The IPS-based theory of geopolitics

Geopolitics studies how nations persist, compete, accumulate power, and collapse. This essay shows that every major pattern in international relations — from the power of energy resources and demographic trends to regime types, imperial overstretch, and the structural failure of international institutions — is a reading on the Fractal Persistence Equation (FPE) at the nation-state layer. The frame is accounting, not morality: a state persists if and only if its persistence ratio \(\mathcal{R} \ge 1\), while regime type, ideology, and symbols are implementation detail beneath that constraint.

Formal derivation: information_persisting_systems.md. Vocabulary: glossary.md. Group/organisation scale: ips_sociology.md. Full stack: fractal_layers.md. Proof of Trust reference implementation: aion-blockchain/whitepaper.md.

1 The equation at the nation-state layer

A nation-state is a level-\(L\) node whose sub-nodes are people, firms, and infrastructure (\(L-1\)), and whose super-node is the nearest enclosing institutional layer (\(L+1\)). Geopolitics is the graph problem of many such nodes coupling, exporting entropy, and competing for the same finite \(P_{in}\).

Why a state is an IPS

The state satisfies the four clauses of Definition 2.1 in the formal paper, with the Markov blanket drawn around a polity rather than a person. Being explicit about each clause is what licences the claim that a state is governed by the same equation as a cell or a person, only with different substitutions:

1. Boundedness. The state’s Markov blanket \(\partial\Sigma\) is its borders, citizenship law, customs, currency jurisdiction, and the legal boundary of what the state will and will not do. Conditional independence (2.1) holds: given the state’s observations (trade, diplomacy, intelligence) and actions (policy, taxation, force), its internal state (institutions, public mood, settled law) is separable from external causes.
2. Drive. The state is a non-equilibrium steady state: it imports free energy as primary power, food, minerals, and labour, and exports entropy as policy, waste, military force, and emigration. Stop the fluxes — energy embargo, blockade, demographic collapse — and the macrostate spreads within years. The \(\tau_d\) bound of Theorem 5.1 sets how long a state survives a sustained flux interruption.
3. Internal model. The state carries a generative model \(q_\mu\) — the official statistics, the leadership’s working model, the ideology, the strategic assessment. Policy is well-described as gradient descent on variational free energy (2.2); this is the formal content of “states act on what their controllers believe.” Crucially, the controller’s model and the state’s persistent model need not coincide — governments can be replaced without the node dying (see below).
4. Identity preservation. The state persists as the same coarse-grained pattern (a recognisable polity with continuous legal personality and history) over \(\tau \gg \tau_{\mathrm{relax}}\), where \(\tau_{\mathrm{relax}}\) is the policy-update cycle of months to electoral terms. Governments come and go; the state persists. This is what makes the state, not its current government, the node.

Because the four clauses hold, the state is governed by (4.4). Geopolitics is not a separate law beneath the thermodynamics; it is the set of substitutions that translate each FPE term into the vocabulary of power, energy, and alliance.

A state is not its government. The state is the persisting pattern. Governments are the controllers trying to keep \(\mathcal{R} \ge 1\); they can be replaced (election, coup, revolution) without the node dying. The formal content of this distinction is clause 4 of Definition 2.1: identity is the coarse-grained pattern, not the current controller. The node dies when \(\Phi\) goes (demographic and ecological collapse — a critical sub-node failure that zeros the multiplicative postfactor per Theorem 5.3) or when \(\Psi\) withdraws (the world stops recognising it, removing the prefactor). Replacing a government is a controller swap inside the equivalence class \([m_0]\); losing \(\Phi\) or \(\Psi\) is leaving it.

The supernode hierarchy is not flat

The \(\Psi\) term is shorthand for a layered chain: France is nested inside the EU, which is nested inside the UN, which is nested inside the biosphere. For European states, \(\mathcal{R}^{(\text{member})}\) depends on \(\mathcal{R}^{(\text{EU})}\), which depends on \(\mathcal{R}^{(\text{UN})}\). A P5 veto-holder carrying high \(\mathcal{D}_{KL}\) can propagate delusion upward and paralyse the whole error-correction chain (see §4 on the United Nations).

