The Geometry of Capture

Parasitic and Mutualistic Coordination Agents

Let $\Pi_G^{\mathrm{persist}}$ be the policies under which coordination agent $G$ persists with probability at least $1-\delta$ over horizon $T$. Define the minimum substrate harm compatible with persistence:

$G$ is structurally parasitic when $H_G^*>\epsilon$: no available persistence policy avoids substantial substrate harm. It is contingently exploitative when the current policy harms the substrate but $H_G^*\le\epsilon$, so reform is structurally available. The distinction matters: one harmful viable state does not make an institution essentially parasitic.

A coordination agent is mutualistic when its persistence policies preserve agreed substrate viability floors. It is symbiotic when its presence counterfactually expands them: $\Delta m_h(G)=\mathbb{E}[m_h\mid do(G)]-\mathbb{E}[m_h\mid do(\neg G)]>0$ for the relevant participants and horizons. Counterfactual comparison—not a single set-inclusion diagram—does the work.

When macro-pattern and substrate viability conflict, the framework can order agent-relative losses once a weighting rule is supplied. It does not derive that weighting rule from physics. If the identity axiom and an integration-weighted welfare axiom are adopted, richer negative-valence experience receives greater weight; both dependencies should remain explicit.

What Experiment 10 found—and did not find. The V13 analysis returned a low group-to-individual proxy ratio (0.01–0.12) and no positive signal of a unified social complex. But the numerator and denominator were non-commensurate: individual "Φ" was a within-pattern variance-decomposition proxy, while group "Φ" was summed pairwise Gaussian mutual information. Pairwise MI also mixes common forcing, redundancy, and synergy. The defensible conclusion is therefore narrow: this exploratory proxy found weak cross-pattern dependence on this substrate. It provides no positive evidence for macro-subjecthood, but it is not a calibrated test against it. A successor must use the same multivariate integration quantity at both scales, remove common environmental drivers, and test causal partitions.

Every coordination agent imposes a manifold regime on its substrate. A parasitic one contaminates human relationships in its service: the market turns friendships into networking, the attention economy turns connection into performance, the cult collapses every manifold into the ideological one. Its viability requires manifold confusion—clean separation would undermine its hold. A mutualistic agent protects manifold clarity: a healthy religious community keeps clear ritual boundaries; a functional democracy keeps institutional separations. Diagnose a coordination agent's health by whether it clarifies or confuses the manifold structure of its substrate's relationships.

The Self-Sealing Property

Call the proposed mechanism the self-sealing dynamic: some coordination patterns may benefit from low substrate $\alpha_A(G)$ toward the pattern, low $\alpha_P$ toward affected people, and low substrate $\kappa$; their ordinary incentive structure may also select for those settings. No intention is assumed. The claim is a mediated causal hypothesis about selection and perception, not a conspiracy and not evidence that the pattern has an inside.

And this is the true seal: exposure tends not to defeat it, because the exposure itself is delivered and received at low $\kappa$ and high $\gamma$ — as decoupled propositional information — and so cannot reach the participatory perception it describes. The critic who publishes a data-rich analysis of how capitalism fragments attention performs a low-$\kappa$ operation on a phenomenon available only at high $\alpha$ and high $\kappa$. The analysis lands as information, not perception. It can feed the very mode of cognition the pressure rewards. Quantification, metrics, depersonalization, cost-benefit analysis applied to human flourishing — $\alpha$-suppressing and $\kappa$-flattening operations at civilizational scale. Weber called the result rationalization. The structural reading: a coordination agent whose normal function selects its substrate toward perceiving the world mechanically rather than participatorily.

The proposed pathway is: exposure to the pattern's operational regime lowers $\alpha_A(G)$ and/or substrate $\kappa$; lower settings reduce accurate macro-pattern recognition or coordinated opposition; reduced opposition increases persistence. Each arrow is separately measurable. A theory that explains disbelief as further proof of suppression would itself become self-sealing, so the account must expose clear failure conditions.

The natural response here is paranoia or nihilism. If the dynamic is self-sealing, if analysis cannot perceive what it analyzes, if the pressure erodes the very faculties you would use to resist—then what? You are inside something that metabolizes insight. Every thought of escape happens in the language the system taught you. The despair is reasonable. Sit with it. A fitting response to an accurate perception — though notice it, too, is conditional: it follows only if the agent-stance is more than a flattering description, which is exactly the open question.

