The Exit Asymmetry: Position, Aeon, and the Institutional Design of Intertemporal Allocation

Cluster: position — aeon — republic — automation-tech — allocation

Extends: 051-pension-rights-stratification-uncertainty-meaning.md (the pension as the canonical intertemporal transfer; underfunding as a deferral mechanism), 041-threshold-archive-subsistence-absolutism-distribution.md (constraint architecture, distribution as the governed layer), 074-operationalization-gap-abstraction-pluralism-war-poetry.md (the operationalization gap — here, the social discount rate as the contested operationalization), 080-labyrinth-proof-aesthetic-disinformation-framing.md (the encoded rule as a frame invisible to formalist audit — here, the algorithm as dead-hand policy), 115-emergence-commons-tragedy-profit-aging-witness.md (the intertemporal commons, aging as horizon-compression), 1285F-outsourced-audit-veto-points-commitment-devices-ostrom.md (commitment devices, veto points, Ostrom), 1280-redemption-journey-armistice-legislature-positions.md (positions as the legislature’s currency), 1253-institutions-are-memories-deregulation-amnesia.md (institution as continuant, the perpetual principal)

Core Claim

A position is an exitable slot. Whoever holds an office, a portfolio, a managerial seat, or a financial stake can — by term limit, retirement, resignation, sale, or death — close the position and walk away before the long-horizon consequences of their in-position decisions arrive. The aeon is the horizon over which those consequences actually land: the perpetual life of the institution, the welfare of cohorts not yet born, the depletion or accumulation of capital stocks that mature over decades. The principal whose interest spans the aeon — the res publica, the trust corpus, the future cohort — cannot exit. It is captive to the territory, the balance sheet, the climate, the encoded rule.

This is the structural defect, and it is sharper than the standard agency problem. Ordinary moral hazard assumes the principal can observe the agent and sanction misconduct. Here the principal whose interest is at stake does not yet exist (future cohorts) or has no standing to act (the institution-as-continuant), and the consequences of the agent’s choices are realized after the agent has exited the position — i.e., after the only contract that could have disciplined them has terminated. Pay-for-performance, the canonical fix, fails by construction: performance is observed when the agent is already gone.

The narrow claim: the position–aeon problem is not a problem of bad values, short-sightedness, or insufficient norms of stewardship. It is an exit-asymmetry problem — the agent can close the position; the principal cannot close the aeon — and it is solvable only by institutional rules that deny, delay, or price the agent’s exit so that the agent’s effective horizon is dragged toward the principal’s captivity. Every workable mechanism below is a species of one move: convert the position from an exitable claim into a partially non-exitable one, and manufacture a present proxy with standing for the absent principal. The republic is the historical name for the institutional layering that does this; automation-tech is the variable that both supercharges the agent’s reach into the aeon and tempts the republic to collapse the very layering that solved the problem.

I. The Temporal Agency Model

Make the misalignment precise, because “short-termism” as a moral complaint is useless for design.

Let the institution generate a consumption/extraction stream c(t) over time. Two valuations of the same stream:

The aeon-principal values ∫ u(c(t)) · e^(−r_a · t) dt over an unbounded horizon, with r_a ≈ 0 (a perpetual entity has no reason to prefer its own decade to a later one).
The position-holder, with tenure [0, τ], values ∫₀^τ u(c(t)) dt + V(τ), where V(τ) is whatever reputational or re-selection continuation value survives term-end. Beyond τ, the agent’s marginal valuation of consequences drops discontinuously toward zero. The position imposes an effective discount rate r_p ≫ r_a, set not by the agent’s character but by the duration of the slot.

The agent maximizing their own objective will:

Harvest — front-load c(t) into [0, τ] (realize gains in-term: the dividend, the headline metric, the ribbon-cutting, the tax cut).
Defer — push costs to t > τ (the deferred maintenance, the unfunded pension liability, the rolled-over debt, the drawn-down aquifer, the technical debt of an encoded rule).

The moral hazard is not shirked effort. It is intertemporal cost-shifting, and it is uncontractable by ordinary means precisely because its signature appears after the contract ends. The whole design problem reduces to: how do you write a contract on c(t) for t > τ with a counterparty who has already left?

