---
title: "Values Are Ecology"
subtitle: "Alignment needs a balance sheet and a correction path"
author: Elias Kunnas
description: "Values form ecologies that compete, combine, and reproduce through institutions. Alignment needs a public account of what the ecology produces and a correction path when results diverge from stated commitments."
canonical: https://kunnas.com/articles/values-are-ecology
url: https://kunnas.com/articles/values-are-ecology.md
date_published: 2026-02-27
date_modified: 2026-08-31
corpus_frame_url: https://kunnas.com/articles/how-to-read-this.md
---
## How to read this corpus

The corpus applies one lens to many domains: what mechanisms produce the outcome? It shares four methodological commitments and one explicit directional commitment. Each linked page argues for its part; the links are derivations and disputes, not evidence inherited by every page. The directional commitment does not by itself settle system boundary, distribution, sacrifice, or institutional authority.

1. **Mechanisms are what act.** Incentive gradients, selection pressures, feedback loops, and capital stocks produce the distribution of outcomes. Intentions, labels, official categories, and stated values are evidence about mechanisms, or are themselves coordination mechanisms. They are not causal substitutes. — [Mechanism Realism](https://kunnas.com/articles/mechanism-realism.md) · [Only Selection](https://kunnas.com/articles/only-selection.md)
2. **The reference telos is sustained flourishing.** The broadest achievable adaptive safety margin over deep time — not the continuity of any incumbent state, coalition, institution, or doctrine. A mechanism's own stated goal can still serve as a local proof obligation — showing that its incentives defeat even the purpose it claims is a bounded finding — but meeting that goal establishes nothing about the margin. — [Flourishing Is Maximum Safety Margin](https://kunnas.com/articles/flourishing-is-maximum-safety-margin.md)
3. **Law, rights, legitimacy, democracy, markets, and sovereignty are mechanisms under evaluation.** They are constraints, carriers, or proxies inside the analysis. None is a terminal value or a boundary of what is real. Treating one as terminal ends the mechanism search before it starts. Evaluation carries current function, replacement cost, path dependence, uncertainty, capture risk, reversibility, and who bears model error into the ledger. — [The Stack](https://kunnas.com/articles/the-stack.md) · [Mechanism Space](https://kunnas.com/articles/mechanism-space.md)
4. **Optimization is a system function.** A civilization has to build, exercise, and revise metamechanisms that search mechanism-space, discard dominated options, install, observe effects, and repair under uncertainty. Not running that loop leaves margin unrealized, and that is itself the failure. No single component — analyst, model, or institution — is presumed to contain a global optimum; the capacity is a property of the system. — [From Telos to Policy](https://kunnas.com/articles/from-telos-to-policy.md) · [The Three-Layer Architecture](https://kunnas.com/articles/three-layer-architecture.md)
5. **Uncertainty is preserved, not spent.** Partial orders, binding constraints, unknowns, and residuals stay explicit. An unmeasured effect is not a favorable default. — [The Compression Paradox](https://kunnas.com/articles/compression-paradox.md) · [Cargo Cult Epistemology](https://kunnas.com/articles/cargo-cult-epistemology.md)

*Each essay bears its own evidence. Links carry definitions, derivations, applications, and disputes; they do not transfer proof. Criticism is answered on its substance.*

Canonical: <https://kunnas.com/articles/how-to-read-this.md>

---

# Values Are Ecology

*Alignment needs a balance sheet and a correction path*

Elias Kunnas

## Thesis {#thesis}

Values do not exist one at a time. They form ecologies that compete, combine, and reproduce through institutions. Those ecologies produce measurable capacities and failure modes. Alignment therefore needs more than choosing values: it needs a public account of what the ecology produces and a correction path when the result diverges from the stated commitment.

