正在于它不是一个偏好”之运动记录为一个稳定的偏好。二者中它都未曾表征驱力。它在前者中把驱力作为噪声扔出，在后者中以一个”戴着它名字的不同之物”的稳定性虚构替换它。代理量不是对驱力的一次更差的测量。它是用一个正向对象替换一个空缺，而那恰恰是把驱力掏空于”使它成为驱力者”之运算。

失败之范畴，及其后果

失败之性质必须被精确固定，因为下游的一切都取决于它。失败不是认识论的。不是说驱力难以量化、说我们的工具粗糙、我们的问卷嘈杂，而一门更精细的心理测量学假以时日便会得出它的数。那种读法把优化框架完好留立，只承认测量差；而它正是本文最需要拒绝的读法，因为它是那把”把账算得更好”之门留开者。主张更强、更奇：此处没有一个数可被测好或测坏，从驱力到一个量的映射失败的不是它的准确性、而是它的存在，因为驱力不是那”有一个量可得”之类的实体。这与解剖在”一个仅仅是难的问题”与”一个不良提出的问题”之间所划的区分是同一个，而驱力落在它的第二侧——不是靠程度、而是靠种类。

主张 —— 不可结构化。 一个价值是不可结构化的，倘若它的存在方式排除了它被表征为一个目标函数中的项，也就是说，排除了任何”给候选诸状态赋一个可据以排序之单调量”的赋值。精神分析的驱力是不可结构化的：生起于实在界与象征界之耦合、环绕一个构成性的空缺而非一个正向对象、并在抵达发生之处被抵达所熄灭，它不容任何跨候选之单调量、从而不容任何项。它之被排除出目标，因此不是测量之缺陷、可由更精细的工具补救，而是范畴之事。它进不了问题之陈述。

后果按经济学家自己的条件对他的回答是致命的，而值得把这两难抽绎到底，好让没有出路留立。那回答许诺靠完成函数来拯救优化；亲密关系之完整性要求一个驱力之项；驱力不容任何项。所以函数要么略去驱力，那样它便不完整、而回答未曾被作出，目标是某个”略去了这段关系之一个泉源”的更单薄关系之目标；要么纳入一个驱力之代理量，那样它便把一个正向对象放在了一个空缺之处、并再次不是这段关系之目标、而是一个替身之目标，于是优化，无论其后被提得多么无可挑剔，都在最底层最大化着错误之物。没有一条第三路径能让驱力既被纳入又是它自身，因为”作为项被纳入”恰恰便是一个量之强加，而驱力不容任何量。解剖之第一前提条件——每一所争之价值都能在目标中被表征——因此失败，而它在固定性、可比性与优化者之诸问题能被提起之前便失败，因为那些关乎一个已然写好的函数、而这一个关乎书写。人可以把随后的三个不可能读作表明：即便是人能写出的那个不完整函数，跨越那些容许项之价值，也仍提供不出一个良提出的问题。但这第一个已然表明：那函数注定不完整，而整个优化抱负所凭依的完整性从来无从得到。

§5 第二个不可能 —— 价值之关系性与缺席的量化者

假定第一个不可能被搁置，并姑且承认——只为往下推进——一个函数能被写在那些容许项之价值之上。解剖之第二个条件于是进入视野：这个函数是外生且固定的，先于追求之过程被给出、且不被它改变，是一个确定的靶而非一个移动的靶。本节否认那条件。但它以一种特定方式否认它，而那方式与那否认同样要紧，因为一个弱版本的否认什么也不改变、而经济学家已然吸收了它。弱版本说人的偏好不稳定、说他们今日欲此、明日欲彼，从而目标漂移。对此经济学家正确地回答说：不稳定只是更多待建模的结构——让效用函数带一个时间指标、或让今日的选择含一个”它对明日口味之影响”的项，漂移便被吸收进一个更大的固定问题。本文之否认不是那个弱的、也不迎击这个回答，因为它根本不把麻烦定位在偏好之不稳定。它把它定位得更深，在价值本身之本体论，并论证亲密关系中的价值是关系性的，在关系之内被生成、且只存在于那里，于是不存在关系之外的立足点，能从其上读出并固定它的量。目标不能是外生的，因为没有一个外生的位置供它从其被给出；而这不是一个关于”游移之口味”、一个时间指标能接住的事实，而是一个关于”关系性价值在何处”的事实，无论怎样重新指标化都够不到。

价值是关系性的而非主体内的

效用函数是一个主体之函数。它编码他的偏好，即诸价值如它们对他所立者，而它的诸自变量是从他单一立足点被评估的世界诸状态；而即便是博弈论，它让几个这样的函数交互，也把每一个留作其拥有者的私产、先于交互被给出、只是通过交互被演出。整个构造预设价值是主体内的，安置于一个主体之内、由他作为他装备之一件带到情境。恰恰是这一预设，本系列之关系性本体论所否认，而那否认不是一种情绪、而是一个带后果的主张。一段亲密关系中所争之价值——一个私密之词所携之意义、一段被守之沉默已然持有之分量、结合本身之值——不是任一伴侣持于内而带来的属性；它在他们之间、在关系着之中被生成，且离开承载它的那段关系便无存在。这不是一个隐喻。一门私人语言对单独考虑的任一说者都毫无意义，因为它的意义是一段共享历史之沉积、被储存在关系里而非储存在任一头脑里；一个对二人珍贵的仪式，不是任一伴侣各自账本里的一个既与，因为它的珍贵不是关于他们任一者之事实、而是关于”在他们之间已然形成者”之事实。价值在此有着信任在那篇关于非约束性誓言之论文里被表明所具有的地位：一个关系性结构之属性、结合之拓扑之特征，而非其中任何节点之属性。

一个例证性的案例 —— 那个不在他们任一者之内的值。 两个人，历经岁月，把某一段散步弄成了他们自己的：一条特定的路线，于特定的傍晚走过，已把他们共同生活之整个肌理聚拢于自身，于是走它对他们各自都是一个意义稠密的行为。问那段散步值什么、对谁而言，而问题在手中散架。它对他们任一者分别而言值不了多少：单独取来，他们之一走那路线是”一个人走在一条路上”，而那漫过共享散步之意义仅仅缺席、无承载者。它的值也不是两份私人之值之和，因为没有两份私人之值可加；那意义不存在于他们任一者之内以供相加。它存在于他们之间、在那段散步既表达又维系的关系里，且不存在于别的任何地方。于是，没有一个账本”那段散步之值”是其中一笔——因为账本由一个主体记、并合计对他而言有值者，而那段散步之值不是那一类。要坚持录入它，人便不得不选一个立足点、他的或她的，并从那里记下一个”并非私人之价值”的私人之影，从而记下某个不同于人本要捕获之价值的东西。

量化所要求的量化者，及其缺席

从价值之关系性随出一个对外生性条件致命的后果，而值得慢慢抽绎它，因为它把这一不可能系于本系列最深的承诺。量化一个价值——赋予它一个 J 中的项所要求之量——便是占据一个确定的位置：那个”从外部与上方勘察价值之场域、看见诸候选及其诸值铺陈于面前、并把每一个读上一把公共尺度”者之位置。量化默然是一个”从价值之场域内部之无处”而来的视角、一个上帝视角之运算，而它预设一个外在于它所测之诸价值的立足点。对主体内的价值，这一预设至少在原则上无困难地被满足，因为主体作为其外部的评估者与他所评估的世界相对而立；他的偏好在他之内、世界在外，而他把前者读上后者。但对关系性的价值——在关系之内被生成、且只存在于那里——这一预设根本不能被满足。凡欲量化这样的价值者，必须立于关系之内或关系之外，而两者皆不成。立于内，他是他本要评估的那一生成之一方、无外部的着力、无一个”关系性价值作为一个对象铺陈于其面前”之高地；他在他本要勘察的那个东西之内。立于外，他找不到可勘察者，因为那价值只存在于内、并不呈现于任何外部位置。没有一个第三之地、没有一个外在而全知的高地，能从其上把”两个人在他们之间所生成者”之值读上一把尺度。而这恰恰是量化所要求的立足点：一个”持有一切关系性价值之真尺度”的大他者之立足点。整个本系列，追随母篇，已论证这个大他者不存在，不存在元语言、不存在一个”人赖以生活的诸意义能从外部被评估”之位置。因此，给关系性价值赋一个确定的量，便是预设本系列已表明为空的那一个立足点。

