Trapped Equilibria

I. Adoption is not a legal move

Take a stable cluster: a tenured department, a regulatory agency, a foreign-policy consensus, a professional guild, a movement. Its members are intelligent. Many peripheral members, asked privately, admit the equilibrium is broken. Funding follows fashion. Methods are known to be broken. The dominant frame is older than its evidence. They see this. They do not adopt.

Two standard explanations dominate, and both miss the structure. Incompetence says the actors are bad at their jobs — staff smarter people and the cluster repairs. Captured clusters attract the most credentialed and articulate people available; the equilibrium holds anyway. Bad-faith hypocrisy says the actors know better and lie in public — expose the dishonesty and the cluster pivots. Exposure tightens it. After 2008, after the replication crisis, after every well-publicized failure that left the responsible cluster intact, everyone knew, and the cluster got more captured.

The structural fact both explanations miss: defection on any one axis is a cliff-edge move on the others. The peripheral insider sees the better frame, sees instantly that adopting it would be suicidal, and routes around it before deliberation begins. The core does not route around the frame; it has already metabolized the frame away. The hypocrite reading is too generous in a different way: it assumes the actor still privately believes the better frame and lies in public. In mature capture, the core can no longer see the better frame at all; the periphery still sees it but cannot afford to act on it. The actor is sincere either way. Three populations sit inside capture differently, and any single account that flattens them contradicts itself.

II. Concentric capture

Four axes specify the local game.

Epistemic. The beliefs and intellectual moves the cluster rewards as competent. To pass as a member, the actor demonstrates fluency in the cluster’s premises and treats them as obvious.

Status. The reputation, prestige, citation graph, and informal hierarchy that determines whose voice carries. Visible loyalty to cluster premises gates standing. Visible defection destroys standing overnight.

Material. Income, role, professional license, grants, contracts, platforms, protection from prosecution or social punishment. Material applies to friend groups, hobby communities, and online tribes as well as formal institutions. The institutional case is one specialization where the stakes are largest and most legible.

Coordination. The dense web of relational expectations — collaborators, peers, neighbors, friends, spouse’s milieu — aligned around shared cluster premises and primed to sanction departure.

The four axes individually are well-described in sociology. What concentric capture names is their bonding into one local game. In mature clusters the axes lock together. A small defection on one is not a small cost on the others — it triggers collapse across all four at once. A scientist who publishes outside the field’s orthodoxy loses the citations, then the grant cycle, then the speaker invitations, then the colleagues who stop recommending her students, then the marriage that was nested in the same intellectual milieu. Each loss feeds the next within months. In ordinary settings, a loss on one axis can be offset by gains on another — lose status, recover through money or a new network. Under concentric capture, the offsets disappear: the axes are bonded, so a hit on one becomes a hit on all.

Define the mechanism cleanly:

Concentric capture is the bonding of epistemic, status, material, and coordination positions into one local game, so that defection along one axis produces cliff-edge losses across the others.

The architecture is scale-free across coordination clusters: dyads, friend groups, family systems, online tribes, subcultures, professions, institutions, civilizations. Bad marriages, recovery communities that have absorbed all four axes, and tenured departments share the same bracing geometry at different scales.

Capture is not purely emergent. The cost cliff stays steep in part because specialized roles maintain it: peer-review gatekeepers, compliance and HR officers, departmental dean-level enforcers, purity-policing reply guys, the orthodox commentators. Some are core members defending their identity outward; others are rent-extractors profiting from the cluster’s need for boundary policing. Either way, the active-policing layer is part of the mechanism. Concentric capture is not an emergent natural law. It is a maintained political order. Active policing keeps the costs steep.

III. Three populations

A single account of the captured insider produces contradictions. The predictive-processing account from cognitive science says the dissenting frame has been pruned away by the brain saving energy — the better view is no longer accessible to the captured insider at all. The preference-falsification account from political science (Kuran) says a dense fraction of insiders has privately abandoned the frame but stays publicly compliant out of fear. Both empirical patterns are real. They describe different populations inside the same cluster.

Core. Identity-fused, status-fused, epistemically fused. For the core, the cognitive-science account fits. The brain saves energy by tuning its predictions to whatever its environment most reliably produces. Inside the cluster, the most reliable predictors of social outcome are the shared frames; the brain takes them as the default model of the world. Holding a dissenting frame generates a steady stream of prediction errors during every social interaction — the brain has to keep correcting against the cluster’s signals. Over months, the brain saves energy by pruning the dissenting pathway. The frame stops being available; metabolic budget has been reallocated away from it. The core defends the cluster sincerely because the alternative is no longer in working memory. (Veissière, Constant, Ramstead, Friston, and Kirmayer formalize this in their extension of the Free Energy Principle to cultural cognition; the citation is in the notes.)

