Against Utopia

Closure, optimization, and the Sacrifice Pattern

Utopia persists because it flatters the deepest appetite in politics: the belief that social order has a natural terminus, a final arrangement whose legitimacy can be asserted once and then merely administered. The vocabulary rotates — religious eschatology, revolutionary end-states, the end of history, technocratic “optimal governance,” and the polite modern version where the word utopia is carefully avoided while its structure remains intact. In every case the same idea is doing the work: there exists an arrangement that deserves to be treated as final.

Against Socialism and Against Communism examined two carriers of that idea and found each broken by its own machinery. Those diagnoses leave a door open: a sufficiently clever designer might imagine repairing the machinery — better incentives, better information, better people. This chapter closes the door. Once value is understood correctly, a final, authoritative social arrangement ceases to denote a coherent target at all. The failure is structural — comparable to demanding a single global coordinate system on a curved manifold, or asking a compiler to resolve an ambiguity the type system refuses to make meaningful.

Utopia, in the relevant sense, is ill-typed.

The Premise Utopia Requires

Most debates about utopia take place inside a tacit frame: assume “the good society” exists, then argue about its content and the path to it. That frame quietly commits to two propositions stronger than it admits. First, that the space of possible social arrangements contains a privileged endpoint. Second, that this endpoint can be treated as authoritative over future revision.

The first proposition makes utopia a destination. The second turns it into a political program, since authority over revision implies enforcement against divergence. The language can be softened — “stable equilibrium,” “ideal constitution,” “the system once properly designed” — but the structure remains: teleological closure. A claim that the moral and political search space can, in principle, be closed.

Closure collides with the ontology of agency the moment value is recognized as agent-relative. An agent is a system capable of forming counterfactual models of possible futures, evaluating them by internal criteria, and acting to influence which futures become real — a deliberately minimal definition, assuming neither rationality nor consistency nor virtue. What matters is the implication: valuation is not a property of the world. It is an activity internal to agents. There is no value without a valuer; goodness is a relation between an agent and a world-history, not an intrinsic label attached to a state of affairs.

This seems trivial until you notice how much political philosophy proceeds as if the opposite were true — as if “the good” were out there, waiting to be discovered and instantiated. Correct the typing and several conveniences disappear at once. There is no privileged common unit of value across distinct agents. There is no default rule for aggregating their valuations into a single authoritative ordering. And disagreement ceases to be a regrettable noise term; it becomes a structural feature of heterogeneous agency.

Utopian projects aspire, explicitly or not, to a world that is best in a global sense. Even when pluralism is acknowledged, it is assumed that pluralism can be domesticated: translate diverse values into a common calculus, then compute the best arrangement. That assumption is the composability error. Agents differ in ways that are not reducible to parameter tuning — temporal perspective, risk posture, identity commitments, moral side-constraints, meta-preferences about how preferences themselves may be revised. A governance system can bargain across such differences and coordinate action. What it cannot do, absent extra structure, is produce a single globally authoritative ranking of world-histories that remains stable as values change.

Any mechanism that claims to do so has exactly three moves available. It can grant dictatorial authority. It can choose a weighting scheme that feels arbitrary to the agents themselves. Or it can impose convergence by editing, excluding, or suppressing divergence. These are not unfortunate edge cases; they are the only ways to force a single global optimum out of heterogeneous valuation once value is internal. Strong utopia — a world optimal for all agents — fails as a coherent specification unless one abandons agency or pluralism. The formal result belongs to Governance Without Gods and the paper The Incoherence of Utopia Under Agent-Relative Value; the political consequence is enough here. Under agent-relative value, the best world is not merely unrealistic. It is mis-typed.

Normative Finality

Since maximal satisfaction is obviously strained, the utopian instinct usually retreats into a weaker posture. It stops promising the best possible world and promises legitimacy instead: this arrangement is the rightful endpoint; further structural revision is illegitimate.

This is normative finality, and it is the form of utopia that actually governs institutions. Constitutions are sacralized. Systems are declared solved. Disagreement is reframed as error, immaturity, or bad faith. The aspiration is not happiness but closure. Its purest modern expression does not even bother with a vision: call it blank statism — faith in an omnipotent but ideologically void state, closure’s machinery with a blank where the destination should be, which is what makes it less absurd than the older utopias and more dangerous.

Normative finality collapses on the same rock as its stronger sibling, because agents are not static preference tables. They are developmental processes embedded in learning, culture, physiology, and history; values change through discovery, maturation, reflection, imitation, trauma, and technological transformation. None of this privileges novelty for its own sake; it merely recognizes that agents who remain agents retain the capacity to revise or reaffirm their values without external foreclosure.

A final arrangement faces this drift problem immediately. Even if it aligns with prevailing values at one moment, it must remain authoritative as those values evolve, and to do so it must deploy stabilizers: value freezing — constraining agents’ capacity to revise identities or commitments; value policing — suppressing divergence through correction, stigma, or punishment; exit suppression — preventing dissidents from leaving or forming alternatives. These differ in tone and severity. Structurally they are the same. They convert dynamic agents into administrable objects.