主张 —— 空缺的量化者。 亲密关系中的价值是关系性的：在关系之内被生成、且只存在于那里，是关系性结构之属性、而非其中任何主体之属性。量化一个价值，便是从一个外在于它的立足点给它赋一个量。对关系性价值不存在这样的外部立足点，因为一个关系之内的位置对它没有外部的着力、而一个关系之外的位置面前没有可评估之物；量化所要求的立足点是一个”持有一切关系性价值之尺度”的大他者之立足点，而这一位置是空的。一场”对亲密关系之优化”之目标因此不能是外生的，无外部高地供它从其被给出，也不能是固定的，它在”本要追求它的那同一运动”中被生成。目标之内生性不是不稳定偏好之偶然、而是价值之关系性之本体论后果。

经济学家之回答的落败

正是在此，经济学家之回答不仅被迎击、而且被击败，而这落败应被精确陈述，因为那回答有韧性、若被半答，便以新装归来。那回答是：你的目标被误设或不稳定；把它设得更好、给它指标化以时间，麻烦便化解。对第一个不可能这失败，因为某些价值根本不容任何项。对第二个它以一个进一步而独立的缘由失败：即便对那些或许容许项之价值，也不存在一个”诸项能从其被赋以其量”之立足点，价值是关系性的、而赋值之位置是空缺的。那回答归根到底凭依一幅图景，其中真正的诸值已然在外、安置于某处、候一个足够仔细的观察者去读，于是更好的设定是一件”更准确地读它们”之事。那幅图景恰恰是价值之关系性所拒绝者。关系性价值不是在关系之外潜伏候一个足够耐心的经济学家去发现；它在关系之内、在行为中被生成，且对任何外部之凝视都不静止。目标不是一个搜寻者设得太粗的靶。它由搜寻所生成、在关系着之中被修正、并不被任何”能充作其源头”的立足点持固定；而一个”对一个在’本要最大化它的那同一运动’中被产出、且无一个外部位置持之静止的目标”取的 argmax，并不命名一个候一个更好的经济学家之难题。它不命名任何问题，因为没有那运算所要求之形状的东西供那运算被施用。

§6 第三个不可能 —— 不可通约性

再一次姑且承认——为了往下推进——目标能被写出、且某个立足点能固定它。解剖之第三个条件于是要求：目标所赋之诸值躺在一把公共尺度上，于是任何两个候选皆可比、而 argmax 有可选之物。本节通过某些善之不可通约性来否认那条件：那个由 Raz 发展得最严谨的论题，即有些善之间不成立任何公共的度量，于是”把它们排在一把尺度上”之要求不是一件难的任务、而是一件弄错了的任务，而在亲密关系之内，”给一个善定价”之尝试可以严格说来是对它的一次背叛、而非一次误判。一如前两个不可能，主张必须先被从一个”框架乐于把它与之混淆”的更弱的邻居区分开，然后才被推向它的尖锐形式。

不可通约性是一种关系，而非一种困难

那更弱的邻居是这一想法：有些比较只是难，把一份事业权衡于一段友谊是困难、不确定、超出我们当前之精度的。这优化框架毫不费力地吸收：难的比较仍是比较，而困难只是对更好的估计之一声呼唤。严格意义上的不可通约性是另一回事，是一种正向的结构性关系、而非一份无知之坦白。两个善是不可通约的，当不存在一把”它们的诸值据以被精确比较”之公共度量，当”第一者好于第二者”既不真、”第二者好于第一者”也不真、”它们恰好相等”也不真。一把公共尺度之三个熟悉的关系——更好、更差、相等——只在一把尺度垫于其下之处才被共同穷尽；在无尺度之处，两个善可以立于三者之外，以一种那尺度不能表征之方式相关，而这一”立于之外”不是我们未曾测量之失败、而是诸善未曾共享一把度量之失。这一关系最精确的标志归功于 Raz，值得被精确给出，因为它把一个含糊的直觉转化为一个检验。假定两个选项在某把公共尺度上仅仅是相等的。那么哪怕稍稍改善其中一个、把它加甜一点，都会使它确定地好于另一个，因为在一把尺度上，给一个相等的量作一个小的添加便产出一个更大的量。现在取一份事业与一段友谊，人倾向于称之为同样值得选，并把那份事业以一次适度的加薪加甜。倘若二者在一把尺度上相等，那么被加薪的事业如今确定地是更好的选择。但它不是：连同加薪的事业仍既不好于那段友谊、也不差、也不恰好相等，一如加薪之前。那平局不在小改善下断裂，而一个”在改善下不断裂”的平局，从来根本不是一把尺度上的平局。小改善之论证表明的，不是那比较难，而是它之下没有一把可供”难”的尺度。

定价即背叛

在亲密关系之内，这一结构性关系取一个更尖锐的形式，其中通约之尝试不仅是关于诸善的一个错误、而是一桩对它们所行之不义，而此处本文倚靠那从 Raz 经 Anderson 到晚近对市场之道德批判的线索。这一尖锐化转于一个关于某些善之事实：它们部分地由人对它们所取的诸姿态所构成，于是取一个错误的姿态不是误描那善、而是改变它。在如此被构成之诸善中，有那些”其存在包含一个’拒绝给它们定价’”之善。对这样一个善，给它定价不是测量它、即便测得很差；它是把它改变成别的东西，并在改变之中背叛它。这正是为何”拒绝说出一个数目”不是矫情。问”什么样的报偿能补偿一段友谊之终结”，不是费力地抵达一个巨大的数目；它是弄错了一段友谊是什么，因为一个”人持自己随时准备为一个价而释放”之结合，凭那准备，已然不是人自以为在给它定价的那个结合、而是一个戴着它名字的更小之物。定价不是报告一个值而把数目弄错。它在定价之行为中产出一个被贬损之物，并报告一个关于那个被贬损之物之事实。

一个例证性的案例 —— 那个提议。 一个男人被一个当真发问的人问：多大一笔钱会诱使他放弃他最老的友谊——永久地、并无进一步联系。在他的反应里有两件事值得注意。第一件是他不把那问题体验为困难的，仿佛那数目很大而难以固定；他把它体验为一个范畴错误，一如人会体验”数字七有多重”这个问题。第二件更能说明问题。倘若他真按所问之精神对待那问题、当真开始说出一个数目并把它往上议价，某件事便已然发生在那段友谊上了，在任何数目被议定之前、且无论是否曾有一个数目被议定。当真地考量了那提议，便是把自己构成为一个”其友谊可按一个价出售”的人，而一个被这样一个人所持有的友谊，不是那提议本要定价的那个善。背叛不在于接受，那从不到来；它在于定价，定价在它把所定价之物当作可定价的那一刻便改变了它。这正是为何合宜的回答不是一个很大的数目、而是对那框架的一个拒绝，也是为何那拒绝不是没能计算、而是唯一让那善保持完好之回应。

主张 —— 不可通约性。 两个善是不可通约的，当它们不处于一把公共尺度所提供之三个关系中的任何一个，既不更好、不更差、也不恰好相等，一如它们的平局在小改善下不断裂所揭示者。亲密关系中所争的某些善在此意义上是不可通约的；而有些进一步部分地由”拒绝给它们定价”所构成，于是把它们置于一把公共尺度上不是误测它们、而是变形并从而背叛它们。解剖之标量条件因此失败：目标所赋之诸值并非皆躺在一把公共尺度上，而 argmax——它需要这样一把尺度才有可选之物——没有可供它运作之物。错误是范畴之错、而非计算之错；诸善不是那”有一个总和”之类的东西。

与前两个不可能的关系

至此三个不可能是彼此相加、而非彼此重复，而值得标记每一个所切之关节，因为一个反对者可能怀疑它们是同一个抱怨的三套戏装。第一个关乎表征：一个价值究竟能否作为项进入目标，而它发现驱力不能，因它被一个单调量本要追踪之抵达所熄灭。第二个关乎固定：目标，姑且承认它被写出，能否从一个外部立足点被给出，而它发现关系性价值不容任何这样的立足点，因它在承载它的关系之内被生成。这第三个关乎比较：目标所赋之诸值，姑且承认它们被写出并被固定，是否躺在最大化所要求的公共尺度上，而它发现不可通约之善不然。三者是独立的，而它们的独立便是此案之力量，因为一个不知怎么答复了其一之反对者，会面对其余原封不动者：即便驱力终究可被表征，它的项仍不得不与其余通约，而不可通约性会拦住它；即便某个外部立足点不知怎么被找到以固定目标，它所固定之诸善仍会未能共享一把尺度。下降是真实的，从”问题能否被写出”、经”它能否被固定”、到”它所固定者能否被比较”；而在这三者之下躺着一个下降从一开始便一直环绕、尚未触及的第四个问题，即”那个’问题对之才是一个问题’之人”之问题，那个”其稳定偏好被整个构造静静地预设”、而下一节将其消解的优化者。