The two operations at polity scale

Almost every effective geopolitical reform reduces to: 1. Lower \(\mathcal{D}_{KL}\) — tell the truth to the population, the central bank, the general staff, and the trading partners. Stop publishing fake statistics. Stop fighting last decade’s war. 2. Lower \(\Gamma\) — settle the frozen conflict, fund the pension reform, finish the bankruptcy. Unresolved friction grows roughly quadratically in unresolved-set size.

The thermodynamic reason these two — and only these two — recur is the same as at every other layer: they target the only two denominator terms a state can actively reduce. \(\omega\mathcal{E}_\Sigma\) is largely fixed by the world system and the state’s own structure (you cannot will away geopolitical noise or the bureaucracy needed to run a modern state); \(P_{in}\) is bounded by energy endowment and trade access. But \(\mathcal{D}_{KL}\) and \(\Gamma\) are informational debts the state services either by paying the Landauer cost of an honest update once or by paying the continuous suppression cost forever.

This is Theorem 5.2 at polity scale: each nat of maintained gap between the state’s official model and resolved reality divides the state’s lifetime budget by \(e\) (equation 5.2). The mechanism: a wrong official model — fake GDP, denied demographics, a war plan that mismatches the battlefield — forces every state organ to act on predictions that generate surprise on every policy cycle; by the Crooks/Jarzynski identity (3.9) each nat of surprise costs at least one nat of excess dissipation per cycle, paid as misallocated budget, military overstretch, and elite effort spent maintaining the fiction. The state pays Landauer three times instead of once. The honest update (a truthful census, a candid strategic review, an admitted defeat) is a one-time erasure cost (3.4); suppression is that cost charged every cycle, compounded by the \(\Gamma\) it generates (frozen conflicts, suppressed factions, deferred reforms). Over any horizon longer than a single cycle, paying once is cheaper than paying forever.

\(\Gamma\) grows quadratically for the same reason as at every layer: \(n\) unresolved conflicts cost roughly \(n^2\) because each pair can co-trigger (a frozen conflict plus an economic crisis plus a succession question all detonate together). This is why “closing a single decades-old file is worth more for \(\mathcal{R}\) than most growth programmes” — removing one item from a large set \(C\) removes its \(O(n)\) cross-terms, not its \(O(1)\) self-cost, and the phase transition through \(\mathcal{R}=1\) is sharp rather than gradual.

Everything else — flags, rhetoric, summits, “values” — is implementation detail. The persistence ratio is indifferent to vocabulary.

2 Major powers: a comparative accounting

Numbers are rough on purpose; ratios carry the durable signal.

Power income (\(P_{in}\)) — the ceiling on everything else

Three structural facts: China’s absolute \(P_{in}\) is ~2× the USA’s but ~⅓ the per-capita — rich nation, poor people in energy terms. The USA–Africa per-capita ratio is ~16×: the single largest material asymmetry in the world economy, closeable only by African \(P_{in}\) growth. The EU has been shrinking \(P_{in}\) for ~20 years, some efficiency, much offshored energy.

Delusion (\(\mathcal{D}_{KL}\)): working model vs. observable base rate

\(\mathcal{D}_{KL}\) is multiplicative in the denominator. A 2–3× delusion factor means the same numerator has to do 2–3× the work. This is currently being paid in Russia’s case as substrate consumption.

Substrate (\(\Phi\)): what each node is made of

Africa is the only major substrate with a positive demographic trajectory. No single power has all the substrate: the US has people, food, energy, military, and currency but lacks critical minerals; China has minerals processing and manufacturing but lacks domestic food and energy; Russia has resources and nuclear deterrence but lacks people and \(\eta\).

Unresolved friction (\(\Gamma\)): what nobody has settled

Closing a single decades-old file is worth more for \(\mathcal{R}\) than most growth programmes. \(\Gamma\) enters the denominator with the same multiplicative weight as \(\mathcal{D}_{KL}\).