The terminal-period theorem

Any finite-horizon position has a last period in which future-discipline is zero. By backward induction, the cooperation that V(τ) sustained in earlier periods unravels: a term-limited official in their final term, a CEO weeks from retirement, a trustee about to hand off — each faces a final round with no shadow of the future. The republic’s standard prophylaxis against entrenchment (rotation, term limits) manufactures terminal periods on a schedule. Rotation is, by itself, a harvest-amplifier. Holding this tension is the rest of the analysis.

II. The Four Failure Modes, as Specific Defects

#	Failure mode	What the agent does	Which exit it exploits
1	Terminal harvesting	Extracts maximally in the last period when `V(τ)→0`	Exit by term-end / retirement
2	Horizon cost-shifting	Defers costs past `τ`; books only in-horizon accounts	Exit before consequences land
3	Patrimonialization	Converts a delegated office into owned, heritable property (prebend, sinecure)	Attacks the rotation that would force exit
4	Dead-hand lock-in	Binds all future holders by irreversible commitment (entrenchment, perpetual contract, encoded rule)	Denies the next holder’s entry; the agent never really exits

Note that 3 and 4 are mirror-images of 1 and 2. Harvesting and deferral exploit the position’s exitability; patrimonialization and lock-in attack the position’s temporariness from the other side — the agent who cannot harvest enough in one term instead makes the term permanent (3) or projects one term’s politics across the aeon (4). A design that only fences harvesting (lengthen accountability) invites entrenchment; a design that only fences entrenchment (force rotation) invites harvesting. The mechanisms must address both vectors at once.

III. The Mechanism Stack

Organized by the move each rule makes against the exit asymmetry.

A. Re-couple the agent’s horizon to the aeon — deny/delay financial exit

M1 — Deferred, clawback-able compensation indexed to realized long-horizon outcomes. Rule: a fixed fraction φ of the position-holder’s total compensation (including pension and post-tenure benefits) converts into a claim that vests only T years after they leave the position and is reduced, pro rata, by the realized long-horizon performance of decisions made in-tenure. Concretely: a finance minister’s pension is partly indexed to the realized 20-year fiscal trajectory; an infrastructure director’s deferred tranche is written down by the realized failure rate of the assets they commissioned; a regulator’s is written down by the realized incidence of the harms they were charged to prevent. Why it works: it synthetically lowers r_p toward r_a by giving the agent a residual claim on year-(τ+T) outcomes. It is the only version of pay-for-performance that survives the agent’s exit, because it observes performance late but keeps the agent’s stake alive late. It directly dissolves the terminal-period theorem: the last in-tenure period is no longer terminal, because stakes extend past it. Design parameters: set T to the consequence-maturation horizon of the decision class (decades for pensions/infrastructure, years for procurement); set φ so the agent’s private gain from deferring a unit of cost past τ is at least offset by the expected clawback on that deferral. If φ is too low, the agent buys out the constraint with the harvest.

M2 — A non-transferable aeon-stake. Rule: require position-holders (and, in the republican case, hold this as a property of the median decisive cohort) to carry a long-duration instrument whose payoff is the state of the institution in the far future — and make it non-saleable and non-hedgeable within the position’s horizon. Why non-transferability is load-bearing: this is the lesson of position-as-financial-instrument. Any aeon-exposure you grant an agent who can trade, they will price and then sell — re-shortening their horizon while pocketing the premium. An aeon-stake that can be hedged is a harvest with extra steps. The constraint binds only if the exposure cannot be closed before the aeon arrives. The mechanism’s whole content is the illiquidity it imposes.

B. Manufacture a present proxy for the absent principal — deny legal exit

The deepest defect is that the aeon-principal is not in the room: future cohorts do not exist to monitor, and the institution-as-continuant (1253) has no standing of its own. Standard agency assumes a principal who can observe and sanction. Here you must fabricate one.