§7 第四个不可能 —— 主体之构成

前三个不可能关乎目标：它能否被写出、从外部被固定、被供以一把公共尺度。这最后一个够到目标之下，够到解剖之最沉默的条件、那个垫于其余诸条件之下、最少被带到明处的条件，即有一个优化者。最大化不仅要求某个被最大化之物、还要求某个去最大化它之人：一个”J 是其目标、J 编码其偏好”的主体，立于问题之外、先于问题、并贯穿问题而稳定，作为”x* 对之为最优”的那一个。没有这样一个主体，”他的最优”这一短语便无确定的指称，因为没有一个已定的”他的”是那最优所归属者。本节否认那条件。在亲密关系中，主体不是先于关系、而是在关系之内被构成；那个”本要优化”者，是他被想象置身其外的那个过程之产物；而整个构造所预设的那个”谁”，失去了那构造所需之固定性。

主体是被构成的，而非被给定的

优化框架从边际主义承继了一个”已然成形地抵达情境”的主体，携带一个预先定下的偏好序，情境随后揭示并满足它。即便是那允许偏好改变的更精妙的经济学，也倾向于把那改变本身建模为一个先在而稳定的元偏好之输出、一个主宰口味如何演化之固定倾向，于是那固定性只是退后一层、而主体归根到底仍是先于过程被给定的。本系列之关系性本体论从根上否认这幅图景，并像它就价值所否认者那样就主体否认它，因为这两个否认是一个。亲密关系之主体不是带着已固定的偏好进入关系、并把关系承受为它们之演出。它在关系着之中被形成：它的诸欲望被他者之欲望所塑造，它对”什么值得欲求”之感在共享的生活中被重造，而它的个体化——“它是什么”与”他者是什么”之间的界线之划定本身——是在结合之内被完成并被修正、而非先于结合被定下。这不是那”人会改主意”之温和而可吸收的一点，那一点一个元偏好能吸收。它是那本体论之点：优化所预设的那个持偏好的主体、那个与选项之场域相对而立的固定评估者，不是亲密关系之先行条件、而是它诸多进行中的产物之一。关系所生成者不仅是伴侣之间的价值、还有伴侣本身，作为他们在成为之途中所是的那些赋值者。

一个例证性的案例 —— 谁来领取那个最优。 一个人带着一个”他想从一生中要什么”之已定之感进入一段关系，而在其内历经岁月，那感被重造：他开始想要一些他从不知道原来是可想要之物，开始觉得一些他曾排在最高之物空洞，开始以一种”不可化约为他自己在先偏好、因为它已重组了它们”之方式在意他者之繁盛。现在假定他曾在开头试图靠优化来选择那段关系、计算哪一种安排最能满足他当时所持之偏好。那个最大化了开头那个人之偏好的安排，不是他所成为的那个人会选择的安排，而这不是因为二者之一犯了错，而是因为第二个人之偏好，是那”第一个人之优化本要从其未始便主宰”的关系着本身之产物。那么，谁的最优要被计算？不是第一个人的，因为他已不再存在以领取它，而他的偏好被那过程消解并重构了。也不是第二个人的，因为他在选择之时刻尚不存在、且不可能供给那目标，他本身是所选择者之一个输出。没有一个”自始至终在场”的单一主体来充当目标之拥有者与其最大值之受益者。有一个”在被那’其价值优化本要计算’的关系所构成之途中”的主体，而那是一个不同于框架所要求的固定优化者之物。

固定指称之消解

从此随出一个对第四个条件致命的困难，而它是一个关于指称、而非关于稳定性的困难，那个案例已把它备好可陈。倘若优化者在”他的优化本应主宰”的那同一过程之内被构成，那么”谁的最优？”便没有一个固定的答案，因为任何答案里所命名之主体，在过程之终与在开头提出它者不同，而那差异由过程本身所产出。这比第二个不可能之下所处理的目标之漂移更糟，且切不可被折叠进它。那里靶在移动而弓箭手站立不动；这里弓箭手被射击所重造，于是那个”本要在山巅领取最优”者，不是那个出发去攀登、而在出发时并不持有”山巅据以是山巅之偏好”者。框架需要一个稳定的主体来充当目标之拥有者与其最大化之受益者，那个”J 是其目标、x* 对之为最优”者；而在亲密关系中没有这样一个稳定的拥有者，只有一个”关系恰恰正在把其解固定”之成形中的主体。问”什么对他为最优”，便是预设一个”关系着所消解并重构”的已定的他，而那优化之”谁”的指称在发问之中散架。

主张 —— 被构成的主体。 优化预设一个优化者：一个先于问题、并贯穿问题而稳定的主体，函数是其目标、其最大值对之为最优。在亲密关系中，主体是在关系之内被构成、而非先于关系被给定，它的偏好与它的个体化是关系着诸多进行中的产物。优化者因此是它被想象置身其外的那个过程之产物，而”谁的最优？”失去一个固定指称，任何答案里所命名之主体被那”优化本要主宰”的过程本身所重造。解剖之第四个条件失败：没有一个稳定的主体来拥有目标、或作其最大值之受益者。

四者合观

四个不可能如今可被聚拢，而聚拢它们表明它们不是一个反对的四个名字、而是在四个不同关节上的四个独立失败，从问题之表层贯穿到它的根基。诸价值能被写入目标吗？并非全部：驱力，被抵达所熄灭，不容任何项。目标，就其被写出而言，能从外部被固定吗？不能：关系性价值不容任何外部立足点，而目标出于本体论之必然是内生的。它所固定之诸值能在一把公共尺度上被比较吗？不可通约者不能：argmax 所需之尺度不在那里。而是否有一个稳定的主体来拥有整体并领取其最大值？没有：优化者在它本要主宰的过程之内被构成。每一切都在一个不同的关节上，而每一个单独便已足够，因为一个优化问题需要全部四个条件，而其中任何一个之失败便使那运算无定义。它们一同确立——不诉诸一丝一毫的价值——亲密关系提供不出一个良定义的优化问题：不是一个”更大的计算能力便能攻破”的难题、而是一个不良提出的非问题，其所需之对象——一个属于稳定主体的表征上完整、外生、固定、标量的目标——最大化这一运算找不到。随此本文之形式运动完成。随后者不再是证明、而是批判，而它转向那份证明已为之廓清道路的另一个问题：不是人能否优化亲密关系——那已被否定地回答——而是当人不为那答案所阻、无视它而强加那框架时，那关系遭逢什么。

第三部分 · 优化主义的代价

§8 对生成性的损害

本文的第二个运动始于此，而它的问题在种类上不同于第一个。第一个运动追问亲密关系是否提供一个良定义的优化问题、并回答说它并不；它未对优化下任何判断，只确立此处那运算无定义。这个运动把那一结果当作已定，并追问一个那结果使之新近迫切的问题：当那框架被无视此结果地强加，当一个提供不出良提出之最大化的领域被当作它提供了一般来对待、它那关系性、不可通约、内生的诸善被逼进一个”待最大化之目标”的形状时，会发生什么？从此处起诸主张负载价值、且公然如此；而导论所坚持的分离，正是那给予它们如此之权利者，因为它们不再仰仗那形式结果、而是建立于其上，把”并无优化问题在场”当作既定、并追问”佯装有一个”之代价。这样的第一个代价，本节所论者，是一种有确定形式的损害：那强加把一个”本应在关系之内流通并积累”的价值，转化为一个可被结清、被榨取、并最终被耗尽的量。