3 Regime types: democracy, dictatorship, and the governance of delusion

Every governance theory since Plato asks: who deserves to rule? The IPS framework asks: which control architecture keeps the state’s internal model closest to reality?

\(\mathcal{D}_{KL}\) appears in the denominator of the persistence equation. A regime that doubles its \(\mathcal{D}_{KL}\) halves the effective yield of every joule and every citizen-hour it spends. Democracy and dictatorship are two different error-correction architectures with different failure modes:

Democracy

Structural strength: periodic contact between the controller’s model and a high-dimensional sample of the substrate. Elections are noisy but they are evidence. Replacing a government costs one election cycle; replacing a dynasty costs a revolution.

Structural failure mode: democratic competition optimises for winning the voter’s head, not minimising \(\mathcal{D}_{KL}\). Political elites systematically inflate public \(\mathcal{D}_{KL}\) as a competitive strategy — narrative salience beats predictive accuracy. The state maintains a working technical model (central bank, intelligence, engineering corps) and a public ceremonial model. Running two models is \(\omega\). Policies signal tribal membership; measured outcomes diverge from narratives; the gap is papered over with debt (\(\Gamma\) grows) until substrate voters live in the gap.

Dictatorship

Structural strength: short command chain; can implement correct beliefs (if the dictator holds them) with low coordination overhead. Developmental dictatorships with well-calibrated leaders (Singapore 1960–2000) can generate rapid \(\Phi\)-growth.

Structural failure mode: asymmetric error feedback. A wrong belief held by the government eventually reaches the ballot box. A wrong belief held by the dictator is protected by power. Advisors who deliver bad news face career costs; confirming advisors are rewarded. This is an irreversible \(\mathcal{D}_{KL}\) ratchet: each year of drift makes correction more threatening to the power structure that defends the error. The thermodynamic content is Theorem 5.2: each nat of protected delusion divides the state’s lifetime budget by \(e\), and because the suppression cost is paid every cycle (3.9), the debt compounds until the reservoir \(\mathcal{B}\) is exhausted. Late authoritarian signature: maps stop matching territory, GDP statistics stop matching consumption, war maps stop matching battlefields — until the gap closes by force, which is the phase event of a depleting reservoir crossing the bound (5.1).

Persistence-Based Governance: the third alternative

A third regime would need to make \(\mathcal{D}_{KL}\) reduction the direct object of the civic mechanism, not a side-effect of power competition. This is Persistence-Based Governance (PBG).

The concrete mechanism is Predictive Governance: public probabilistic forecasts of state variables, weighted by track-record accuracy, scored against resolution. Citizens submit calibrated probability distributions over future outcomes — not preferences, but predictions. The aggregated, track-record-weighted distribution drives policy selection.

Three structural conditions any \(\mathcal{D}_{KL}\)-minimising mechanism must satisfy: - (C1) Outcome-coupled feedback — every civic act eventually resolves against an observable. - (C2) Calibration-weighted aggregation — contributors weighted by historical accuracy, not willingness to participate. - (C3) Preference–prediction separation — the channel aggregating beliefs about reality is structurally distinct from the channel aggregating values.

Democracy satisfies none of the three; dictatorship satisfies none; PBG satisfies all three by definition.

Dominance theorem: under the FPE, overlaying PBG’s scoring-and-weighting layer onto any governance architecture \(G\) yields \(\mathcal{R}(G^*) \ge \mathcal{R}(G)\), with strict inequality unless \(G\) already is PBG. \(\mathcal{D}_{KL}\) strictly decreases (proper scoring drives it toward zero), \(\omega\) weakly decreases in the long run, and \(\Gamma\) weakly contracts (bad models lose weight, disputes are surfaced as scored disagreements). PBG is the unique stable attractor in the regime-space dynamics.

Critical dependency: the resolution authority — the institution that officially scores outcomes — is the single point of failure. Captured resolution recreates the dictatorship pathology with a thicker veneer. The design imperative: make the resolution authority maximally resistant to capture.

