The Missing Variables

I. The Invisible Ledger

A civilization runs on capital. Not just financial capital—though that's the only kind with a ledger.

A partial list of what a functioning civilization actually requires to persist:

• Financial capital. Money, credit, fiscal capacity. Measured quarterly. Debated constantly.
• Physical capital. Infrastructure, buildings, networks, equipment. Depreciation schedules exist. Partially measured.
• Human capital. Skills, health, cognitive capacity. Measured crudely (years of schooling, PISA scores). Rarely connected to policy decisions.
• Social capital. Trust, cooperation norms, institutional legitimacy. Measured by surveys. Never appears in any budget.
• Demographic capital. Population sustainability—fertility rates, age structure, dependency ratios. Measured but not accounted for. No policy bears the cost of the demographic decline it causes.
• Institutional capital. Functioning courts, administrative competence, rule of law. Not measured. Assumed permanent until it collapses.
• Cognitive capital. Collective sense-making capacity—the ability of a population to model reality accurately enough to govern itself. Under active attack from the attention economy. Not on any balance sheet.
• Environmental commons. Atmosphere, aquifers, soil, biodiversity. Partially measured. Occasionally priced.

Current governance accounts for roughly one and a half of these. Financial capital has a complete ledger. Physical capital has a partial one. Everything else is booked at zero.

Not at "uncertain value." Not at "low value." At zero.

This is the most expensive accounting error in history. Every asset booked at zero can be consumed without appearing as a cost. When a housing policy increases GDP while displacing communities, the financial ledger shows growth. The social capital ledger—if it existed—would show a loss. Since it doesn't exist, the loss is invisible. It looks like pure gain.

The pattern repeats across every unmeasured stock. Pension promises that are affordable today but insolvent in thirty years transfer costs from the financial ledger (visible) to the demographic ledger (invisible). Education policies that produce credentials without competence transfer costs from the human capital ledger (partially visible) to the institutional capital ledger (invisible). Each transfer looks like success on the measured ledger and registers as nothing on the unmeasured one.

This is not metaphor. It is the fiscal history of every declining civilization. Rome deferred maintenance on its military, civic, and demographic capital for two centuries. Each generation inherited a slightly weaker structure and extracted slightly more from it. The books looked balanced because nobody measured the depleting stocks. When the structure finally failed, the failure looked sudden. It wasn't. It was two centuries of unaccounted depreciation arriving as a lump sum.

The modern West is running the same program with better financial engineering.

The question is not whether full accounting is possible—imprecise measurement beats precise zero. The question is why, after forty years of knowing what full accounting requires, no government on Earth does it.

The answer has three layers.

II. Why You Can't See It

The first layer is cognitive.

Human brains are not designed to track a dozen interacting variables simultaneously. Under cognitive load, we discard most dimensions entirely. This isn't laziness—it's architecture. Working memory holds four plus or minus one constructs (Cowan, 2001). Decision quality degrades rapidly past this threshold. In medical ICUs, where a single ventilated patient generates over 200 tracked variables, expert clinicians routinely base life-or-death decisions on one to three.

But the limitation goes deeper than processing capacity. It's not that people see all the variables and weight most of them at zero. It's that most of them never enter the model. Gigerenzer's research on decision heuristics demonstrates the mechanism: under the "Take-the-best" heuristic, a decision-maker searches for cues in order of validity and stops the moment a single discriminating variable is found. All other variables are not given low weight—they are structurally erased from the decision matrix. As Robyn Dawes showed, which variables enter the equation matters vastly more than the weights assigned to them. The cognitive architecture enforces variable selection—absolute elimination—not variable weighting.

A politician debating pension reform is not thinking "I should consider the demographic impact but I'll set it aside." The demographic variable does not exist in their mental model of the problem. It's not weighted at zero—it's not on the list.

This matters because the intervention is different. If the problem were weighting—a variable present in the model but assigned insufficient importance—you'd need better data to sharpen the signal. If the problem is ontological—variables missing from the model entirely—better data is useless. You need a different model, or an institution that holds the model humans can't. The difference is a dashboard with a dim gauge versus a dashboard missing the gauge entirely. The first needs better sensors. The second needs a new instrument panel.