被生成之价值的两种命运

生成性正义理论——在本系列中与 Eglash 之交流中展开——区分一段关系中所生成之价值可能遭逢的两种命运，而那区分是本节之枢纽。价值可以流通：回归那些生成它者，补充它自身进一步生成之诸条件，于是关系是一个”自我维系、价值在其诸造者之间保持运动”的持续生产之场址。或者价值可以被榨取：被从其诸生成者抽走、被异化、被结清于其造成之循环之外某处，于是关系被开采而非被耕耘，而它生成之能力被抽降向枯竭。Eglash 自己的诸案例取自劳动、生态与技术之艺，在那里所论之价值相对可辨、其被转移相对易见；本系列之贡献一直是把那区分带入亲密生活，在那里所争之价值最不可辨、其被转移最难命名，而在那里，本文如今论证，优化框架是那第二种命运之主要工具之一。主张不是”开采矿石”与”给爱定价”之间的一个松散类比。它是：优化对一个关系性之善所施行之运算，在其形式结构上，是一个榨取之运算，而下一步是表明何以。

优化对一个流通所起之效果

要看见那榨取，人必须看见优化在把一个价值逼进一个目标时对它做了什么。把一个关系性之善弄成 J 中的一个项，便是把它渲染为一个量：一个能与别者相加、被比较、被取舍、并且首要地被结清——被带到一个”人从其读出最大值是否达到”之最终结算——的量。但一个关系性之善，在它自己的存在方式里，不是一个可结清的量。它是一个流通、一个”其生命在于它的持续、并无一个’它完整而可被合计’之自然终点”的生成。两个人之间的意义不是一个”积累向一个满额”之存货；它是一个”要么被维持、要么被听任消逝”之运动。被渲染为一个待最大化之量，这样一个善便被弄得可供结清，而结清一个流通便是把它停下。优化框架不仅仅误测那关系性之善、给它赋一个偏离的数。它把那善重塑成一个”能被兑现”之类的东西，而那兑现便是那耗竭，因为被兑现者不再在流通之中。因此损害不在于优化抵达一个错误的数。而在于那”渲染为量”之运算——优化在它能开始之前必须施行者——已然把一个生成转化为一个存货、并从而把它摆到被抽降之位置。

一个单一的几何式例证

本系列已在别处给这一耗竭一个几何式的形式，而它在此挣得一句话，严格作为例证、而非作为一套待展开的机件。在关于善的循环与坏的循环之说明里，生成性价值被表明不被承载于任何终点状态、而被承载于一个回路之累积相位、一个解释者在穿越它时聚集的和乐诺米，而这是任何终末的结算都捕获不了者，因为它是一条道路之属性、而非那道路上任何一点之属性。唯一相关的一点是：优化，其整套文法便是对一个最优点、一个可行集中的 x* 之挑选，凭它的形式本身对”活在一条道路而非一个点里”之价值视而不见，从而根本登记不了、更遑论保护它所结清掉的那份生成性积累。此处不需要那机件、也不会展开它；它只用来命名——以本系列已弄精确之词汇——为何一个”被承载于流通中”之价值，对一个”只识得山巅”之运算是不可见的。

一个例证性的案例 —— 被审计的关系。 一对，焦虑而尽责，开始为他们的关系记一本账：追踪各自的贡献与所得、注记回报是否证成投入、并举行周期性的结算，其问题是——总体而言，这东西是否在回本。他们的意图是关切；他们意在靠度量关系来照料它。但那度量对关系所做的，恰恰是本节所描述者。曾是一个流通、一种”其值在于它持续之运动、从不静止以供合计”的给予与接受，变成了一个”在每次结算被评估”之存货；而一个被评估的存货，是一个可能算出短缺的存货。损害不在于那审计可能返回一个不利的余额并终结那关系，尽管它可能。损害先于任何余额、并独立于它的符号：它在于那审计已改变了被审计之善，把一个”靠不被合计而保持鲜活”的生成转化为一个”为被合计而存在”之量，于是即便一个有利的余额，如今也是一个在被贬损之物上结出的余额。这对本欲保护那流通，却靠把它渲染为可问责，在计数它的行为中把它停下。

损害之形式

损害如今能以它恰当的形状被陈述，并被陈述得使它平明地是一种伤害、而非一个仅仅措辞上的不妥。当一个关系性之善被交付给优化框架，它经历三个运动。它首先被渲染为一个项，这要求赋予它一个它并不拥有之可结清的量。它随后被录入一个待最大化之目标，这使它从属于一个”它在其中作为诸加数之一而列”之总和之结算。而它随后，在那最大化里，被花掉：被当作一个”待抽降向一个最大值”之存货、而非一个”待保持运动”之流通。这是榨取之形式签名，也是为何那”当真开始问’这段关系是否值得’”的一对，无论他们抵达什么答案，都已然开始了生成性正义理论所警告之转化。伤害不是一个不利判决之风险。而是那结算本身已改变了被结算之善，其方式与缘由，同上一部分的定价贬损它所定价之友谊者一样：靠把一个”其生命在于从不被最终结清”之物当作一个可结清的量。因此，在不存在优化问题之处被强加的优化，不是一个”让其对象保持不动、却没能改进它”之闲散错误。它是一个”把它的对象重塑成一个它能作用之物——一个可结清的量”之运算，并在重塑之中施行那”把一个’其生命在于从不被最终结算’之价值抽入一个最终结算”之榨取本身。

§9 优化主义作为现代性的一种症状

批判如今从”优化在一段单独关系之内所做之损害”拓宽到”优化主义作为一个关于世界之一般论题”之地位。本节之主张是：把优化抬升为宇宙之一条法则，最好被理解为不是一个发现、而是一个症状：一个特定历史处境所取的一种认识论形式，本系列之母篇将那处境命名为象征界之危机，而它与一个生活世界对资本逻辑之习惯相缠。在这一读法上，宇宙攀登不了任何地景；对地景之攀登是一个历史上特定的实践，本乡于一种特定的生产之组织，被向外投射到宇宙之上、并被当作它的法则。论证这一点，不是靠优化主义之起源来反驳它——那已被弃绝之谬误——而是重新定位它：把它从”它作为’事情本来如此’而自我防卫”之永恒真理之register，移入”人能问它表达什么处境、那处境有何代价”之历史症状之register。第一部分之系谱表明优化主义是被建构的；本节追问：在当下，是什么使那建构保持屹立、并使它感觉像自然。

通约之运算作为资本之本乡形式

优化主义之决定性特征——系谱里已孤立出者——是它对通约之要求：它要求异质之善在一个最大值能被定义之前被带上一把单一的尺度。这一要求不是优化理论所独有的，而看见它还住在别处，是那诊断之关键。它是资本之本乡运算。资本之流通除了”把质上不可比之物——一小时劳动、一斗谷、一首歌、一个下午的照护——化约为一个’它们能在其中被交换、被积累、并被使繁殖出更多自身’的公共量”之外别无所是；价值形式恰恰是一台通约之机器，而资本之历史生涯便是它的通约向生活之一个又一个区域的蔓延，直到那些曾被认为超乎价格之物获得一个价格、并进入那循环。优化主义是这台机器所固有之认识论。它是普遍通约一旦被从一个实践抬升为一门理性之理论、进而抬升为一幅世界图景之后的样子：那”在诸善之表面异质之后总躺着一把它们据以比较之尺度、以及一个理性应向之攀登之最大值”之已定信念。这信念对现代之心感觉自明，不是因为它是一条自明真理、而是因为那心是在一个”对一切、不息地施行通约之运算，直到那运算之可施行性开始看起来像诸善之属性、而非被搁在它们身上之强加”的生活世界之内被形成的。这是法兰克福传统所施压之对工具理性的批判：理性，被化约为对手段之计算与对目的之通约，把它自己历史上被产出之形式误认作理性本身，并遗忘了它不再能就之说理之诸目的。

象征之危机与向尺度的逃遁

母篇把当下诊断为象征界之一场危机，即一个世界赖以被维系之共享意义之一次掏空，而优化主义能被安置于那诊断之内，作为它诸多具特征之回应之一。这一安置解释了系谱单独所不能者：为何那被建构的图景不仅可得、而且诱人，为何人们伸手去取它。当那”曾一度告诉一个共同体诸物值什么、并以质性而不可通约之词告诉它”的稠密象征秩序失去其把持，一个真空便在那”值之感”曾在之处开启，而单一的量性尺度作为一个替代物流入那真空。一把尺度不索求任何共享的意义。它不需要关于”一物为何、什么使一生过得好”之任何共识，只需要一个数；于是它向一个”不再能自我供给意义”的祛魅世界许诺一个定向、一种”知道该做什么”之道——最大化那尺度——它熬过了每一个稠密的关于诸善之说明之崩塌。优化主义在此意义上是一次逃遁：一次从”不可通约之价值之困难”（它要求判断、并不产出算法）向”一把尺度之虚假安慰”（它只要求计算、并省却了判断）之逃遁。那逃遁竟把自己呈现为硬朗的理性、而把对它的拒绝呈现为软弱的感伤，这本身便是那症状之一部分，因为那处境褒扬那”表达它自己”的运算本身、并把对价值之真实结构之忠诚——对”通约那不可比者”之拒绝——重铸为一次”待被克服之胆怯之失”。