Periphery. The hidden fraction of insiders who hold the better frame privately or semi-privately but find public action too costly. They pay through burnout, cynicism, detachment, absenteeism, private jokes, anonymous browsing, exit fantasies, covert support for outsiders. Timur Kuran’s account of preference falsification fits this group: under sustained social pressure, public expression and private belief decouple, and direct surveys cannot detect the hidden fraction because every respondent has the same incentive to express the cluster’s frame. Indirect methods — anonymous behavior, exit signals, absenteeism, parallel-media engagement, overlapping traces — are designed to find what surveys suppress. Jiang and Yang’s 2016 study of a 2006 Shanghai political purge isolates the periphery profile cleanly: respondents’ expressed political support increased while emotional well-being and confidence-related measures moved in the opposite direction, with the effect concentrated in wealthy, educated, public-sector employees. High capture, high interior dissent, low behavioral signal under ordinary survey conditions.

Boundary actors. Pre-decoupled on one or more axes. Independent income, cross-domain identity, weak social embedding, low sensitivity to status pressure, outsider profession, alternative prestige network. They see and act earlier because the cluster’s bonding has not set around them. The override-capable minority is small. Its position is structural, not moral.

The three populations are structural positions, not permanent types. Individuals move between them over time and across situations. Young entrants often begin as periphery and harden into core over a decade of incentive alignment as the dissenting frame gets pruned. The same actor can read as core in a faculty meeting and as periphery in a private conversation with trusted friends. The model describes positions; people occupy them and can move.

The split predicts the dynamics of rupture. When triggers hit, the core defends with full conviction, the periphery flips — private dissent becomes public action and provides the transition energy — and boundary actors absorb defectors into the alternative substrate. A cluster with a thin periphery hardens; a cluster with a dense periphery and ready substrate jumps.

IV. Where this differs from existing theories

The Core/Periphery/Boundary split is why one framework can hold both kinds of evidence at once: the cognitive-science finding that captured insiders sincerely cannot see the better frame, and the political-science finding that they often hold it privately while suppressing it publicly. Without the split, the two findings appear to contradict each other. With it, each describes its own population.

One more parallel. For some captured equilibria, the simplest explanation is public choice: actors are paid to defend the frame, and the rent is enough to explain the behavior — no metabolic story needed. Banking deregulation through 2008 fits this reading better than a multi-axial cognitive trap. Concentric capture and rent equilibria are parallel mechanisms; some clusters are held by one, some by the other, some by both. The framework here is most useful where rent alone is too thin to explain the depth of the lock-in, and where reaching the cluster from outside fails even when nobody is being directly paid.

V. Where the trap forms — and where it does not

Concentric capture requires axis monopolization. The trap forms when a single cluster controls all four axes and external feedback is slow. It dissolves when any axis stays liquid. Call the trap-forming case rigid-solid and its inverse viscous-fluid.

Six conditions produce the rigid-solid case:

1. Epistemic monopoly. The cluster controls what counts as expertise.
2. Status monopoly. The cluster controls prestige and reputational survival.
3. Material monopoly. The cluster controls jobs, grants, licenses, platforms, access.
4. Coordination monopoly. The cluster controls the social and professional network needed to function.
5. Slow external feedback. Reality does not quickly and publicly punish false frames.
6. No ready alternative substrate. Exiting means falling into a void — no alternative employer, no alternative prestige network, no alternative venue for the dissenting frame, no community to land in.

The viscous-fluid case has the inverse signature: multiple non-overlapping networks, fast empirical feedback that re-prices status and material rewards, transferable status across networks, liquid labor markets, and external judges that don’t answer to the cluster — markets, courts, and code that runs or fails to run.

Rigid-solid territory: academia, public bureaucracies, credentialed professions, legacy media, party machines, regulatory and public-health clusters, isolated movements. Viscous-fluid territory: software engineering during the Agile and Deep Learning shifts, early-stage startups, financial-trading roles where P&L is binary, technical-empirical domains generally. The framework applies most strongly where empirical feedback is slow, legitimacy is internally gated, and all four axes are monopolized by the same cluster. It applies weakly where reality re-prices fast.

If Smolin’s diagnosis of string-theory hegemony in The Trouble with Physics is even partly right, theoretical physics supplies a clean rigid-solid case study. Smolin documents the mechanism: tenure required ten to fifteen expert evaluations from the dominant cluster, ranking pressure punished deviation from consensus, patron-client networks gated grants and citations, and the field reported a “loss of a generation” of theoretical physicists who could not get hired outside the dominant program. All six rigid-solid conditions hold simultaneously. In fundamental theory, empirical feedback takes decades. No alternative substrate existed for non-string research. The four axes braced each other tightly enough to produce capture without anyone designing it.

The viscous-fluid contrast is software. When the empirical-feedback domain is fast and external (the system runs or it does not), incumbent paradigms collapse on a timescale of years rather than generations. Agile displaced waterfall when teams using it shipped working software; deep learning displaced symbolic AI when networks beat hand-crafted features on benchmarks the field already accepted. No persuasion was required of the previous core. The cost surface changed.

Technical domains are not immune to capture. Agile ossified into a credentialed industrial complex once consultancies, certifications, and prestige networks attached to it. Web3 and crypto sustained rigid-solid traps for years despite negative empirical feedback: near-zero interest rates kept token prices and investment flows artificially elevated, suspending the feedback-repricing that normally dissolves technical capture. The viscous-fluid claim is not that software is uncapturable; it is that fast external feedback, when present and not suspended by capital glut or platform monopoly, re-prices the axes faster than capture can set.