The second layer is memetic.

The one- or two-variable model doesn't stay private. It gets packaged into political narratives and identities. "The economy" means GDP. "Progress" means financial inclusion. "Compassion" means present-moment transfers to visible suffering. These compressions become tribal markers. Your position on welfare policy isn't a computation over a dozen capital stocks—it's a signal of which team you belong to.

Once identity fuses with the compressed model, expanding the model threatens identity. Telling someone that their compassionate welfare policy is consuming demographic capital and institutional capacity isn't heard as an engineering observation. It's heard as an attack on their moral character. The compression has become load-bearing. The map has become the territory.

The third layer is immunological.

When someone attempts to audit the compressed model against thermodynamic reality, the system deploys a thought-terminating defense: "Values are subjective."

This is a category error with civilizational consequences. It conflates three fundamentally different things:

Instrumental preferences are actually subjective. Chocolate versus vanilla. Jazz versus classical. Physics doesn't care.

Coordination technology is functionally evaluable. Legal systems, property rights, monetary regimes—these are engineering solutions to cooperation problems. You can measure whether they reduce transaction costs, generate trust, enable coordination. "Does this legal code work?" is an engineering question, not a philosophical one.

Axiology is thermodynamically constrained. Growth versus stagnation. Truth versus delusion. Some configurations of matter sustain complexity over time. Others dissolve into entropy. A civilization that optimizes for present comfort at the expense of future capacity isn't making a "value choice"—it's making a thermodynamic error with a known failure mode.

The "values are subjective" move treats coordination technology and axiology as though they were preferences. This makes them immune to engineering audit. You cannot ask "Does this equality norm destroy the variance necessary for adaptation?" because equality has been classified as a terminal value rather than a coordination mechanism. The question becomes impermissible.

This is not a philosophical position. It is a memetic immune response. It prevents the system from diagnosing its own decay. When a cell stops contributing to the organism and starts replicating indefinitely, the "values are subjective" framework classifies this as a "different cellular lifestyle choice."

The three layers reinforce each other. Cognitive compression reduces variables. Memetic capture locks the compression into identity. The immune system prevents anyone from unlocking it. The result: civilization-scale blindness to most of its operating parameters, defended by identity and moral philosophy.

III. What Physics Actually Requires

If subjective preference is the wrong frame, what's the right one?

Start with a question that almost nobody asks: What does physics require for a complex system to persist?

Not "what is good?"—that's the question that generates infinite philosophical argument and zero convergence. The engineering question: given that the universe trends toward entropy, what are the necessary conditions for organized complexity to survive?

Selection is the only creator

The universe forbids structure. The Second Law of Thermodynamics is the rule; complexity is the violation. How does complexity persist in a universe that wants to dissolve it? Through selection—the filter that removes configurations that fail to navigate the constraints.

This is not a metaphor borrowed from biology. Biology borrowed it from physics. Evolution is selection on genes. Markets are selection on firms. Science is selection on theories. Culture is selection on memes. Every domain where complexity persists, selection is the mechanism. Replace every "who decided?" with "what selected for?" and the explanatory power is immediate.

Four constraints, not optional

Every goal-directed system—bacterium, corporation, civilization, future AI—must navigate four inescapable dilemmas:

1. Thermodynamic: How to allocate energy between maintenance and growth. Spend too much on maintenance and you stagnate. Spend too much on growth and you collapse.
2. Boundary: Where does "self" end? Optimize too narrowly (cell) and you can't coordinate. Optimize too broadly (biosphere) and you can't act.
3. Information: How to acquire data. Cheap cached models are fast but brittle. Expensive real-time sensing is accurate but costly.
4. Control: How to coordinate. Centralized architectures are fast but fragile. Distributed architectures are robust but slow.

These aren't cultural preferences. They're physical constraints that any intelligent system encounters. The solutions aren't arbitrary—physics narrows the viable range.