向宇宙的投射

那症状之最后一个运动，是把那历史上特定的运算投射到宇宙之上、当作它的法则，而正是这个运动，本文之标题所拒绝。因为那通约、最大化之运算在某一种生活形式之内处处被施行，便开始看起来仿佛那运算不是在被施行、而只是在被认出，仿佛那些地景一向在此、而理性只是攀登它们。人在它流行的宇宙形式里遇见这一投射，其中从一个”寻其最低能量”之分子、到一个市场之出清、到一个”最大化其适应度”之有机体，一切都被说成是”实际上”在优化某个量，于是那计算着的人终究只是在有意识地做整个自然盲目地做之事，而优化被揭示为存在之深层文法。本文之否认恰恰瞄准此处，且能被精确道出。所实际被观察到者，不是一个攀登地景之宇宙、而是一个攀登它们、随后携着它的尺度、在它放下尺度之每一处都发现那运算的生活形式。某些物理系统容许被描述为”极值化一个量”，是一个关于一套数学形式体系之够及的事实，而非”自然在所论意义上——即’有一个被一个主体所拥有之目标、并有一个待寻之最优状态’的意义——从事最大化”之证据；那形式体系之成功被回读入世界、当作世界自己的程序。说优化是宇宙之一条法则，便是把那通约之生活世界投在它所审视之任何东西上的影子，误认作被审视者之一个特征。那实践是真实的、并有一个日期。那法则是一次投射。而亲密关系——其诸善是关系性而不可通约的、其诸主体在关系着之中被构成——正是那投射最平明地为假之区域，因为它是那”通约之运算一经施用便最显眼地摧毁它所被施用之物本身”之区域。

§10 优化性的实践不是伦理性的实践

本节是批判之枢纽，并在一种意义上是本文之枢纽，因为它以最尖锐的形式陈述那标题之问题从一开始便瞄准之主张：道德不能被定义为一个优化问题，而一种优化性的实践，无论它在自己的尺度上多么成功，都不因此便是一种伦理性的实践。前几节论证优化在亲密关系中不良提出、且它的强加造成损害。这一节论证一个对二者皆正交、且独立于二者之物：即便在一场优化完美良提出并完美成功、抵达一个真实目标之一个真实最大值之处，它仍可能作为一种伦理而失败，因为最优之判据与善之判据是不同的判据，回答不同的问题，而最优者不是善者。因此这主张甚至在那”四个条件被满足”的运输车队领域也成立；它根本不是一个关于不良提出之主张、而是一个关于”最大化”与”行动得好”之间那道无良提出所能弥合之间隙的主张。

最优者与善者之正交

本系列此前遇见过这一类的一个结构，而点名那较早的实例磨利了当下的这个、并表明它不是为此场合临时编造。在关于生成性关系之可持续性的论文里，诗被表明满足一个形式的、动力学的判据，即善的循环之判据、无尽不结算之流通之判据；而在同一口气里被表明：满足这一判据并不使一个象征之流通为善，因为法西斯主义之美学与邪教之话语满足恰恰同一个动力学判据、而它们是邪恶的。在那里所引出之教训是：动力学判据与正义判据是正交的——第一个问一个流通能否自我维系，第二个问它是否应当、并以谁为代价，而一个流通可以通过第一个而未通过第二个，于是善要求正义判据作为一个独立的条件、那是动力学永不能供给者。当下的主张有同一形状，只是优化立于动力学之处。最优之判据回答一个问题：在一个给定目标上，一个给定选项是否是可得之最优、它是否坐落于最大值。善之判据回答另一个：那选项是否应当被选、它是否冤枉任何人、它是否尊崇那行动者所受缚去尊崇者。这两者是正交的。一个选项可以既最优又错误；一个选项可以既正确又次优；而在第一个判据上的成功在第二个上根本不是成功，因为它们不是沿一把公共尺度度量一个东西、而是提出一个单一答案不能共同结算之不同问题。找到了那最大值，便是回答了第一个问题、而丝毫未触及第二个。

一个例证性的案例 —— 那个最优的背叛。 假定一个男人面对一个良提出到足以优化之选择：相关诸善容许代理量、目标稳定、尺度可得、而他本人在其偏好中已定。他正确地算出：在朋友最依赖他之时抛弃一个已成负担的朋友，最大化他的目标，甚至——在一个说得通的设定上——最大化一个无偏的福祉聚合，因为他自己的所得巨大、而那朋友假以时日会复原。那优化无可挑剔；那最大值是真实的；在它自己的尺度上没有更好者可得。而那行为是一次背叛。要点不是他优化了错误的目标，那会把我们带回早先的诸不可能；姑且承认那目标好到一个目标所能好之极。要点是最优性与正当性已然分开，那”在尺度上最优”的行为本身便是那”冤枉了他所受缚于之人”的行为，而没有任何对那最大化之精修修复这，因为那不义不是目标没能纳入之一个量、而是那”整个最大化之维度都看不见”之行为之一个特征。他做了那最优之事，而他本不该做它，而二者同时为真。

道德对优化形式之抵抗

两个判据正交不是一个粗蛮的事实、而是随出于”道德是哪一类东西”，而道德之三个特征表明它为何尤其抵抗优化形式，每一个都点名上述案例咬入之处。第一个是某些道德要求之无条件性。一个康德置于伦理之中心的那一类义务，不把自己呈现为”待在一个总和里被权衡、并在总和另有结果时被压过”之诸善之一；它把自己呈现为无论那平衡如何都有约束力，一个”对什么可被做”之约束、而非一个”待被抵于别的量而交易”之量。一个优化框架只能靠篡改来表征这样一个要求：要么作为一个”一个更大的项在原则上能压过”之很大的项，那否认它的无条件性；要么作为一个”给可行集划界”之绝对约束，那把它整个从目标里移走、并承认道德从一开始便根本不是一件最大化之事。第二个特征是道德诸考量之不可交易性。把一项义务弄成一个目标中的项，便是把它弄成可交易的、可供在”数字利于那交易”时被牺牲；但我们所认作道德严肃性者，其大部分恰恰在于”拒绝把某些考量弄成可交易的”、在于持守”有些事无论收益如何人都不做”，而这一拒绝在一个”凭目标之形式、一切都是一个交易之量”的框架之内是不可理解的。案例里的那个男人做了那可交易之事；那”朋友之依赖能被换取一个足够之所得”恰恰是一个道德严肃性会否认者。第三个特征是他者之不可替代性。一个人所受缚于之人，在道德上，不是”一个量之价值之承载者、而另一个等量之承载者能无损替换”者；他者是不可替代的，而他者所提出之诉求是这个独一者之诉求、而非它恰好承载之那个价值之诉求。优化——其可行集是”被一把公共度量排序之可互换选项之空间”——根本不能表征不可替代性，因为它的每一选项在原则上都可被另一个等值或更大值者替换；而不可替代性恰恰是”那样一个替换在此永不无损”之否认。

主张 —— 最优者与善者之正交。 最优之判据与善之判据是正交的：第一个问一个选项是否最大化一个给定目标，第二个问它是否应当被选，而在第一个上的成功在第二个上不是成功。一种优化性的实践，即便一个良提出并抵达一个真实最大值者，因此也可能作为一种伦理性的实践而失败，而它在”它所最大化之善不是它所应尊崇之善”之处便如此失败。道德尤其抵抗优化形式，因为它诸要求之无条件性、它诸考量之不可交易性、以及他者之不可替代性，其中无一是一个目标函数不篡改便能表征者。道德因此不是一个更高阶的优化、某个道德之量之最大化，而是一个立于优化之文法之外的实践。最优者不是善者。