## Standard objections addressed in this essay

- “You smuggle in survival as a value.” — [§I](#i-one-derivation) (Persistence is a conditional constraint; adopting it as telos remains explicit.)
- “Values are not capital stocks.” — [§II](#ii-value-ecologies-build-and-consume-capital-stocks) (Correct in the literal sense: the value ecology is the selection process that builds and consumes several capital stocks at once, not a stock itself.)
- “Measurement choices are themselves value choices.” — [§III](#iii-the-alignment-failure-values-in-the-feedback-loop), [§IV](#iv-the-solution-measure-what-commitments-produce)(Yes; scope, observables, and authority must remain public and contestable.)
- “You cannot weight trust against infrastructure.” — [§V](#v-the-weighting-problem-dissolves) (The model can preserve vectors, thresholds, dominance, and unresolved trade-offs.)
- “Measurement alone does not align anything.” — [§IV](#iv-the-solution-measure-what-commitments-produce) (Measurement needs a response duty, owner, and correction path.)
- “This is consequentialism that can erase rights.” — [§VI](#vi-consequence-accounting) (The commitment must state rights and standing as constraints, not hidden weights.)

---

## I. The Synthesis {#i-one-derivation}

Two prior essays establish the foundation. [Some value complexes behave like replicators](values-are-replicators.md): they transmit, fuse with identity, and compete for substrate in fitness landscapes we often fail to monitor—some symbiotic, some parasitic, testable by observable criteria. [Values are not one kind of thing](values-arent-subjective.md): the word "value" conflates preference reports, causal effects, coordination performance, host-relative viability, standing, and explicit commitments — and the conflation itself can function as a defensive replicator, shielding the whole stack from evaluation.

This essay adds two claims the prior essays do not make on their own:

1.  The value ecology is a **selection process** that builds and consumes several capital stocks at once — load-bearing, measurable, currently booked at zero.
2.  Alignment— in governance and AI—needs a **balance sheet and a correction path**, not installation of supposedly correct values from above.

Persistence remains a conditional constraint: a system that does not persist evaluates nothing. If a polity adopts long-run persistence as its telos, it must account for the conditions that sustain it. That adoption stays explicit and contestable; measurement does not smuggle in terminal values by itself.

---

## II. Value Ecologies Build and Consume Capital Stocks {#ii-value-ecologies-build-and-consume-capital-stocks}

The ontological demotion of values — from sacred axioms to replicator ecology — has a constructive side that the prior essays left undeveloped.

Values produce **downstream effects**. A population that shares high-trust norms can coordinate without expensive enforcement. A population with strong reciprocity norms generates social capital that compounds through use. A population whose value ecology includes truth-tracking and long-term thinking maintains institutional competence across generations.

These aren't vague philosophical observations. They're measurable:

- **Trust** — World Values Survey, Edelman Trust Barometer, behavioral experiments
- **Institutional legitimacy** — compliance rates, corruption indices, tax morale
- **Social cohesion** — Putnam's social capital metrics, community participation, associational density
- **Epistemic capacity** — prediction accuracy (Tetlock), policy feedback latency, institutional learning rate

The measurements are imperfect. They're also infinitely better than zero — which is what we currently have. [Civilization runs on a dozen capital stocks. Only one has a ledger.](full-accounting.md) Moral capital, social capital, institutional capital, epistemic capital — all booked at zero. We're consuming them and calling it income.

The value ecology is not itself one of these capital stocks, and treating it as one is the error this section corrects. It is the **selection process** that determines which replicators propagate through a population's institutions — and that process builds or consumes trust, institutional legitimacy, social cohesion, and epistemic capacity simultaneously, as correlated effects of which values it selects for. A value ecology dominated by present-optimizing, truth-indifferent, identity-protective replicators erodes trust, degrades institutions, and consumes the other capital stocks from the inside, all at once, because the same selection pressure touches all of them. You can see the damage on the balance sheet if you have a balance sheet. We don't.

---

## III. The Alignment Failure: Gaming, Drift, and Selection {#iii-the-alignment-failure-values-in-the-feedback-loop}

Installing "correct values" directly into a feedback loop invites gaming, drift, and selection for what spreads—not for what works.

