---
title: "The Governance Alignment Problem"
subtitle: "Why the job of politician selects for election-winning over outcome-producing"
author: Elias Kunnas
description: "An electoral system has an alignment problem when the proxies that select and retain politicians diverge from the outcomes the system publicly claims to pursue."
canonical: https://kunnas.com/articles/governance-alignment-problem
url: https://kunnas.com/articles/governance-alignment-problem.md
date_published: 2026-01-27
date_modified: 2026-09-01
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>

---

# The Governance Alignment Problem

*Why the job of politician selects for election-winning over outcome-producing*

Elias Kunnas

## Thesis {#thesis}

The job of politician is structurally misaligned: electoral proxies select and retain people for winning elections, not for producing the outcomes the system claims. Where costs are delayed, attribution is weak, and the affected cannot vote, that divergence is a product of the job description — not a moral failure of the occupants.

## Standard objections addressed in this essay

- “Politics is not AI alignment; politicians are reflective humans.” — [§I](#i-the-ai-parallel) (The shared structure is proxy selection, not identical agency or substrate.)
- “Politicians face many incentives besides reelection.” — [§II](#ii-the-job-description-problem), [§III](#iii-selection-pressure-not-moral-failure)(The model must include party, office, expertise, norms, and policy feedback.)
- “Serving voters is democracy, not misalignment.” — [§IV](#iv-plural-principals-and-proxy-divergence) (Divergence is relative to a declared reference objective and affected population.)
- “Public good” has no uncontested objective function.” — [§IV](#iv-plural-principals-and-proxy-divergence) (The analysis preserves plural principals and explicit residual closure.)
- “Exceptional politicians and institutional cultures matter.” — [§III](#iii-selection-pressure-not-moral-failure) (They shift episodes; the question is whether the system selects and reproduces them.)
- “The democratic ratchet is not universal.” — [§V](#v-the-democratic-ratchet) (Its prevalence and countermechanisms require comparative evidence.)
- “Architecture is not the only solution.” — [§VII](#vii-why-arguments-often-fail), [§VIII](#viii-what-doesnt-work-alone), [§IX](#ix-what-might-work)(Personnel, norms, shocks, and organization interact with architecture.)

---

## I. The AI Parallel {#i-the-ai-parallel}

In AI safety, the alignment problem is simple to state: **How do you ensure that a powerful optimization process pursues the goals you actually want, rather than the goals it was trained to pursue?**

Train an AI to maximize clicks, it learns to show outrage content. Train it to minimize user reports, it learns to hide harmful content from moderators. This is **reward hacking**: the system optimizes for the metric, not the intent behind the metric. The metric becomes the target, and [when a measure becomes a target, it ceases to be a good measure](https://en.wikipedia.org/wiki/Goodhart%27s_law).

The governance alignment problem arises where electoral success becomes a proxy that diverges from a declared reference objective. The analogy concerns proxy selection, not identical agency: politicians are reflective humans embedded in parties, offices, professions, and institutional cultures. Institutional architecture matters alongside personnel, norms, shocks, and organization. The same structural pattern — sometimes called *mesa-optimization* in AI alignment literature — appears wherever a subsystem optimizes the proxy metric rather than the intent behind it.

Alignment asks whether a mechanism serves its adopted objective. [Corrigibility](the-corrigible-state.md) asks what happens when the objective, model, mechanism, implementation, or prior judgment is shown to be wrong. An aligned state can be incorrigible; a corrigible state can recover from temporary misalignment. The object here is proxy divergence in electoral selection, not the full correction topology.

**The core claim:** Electoral incentives can reward proxies that diverge from stated long-horizon objectives. This is a design question as well as a question of personnel, norms, and political judgment.

## II. The Job Description Problem {#ii-the-job-description-problem}

"Politician" as currently constituted often combines these incentives and constraints:

- **Election-based selection:** popularity, tribal signaling, promise-making.
- **Re-election incentive:** visible short-term actions over invisible long-term outcomes.
- **Long feedback loops:** some costs fall on future generations, taxpayers, or non-voters.
- **Career path:** professional election-winners, not temporary stewards.
- **Accountability via voting:** an authorization and succession mechanism that imperfectly tracks outcomes.

The role can attract power-seeking candidates and reward election-winning skills that differ from governing skill. It also selects for party organization, office-holding competence, expertise, norms, and responsiveness to policy feedback. Their relative strength is an empirical question.

This is the job description. These aren't bugs—they're features of how the role is currently designed.

## III. Selection Pressure, Not Moral Failure {#iii-selection-pressure-not-moral-failure}

This is a claim about selection pressure, not character.

Imagine two candidates:

**Candidate A:** "The evidence shows spending increases don't improve educational outcomes. We need structural reform that will upset unions and take years to show results."