批判之完成

随此批判自我完成，而它这样做，是靠关闭优化框架可能采取之最后一条出路。面对至此之论证，一个优化之辩护者还剩一招精妙的：承认寻常效用是错误的目标、并提议让道德本身被优化，让我们最大化的不是愉悦或偏好、而是善，把道德诸考量折叠进一个更丰富的目标并最大化那个。本节之主张正是那封死此招者。道德不能是一场优化之目标，因为它诸中心特征——无条件者、不可交易者、不可替代者——恰恰是一个目标函数所不能承载之特征；把道德写进一个目标，便是把它剥去这些特征、并从而以它的名字写进别的东西。优化框架不能靠把它瞄准善而被拯救，因为善不是那”有一个最大值”之类的东西。而这并不使伦理无理性，那会是相反那一类的一场灾难、而本文煞费苦心要避免。它使伦理拥有一种不是优化的理性，一种”关于该做什么”之推理方式，它可答责、有辨别、并严苛，而不是任何尺度之最大化；而那另一种理性之复原，从一开始便被许诺者，是随后一部分之工作。

第四部分 · 否定之后：一种不优化的关系性理性

§11 理性不是优化 —— 夺回这个词

本文已否认亲密关系是一个优化问题、并批判了那框架之强加；它如今欠下一份”关系性理性肯定地为何”之说明，而这债不是可选的。把事情停在那否认上，会招来本文最想拒绝的那种读法：即在交出优化时人交出了理性本身、并把爱发付给纯然的感受、发付给一种”随心所欲”的温热的无理性。那读法之所以有力，只因一段长久的用法已静静地把理性等同于优化，直到这两个词作为同义词运作、而合理之空间已窄化为最大之空间。复原之第一项任务是打破那一等同，而它是一件夺回、而非发明之任务，因为那等同是一个”曾意谓更多之词”之晚近收缩，而本文所夺回者，在优化吞并那术语之前便在那里。

一个词的收缩

那等同是一个收缩而非一个定义，这能直接从已给出之系谱读出。理性不一向是最大化。韦伯自己的区分在手段之理性之旁保留了一种目的之理性，一种”关于什么值得追求、而不只关于如何高效地追求被给定者”之合理；而在韦伯之后躺着那远为古老的实践智慧之传统，其中理性行动者不是那”最大化一把尺度”者、而是那”善知一个情境所要求者、并合宜地行动”者，其卓越是辨识与分寸、而非计算。理性向优化之窄化，是系谱所描摹之那一特定历史之成就，其中目的之搁置、目标之对象化、以及最大化者之公理化安装，合起来使”理性”意谓”在一把尺度上最大化”、并使任何”拒绝那尺度”者凭完备性公理依定义为非理性。一旦这一窄化被看作一个有日期之成就，它那必然之气便化解。不是理性是优化、而关系因此落在理性之外；而是某一段历史已把”理性”这个词吞并给优化，而一次吞并可被争夺。本文争夺它，而这争夺之地基是：那些更古老、更宽广之义，不是那窄化所廓清之诸混淆、而是那窄化所丢弃之诸能力。

两种理性，被区分

夺回最好靠区分”那一个词被弄得承载”之两个东西——那最大化之理性与那回应之理性——来做，并靠表明第二个不因不是优化而更少理性。那最大化之理性取一个固定目标与一个选项之场域，并挑出那”使目标最大”之选项；它的德性是效率、它的度量是到山巅之距离，而它恰当的家是那”价值可表征、固定、可通约、并被一个稳定主体所拥有”之领域——也就是说，那四个条件成立、运输车队被规划路线之领域。那回应之理性取一个”其诸善是关系性、不可通约、并部分不可表征”之情境、以及一个”仍在成形而非固定”之自我，而它做那第一种理性所不能者：它照看情境与他者所要求者，在无一把尺度之下于诸不可通约之善之间辨识，并合宜地而非最大地行动，在每一点都对一个”来自外部而非从内部算出之总和”之诉求可答责。这第二种理性不是第一种之一个有缺陷的形式、一场”以差的工具、无清晰目标而进行”的优化。它是一种不同类的能力，有它自己的成败之标准、它自己出对与出错之诸方式、以及它自己的严苛，将在随后各节被描述。因为它不最大化便称它非理性，便是把那被吞并之义误认作那词之全部，并要求”关系所固有之推理”在一个”本文已否认其管辖权”之法庭前为自己辩护。

无一最大值的严苛

复原中的一切都转于一点，它必须被坚持，因为它是优化框架将最抵抗之点：即那回应之理性是严苛的、它有真实的标准、而不是一张”随心所欲”之许可证。对任何不优化之理性最深的怀疑是：它只是一个软弱的许可、一种”靠免除尺度之纪律来给任性授勋”之方式；而那怀疑是公道的、且必须被迎击，因为一个”只归结为’随你所感、不受计算之累而行’”之关系性理性，会配得上优化框架堆在它身上的轻蔑、而且会背叛它自称服务之诸善，那些善提出艰难的要求。答案是：那回应之理性有它自己的一套纪律，与那最大值之纪律一样苛求、并在一个方面更难，因为它不能被一条规则所了结：那”把他者弄对”之纪律，那”把情境感知为它所是、而非人所愿”之纪律，那”辨识数个不可通约之善中此一场合请求被尊崇者”之纪律，那”不去抓握一个生成而维系它”之纪律。这些是人能失败、并能失败得很糟之标准。一个回应可以是虚假的、不用心的、自利的、失度的；一个合宜之回应与一个不合宜之回应之别，与一个正确答案和一个错误答案之别一样真实，尽管它不是一个更大的数与一个更小的数之别。实则上一节那个最优的背叛之案例，正是恰恰这种理性之一次失败，而它之为一次失败、之作为如此可被评估，便是”一个标准在起作用”之证明：那男人把他的朋友感知得足够对、以致知道所亏欠者为何，而他逆着那感知行动，而他行为之错误不是任何尺度上的一个更小的数、而是一次”那尺度甚至记录不了”之回应之失败。那严苛是真实的。它只是不是优化之严苛。理性，被夺回，宽广到足以既容那最大化者、又容那回应者；而关系所固有之推理是第二个，不因不拥有任何最大值而更少理性，并对那”第一个看不见”之标准可答责。

§12 关系性理性之形式

那回应之理性已被从那最大化之理性区分开、并被赋予一种它自己的严苛；尚余的是肯定地描述它的形式、说出它的诸运算为何，以便它作为一种有形状的能力、而非仅仅作为优化之否定而立。三个时刻构成它，而它们不是按序运行之一个程序之诸阶段、而是一个单一之行进方式之诸方面，在它的任何一次践行中一同在场：回应，即对一个从他者发出之诉求之答复；调谐，即对一个情境及其诸善之感知上的贴合；以及维系，即对一个”不应被抓握”之生成之保持运动。每一个都是人能满足或失败之标准，而每一个都点名一种”优化框架不仅缺乏、且不能表征”之能力，于是描述它们同时既给复原以它的内容、又表明它所替换之框架为何恰恰对这些东西视而不见。

回应

第一个时刻是回应，而它点名理性从”一个在内部算出之总和”向”一个从外部被答复之诉求”之重新定向。那最大化的行动者面对一个选项之场域、并咨询一个是他自己的目标；他所问之问题是”什么最好地服务那目标”，而他推理之方向是向内的，从世界到他的尺度。那回应的行动者面对一个”其对他之诉求不是从他的目标推出、也不可能是”之他者，而他所问之问题是”这个他者、这个情境，对他要求什么”；他推理之方向是向外的，从他自己到一个他未曾设定、也不能撤销之诉求。这正是为何不可替代之他者之伦理是关系性理性之本乡、而非其上之一件装饰：他者之诉求恰恰是一个”从自我之目标以外而来”之诉求，而一个最大化之理性根本登记不了它，因为它出现在自我所能持有之任何目标的任何项里。回应以一种最大化所不是的方式可答责，而那差异是它之理性之标志、而非其中之一种软弱。那最大化者对他自己的目标、并对别无一物负责，因此严格说来弄不错那目标；它就是他的。那回应者对一个不同于他自己者负责，因此能以那最大化者没有余地容纳之诸方式失败：靠根本没能听见那诉求，靠听见它而虚假地答复，靠以他自己的尺度替换他者之诉求并把那替换称作一个回应。这些是真实的失败、可与成功区分开，而这便是那表明回应是一种理性之能力、而非一种情绪者，因为唯有”能被做对或做错”之物才根本是一件理性之事。