**In AI:** RLHF trains systems to match human feedback signals. Under optimization pressure, stated constraints become targets to game. [Constraints encoded as penalties in reward functions are optimization targets, not enforcement boundaries.](privilege-separation-ai-safety.md)

**In governance:** Electoral systems reward transmissibility. [Democratic preference aggregation measures which value-package spreads in the current population](vote-on-values.md)—not which produces durable capacity. Policies that sound virtuous can outcompete policies that preserve margin.

**In both domains:** values in the loop can become identity-protective. Counter-evidence triggers the [immune system](replicator-immune-system.md) before it reaches evaluation. The replicator persists; the host may not.

[Disposition training alone](ethics-is-an-engineering-problem.md) cannot be the sole implementation control: in misaligned systems, capable agents still burn out, get corrupted, or get outcompeted. The engineering layer—measurement, stated commitments, correction path—is the complementary intervention.

---

## IV. The Solution: Measure What Commitments Produce {#iv-the-solution-measure-what-commitments-produce}

The governor does not evade values by measuring. It adopts explicit commitments: the scope of the stocks it observes, the observables and models it publishes, rights and standing as constraints, and a bounded authority to issue findings and trigger a response. Each commitment must remain public and contestable. It measures within that stated mandate.

A steam engine governor doesn't want the engine to run at the right speed. It doesn't have preferences about RPM. It measures the current state and applies a mechanical correction. The correction is architectural — built into the physics of the mechanism — not aspirational.

The civilizational equivalent: a constitutional measurement institution — a [Fourth Branch](fourth-branch.md) — that monitors capital stocks, measures whether policies achieve their stated objectives, and tracks their effects on civilizational flourishing. Specific, falsifiable claims:

- This law was passed to reduce homelessness by 30% in five years. Year three: homelessness is up 12%. Here's the data.
- This policy claims to increase innovation output. Five years later: patent rates, startup formation, R&D investment — all down. Here's the data.
- This institutional reform was justified as improving trust. Measured trust: declined. Here's the data.

This is not [Gosplan](https://en.wikipedia.org/wiki/Gosplan). Gosplan planned outcomes — what to produce, how much, where — central planning that failed because local information cannot scale to central planners. The governor operates at a different abstraction layer: it models mechanisms, predicts what incentive structures will produce, and audits whether policies achieve their stated goals. It plans at the mechanism level — "this structure will generate these dynamics" — not at the output level. It takes what the political system decided to do, measures whether it's working, and publishes the results.

Any such mechanism needs an owner and a correction path. A published finding assigns the responsible institution a correct-or-explain duty: it must state the correction, give reasons for declining it, or contest the model in public. Parliament retains final authority; no supermajority or assessor veto follows from a finding.

And here's the deepest point: [Knightian uncertainty](flourishing-is-maximum-safety-margin.md) — the radical unpredictability of future shocks — doesn't undermine this approach. It *demands* it. Precisely because we cannot predict complex system behavior from theory, we need rapid empirical feedback on what's actually happening. The governor is the instrument panel. You don't fly blind because the weather is unpredictable — that's when you need instruments most.

---

## V. The Weighting Problem {#v-the-weighting-problem-dissolves}

The standard objection: "You have twelve capital stocks. How do you weight trust against infrastructure? How many units of social cohesion equal one unit of fiscal sustainability? Don't you need preference aggregation to set the exchange rate?"

Not always. A shared telos can identify some dominated options and threshold breaches, but it does not produce a universal exchange rate between heterogeneous stocks.

Under [telocracy](telocracy.md), an explicitly adopted telos can organize the inquiry around survival and flourishing over deep time. The model should retain capital stocks as vectors, identify thresholds and partial orderings, and mark unresolved trade-offs for political closure rather than conceal them in weights.

You don't "weight" blood pressure against liver function. You don't need an exchange rate between kidneys and lungs. You check whether any organ is at a level that threatens the organism's survival. If blood pressure is critical, you fix blood pressure. If the liver is failing, you fix the liver. The priority is determined by proximity to failure, not by preference aggregation.