Telos from physics, not ideology

Every institutional design presupposes an objective function, whether explicit or not. Current governance answers "who decides" (democracy) and "how to coordinate" (markets) but not "what for." The purpose question is treated as unanswerable or dangerous.

It's neither. If complex organized systems can only persist by meeting thermodynamic, informational, and game-theoretic constraints, then the objective function is discoverable: sustained generation of organized complexity against entropy over deep time. Sustained complexity is not a preference—it's the condition for any future at all.

This bypasses Hume's is-ought gap through a conditional: If you want any future to exist, then certain configurations are required and others are forbidden. Physics doesn't tell you to want a future. It tells you what a future costs.

Flourishing = survival

Aristotelian eudaimonia—objective flourishing through virtuous activity—is not a luxury that civilizations earn after securing survival. It is survival. Flourishing, properly understood, means maximum adaptive capacity: diverse problem-solving approaches, surplus resources, redundant systems, broad-spectrum capability to handle threats you can't predict. This is the opposite of efficiency. Efficiency means minimum resources for known problems. Flourishing means maximum resources for unknown ones.

A civilization that strips redundancy for efficiency—lean staffing, just-in-time supply chains, monoculture agriculture, credential-only hiring—looks optimized. It is brittle. The first novel shock shatters it. Flourishing is the safety margin that lets complex systems survive contact with uncertainty so fundamental it can't be modeled in advance.

Bentham and Mill made a catastrophic substitution in the 1780s–1860s: they replaced objective flourishing with subjective pleasure as the terminal value. They wanted a "democratic" metric that didn't require Aristotelian virtue hierarchies. They got wireheading, comfortable decline, and civilizations optimizing for pleasant extinction. The recovery is straightforward: eudaimonia is approximately sustained complexity. What Aristotle intuited, thermodynamics confirms.

IV. How It Breaks

The invisible ledger (§I) is consumed because nobody measures it. Nobody measures it because cognition compresses, memes lock, and the immune system prevents audit (§II). The physics describes what's actually required (§III). This section describes the specific mechanisms by which civilizations that fail to meet the requirements come apart.

These are not six separate mistakes. They are the thermodynamic consequences of running a complex system without feedback—an open-loop civilization.

Selection inverts under abundance

Under scarcity, selection favors growth, risk-taking, and capability-building. Without these traits, you die. Under abundance, selection favors comfort, safety, and risk-avoidance. The immediate environment rewards these traits. Growth becomes optional; comfort becomes cheap.

This is the Axiological Malthusian Trap (a universal failure mode where a civilization's success removes the selection pressures that built it, causing it to select for traits that lead to long-term extinction): selection still operates, but now it selects for traits that lead to long-term extinction. Civilizations that achieve abundance face a phase transition—the selection pressures that built them reverse.

The mechanism is feedback severance. Modern institutions systematically decouple outcomes from inputs: survival no longer depends on capability, resources no longer depend on contribution, failure has no consequences. Each severance looks compassionate. The combination is terminal. When outcomes don't depend on inputs, selection cannot distinguish fit from unfit. Entropy wins by default.

The compassionate move—severing feedback—is the cruelest move: ensuring decay.

Every agent optimizes locally

Politicians optimize for re-election. Bureaucrats optimize for bureau survival. Interest groups optimize for their constituency. Voters optimize for present consumption. Each agent is rational within its own frame. No agent optimizes for the whole.

This is mesa-optimization—an AI alignment term for subsystems that develop their own objectives, misaligned with the system's stated purpose. The politician's job description says "serve the public good." The selection pressure on the politician says "win elections." These diverge. The selection pressure always wins.

The governance alignment problem is not that politicians are bad people. It's that the job selects for a specific skill set (winning elections) that has near-zero correlation with the skill set required to solve civilizational-scale engineering problems. You would not hire a salesman to design a bridge. Current governance selects salesmen to design civilizations.