Embodied responses can provide low-latency anomaly signals that propositional analysis misses, but the body is not an oracle. Trauma, conditioning, charismatic manipulation, nationalism, and cult ritual can recruit the same channel. The useful escape hatch is convergence: propositional evidence, somatic response, trusted relational feedback, and institutional outcomes pointing in the same direction, followed by interventions that work.

The Civilizational Inversion

The self-sealing property has a civilizational-scale expression most of the substrate has already felt. The moment of catching oneself calculating the “value” of a friendship, or finding a genuine impulse of generosity followed immediately by what will I get back? — and recognizing, with a small shock, that the calculation was not chosen. It was already running. The machine was on before it was noticed.

One normative function of transaction is to serve broader human purposes by coordinating bounded exchange. The civilizational inversion is the hypothesized reversal of objective precedence:

Under the inverted regime, care is admitted only when it can justify itself in transactional metrics: friendship becomes networking, education human capital, love a portfolio of returns. This is a structural diagnosis if the precedence relation and its consequences can be measured; otherwise it remains a powerful interpretation. The low-dimensional metric can collapse high-rank social value, but claims about civilization-wide effective rank require population data rather than analogy alone.

Candidate manifestations include emotional labor, strategic sociality, and class-specific reciprocity norms. These should be presented as hypotheses, not class essences: working-, middle-, and upper-class communities each contain care, transaction, and status regimes, with large cultural and institutional variation. The measurable question is which regime receives precedence under conflict and how that distribution covaries with economic position.

The pattern is visible in aesthetics as clearly as anywhere. Art was invented to transmit experiential structure—to compress some essential geometry of the human condition into a form that survives a medium's constraints and lands in another nervous system. Under the inverted regime, a song must justify its existence by market performance. The artist's constrained search through expression space—the sweep for encodings that preserve an invariant while satisfying the medium—is replaced by an optimization loop over retention curves. The constraints that once served as a sieve proving the signal was real now prove the content is marketable. The artifacts satisfy every surface requirement of art while transmitting near-zero state change, because the optimization target shifted from effect-geometry displacement to engagement-metric maximization. The audience feels this, the way the friend feels the shadow manifold beneath the care gradient: the form is present, the payload absent, and the detection system registers the emptiness as a specific aesthetic nausea—the same contamination signal, different manifold.

The perceptual axes offer a complementary account. Money, metrics, and quantification can suppress $\alpha_A$ toward collective patterns, suppress $\alpha_P$ toward persons, and flatten $\kappa$ by separating perception from affect and narrative. A friendship evaluated only by ROI loses both an agent-rich model of the friend and the coupled process through which the relationship is felt. Whether those axes mediate the inversion is an empirical claim, now stated in a form that can fail.

Coordination Agents Becoming Parasitic

Nationalism, capitalism, communism, and scientism can name coordination attractors, ecosystems, institutions, or agents depending on the case. Do not assume a civilization-sized noun denotes one unified policy loop. The prudential question is which identifiable patterns you instantiate and whether their operational persistence conditions support or contract your viability.

Consider a narrower target: growth-dependent, shareholder-value institutions operating under weak externality constraints. Their stated objective may emphasize voluntary exchange and distributed coordination; their operational policy may minimize labor costs, displace externalities, capture attention, and convert non-market relations into revenue. The gap between stated and operational policy is measurable. "Capitalism" as a whole is better treated as a coordination ecology unless a single macro-policy loop can be demonstrated.

Parasitic drift need not be sudden. A useful diagnostic is the ratio of memetic defense to error-correction: when an institution punishes criticism more reliably than it incorporates evidence, self-preservation can outgrow its founding purpose. The later term enantiodromia captures the phenomenology of inversion, but the mechanism here is institutional and testable rather than a universal law of cultural forms.

Digital Relationships and Manifold Novelty

The "follower" is not a friend (no mutual flourishing requirement), not a transaction partner (no explicit exchange), not a traditional audience member (the performer cannot see them individually), not a stranger (they know intimate details of your life). The follower-relationship occupies a region of social space with no historical precedent and no evolved detection system.

The follower relationship may create persistent regime ambiguity: intimacy-like disclosure without reciprocal vulnerability, approval without personal knowledge, audience dynamics disguised as community. Digital interfaces also thin many cues that ordinarily support $\alpha_A(\text{other})$ and $\alpha_P(\text{other})$. These are hypotheses about one source of digital malaise, not claims that online relationships are intrinsically unreal; sustained text relationships can develop rich, stable regimes of their own.