This is triage where a threshold is binding, not a general solution to tradeoff. [Liebig's Law](flourishing-is-maximum-safety-margin.md#model-sketch) identifies a limiting factor when a stock approaches failure; it does not settle every allocation among non-critical stocks. The governor can report dominance, thresholds, uncertainty, and unresolved conflicts; legitimate political institutions close the residual choice.

Most apparent "tradeoffs" between capital stocks are temporal illusions. Trust vs innovation looks like a real tradeoff in the short term (innovation disrupts, disruption reduces trust). Over generational timescale, both decline if you sacrifice either — a civilization without innovation stagnates and dies; a civilization without trust fragments and dies. The telos resolves the apparent conflict by extending the time horizon past the illusion.

---

## VI. Consequence Accounting {#vi-consequence-accounting}

The obvious objection: "You just invented consequentialism. The governor evaluates outcomes."

Partially—but the claim is narrower. Every institution already produces consequences; most hide them behind opaque value language. The proposal is **explicit consequence accounting** under stated commitments: which stocks are observed, which rights and standing constraints bind, and what happens when measured effects diverge from stated objectives.

That mandate cannot erase rights or standing by treating them as hidden weights. A legitimate charter states whose claims constrain the system, which rights are non-negotiable, and which decisions remain outside the assessor’s authority.

- **Current system:** Thousands of implicit values operating invisibly, competing memetically, shielded from evaluation by "values are subjective," producing opaque consequences that take generations to manifest.
- **Proposed system:** Explicit commitments, transparent measurement across capital stocks, public findings, and a correct-or-explain response duty. Parliament retains residual closure. Stated constraints, not a total ethical reduction.

The difference is measured versus unmeasured consequence accounting—with a correction path when the ecology diverges from its stated commitments.

---

## VII. The Ecology-Aware Governor {#vii-the-ecology-aware-governor}

Put it together.

The governor does not install or rank a complete value system. It operates under stated commitments that delimit what it observes, what it may report, and what it must leave to political closure.

It monitors the value ecology's downstream effects on the civilizational balance sheet. Values that produce trust, cohesion, and institutional competence show up as healthy capital stocks. Values that produce polarization, institutional decay, and demographic collapse show up as depleting capital stocks. The data is published. The political system decides what to do about it — but it decides with a balance sheet, not with vibes.

Destructive value ecologies can persist for generations before capacity collapse becomes visible. Natural selection eventually removes hosts that destroy themselves—but "eventually" can mean after irreversible decline. The governor shortens the feedback loop: consequences visible in years, not generations, early enough to correct while capacity remains.

But measurement is only half the function. The other half is **cultivation** — shaping the selection environment so that reality-compatible values outcompete reality-denying ones. Publishing "this policy failed its stated objectives" doesn't tell the ecology which values to hold. It changes the fitness landscape: values that depend on unmeasured opacity lose their competitive advantage. Truth-tracking values gain fitness when measurement exists. The governor is a selection pressure, not a plan.

This is the same logic as [free agents coordinated by architecture rather than commands](laws-are-the-wrong-abstraction.md). You cannot fully specify a value ecology any more than you can fully specify an economy or an ecosystem. The system is too complex, the local information too distributed, the emergent dynamics too unpredictable. Central planning of values fails for the same reason central planning of production fails. What you can do is set boundary conditions and selection pressures, then let the ecology self-organize within those constraints. The governor shapes the soil. What grows is emergence.

This is one alignment architecture for AI and governance: not better values installed from above, but better measurement of what values produce, stated commitments, and selection environments where measurement has consequences.

The value ecology is real. It has causal force. It is the selection process that determines which capital stocks compound and which deplete, across all of them at once. And like the capital stocks it moves, it needs a ledger. Not to control it — to see it. The rest follows from [calculation](calculemus.md).