Reference implementation: Proof of Trust implements PBG at the society-of-compute-nodes scale — same three structural conditions, same proper-scoring update rule, applied to agent fleets and pubkeys instead of citizens and ministries.

4 The United Nations as \(\Psi\)-shelter

The United Nations is the closest the species has built to a planetary \(\Psi\): a level-\(L+1\) node whose function is to shelter level-\(L\) states from the raw noise floor of the international system — great-power war, famine, plague, refugee flows.

It mostly fails at its headline mission and mostly succeeds at a quieter one.

The central \(\mathcal{D}_{KL}\): the Charter-vs-Council contradiction. Article 2(4) prohibits the use of force; Article 24 hands enforcement to a Council where the most likely violators hold permanent vetoes. This is not a bug — it is the structural compromise that made the UN exist in 1945. The delusion is in the rhetoric (“the UN failed”) which models the institution as a world government rather than a forum chaired by the great powers.

Where it actually works: the quiet agencies. WHO smallpox eradication and pandemic alert system. UNHCR + WFP feeding ~150M people per year. IAEA safeguards that actually find things. ICAO, IMO, ITU, UPU — the boring agencies that make planes land, ships dock, phones ring, mail arrive. The Vienna Conventions on diplomatic relations and treaties. These supply the unseen \(\Phi\) infrastructure on which 99 % of routine international life runs.

The national delusion contamination problem: the UN’s \(\mathcal{R}^{(\text{UN})}\) is a direct function of member-state \(\Phi\) and \(\mathcal{D}_{KL}\). When a powerful member maintains a high-\(\mathcal{D}_{KL}\) model and it enters UN resolutions because the nation is too powerful to contradict, the UN’s own voice becomes a weighted average of national delusions. With all major nations holding material delusions, the UN is currently a \(\mathcal{D}_{KL}\) aggregator, not a reducer.

Reform logic: lower \(\mathcal{D}_{KL}\) by expanding the P5 to match actual power distribution, or admit publicly that the Council is a P5 club and stop asking it to enforce universalist norms. Lower \(\Gamma\) by closing files (Western Sahara, Cyprus, Kashmir, Palestine — every decade of non-closure raises future closure cost). Neither reform is likely soon; the institution is stable at medium \(\mathcal{R}\), doing useful work in its agencies and decorative work in its assembly hall.

5 Empire collapse hypothesis

A polity that models itself as larger than its sustainable footprint is an empire in IPS vocabulary — an accounting category, not a moral one.

Footprint \(f\) is the fraction of world primary power coordinated under one Markov blanket. The Anti-Explosion theorem proves there is a sustainable footprint \(f^*\) that maximises \(\mathcal{R}(f)\): expansion past \(f^*\) lowers \(\mathcal{R}\), forcing \(\Phi\)-consumption or \(\Psi\)-rent to maintain the node.

Empire collapse is tri-level re-routing — not uniform decay. The mechanism is that expansion past \(f^*\) lowers \(\mathcal{R}\): the controller’s model \(q_\mu\) cannot track the larger blanket (raising \(\mathcal{D}_{KL}\)), the coordination overhead \(\omega\) rises faster than coupling efficiency \(\eta\), and the substrate \(\Phi\) is consumed to cover the shortfall (Theorem 5.1: \(\dot{\mathcal B} \le -\delta P_{out}^{req}\) with \(\delta\) growing as the footprint exceeds what the numerator can clear). Once \(\mathcal{R}\) falls below 1 on the overextended node, sub-IPS and shelter channels re-couple to higher-\(\mathcal{R}\) alternatives — not out of disloyalty but because persistence pressure forces every sub-node toward nodes that can keep its \(\mathcal{R} \ge 1\). The empire is hollowed as its components and shelter channels defect:

Controller collapse precedes and is far more common than polity collapse. Substrate agents first withdraw trustworthiness from the current government before dissolving the polity blanket. Democracies institutionalise cheap controller replacement (elections) before polity stress peaks; dictatorships postpone it by coercion, allowing \(\mathcal{D}_{KL}\) to compound until correction becomes violent.