M3 — A chartered Posterity Trustee with standing, a measurable mandate, and an adversarial incentive — not an advisory commission. The existing “future-generations commissioner” offices (Wales, Hungary, others) fail for an identifiable structural reason: they are advisory. An adviser has no lever; the position-holder routes around them at zero cost. Fix the three defects:

(a) Standing, not advice. A suspensive veto over major long-horizon commitments that forces a recorded supermajority override. This does not block the polity; it converts an invisible intertemporal transfer into a visible, attributable, roll-call political act. (The transfer’s power was always its invisibility on the position’s accounts — cf. 080: the frame is invisible to the fact-checker. The veto forces the frame into the record.)
(b) A mandate as an objective function, not a sentiment. “The interests of the future” is undisciplineable discretion. Specify it: a non-declining per-capita capital rule (Hartwick) over a named vector of stocks — fiscal, infrastructural, ecological, institutional/knowledge. The Trustee defends a number, not a mood.
(c) An adversarial incentive. The Trustee’s own compensation is tied inversely to realized long-horizon capital depletion. The guardian personally loses when the aeon is harvested. This is what makes the watchdog bite. Who guards the guardian (the second-order agency problem): a long, non-renewable term (kills the re-appointment capture channel), strong removal protection (insulates from the rotating deciders), the narrow measurable mandate (limits discretion), and the inverse-depletion stake (aligns the guardian’s purse with the aeon). Non-renewability is doing the work that term limits do for ordinary office, without re-importing the harvest problem, because the Trustee custodies but does not allocate (see §V).

M4 — Give the aeon a litigant: generalize the trust. Rule: vest the institution-as-continuant (or a designated trustee) with legal personality and standing to sue sitting position-holders for breach of an intertemporal fiduciary duty — a duty owed to the perpetual entity, explicitly distinct from the interests of the current beneficiaries. Why the trust is the template: private trust law already solves a position–aeon problem cleanly. A trustee holds a temporary, fiduciary position over a corpus they cannot harvest, because the remainder interest belongs to parties not yet ascertained, and they can be sued for favoring living beneficiaries over remaindermen. The republic should steal this architecture wholesale: the magistrate as trustee, the citizenry-across-time as the class of beneficiaries-plus-remaindermen, breach as actionable. This converts the absent principal into a present litigant — the cleanest possible denial of the agent’s legal exit.

C. Make the deferral visible — strip the agent’s accounting exit

Cost-shifting is possible only because position-accounting uses the position’s horizon. Change the books.

M5 — A mandatory aeon-horizon balance sheet at a legislated discount rate. Rule: every major allocation decision ships with a published intertemporal balance sheet computed at a social discount rate fixed in statute, low, and uniform — not chosen by the deciding agent. Why this is the crux: the discount rate is the operationalization gap of 074. Everyone endorses “weigh the future”; the entire intertemporal politics hides in the number. A position-holder free to pick r will pick a high one, under which deferral looks efficient and harvest looks prudent. Constitutionalizing r (the Stern/Ramsey dispute, resolved by statute rather than by whichever technocrat is in the seat) removes the single most powerful laundering instrument the agent has.

M6 — A capital-maintenance rule with automatic triggers. Rule: define the long-lived capital vector; require aggregate per-capita capital be non-declining; on breach, fire a non-discretionary corrective — a binding sinking-fund contribution, a freeze on new long-horizon commitments — rather than a recommendation that some future holder act. Why automatic: a discretionary corrective just relocates the decision to the next short-horizon agent, who has the same incentive to defer it again. The trigger must bind without a position-holder choosing to pull it. (This is the commitment-device logic of 1285F: the rule’s value is that no one in the moment can decline to honor it.)

D. The automation-tech layer — the decisive new variable

Automation does two opposite things to the position–aeon relation simultaneously, and conflating them is the central contemporary error.

It amplifies the agent’s reach. A decision made inside one term, once encoded in deployed infrastructure (an eligibility algorithm, a pricing rule, an enforcement-targeting model, an autonomous weapons authorization, a smart contract), propagates across the aeon without the forced periodic reconsideration that human administration imposed. Every human administrator must decide afresh each morning; an algorithm re-executes one position’s politics, identically and invisibly, forever. Automation is the dead-hand problem (failure mode 4) with the hand made of silicon — and the silicon hand never tires, never retires, and is never voted out.
It decouples function from human position-holders. If the institution’s allocative work is done by persistent automated systems, the meaningful “position” is no longer the rotating office — it is ownership of and control over the automation. The agency problem migrates from “officeholder vs. polity” to “whoever owns the persistent allocator vs. everyone captive to it,” and that owner faces no term limit at all.

Three mechanisms answer this.