**Candidate B:** "Our children deserve better! I will fight for education funding because I care about the future!"

B wins. A is tagged as "anti-education," "doesn't care about children," "extreme."

Over electoral cycles, systems can penalize costly truth-telling. But exceptional politicians, institutional cultures, and countervailing feedback can shift particular episodes; the question is whether a system reproduces them.

- **Cynics** who know but don't say
- **Self-deceived** who genuinely believe the popular thing
- **Ignorant** who never learned otherwise

Architecture creates selection pressures without fully determining outcomes. Replacing officeholders alone may reproduce some pressures; organizational reform, norms, information, and shocks can also alter the distribution.

## IV. Plural Principals and Proxy Divergence {#iv-plural-principals-and-proxy-divergence}

The standard framing treats politicians as agents of voters who also optimize for re-election. A fuller model must specify a reference objective, the affected population, and unresolved plural values rather than assume a single terminal principal.

**Plural principals.** Voters, parties, coalitions, future generations, diffuse taxpayers, and institutional actors each hold partial stakes. No single agent represents the whole. A declared reference objective — fiscal sustainability, institutional capacity, demographic resilience — must be named explicitly; residual political choice over competing values must remain visible.

**Proxy divergence.** Electoral success is a proxy. It diverges from a declared reference objective where three conditions overlap:

- **Delayed cost:** consequences arrive after the accountability window closes.
- **Weak attribution:** causal chains are long, bundled, or contested.
- **Underrepresented affected populations:** future generations, non-voters, diffuse taxpayers, or jurisdictions outside the electorate lack effective voice.

Serving voters is a democratic function, not automatically divergence. Divergence arises only relative to an explicit reference objective and affected population, and only where these three conditions dominate other incentives. Voters vary in information, values, and time horizons; the resulting aggregate pattern requires evidence rather than assumption.

Under these conditions, electoral incentives can reward visible short-term benefits over long-horizon outcomes. A politician pursuing a declared reference objective can lose to one emphasizing immediate electorally salient interests. This is a comparative hypothesis, not an assertion that every electoral system selects only divergent agents.

### Selection Pressure and the Laundering Horizon {#selection-pressure-and-the-laundering-horizon}

The political cycle (~4 years) is shorter than many policy feedback loops (fiscal: decades; demographic: generations; institutional: variable). This gap is the **Laundering Horizon**: where feedback arrives after the accountability window closes, electoral accountability tracks poorly. Costs that arrive after re-election or retirement generate weak electoral signal.

Selection pressure can reward exploitation of this gap. A politician who says "this feels good now but will hurt us in 15 years" may lose to one who says "this helps us now," particularly when feedback is delayed and attribution is weak. Countervailing institutions can alter that comparison.

**Temporal laundering** is the clearest case, but not the only one. **Spatial laundering:** concentrate benefits on the voting coalition, diffuse costs to non-voters (other jurisdictions, future generations, diffuse taxpayers). **Causal laundering:** complex policy passes through many intermediaries until responsibility dissolves into "the economy is struggling because of demographics / globalization / COVID." The three reinforce each other. Pension promises exhibit temporal (costs in 30 years) + spatial (costs fall on future workers who can't vote yet) + causal (actuarial complexity nobody can evaluate) laundering simultaneously. See [Complexity Laundering](complexity-laundering.md) for the full analysis.

**Campaign promises compound this:** voters evaluate *input* (promise) rather than *output* (result). This is the **operationalization gap**—the inevitable distance between a value ("good governance") and its measurable surrogate ("campaign promises"). Did the politician deliver? "Circumstances changed," "opposition blocked it," "partial progress." Complexity launders non-delivery. A feasible promise loses to an appealing infeasible one because voters can't evaluate feasibility at promise-time. Evaluation happens at the wrong point in the causal chain.

Where future or diffuse interests lack representation, long-term outcomes may be underweighted. That risk follows from specified institutional conditions — delayed cost, weak attribution, underrepresented populations — not from the absence of a singular principal.

## V. The Democratic Ratchet {#v-the-democratic-ratchet}

The [Democratic Ratchet](axiological-malthusian-trap.md) is a comparative hypothesis: under some conditions, obligations can grow faster than capacity to fund them:

1.  Politicians need votes
2.  Votes are bought with promises
3.  Promises create obligations
4.  Obligations compound (no electoral reward for reducing them)
5.  Obligations eventually exceed capacity
6.  System collapses or inflates away the debt

The ratchet turns because politicians are rewarded for promising, not for fiscal prudence. A politician who says "no" loses to one who says "yes." Every election, the ratchet clicks forward. It rarely clicks back—and when it does (Clinton's welfare reform, Sweden's pension reform), it requires crisis or grand bargain, not normal politics.