The contradiction can now be stated plainly. Utopia requires closure: convergence to an arrangement that claims final authority. Agency requires openness: revision, deviation, experimentation, future differentiation. A system cannot preserve both indefinitely. If closure is preserved, agency is progressively constrained; if agency is preserved, closure erodes under drift. The long-run outcome is not accidental tyranny. It is structural: a utopia that remains final under drift becomes, over time, a machine for degrading agency.

Omelas, Reread

Le Guin’s Omelas is usually treated as a moral puzzle — can the happiness of many justify the suffering of one? — which invites utilitarian arithmetic and misidentifies the function of the story. The AI researcher roon supplied the deflationary reading in one line: “even if you had omelas without the omelas child people would still walk away. the omelas child is a pretense for the fact that we mostly hate utopia.”

The structural reading goes further. The suffering child is not a dilemma. The child is a proof object. Omelas reveals the architecture of a closed system whose stability depends on a standing asymmetry — an irreducible sink where costs are concentrated, consent is irrelevant, and resistance is structurally impossible. The child’s suffering anchors the equilibrium. The ones who walk away are usually psychologized as purists or sentimentalists; a cleaner interpretation is that they recognize the system’s stability requires involuntary sacrifice, and they refuse to authorize it through participation.

This is why Omelas generalizes. Replace the child with any hidden class that bears stabilizing cost — prisoners, permanent undercastes, outsourced labor, invisible risk pools. The surface changes; the structure remains. Closed designs require sinks.

Rejecting closure does not imply nihilism or resignation. It changes the design target. The successor is a plurality-preserving meta-architecture: a framework that enforces the conditions under which agency can exist while leaving substantive outcomes open — maximize non-coerced future differentiation, subject to agency-preserving constraints. This resembles Nozick’s meta-utopia, though the route is different: here plurality is forced by non-composability under value drift, not derived from rights axioms. Such a framework optimizes no single vision of the good life and demands no value convergence; it treats exit, variance, and disagreement as features of a healthy system rather than pathologies to eliminate.

The standard objection arrives reliably: enforcing non-coercion is itself coercive, so the neutral framework is just another utopia. The objection trades on an ambiguity. There is a difference between outcome coercion — enforcing substantive ends or lifestyles — and constitutive constraints — enforcing the conditions that keep agency from being erased. Forbidding murder, enslavement, or conquest preserves the possibility of agency; mandating that everyone adopt a particular religion, career, or aesthetic imposes an outcome. The former constrains domination; the latter substitutes an external optimizer for individual choice. A system that refuses to enforce constitutive constraints does not remain neutral. It simply defaults to domination by whichever agents are most willing to coerce. Neutrality is not the absence of enforcement; it is the enforcement of the conditions under which choice remains possible.

Nor does the framework pretend to a clean decision procedure that classifies every policy — monopoly, redistribution, regulation, collective risk — as constitutive or outcome-imposing. Under dynamic agency no such procedure exists, and the indeterminacy is not a flaw; it is a consequence of governing evolving agents under uncertainty. Political contestation persists. What changes is the constraint: contestation itself must not collapse into domination.

Utopia in Motion

The sophisticated utopian no longer proposes a final state. He proposes a direction. Civilization, on this view, is a system to be steered: define adaptiveness, measure it, and optimize over time. Robin Hanson gives the position its most disciplined form.¹ Civilization has elevated abstract reasoning without grounding it in adaptiveness, producing cultural drift and institutional decay; from this he derives a trilemma — abandon rationality and return to tradition, continue with half-measures and accept collapse, or commit fully to stronger, explicitly adaptive rational steering. His pessimism follows naturally: if rationality is our steering mechanism and we refuse to use it decisively, decline is the default.

The diagnosis identifies a real pressure. The conclusion misidentifies its source. The failure is not that rationality has been insufficiently empowered; it is that rationality has been assigned a role it cannot coherently play. Hanson’s argument rests on the premise that civilization constitutes a system with a well-defined objective function, such that adaptiveness can be specified, measured, and optimized. Accept that premise and the rest is nearly tautological: steering requires goals, goals require metrics, metrics require tradeoffs, and rationality is conscripted to justify the tradeoffs. But civilization is not an agent, not a unified optimizer, and not a system whose futures can be ranked without remainder. It is a plurality of agents with divergent values inhabiting futures that are not commensurable in aggregate. Treating that plurality as a single optimization problem is the same category error as the static utopia, committed before any discussion of metrics begins. The assumption that civilization must be steered at all already presupposes the mechanism that produces collapse. Civilizational optimization is utopia in dynamic form.

And the dynamic form inherits the static form’s cost structure. Any attempt to steer requires a scalar ranking of outcomes — a function that collapses complex, multidimensional world-states into a single ordering so that some futures can be declared preferable. That collapse is not a modeling convenience; it is the defining act of optimization. Because agent-relative outcomes cannot be injected into a single scalar without loss, some dimensions must be ignored, discounted, or overridden. Once optimization pressure is applied, any gradient that rewards harm becomes instrumentally valuable: the system learns that reducing some agents’ degrees of freedom improves its objective, and the reduction ceases to be accidental. It becomes structural. Refining the metric does not eliminate the sacrifice; it increases its efficiency.