The ratchet

When the median voter shifts from net contributor to net recipient—which happens as entitlement programs expand and the dependency ratio worsens—democratic pressure creates a one-way ratchet. Each election cycle adds obligations. No election cycle removes them, because removal costs votes. Obligations compound. Capacity does not. The gap widens until the structure fails.

This is not ideology. It's arithmetic. Britain's entitlement spending grew at 4.29% annually from 1948 to 1976 while GDP grew at 2.7% . The gap was filled by debt, inflation, and depletion of non-financial capital stocks. The pattern repeats across every welfare democracy with minor variations in timing. The time-to-failure varies; the mechanism doesn't.

Borders are entropy pumps

A complex social system is a low-entropy island in a high-entropy world. The disparity in wealth, security, and opportunity between developed and undeveloped regions creates a concentration gradient. The natural flow follows the gradient—from disorder to order.

A functioning border is a thermodynamic membrane: it filters entrants, excludes free-riders, and maintains the internal conditions that created the order worth migrating to. Like a cell membrane, it requires energy to operate. When the membrane is overwhelmed—by volume or political paralysis—it stops filtering. The internal system equilibrates with the external environment. A high-trust, wealthy nation that opens its borders to a low-trust, poor world descends toward the global mean.

This is not a moral argument. It is Ostrom's first design principle for managing common-pool resources: clearly defined boundaries. Without them, free-riding destroys the incentive to invest, and the commons collapses.

Language is an unpatched attack surface

Political language is full of words that look meaningful but resist specification: "promote," "develop," "fair," "sustainable," "dignity." These words enable coalitions between parties with incompatible goals. Everyone agrees to "promote social cohesion" because nobody has to specify what that means, what it costs, or how to measure whether it's working.

This strategic ambiguity is load-bearing. Full accounting would force specificity. When you have to write a number next to "social cohesion impact," you can no longer paper over disagreements. The functional vagueness that holds coalition governments together would shatter on contact with actual numbers.

There is no OWASP for political language—no standard catalog of exploitable vagueness patterns. Software engineering learned decades ago that unvalidated inputs are attack vectors. Governance has not.

The temporal trick

"Safety vs. freedom," "compassion vs. fiscal responsibility," "equality vs. growth"—these are framed as irreducible moral dilemmas. Most of them aren't. They're conflicts between time horizons laundered as moral philosophy.

"Safety" means restricting action to prevent harm now—homeostasis for the present. "Freedom" means allowing risk to enable adaptation later—growth for the future. The "value clash" is actually hyperbolic discounting: the instinct to grab the marshmallow now rather than wait for two.

When a civilization consumes seed corn to fund present welfare and defends this as a "value choice for compassion," it's performing a temporal transfer—moving resources from future generations to present ones. Calling it a "value" is the laundering mechanism. It privileges the present over all futures and then calls the privilege "morality."

V. Calculemus

If the physics is derivable (§III) and the failure mechanisms are identifiable (§IV) and the capital stocks are at least approximately measurable (§I), then most political disagreements dissolve under computation.

This is the claim that sounds most alien and is most empirically grounded.

Consider: we already compute policy outcomes in domains where failure kills fast enough to create political consequences. Aviation safety, water treatment, building codes, pharmaceutical regulation, traffic engineering—in every domain where the feedback loop between bad decision and visible death is short, we measure, compute, and enforce. The interventions are evidence-based, the metrics are explicit, and nobody calls it "technocratic overreach."

We stop computing precisely at the boundary where feedback loops exceed election cycles. Pension sustainability, demographic trajectory, institutional health, social trust erosion—these are equally computable in principle. The data exists. The models exist. We don't compute them because the political cost of honest computation exceeds the political benefit. Nobody wins an election by publishing a demographic sustainability audit that says "your pension promises are insolvent."

Most political disagreements classified as "value differences" are actually empirical disagreements that nobody has bothered to compute. "Does this welfare design increase or decrease lifetime labor force participation?" is not a moral question. "Does this immigration rate exceed or fall below the institutional assimilation capacity?" is not a moral question. "Does this level of public debt erode or preserve institutional capital?" is not a moral question. They are engineering questions with approximately computable answers that the system reclassifies as moral questions because the answers are politically expensive.