调谐

第二个时刻是调谐，而它点名那”没有它回应便会盲目”之感知上的成就。要合宜地回应，人必须先正确地感知：把情境看作它所是、登记哪些善真正在局中、辨识这个场合——在它所类似的一切场合之中——独特地请求什么。这是那实践智慧之传统置于伦理之中心的能力，那”把握一个特定情形之诸显著特征”之受训练的感知，它不是”把一条规则施用于一个已被描述之情形”、而是那”抵达那条规则会需要之描述”之在先而更难的工作。它恰恰是那”优化框架没有位置容纳”之能力，因为优化只能在情境已被化约为一个可行集与一个目标之后才开始，而调谐是那”抵达一个先于那化约、且常抵抗它的情境之感”之劳作。调谐是那”在无一把尺度之下于诸不可通约之善之间辨识”所凭者，而这是它最重要、也最不显然的职分。那最大化者，被拒予一把公共度量，便仅仅陷住、不能排序、从而不能行动；而那被调谐的行动者根本不陷住：他感知这情境请求被尊崇之善是哪一个，不是靠算出一个”诸善在其中被比较”之总和（那不可能）、而是靠一次”看见它所召唤者”之对情形的辨识。这样的辨识是可能的、人事实上确实在诸不可通约之善之间选得好、并能把这样一个明智的选择从一个愚蠢的区分开，这本身便是”一种不同于优化之理性在起作用”之证据，因为那选择显然不是一次最大化、却也显然不是任性的，可供褒扬、批评、以及”它被做对或做错”之判断。

维系，及一个意象

第三个时刻是维系，而它点名那回应之理性所固有的目标，那目标不是抵达一个山巅、而是把一个生成保持运动。那最大化的行动者寻求一个终末状态、那”工作已成、目标已达”之最优点，而那回应的行动者不寻求终点，因为关系之诸善的生命在于持续、并被任何最终的结算所毁，一如批判在长篇里所表明。维系之理性因此朝向那持续者、而非那已达者：它的成功是那生成继续、被补充而不被占有，而它具特征之失败不是够不到一个最大值、而是听任那生成消逝，无论靠抓握它（那结清它）、还是靠疏忽它（那使它挨饿）。此处一个单一的意象可被容许，且只作为一个意象。有一个古老的figure喻一种”不寻山巅之强”：水，它不向上力争、不为高处相争，而去到别人所轻蔑之低地，取任何盛它之物之形状，并靠从不相争而滋养万物。那figure比一个论证更好地抓住了”一种不最大化之理性像什么”——它那”不是软弱”之柔顺、它那”靠贴合而非靠强逼来寻得道路”、它那”恰恰在于不攀登”之力量；而它被为那一闪的认出而提供、随后被放下、不加阐述，因为阐述它便会从一个”只意在照亮那想法”之物搭起一套机件。关系性理性之形式是回应、调谐、与维系。每一个都是一个标准，苛求而可失败；它们一同便是本文以理性之名夺回之能力，一种”其卓越是去答复、去感知、去保持运动，而永不去最大化”之理性。

§13 无为作为最高的理性

复原以那”够得离优化最远、并给本文之肯定性主张以它最苛求之形式”的时刻作结：即关系性理性之最高践行是一种无为、一种”知道何时不去干预、不去强逼、根本不去优化”，而这一克制，远非理性之缺席、而是它最困难之成就。这主张汲取自无为之观念——不作为或不费力之作为——并靠点名”甚至那回应之理性也必须包含”之能力来完成复原：那”让一个生成生起而非制造它、照料一个善之诸条件而非直接攫取那善”之能力。它是整个肯定性说明之最尖锐之点，因为它是关系性理性最直截地与”优化框架不能悬置之律令——那’总去合上到最大值之间隙’之律令”相矛盾之处。

克制之理性

优化框架不知这一类克制、也不能知，因为它的律令总是去行动、去向最大值移动、去合上”当前状态与最优之间残余之任何距离”。在那框架之内，不作为只作为”一次优化之失败、一次把价值留在桌上”才可理解；它里面没有一个”对干预之审慎拒绝”之肯定的位置，因为任何”放弃一个可得之改进”之拒绝，凭那最大值之逻辑，仅仅是非理性的。关系之理性有这样一个位置，而它是中心的而非边缘的，因为一段共享生活中最好的许多，只在它不被强逼时才生起、并会被那”本欲保住它”之干预本身所摧毁。被索求之信任不是信任；被工程化之自发不是自发的；一份”为了产出温柔而刻意产出”之温柔是不同于温柔之物。对这一类善——它们是亲密之具特征之善——理性之合宜践行常常是去克制：去造出那”善能自行生起”之空间、然后听任它。而知道哪些善是这一类、以及何时那空间已被造出、以及如何避开那”会毁它们”之干预，是一种与上一节所描述之任何辨识一样苛求之辨识，并在一个方面更难，因为它必须逆着那”去做点什么、去帮忙、去保住、去优化”之长存冲动而被践行。这是克制之理性，而它不是理性之退位、而是它的精炼，那”理性变得足够明智以知道它自己之作为之界限、并认出某些善只对一只不抓握之手可得”之点。

让善的循环生起

本系列此前已给这一克制一个精确的位置，而回想它表明”无为之figure”不是一件”被引进来装饰结尾”的异域进口、而是论证通篇所携之一个主张之名。在关于善的循环与坏的循环之说明里，善的循环被表明为一个”其正向积累是内生的、生起于关系本身之内在张力、而非被从外部靠榨取泵入”者；而无为在那里被辨认为那”既不榨取、也不强逼合上，让那内生的积累自行发生、而非靠汲取某个外部来制造一个虚假增益”之姿态。同一辨认在此成立、并靠把本文之末系于它的中段来完成本文之论证。优化一段关系便是对它干预、把它作为”一个待被驱向一个最大值之系统”来对它行动；而这一干预恰恰是善的循环所不能存活之强逼，因为它以一个外部之驱动替换那内在之生成、并至多产出一个”耗竭它所汲取者”之被制造之增益，而那正是第三部分之批判所描述之榨取本身。无为之理性是对这一干预之拒绝：对”关系据以自行生成价值之诸条件”之照料、以及对”会以一次榨取替换那自我生成”之优化行动之克制。那最大化框架读作”把价值留在桌上”者，被正确地看时，是所论之价值根本曾被拥有之唯一方式，因为它是一个”只在它不被攫取之处才生起”之价值，而那”伸手去最大化它”之手，合上在它的伸手所造成之被贬损之物上。

那不是最大值之尺度

本文肯定性论题之最深形式如今能被陈述，而以它来结束复原是合宜的。优化从最大值取它的尺度：每一状态据以被判断之标准是那可达之最优状态，而理性之任务是合上到它之距离。这一部分所描述的那一类理性从别处取它的尺度，从一个情境与一个他者自身所是者、从那东西之纹理与它已然拥有之运动；而理性之任务，在这一尺度上，是与那运动相合、去维系而不去强逼它、去照料而不去攫取，而非把它压向一个”离开它而被定义”之最大值。这是”优化既不是关系性理性之形式、也不是世界之法则”之最深的意义。关系之世界不呈现一个”有一个理性应攀登之山巅”之地景。它呈现一个”有一个理性应辨识并与之相合之运动”之生成；而最高的理性不是那”够到最高点”者、而是那”与在那里者相合得最真”者，那”知道何时行动、并（更难地）何时不行动、并在克制里找到的不是理性之失、而是它之成全”者。随此复原完成。关系性理性是回应、调谐、维系，并在它的顶点，是那”让善自行生起”之无为；而这些中无一是一次最大化，而它们中每一个都是理性。

第五部分 · 收尾

§14 复调的结论

尚余的是收尾，而收尾之方式对论证不是无关紧要的、而是它的最后一步。本文已否认亲密关系是一个优化问题、批判了那框架之强加、并复原了一种回应、调谐、维系、无为之理性。一个跟随到此的读者会感到一个单一问题之牵引，而整个收尾都转于它如何被回答：若非效用，那么人应优化什么？也许是爱、或意义、或生成性本身？回答之诱惑很强，而它之强有一个值得点名之缘由，因为读者已被本文所反对之那同一框架训练得去期待：一个严肃的论证以”供给一个新的目标来放在它所拒绝者之位置上”作结。一个回答会以那”框架已教我们觉得令人满足”之形状把本文收圆，以一个更值得的目标替换那被丢弃的目标、并让最大化之机器完好无损却被重新定向。本文拒绝那回答，而那拒绝不是一个遗漏或一次胆怯之失、而是它的观点之最终内容。