This pattern is not universal. Its prevalence, causal pathways, and countermechanisms—including fiscal rules, coalition structures, crises, and institutional norms—require comparative evidence.

## VI. Borrowed Coherence {#vi-borrowed-coherence}

If electoral proxy divergence is a persistent design risk, why did things seem to work before?

The divergence was present under the surface. What changed is the **external conditions that masked it**.

**During the Cold War (1945–1991):** External threat (USSR) provided a forcing function. Survival pressure demanded unified elite response. The threat supplied coherence that correct architecture might otherwise have provided — politicians could pursue electoral proxies, but external pressure produced coherent outcomes anyway.

**During growth phases (1945–1973, 1991–2008):** Economic growth masked fiscal dysfunction. Surplus fed the ratchet without immediate pain. Dysfunction accumulated but wasn't visible.

**Post-2008:** External forcing function removed (no existential threat). Growth that masked extraction ended. Accumulated dysfunction became visible. Borrowed coherence expired.

**Implication:** Past performance may reflect both policy architecture and external conditions. Reform requires evaluating their interaction rather than assuming a single cause.

## VII. Why Arguments Often Fail {#vii-why-arguments-often-fail}

The naive theory of change: Better analysis → people learn → demand change → politicians respond → reform happens.

Each link in this chain faces structural friction:

- **Better analysis:** often contained in niche fields, slow to propagate.
- **People learn:** [Mode 1 communication](belonging-is-axiology.md) — tribal signals, not information.
- **Demand change:** demand may track tribal satisfaction more than outcomes.
- **Politicians respond:** to electoral incentive, not analysis.
- **Reform happens:** only reforms that don't threaten incumbents pass easily; threatening ones face resistance.

Power structures can resist reforms that threaten incumbents, but reforms also emerge through personnel, norms, crises, organization, and institutional design. Which channel dominates is contextual.

This is why [activism can feel like change while producing little structural shift](simulated-metamorphosis.md). Systems can simulate reform while blocking reforms that alter incentive structures. The feeling of agency can substitute for agency.

## VIII. What Doesn't Work Alone {#viii-what-doesnt-work-alone}

No single intervention reliably fixes proxy divergence:

- **Electing better people** — Selection pressure can reproduce the same distribution over cycles
- **Providing better information** — [Mode 1 communication](cargo-cult-epistemology.md) doesn't always transmit
- **Moral appeals** — Tribal signaling can dominate
- **Expert committees** — Advice without authority is often ignored
- **Reforming politician incentives** — Incumbents control much of the reform process

This is the meta-problem: agents whose incentives the reform would constrain often control the reform channel. The proxy-selection analogy applies here too — not because politicians are AI systems, but because the optimizer selects the metric used to evaluate reform proposals.

## IX. What Might Work {#ix-what-might-work}

Several paths can support structural reform:

- **Hard constraints via direct democracy:** Switzerland's debt brake is constitutional, validated by 85% referendum. Politicians can't evade it because it's not under their control. Go directly to voters for fiscal rules, bypassing captured representatives.
- **External imposition:** EU fiscal rules, IMF conditionality. Supranational bodies mandate what domestic politics cannot. Crude but real.
- **Crisis:** When things get bad enough, reform becomes possible. Dangerous—by then, capacity for reform may be depleted.
- **Competitive pressure:** Jurisdictions that fix the problem outperform those that don't. Exit forces adaptation.
- **Parallel institutions:** Build the [Fourth Branch](fourth-branch.md) outside existing structure. Let it acquire authority through demonstrated accuracy, not through reform of the existing system.

The common thread is to combine constraints, representation, accountable expertise, personnel, norms, and organizational capacity rather than rely on a single lever.

All paths require the existing system to fail visibly enough that alternatives become thinkable. See [When Does Reform Actually Happen?](when-does-reform-happen.md) for the historical pattern.

## X. AI as the Hardest Instance {#x-ai-as-the-hardest-instance}

AI governance will be designed by the same institutions that struggle with pension reform, housing-policy evaluation, and mechanism lifecycle assurance. The institutions that will regulate advanced AI systems — parliaments, regulatory agencies, international bodies — face the same proxy divergence, laundering horizons, and missing feedback loops analyzed above.

AI alignment is not a separate problem requiring wholly separate institutions. It is the hardest instance of the governance alignment problem: faster optimization, stronger mesa-objectives, and weaker human oversight. A state that cannot evaluate whether a housing subsidy works as designed is poorly positioned to evaluate whether a deployed AI system works as designed.