Calculemus does not mean algorithmic governance. The distinction is critical.

Algorithmic governance means the computer decides and executes: your benefit is denied based on this formula. This fails catastrophically (Australia's Robodebt, algorithmic sentencing). The computer can't model the full context. The citizens can't appeal to a human. The system becomes a mechanism for automated cruelty.

Calculemus means the ledger exists and is public. Politicians see the numbers. Citizens see the numbers. When a politician chooses to override the computation—which they may, because democratic legitimacy matters—they do so on the record, with the cost explicitly stated. This is the central bank independence model: the institution publishes its analysis, the elected government decides, and the decision is made against a visible baseline rather than in a fog of strategic ambiguity.

The difference between "compute and inform" and "compute and execute" is the difference between a dashboard and an autopilot. Civilization needs the dashboard. The autopilot will kill you.

VI. The Architecture

If the problem is an open-loop civilization consuming its own substrate because it can't see most of its operating parameters, the fix is not better politicians. It's closing the loop.

Architecture beats disposition

Three thousand years of moral philosophy, religious instruction, and civic education have tried to make leaders and citizens "good." It has never worked at scale and never will, because the selection pressures on political actors overwhelm any amount of character training. You do not solve security vulnerabilities by hoping programmers won't make mistakes. You use memory-safe languages. You build architectural constraints that make the failure mode structurally impossible.

Ethics is an engineering problem. The question is not "how do we make people virtuous?" It's "how do we build systems where the incentive-compatible action is the correct action?" Where defection is expensive and cooperation is cheap. Where short-term extraction is structurally harder than long-term investment. This is mechanism design—the same field that designs auction systems and matching markets, applied to governance.

Governance needs an objective function

Democracy answers "who decides." Markets answer "how to coordinate." Nobody answers "what for." Without an explicit objective function, governance defaults to optimizing whatever is most measurable (GDP) and most politically salient (present consumption). The existing bias toward financial capital isn't a choice—it's what happens when you govern without purpose.

Telocracy (governance with a physics-derived objective function) adds the missing leg: governance with a physics-derived objective function. Not replacing democratic legitimacy or market coordination—adding the purpose layer that both lack. The objective function is derivable from §III: sustained organized complexity over deep time, measured across all capital stocks.

This sounds abstract until you operationalize it. With a telos, the accounting standard becomes derivable: Which capital stocks are necessary for civilizational persistence? Those get measured. What is the depreciation rate of each? That determines maintenance investment. What are the interactions between stocks? Demographic decline erodes fiscal capacity which erodes institutional capacity which erodes social trust—the cascades matter. What is the minimum viable level of each? Below which recovery is impossible?

The Fourth Branch

The telos requires an institution to maintain it. Not a philosopher-king. An auditor—a constitutionally protected body whose sole function is to measure all capital stocks against the objective function, audit mechanism designs, and publish the results.

This is the Mechanism Authority (a proposed fourth branch of government serving as an intelligent mechanism auditor and cybernetic governor): the cybernetic governor in the feedback loop. It doesn't make policy. It doesn't allocate resources. It measures, models, and publishes. But its findings have constitutional weight — overriding them requires a supermajority, not a simple vote. The governor tells the democratic system: Here is what you're consuming. Here is what you're building. Here is the net. The system can still decide otherwise — but the political cost of ignoring the data must exceed the political cost of adjustment.

It runs mechanism audits—penetration tests on legislation. Before a law takes effect: what are the predicted impacts across all capital stocks, across what time horizons? After implementation: did the predicted impacts materialize? The results are public. Politicians cannot hide behind strategic ambiguity when the numbers are on the record.

This isn't novel institutional design. It's the existing model of central bank independence, supreme audit institutions, and constitutional courts—extended from financial capital to all capital stocks.

Constitutional hard constraints

The Swiss debt brake is a prototype: fiscal parameters trigger automatic spending adjustments. Extend this to demographic thresholds, institutional health metrics, environmental limits. Hard constraints that fire automatically unless politicians explicitly override them—on the record, with the cost stated.