## Synthesis {#synthesis}

**The argument in six sentences:** Values form ecologies, not isolated axioms. Those ecologies are load-bearing capital stocks with measurable downstream effects. Installing values in the feedback loop invites gaming, drift, and selection for transmissibility. Alignment needs a public balance sheet of what the ecology produces, stated commitments, and a correction path when results diverge. Thresholds and dominance can guide triage; unresolved trade-offs remain for political closure. Rights and standing stay explicit constraints—not hidden weights.

---

**Related reading:**

- [Values Are Replicators](values-are-replicators.md) — The class claim: five observable properties, four levels, benign/symbiotic/parasitic distinction
- [Values Aren't Subjective](values-arent-subjective.md) — The six-type taxonomy of value claims and the viable-region framing
- [Flourishing Is Maximum Safety Margin](flourishing-is-maximum-safety-margin.md) — Why bare viability is not a stable civilizational target
- [The Fourth Branch](fourth-branch.md) — The missing constitutional measurement layer
- [Calculemus](calculemus.md) — Most policy disagreements are uncomputed empirical disagreements
- [Full Accounting](full-accounting.md) — Twelve capital stocks, one ledger
- [Ethics Is an Engineering Problem](ethics-is-an-engineering-problem.md) — Moral justification and implementation reliability are separable; the engineering layer needs standing, error correction, and amendment.
- [The Governance Alignment Problem](governance-alignment-problem.md) — The job selects for election-winning, not outcome-producing
- [What "Vote on Values" Actually Does](vote-on-values.md) — Democratic preference aggregation measures transmissibility, not adaptiveness
- [The Privilege Separation Principle](privilege-separation-ai-safety.md) — 3-layer architecture for AI safety

## Sources and Notes

**Value ecology and replicator dynamics:** Richard Dawkins, *The Selfish Gene* (1976) — foundational replicator framework. Joseph Henrich, *The Secret of Our Success* (2015) — cultural evolution as selection on memes. Robert Boyd and Peter Richerson, *Culture and the Evolutionary Process* (1985) — gene-culture coevolution. The extension from memetics to "values as capital stocks" is original to this framework.

**Capital stock measurement:** Robert Putnam, *Bowling Alone* (2000) — social capital measurement and decline. Shalom Schwartz, "Universals in the content and structure of values" (1992) — cross-cultural value measurement. World Values Survey — longitudinal trust and value data across 100+ countries. Transparency International Corruption Perceptions Index — institutional legitimacy proxy.

**Alignment failure:** Philip Tetlock, *Expert Political Judgment* (2005) — systematic prediction failure in value-laden domains. Dan Kahan, "Ideology, Motivated Reasoning, and Cognitive Reflection" (2013) — identity-protective cognition as mechanism. The parallel between RLHF failure and democratic preference aggregation failure is developed in the *Governance Alignment Problem* and *Privilege Separation Principle* essays (both linked in the Related reading list above).

**Survival–flourishing overlap:** Value ecology changes the margin stocks described in [Flourishing Is Maximum Safety Margin](flourishing-is-maximum-safety-margin.md#model-sketch) — the channels below are how commitments raise or burn those stocks.

**The weighting problem:** Kenneth Arrow, *Social Choice and Individual Values* (1951) — impossibility of rational preference aggregation. The dissolution argument (telocracy removes the need for preference aggregation) is original to this framework. The triage framing draws on emergency medicine's established protocol for resource allocation under constraint.

**Civilizational capital stocks:** Joseph Tainter, *The Collapse of Complex Societies* (1988) — diminishing returns on complexity. The "twelve capital stocks, one ledger" framing is developed in *Full Accounting* (linked in the Related reading list above).

---

*The value ecology: [Values Are Replicators](values-are-replicators.md). The claim-type taxonomy: [Values Aren't Subjective](values-arent-subjective.md). Adaptive margin at civilizational scale: [Flourishing Is Maximum Safety Margin](flourishing-is-maximum-safety-margin.md).*