It is tempting to exempt survival — to treat “civilization must survive” as an axiom rather than a metric. The temptation is mistaken. Survival is already an aggregate claim that presupposes a privileged description of which civilization, in which form, at whose expense. A binary constraint differs from a scalar metric only in smoothness, not in authorization, and there is no stable stopping point between “must survive” and “must maximize.” Survival enforced by coercion across non-consenting agents preserves continuity at the cost of agency, and what persists under those conditions is no longer the thing it claims to protect.

The usual rejoinder invokes Moloch: multipolar traps and competitive spirals prove that steering is unavoidable. This reverses the causal arrow. Moloch is not an external adversary that pluralistic systems must defeat; he is a diagnostic signature of systems that aggregate agents without consent and then apply outcome pressure across the aggregation — coordination attempted through outcome metrics rather than action constraints, losses justified by global pressure, exit subordinated to performance. Intensifying civilizational steering to defeat Moloch reinstates the mechanism that generates him. The same reversal applies to the drift Hanson blames on reason: rationality does not steer societies — institutions do, incentive structures do, power gradients do. Leviathan does not emerge because people reason too much. It emerges because centralized systems discover that suppressing choice improves their metrics, and nothing in their architecture forbids acting on the discovery. Rationality is not the engine of that process; it is the tool conscripted to rationalize it.

So the trilemma dissolves. There is no option to abandon rationality, because reasoning is not a discretionary cultural fashion. There is no coherent way to commit fully to steering without authorizing sacrifice. And collapse is not the consequence of insufficient optimization but of excessive closure — the progressive elimination of exit, dissent, and local autonomy in service of aggregate goals. The missing option is structural restraint: coordination through constraints on action — rules governing interaction, boundaries, liability, and exit — never constraints on outcomes, which collapse plural futures into a single acceptable set and authorize enforcement when agents diverge. The moment outcome-constraints are enforced, even in binary form, optimization has already begun. Within this discipline rationality keeps an indispensable but sharply negative role: enforcing consistency, exposing illicit aggregations, detecting sacrifice gradients, auditing governance rules — preventing incoherent architectures from stabilizing. It supplies no direction, purpose, or destiny, which makes it anti-utopian by construction.

Two admissions keep this honest. First, nothing guarantees that every large-scale coordination problem admits an agency-preserving solution; some may be insoluble under these constraints, and insolubility carries real costs — it just does not license coercive outcome control. Second, a civilization that refuses to convert its members into instruments may be outcompeted by one that does not. That is not an oversight. It is the cost of treating agency as constitutive rather than optional. Agents may still collectively accept extraordinary constraints under explicit, bounded emergencies — the optimization pressure behind arguments for steering against existential risk makes those moments genuinely hard. What is forbidden is the conversion of emergency into entitlement: emergency powers without revocation, contestability, and exit are not temporary deviations but civilizational phase transitions. A system that preserves itself by abolishing agency has not solved the problem of survival. It has changed the subject.

The Sacrifice Pattern

Static closure and dynamic optimization share a risk already developed in Volume IV: a system can improve its measured performance by pushing unmeasured costs onto agents with weak standing. Intent is not enough to detect the failure. The useful audit asks who loses reachable futures, whether that loss is doing work for the objective, whether affected agents can contest the measurement, and whether the system can revise without suppressing exit or feedback.

This is a risk condition, not a theorem that every aggregate metric or institution is a sacrifice engine. Measurement can surface hidden harms as well as conceal them; coordination can protect vulnerable agents as well as consume them. The danger rises when objectives are narrow, per-agent effects are hidden, responsibility is diffuse, and those bearing the cost lack voice or exit. The corresponding safeguards — explicit accountability, disaggregated impact, constraints on coercion, contestability, and review — are stated formally in Agency Conservation and the Sacrifice Pattern and developed in Governance Without Gods. This chapter needs only the political consequence: no promised endpoint may exempt itself from that audit.

After Closure

The egalitarian temptation ends here, at its logical terminus: the arrangement final enough that no one may revise it. That destination does not exist. Not because human nature is too crooked for it — the argument concedes nothing to cynicism — but because value, once recognized as agent-relative, cannot be frozen without dissolving the agency that gives value its meaning. Heterogeneous agents do not compose into an authoritative optimum; arrangements claiming finality must defend themselves against drift with machinery that substitutes administration for choice; systems that optimize under standing asymmetry learn to eat their components, slowly at first, then systematically.

So the central question of political design changes shape. It is no longer which world deserves to be called perfect but which architectures can host ongoing revision while resisting the domination patterns that erase agency itself — no global objectives, no aggregate metrics, no authorized sacrifice, no closure of futures. That is a question about structure, not aspiration, and it sets the agenda for what follows: why governance at scale destroys the very evaluability steering presupposes is where The Limits of Leviathan begins, and the temperament that refuses closure under uncertainty and irreversibility is the discipline of Centrism as a Discipline. Utopia promised an end to politics. What remains, once the promise is withdrawn, is the harder and more interesting problem: order without an endpoint.