VI. How rupture happens

Captured equilibria break under recognizable patterns. None of them is necessary; several typically combine.

1. Repricing shock. The old material or status payoff collapses. Russia’s 2022 invasion repriced Finland’s decades-long policy of strategic neutrality (Finlandization) in weeks; the formal NATO application followed within three months. Stagflation repriced Keynesian consensus in a decade.
2. Legible scandal. Outsiders can see failure without cluster mediation. Theranos’s technology either worked or did not; John Carreyrou’s Wall Street Journal reporting was a binary collision the cluster could not absorb. Weinstein’s pattern, once documented, was legible to non-specialists.
3. Elite split. High-status actors defect and bring networks. Mont Pelerin elites carried the 1970s neoliberal turn. The FTX collapse fractured EA elite cohesion.
4. Alternative substrate. Defectors have somewhere to land. The Mont Pelerin Society spent thirty years building departments, journals, think tanks, and policy pipelines before stagflation made the alternative operational.
5. Generational replacement. Lower sunk costs in younger entrants. Planck’s principle: science advances funeral by funeral.
6. Empirical-feedback domain. Reality rewards the new frame quickly. Agile and Deep Learning ruptured their incumbents through this channel without needing any of the others.

A distinction that frequently gets lost: rupture is not replacement. Rupture is the breaking of the existing equilibrium. Replacement is the installation of a coherent successor. The Soviet collapse was rupture without replacement: the cluster died into a void, and the successor states spent two decades stumbling through partial reconstructions because no Mont Pelerin equivalent had been built. Alternative substrate is the condition for productive replacement, not for collapse. Clusters can fail catastrophically with nothing waiting to receive the population.

Brief case readings:

• Finlandization 2022. Repricing shock plus legible scandal plus elite split. NATO membership was a substrate already engineered through decades of partnership-for-peace integration; the trigger revealed it.
• Theranos. Legible scandal in a binary domain. Journalist and legal substrate already operational.
• Mont Pelerin to Reagan/Thatcher. Pre-built substrate plus stagflation trigger plus elite defection. The cleanest case of pre-positioning paying off.
• #MeToo. The cost surface shifted. Legal substrate (changed receptivity to harassment claims), media substrate (post-Weinstein willingness to publish), and status substrate (reputational cost of accusation) all moved at once. Uncoupled awareness tends to harden; awareness coupled to a changed cost surface can rupture.
• 2008 financial crisis. Failed rupture of the macroeconomic cluster. Costs externalized to the public via bailouts; narrative absorbed via “liquidity event.” The energy did not vanish. On one plausible reading, it transferred laterally to civilizational meta-trust and surfaced in the 2016 populist ruptures — the causal chain is contested, but the meta-trust degradation is documented. When triggers fail to break the specific cluster, the energy weakens the meta-equilibrium that protects it.
• Soviet collapse. Rupture without coherent replacement; the cluster died into a void.

VII. Strategy for moving a captured equilibrium

These are conditional patterns for someone trying to shift a captured cluster from outside or from its periphery. They follow from the population structure and the rigid-solid conditions above. They are not universal laws, and none of them is about persuading the core.

Skip the core. The core is sincere; the dissenting frame is operationally absent for them. Persuasion aimed at the apex feeds the cluster instead. Every failed engagement becomes evidence that the challenger is unserious. Apex actors reached the top by perfectly optimizing for the existing axis lock. Their identity, status, and network rest on the equilibrium. They are the most-constrained position, not the most-free.

Uncoupled awareness tends to harden. Public exposure of equilibrium failures triggers an immune response: defensive narratives, in-group rallying, escalated purity tests. Exposure works only when paired with a changed cost surface. The Weinstein case in §VI shows the contrast.

Target the periphery and boundary actors. The periphery already holds the better frame and is blocked from acting; boundary actors are pre-decoupled and can move first. The strategic goal is to provide a landing zone for the hidden fraction. The cluster grows the hidden fraction automatically with every year the equilibrium holds past its sell-by date.

Lower the cost of exit. Provide status substitutes (alternative prestige networks), material substitutes (income, role, license), epistemic substitutes (legitimate venues for the dissenting frame), and coordination substitutes (a community that absorbs defectors without isolation).

Pre-position substrate. The Mont Pelerin pattern works when the incumbent equilibrium relies on a physically or mathematically forced trajectory: demographic collapse, fiscal arithmetic, energy throughput, thermodynamic constraints. Cultural and aesthetic captures can mutate indefinitely; physical and demographic constraints cannot. Mont Pelerin bet against central planning, which had a thermodynamic problem. The bet against fashion is the wrong bet.

Wait for triggers; do not wait passively. Pre-position the substrate. Amplify readiness for the trigger when it comes: make the cluster’s failures more visible to outsiders, structure penalties so the cluster cannot externalize them, support elite defection where it appears. Triggers are not creators of alternatives; they are revealers of alternatives that have already been built.

VIII. Close

The captured insider does not refuse the better frame. The frame is missing from her board. The repair task is not to explain harder; it is to change the board, or build another one.