M7 — Sunset-by-default for encoded policy (“no dead-hand silicon”). Rule: any allocation rule embedded in deployed automation expires automatically after a fixed interval and may not re-take effect without affirmative re-authorization by a live deliberative body. Why: automation’s specific harm is that it removes the forced reconsideration human administration imposed. Sunset reinstalls it. Design parameter (the key one): the sunset interval must be ≤ the term of the position that authorized the deployment, so that no position can bind beyond its own horizon merely by routing policy through code. This is the encoded analogue of M3’s anti-lock-in: a position may reach only as far into the aeon as its own term.

M8 — Interpretability and an “automation impact statement” as conditions of deployment. Rule: a persistent automated allocator may be deployed only if (a) its decision rule is legible enough to be audited against the legislated objective function (M5/M6), and (b) it ships with a published statement of the long-horizon distributional consequences it encodes. Why: the encoded rule launders a political allocation choice into an apparently neutral technical artifact. This is exactly 080’s structural blind spot, transposed: the allocation rule is a frame, and a frame is invisible to a formalist audit unless interpretability is mandated as a deployment precondition. The fact-checker could verify the algorithm’s outputs and still never see the intertemporal transfer the algorithm encodes.

M9 — A public golden share over the core persistent allocation layer. Rule: automation that performs allocation the polity cannot exit and that runs across the aeon must be held under a public charter or golden share preserving the polity’s standing right to inspect, modify, and switch it off. Why: if automation makes a position permanent by making it ownable, then the republic’s classic remedy for “a delegated office has congealed into heritable property” applies — re-publicize the prebend. The res publica reclaims the sinecure. The incentive subtlety: private operators resist precisely because the value of owning persistent allocation infrastructure is its perpetuity — the very thing the aeon-principal cannot afford to alienate. The golden share prices that perpetuity back to the public instead of letting one position-holder capitalize it privately and forever.

IV. The Republic as the Layering That Solves It

Res publica — “the public thing” — is the name for a principal that cannot exit: a perpetual entity that outlives every officeholder. The republican apparatus is not one mechanism but a layering, and the layering is the whole point:

Rotation (term limits, election, separation) fences entrenchment and patrimonialization (failure modes 3–4) — but manufactures terminal periods and so worsens harvesting and deferral (1–2).
Perpetual fiduciary institutions (the trust, the constitution, the sinking fund, the central-bank-style insulated mandate) hold what must not be harvested — but concentrate power and so risk entrenchment if they also decide.

Neither layer alone works. Rotation alone makes every agent a harvester; perpetual tenure alone makes every agent a potential usurper. The resolution is to decouple custody from decision:

Keep the deciders short-tenured and replaceable (rotation, to prevent entrenchment and capture). Keep the aeon’s capital held by a perpetual, non-rotating fiduciary (trustee/guardian/sinking-fund, to prevent harvesting), armed with the suspensive veto (M3) over the deciders’ long-horizon commitments. The rotating magistrate allocates; the perpetual res publica custodies and can refuse.

The republic at its best is exactly this: rotating magistrates plus a perpetual public thing, with the magistrate forbidden to alienate the corpus. The Roman cursus honorum paired annual, collegial, term-limited magistracies (anti-entrenchment) with a perpetual Senate and a treasury treated as inviolable (custody). The design failure that ended it was not too much rotation but the fusion of custody and decision in a single permanent position — the principate, where one office held both the deciding and the corpus.

The automation warning falls out directly. Automation tempts the polity to collapse the layers — to let one persistent system both hold the assets and make the decisions (the “optimize allocation end-to-end” pitch). That is the principate in code: a permanent position that both custodies and allocates, owned by whoever owns the system, accountable to no rotation. M7–M9 exist to keep custody and decision separate even when both are automated: sunset forces the deciding layer to face re-authorization (it stays rotating); the golden share keeps the custodial layer public and switch-off-able (it stays fiduciary).

V. The Allocation Function, Reframed

“Allocation” is usually modeled spatially — who gets what, across persons. The position–aeon lens reframes it temporally: allocation is the distribution of c(t) across time, and the politically decisive transfers are the ones that cross the boundary at τ — from after-the-position to before-the-position, from the captive aeon-principal to the exiting agent. 041 located distribution as the governed layer; the claim here is that its most ungoverned axis is the temporal one, because the cross-time transfer is invisible to the accounting (M5 fixes the books), unrepresented in the room (M3/M4 fix standing), uncontracted on the agent (M1/M2 fix exit), and now encodable into infrastructure that outlives the encoder (M7–M9 fix the silicon hand).