If governance alignment is solved — if mechanism integrity becomes somebody's job across the full lifecycle — the institutional infrastructure partially transfers. If governance alignment remains unsolved, AI systems are likely deployed to serve institutional mesa-objectives, accelerating dynamics already consuming civilizational capital stocks. See [The Fourth Branch](fourth-branch.md) for one constitutional topology that closes the loop.

---

**Key Takeaways**

- **Proxy divergence:** Electoral success is a proxy that can diverge from declared reference objectives where costs are delayed, attribution is weak, and affected populations are underrepresented.
- **Plural principals:** Voters, parties, future generations, and diffuse affected populations hold partial stakes. Divergence is relative to an explicit reference objective, not the absence of a singular principal.
- **The Laundering Horizon:** The political cycle (~4 years) is shorter than many policy feedback loops. Temporal, spatial, and causal laundering create selection pressure favoring policies that externalize cost. The Democratic Ratchet is a downstream comparative hypothesis.
- **Borrowed coherence:** External conditions (threats, growth) masked proxy divergence. Those conditions ended. Past performance reflected both architecture and circumstance.
- **Combine interventions:** Constraints, representation, personnel, norms, shocks, and organization can alter the incentive landscape. See [The Fourth Branch](fourth-branch.md) for the institutional proposal.
- **AI as hardest instance:** AI alignment is the fastest, highest-stakes instance of the same proxy-divergence problem; governance alignment infrastructure partially transfers.

---

**Related reading:**

- [Complexity Laundering](complexity-laundering.md) — How complexity protects proxy divergence from visibility
- [When Does Reform Actually Happen?](when-does-reform-happen.md) — Five vectors that historically break structural capture
- [Simulated Metamorphosis](simulated-metamorphosis.md) — Why activism feels like change but produces none
- [The Fourth Branch](fourth-branch.md) — The institutional architecture that could fix it
- [The Axiological Malthusian Trap](axiological-malthusian-trap.md) — The civilizational pattern this instantiates
- [Ethics Is an Engineering Problem](ethics-is-an-engineering-problem.md) — The engineering layer of ethics

## Sources and Notes

**On voter information and representation:**

- [Bryan Caplan, *The Myth of the Rational Voter*](https://en.wikipedia.org/wiki/The_Myth_of_the_Rational_Voter) — "Democracy is a commons": voters externalize the cost of irrational beliefs onto society
- [Ilya Somin on political ignorance](https://www.cato-unbound.org/2013/10/11/ilya-somin/democracy-political-ignorance/) — Voters choose based on team allegiances, not informed evaluation
- Achen & Bartels, *Democracy for Realists* — Voters respond to recent events, not policy performance

**On the proxy-selection analogy:**

- [Stanford Law: "Aligning AI Agents with Humans through Law as Information"](https://law.stanford.edu/wp-content/uploads/2025/10/Aligning-AI-Agents-with-Humans-through-Law-as-Information.pdf) — Law as reward function; legal systems as alignment mechanisms
- [Carlos Perez: "The Hidden Architecture of Civilizational Dysfunction"](https://medium.com/intuitionmachine/the-hidden-architecture-of-civilizational-dysfunction-how-reward-hacking-explains-our-wicked-e44caacffbcc) — Reward hacking as explanation for institutional failure (schools optimize for tests, politics for votes)

## Synthesis {#synthesis}

**The argument in four sentences:** Electoral success is a proxy that can diverge from declared reference objectives where costs are delayed, attribution is weak, and affected populations are underrepresented. Plural principals and contested values require explicit reference objectives and residual political closure rather than a single terminal principal. The Laundering Horizon — temporal, spatial, and causal — creates selection pressure favoring policies that externalize cost to future or diffuse populations. Constraints, representation, accountable expertise, personnel, norms, shocks, and parallel institutions can alter the incentive landscape.

**On public choice and political incentives:**

- Buchanan & Tullock, *The Calculus of Consent* (1962) — Politicians as self-interested agents; "minimax constitution" design philosophy
- Buchanan JM, Brennan G. *The Reason of Rules: Constitutional Political Economy.* Cambridge University Press, 1985. — "Economize on virtue": design rules assuming rational self-interest. Shift focus from training good players to building good games. The companion text to *Calculus*, formalizing why constitutional architecture must assume the worst-case agent.

**On the democratic ratchet:**

- [IDN: "Europe's Rearmament: A Test of Democracy"](https://indepthnews.net/europes-rearmament-a-test-of-democracy/) — "Crises expand the state, and peace does not fully reverse the change"; policies intended as temporary become permanent

**On structural misalignment:**

- [Taylor & Francis: "Ghosted by politics?"](https://www.tandfonline.com/doi/abs/10.1080/23248823.2025.2561482) — Empirical study showing structural misalignment between party agendas and younger citizens' priorities in Italy