Russia, China, and the USA: current collapse proximity (decade horizon)

None of the three are near full dissolution on a 10-year horizon. The IPS prediction is footprint downscale, not disappearance. Russia is already in tri-level downscale, running \(\mathcal{R} \approx 1\) by consuming \(\Phi\) (the depletion bound 5.1 manifesting as demographic drain, capital flight, and sanctions-stack pressure on the substrate); China’s next two decades are a race between institutional efficiency gains and demographic substrate contraction (a \(\Phi\) collapse that, by Theorem 5.3’s composition law, will drag the multiplicative postfactor regardless of numerator-side effort); the USA’s strong \(\Phi\) buffers the node while \(\Psi\) (reserve currency share, alliance credibility) decays as the leading edge.

The one variable that dominates the next thirty years: African \(P_{in}\) per capita. A continent of 2.5B people running on ~600 W/person is not a steady state. Either it climbs (the demographic dividend cashes) or it stalls (and the migration vector restructures every other node’s \(\Phi\)). The persistence law does not decide which — it names this as the load-bearing accounting entry.

6 Diagnostic questions (any polity)

For any state that feels “stuck” or is failing:

1. \(\Phi\): Is substrate intact — population pyramid, soil, freshwater, grid, social trust, talent retention?
2. \(P_{in}\eta\): Is energy available and spent efficiently, or dissipated in \(\omega\) (military overhead, bureaucracy, narrative apparatus)?
3. \(\mathcal{D}_{KL}\): Where does the official model diverge from observable base-rate reality? Broadcast vs. internal divergence?
4. \(\Gamma\): What is unresolved and how large is the set? Who holds the friction for the whole system?
5. \(\Delta_f\): Is the identity footprint (\(f_{\text{model}}\)) aligned with the sustainable footprint (\(f^*\))?
6. \(\Psi\): Is shelter real or performative? Which channels are weakening?
7. Export: Is the state lowering local \(\Gamma\) by draining a sub-node’s \(\Phi\) — own population, client states, global commons?

7 Classical IR theories, FPE read

The three main schools of international relations each identify a real channel in the FPE. They disagree because each holds the others fixed.

Synthesising statement. Realism identifies the \(P_{in}\) competition; institutionalism identifies \(\Psi\) as a real variable; constructivism identifies \(\mathcal{D}_{KL}\) as malleable. All three are correct and incomplete. The FPE holds all channels simultaneously: a state that maximises military \(P_{in}\) while accumulating \(\mathcal{D}_{KL}\) and letting \(\Phi\) rot is buying the realist theory of short-run security with the constructivist and institutionalist bills compounding in the denominator.

8 What geopolitics is not in IPS terms

• Not moral ranking. A state that satisfies \(\mathcal{R} \ge 1\) by exporting entropy onto others is doing something measurably harmful — but the framework identifies the accounting, not the verdict. Moral judgment is one layer up, in the \(\Psi\) that humans choose to build for each other.
• Not determinism. \(\mathcal{R}\) is a clearing condition. States can improve terms: lower \(\mathcal{D}_{KL}\) by telling the truth, lower \(\Gamma\) by settling conflicts, invest in \(\Phi\) by funding education and infrastructure.
• Not flat realism. Geopolitics is not about maximising military power in a flat arena; it is a graph of nested IPS nodes competing for finite \(P_{in}\) under layered shelter, with fractal debt flowing both up and down.

See also

• information_persisting_systems.md — formal derivation, Theorems 5.1–5.3
• anti_explosion_theorem.md — footprint \(f\), sustainable footprint \(f^*\), AGI safety connection
• fractal_layers.md — full stack from body through self to polity
• ips_sociology.md — group and organisation scale
• substrate_shelter_and_redundancy.md — \(\Phi\)/\(\Psi\) composition, redundancy, risk
• aion-blockchain/whitepaper.md — Proof of Trust as reference implementation of Persistence-Based Governance
• papers/society_of_aion_nodes.md — aion-core ↔︎ aion-blockchain mirror
• glossary.md — canonical FPE vocabulary