The pension (051) is the cleanest specimen: a promise whose funding is allocated by a position-holder whose term ends decades before the promise matures. Underfunding is not an accident or an actuarial error — it is horizon cost-shifting executed perfectly, a textbook deferral that books a benefit now (lower contributions, the avoided tax) against a cost that lands after exit. Every mechanism in §III is, in the pension case, already partially instantiated somewhere (funding triggers, fiduciary duty, actuarial disclosure) — which is exactly why the pension is the right proof-of-concept: the position–aeon problem is solvable, it is solved in fragments, and the design task is to generalize the fragments into the stack.

VI. The Counter-Frame

The strongest objection: the mechanisms above all entrench the present’s judgment about the future and thereby commit the very dead-hand sin (failure mode 4) they claim to fence. A legislated discount rate (M5), a Hartwick capital rule (M6), a non-renewable Posterity Trustee with a veto (M3) — these are one generation’s theory of what posterity will want, frozen into binding rules and imposed on cohorts who never consented and may value entirely different stocks. The future is not a captive principal with fixed interests; it is an open principal whose preferences we cannot know. Binding it “for its own good” is paternalism across time, and it forecloses the future’s own deliberation as surely as harvesting forecloses its resources. Jefferson’s “the earth belongs to the living” is the sharp form: each generation has the same right to rule itself that we claim, and our intertemporal fiduciary duty is itself a dead hand.

This objection is substantial and partly correct. Two responses bound it:

Procedural vs. substantive lock-in. The objection bites hardest against substantive commitments (you must value these stocks, at this rate, forever). It bites far less against procedural ones (you must reconsider; you must see the transfer; you must re-authorize the encoded rule). M7 (sunset-by-default) and M3’s suspensive veto (override available, but on the record) are deliberately procedural: they do not tell the future what to choose; they force the future’s deciders to choose afresh and visibly, which is the opposite of dead-hand. The design rule that falls out: prefer mechanisms that re-open deliberation over mechanisms that pre-empt it. A sunset clause is anti-entrenchment; a perpetual capital floor is entrenchment. The stack should lean procedural, and the genuinely substantive pieces (M6’s capital floor) should themselves carry sunsets.
The asymmetry of irreversibility. The two errors are not symmetric. Over-binding the future (M6 too strict) is reversible — a later cohort can amend a statute, override a veto, let a rule sunset. Harvesting the future (the unfenced status quo) is often irreversible — an extinct species, a collapsed aquifer, a defaulted promise, a foreclosed capability do not come back when the next cohort changes its mind. Where errors are asymmetric in reversibility, the design should tilt toward the reversible error. That argues for the procedural core of the stack and against both extremes: against unfenced harvesting (irreversible) and against hard perpetual substantive floors (entrenching). The future’s best protection is not our theory of its good but our refusal to make exits — ours or its own — too cheap.

VII. What Remains Unresolved

The stack denies the agent’s exit (M1–M2), manufactures the principal’s presence (M3–M4), reveals the deferral (M5–M6), and governs the silicon hand (M7–M9), held together by the republican custody/decision split (§IV). Two things it does not settle, flagged honestly:

The proxy-principal’s fidelity is unverifiable in real time. M3’s Trustee defends a measured capital vector — but the choice of vector and weights is itself a present judgment about the future’s values, and a captured or merely mistaken Trustee can defend the wrong number with full procedural legitimacy. The inverse-depletion stake aligns the Trustee with the measured stocks, not with the future’s actual interests, and the gap between them is exactly the operationalization gap (074) we cannot close from inside the present. We can make the transfer visible and the deciding rotational; we cannot make our proxy correct.
Illiquidity has a price the polity may refuse to pay. M1–M2 work by imposing non-exitability — deferred clawbacks, non-hedgeable stakes, switched-off-able infrastructure. But non-exitability is costly to the agent and the polity must compensate for it (higher nominal pay for deferred risk, lower private investment in locked infrastructure) or the best agents decline the position and the best capital avoids it. The position–aeon problem is not eliminated by the stack; it is priced. How much the captive principal should pay to shorten the agent’s exit — the term structure of that price — is the live design question, and it is not derivable from the diagnosis. The abstraction (“align the exit horizon with the principal’s captivity”) is shared; the operationalization (how much illiquidity, at what compensating price, for which positions) is contested, and the contest is where the politics lives.