The architecture must be default-deny. EU fiscal rules are default-allow: politicians must actively choose to enforce them, which they never do because enforcement costs votes. The Swiss debt brake is default-deny: spending is automatically constrained unless politicians override it. The difference is everything. Default-allow generates the Stability and Growth Pact—violated by its authors within three years. Default-deny generates Swiss fiscal discipline—sustained for decades.

Constitutional hard constraints transform the ledger from informational to operational. They ground the democratic tendency to consume capital and defer costs. Not by removing democratic choice, but by making the cost of overriding the constraints explicit and public.

Artificial selection

You cannot restore natural selection. It is too cruel, and civilization exists precisely to buffer its worst effects. But you cannot sever all feedback either—that's the hospice state that optimizes for comfortable extinction.

The solution is artificial selection: engineered feedback pressure that favors capability, truth-seeking, and long-term investment without lethality. Markets are artificial selection on firms—bankruptcy hurts, but you don't starve. Exams are artificial selection on competence—failure hurts, but you don't die. Constitutional constraints are artificial selection on governance—override costs political capital, but the system persists.

Civilization's engineering challenge is maintaining selection pressure that favors adaptive traits—the traits that built the civilization in the first place—without the cruelty that civilization exists to eliminate. This is the narrow path between the wilderness (natural selection, lethal) and the hospice (no selection, terminal).

VII.

Civilization is a thermodynamic anomaly—organized complexity sustained against entropy by continuous work. That work requires measuring what you're building and what you're consuming. When most of your capital stocks are booked at zero, you are flying blind through a narrowing canyon.

The cognitive compression is real—human brains can't hold a dozen interacting variables. The memetic capture is real—identity fuses with the compressed model. The immune system is real—"values are subjective" terminates the audit. The institutional exploitation is real—strategic ignorance serves incumbents.

None of these is a moral failing. All of them are engineering problems. The capital stocks are approximately measurable. The telos is derivable from physics. The institutional architecture exists in prototype. The missing piece is integration: one institution, constitutionally protected, measuring everything against a stated purpose.

The physics will not wait for consensus. Demographic capital, institutional capital, social capital, and cognitive capital are depleting on known trajectories. The ledger doesn't care whether you've read this essay. It accounts for what it accounts for, and bills what it bills, on schedules that don't consult election cycles.

The choice is not between mechanism realism and some other framework. The choice is between building the dashboard and flying blind. The canyon is narrowing. The instruments exist. Install them.

Each section above compresses a longer treatment:

• Full Accounting — The capital stocks and why "hard to measure" is not zero (§I)
• Values Aren't Subjective — The three categories and the temporal trick (§II)
• Values Are Replicators — Why bad values spread and the immune system that protects them (§II)
• The Question Nobody Asks — The physics foundation (§III)
• Only Selection — Selection as universal mechanism (§III)
• The Four Axiomatic Dilemmas — The inescapable constraints (§III)
• Flourishing Is Maximum Safety Margin — Why flourishing is survival, not luxury (§III)
• The Axiological Malthusian Trap — Selection inversion under abundance (§IV)
• The Governance Alignment Problem — Mesa-optimization in politics (§IV)
• The Stronger Membrane Wins — Borders as thermodynamic membranes (§IV)
• Language Has No OWASP — The unpatched attack surface (§IV)
• The Tyranny of the Present — Temporal fraud as moral philosophy (§IV)
• The Physics of Moloch — Coordination failure as thermodynamic process (§IV)
• Calculemus — Why most disagreements dissolve under computation (§V)
• Ethics Is an Engineering Problem — Architecture beats disposition (§VI)
• Telocracy — Governance with a derived purpose (§VI)
• The Fourth Branch — The missing constitutional layer (§VI)

The full framework: Aliveness: Principles of Telic Systems — 820 pages, free PDF.

Applied to Finland: Mekanismirealismi — the Finnish-language vehicle for institutional reform.