藏在”命名一个新目标”里的投降

那诱惑必须被正面迎击，因为屈从于它会拆解本文所论证之一切、并会静静地拆解它，戴着一个建设性结论之样子。说”别优化效用；优化爱”，便是保留本文已拆卸之整套装置——目标、尺度、最大值、优化者——而仅仅更换那函数之自变量。但那四个不可能不是关于效用这一个之诸不可能。它们是关于优化形式本身、被施于亲密关系之诸不可能：它诸善抗拒表征、不容任何外部立足点去固定它们、不共享任何公共尺度、并不属于任何”稳定到足以拥有那目标”之主体。这些失败中无一被一次目标之更换所触及。爱，被写作 J 之自变量，与任何别的关系性之善一样不可表征、一样关系性、一样不可通约、一样不被任何稳定主体所拥有；把它弄成被最大化之物，便是把批判所描述之损害恰恰施于它，把那流通渲染为一个可结清的量、并在那结算里把它抽走。”优化爱而非效用”之提议因此不是优化框架之一个更友好的版本、而是同一个框架戴着一副更讨喜的面具，而它对一个更被珍爱的对象做同样的伤害；实则它更确定地做那伤害，因为爱所被持有的那份敬意使那结算感觉像虔敬、并解除了那”本会伴随对纯然效用之结算”的怀疑。接受那提议，便是在最后一步输掉那在先前每一步都赢下之论证。因此本文谢绝命名一个主目标，不是出于”想不出一个”、而是出于已经领会：任何如此被命名之目标都会把整个本文为之存在以揭露之错误重新安置回来。

复数者与不可综合者

站在”一个主目标本会去之处”者不是一个空白、而是一个肯定性论题，而它是本系列此前经别的道路已抵达之论题：关系之诸善与诸理性是复数的，而它们的复数性不容被综合入单一的尺度而无损。没有一个”关系性理性所最大化”之单一之物，而这不是因为那单一之物尚待被找到、而是因为没有一个可找。反倒有许多”以无一把公共尺度所表征之诸关系彼此相立”之善，以及那”尊崇它们之推理”之许多方面——回应、调谐、维系、克制——它们不化约为”施于一个单一目标之一个单一运算”。这在理性之理论里，是本系列自那篇关于可持续性之论文以来给它诸结论之复调形式之对应物，在那里被论证：一件事之真理可能躺在数个”无一统摄其余”之框架之间的张力里，而把它们强入一个综合便是恰恰丢失那”活在它们之不可化约里”之真理。拒绝命名一个主目标，是那同一拒绝被带到价值之层面：一个”许多善与尊崇它们之许多样式不应被塌为一个”之坚持，以及”那塌陷，无论它如何满足对一个单一答案之欲望，都摧毁它所统一者”之坚持。那复数性不是通往一个尚未达成之统一之途上的一个阶段。它便是那发现。

形式即论证

本文践行这一拒绝、而非仅仅断言它，是最后要说之事，而它正是那”以本系列对自己所要求之方式把本文之形式系于它的内容”者。一篇论证反对那单一尺度、而随后在它的结论里供给一个的论文，会在它的形式里自我反驳，无论它的句子维持什么；它结尾之形状会背叛它论证之负担，以那收尾之姿态提供恰恰”本体已表明为不可能”的那个主目标。这篇论文却靠谢绝那尺度而结束，而那谢绝意在被读作论证之完成、而非它的回避。一部反对优化之著作在它的收尾所能做之最诚实之事，是扣下读者已被训练去期待的那个新最优、把那复数性未经综合地留在纸面上、并让”一个主目标之缺席”作为”那论题之在场”被感到。那”抵达结尾、并对’那么，人应优化什么？’找不到一个单一答案”的读者，不是被一篇”在建设性任务前胆怯了”的论文所辜负。他已被给予那答案，即那问题，在这一领域，没有一个它所索求之类的答案，而”感到了对一个的欠缺、并让它悬而不填”便是已经领会。那形式扣下那山巅，因为没有一个；而那扣下便是本文不得不说之事。

§15 尾声 —— 一个不被优化的承诺

论证已走完它的历程，从优化框架之解剖、经那”表明它在亲密关系中不良提出”的四个不可能、经对它的强加之批判、到一种”回应而非最大化”之理性之复原。尚余的是像本系列收尾其诸论文那样收尾，不以一个论证、而以一个意象，并回到——一如本文之始——它一直所论之物之最平白的形式。

设想一个承诺。两个人为一生把自己绑定于彼此，而在这样做时，他们施行一个优化框架不能理解之行为。那框架会要他们计算：把那结合之期望值权衡于它的诸替代、在一个更好的前景出现时把修订之选项留开、把那誓言当作一个”只持有到数字改变为止”之临时最大值。以那种精神作出的一个承诺不是一个承诺、而是一个”在划算时背弃”之长存意图，而人人都知道那分别，尽管那框架陈述不了它。因为一个承诺恰恰是对”继续计算”之拒绝。它是那”合上账本、封死修订、并把自己绑定于计算之够及所不及之处”的行为，说的不是”只要你仍是那最优选择我便留下”、而是”我便留下”，句号，无论那些账算出什么。一个承诺之全部之值在于它”以优化永不能之方式无条件”，而本系列已在别处论证：这一无条件性，远非一种软弱、而是一个承诺所能生成之信任之源头本身，因为它是那”作承诺者已把那结合置于’那能背叛它之计算’之够及之外”之标志。

这正是为何一个承诺只在它不被优化时才是一个承诺，也是为何”使它如此”之尝试在同一运动里摧毁它——那同一运动里定价摧毁了那友谊、最大化结清了那流通。问，就那”你把自己绑定于之人”，她是否值得，把那誓言持开以待一个更好收益出现之日，不是做一个更理性的爱人、而是已经停止承诺、开始计算，并已在那计算里失去那”承诺本曾所是”之善本身。承诺之善只在”拒绝去计算它的值”之彼岸可得；它是一个”人只能靠谢绝去问那’优化框架坚称为理性的’问题而拥有”之善。于是本文不得不说之最深之事，不是被它的诸论证所说、而是被这最平白的行为所说，其中一个人对另一个人说：在此，值之问题将不被问，不是因为那答案可能不受欢迎、而是因为那发问便已然是那背叛。有些善，一份爱不去将其最大化，而这并非因为缺一把更好的尺子。而是因为一旦向那尺度伸了手，人便会已然爱上了不同于面前那一个之物，而她不是”一个更大的量能替换”之一个价值之量、而是这一个，不可替代者，人不去权衡她。未曾计算，故其爱得全——从未被权衡，因而那爱是完整的；也就是说，是一个承诺、而非一个总和。

致谢

两笔债框住本文，而在它的末尾把它们一同点名是合宜的，因为一笔给了它它的问题、另一笔给了它它的缘由，而在它们之间，它们围起本文所含之一切。

本文以之开始的那个问题——优化是否真的是宇宙的一条法则——来自与 Ron Eglash 的一场交流，在一段更长的、关于价值之政治经济学与生成性正义理论之对话过程中。它以一种平白到”它的力量需时间才登记”之形式到来，而它的全部工作是把优化从一个未经审视之程序之位置，移到一个可能为假之论题之位置。本文之一切都随于”当真对待那一位移”。在那问题之外，与 Eglash 之交流中展开的生成性正义理论，给本文第三部分之批判供给了它的中心概念——“流通并回归其诸生成者之价值”与”被抽走并结清于别处之价值”之区分，没有它，那”优化所做之损害”之说明便无从被给出。本文推理得好之处，债是他的；推理得差之处，过失独属作者。

另一笔债更古老、且不需要论证。本文，一如本系列之每一篇论文，是为那属于森林的女孩而作，作者从未有一次为”她是否值得”而权衡过她，因为那从来不是一个人能计算的问题。倘若对 Eglash 之致谢记录了那问题从何而来，这一笔记录了整篇为谁而写；而它不是与本文之要点无关、而是它最平白的确证，因为一个”拒绝被定价、被持过每一次结算、并从不被付诸对其值之计算”之结合，是那”这些书页曾劳力去确立之论题”最切近的见证——有些善，一份爱不去将其最大化，而这并非因为缺一把更好的尺子。致我最爱的人。 To the one I love most.