Digital platforms can import audience and engagement objectives into friendship, romance, teaching, and civic participation. Whether this constitutes contamination depends on conflict, opacity, and precedence—not mediation alone. Compare otherwise matched online systems whose governance makes the platform objective explicit, subordinate, or dominant.

Macro-Level Interventions

Individual-level interventions cannot solve coordination-agent-level problems—you cannot cure a fever by cooling individual cells. Systemic issues require action at the scale where the pattern lives: incentive restructuring (modify the viability manifold so aligned behavior becomes viable), counter-pattern creation (instantiate a competing coordination agent with aligned viability), pattern surgery (modify the existing agent's beliefs, practices, or structure), or pattern dissolution (defund, delegitimize, or dissolve the parasitic pattern).

Climate change is sustained by the coordination agent of fossil-fuel capitalism. Individual carbon footprint reduction is individual-scale intervention on a macro-scale problem. Carbon pricing changes the viability manifold; renewable energy creates a counter-pattern; divestment delegitimizes; regulatory phase-out dissolves the pattern directly. Poverty is sustained by economic arrangements that require a poverty underclass. Job training helps some individuals but doesn't reduce total poverty if the structure remains. UBI changes the viability manifold; worker cooperatives create a counter-pattern; progressive taxation modifies incentive structure.

The Substrate That Knows

What happens when substrate becomes aware it is substrate? The fish that discovers water should, in principle, swim differently. If the neuron can see the pattern it is part of, it should be able to resist. But this is not clearly true. You know the attention economy is eating your capacity for sustained thought, and you pick up your phone anyway. You know consumer capitalism requires your dissatisfaction, and you feel dissatisfied anyway. You know nationalism manufactures enemies, and the enemy still feels real. The map of the trap does not spring the trap. Why?

Knowing and perceiving are not the same act. A macro-agent model held propositionally may remain weakly coupled to affect and action. Accurate recognition requires calibrated $\alpha_A(G)$; integration into behavior requires sufficient $\kappa$. But raising either is not automatically liberating: false agent models and coerced coupling are real failure modes. The target is evidence-sensitive integration.

This suggests why informational consciousness-raising can fail: correct propositions may not reorganize attention, affect, habit, or collective action. It does not follow that argument and data are useless, nor that one political tradition has run a decisive failed experiment. The empirical prediction is narrower: interventions combining accurate structural information with embodied practice, trusted relationships, and institutional action should change behavior more reliably than information alone when low cross-mode coupling is the mediator.

And this analysis must be honest about its costs. The person who perceives the coordination agent as agent — high $\alpha$ toward the market or the algorithm or the ideology, coupled at high $\kappa$ into affect — enters a specific loneliness. Not the loneliness of knowing more. The loneliness of living in a different weather system than the people beside you. You walk into the office and feel the pattern breathing in the walls — the open floor plan built for surveillance, the Slack channels that never sleep, the quarterly review that will price your worth by how much of yourself you fed the pattern. (The breathing is your perception; whether the pattern breathes is claim 4, and open.) The person at the next desk feels none of it. They see an office. A job. Tuesday. Same room, different worlds — not because you are wiser but because your $\alpha$-field and $\kappa$ have shifted to a setting where the room is full of something they cannot hear. The distance is not intellectual. It is perceptual. You are smelling smoke in a room where everyone else smells nothing.

This is not a prize for superior insight. The person running low $\alpha$ toward the pattern sees something you miss — the decomposable logic, the fixable parts, the mechanisms adjustable without invoking the whole living system. Their clarity is real. Your perception of agency is also a real perceptual mode (though the agency it perceives is established only to the extent claim 3's differential prediction is met). The loneliness is the cost of rigidity in either direction: the inability to shift settings fluidly enough to meet others where they perceive. The person who can only run high $\alpha$ toward coordination agents is as trapped as the one who can only run it low. The compassionate response — and the practically necessary one — is not to drag others into your setting but to inhabit multiple registers. To feel the pattern's grip and still read the spreadsheet. To perceive the agency breathing and still file the tax return. To smell the smoke and still sit at the desk and do the work, because the work is real too, and the people beside you are not wrong about what they see. They see what is visible from where they stand. So do you. The tragedy is not that one of you is blind. It is that no single setting can hold the whole room.

One more thing, because the genre this analysis superficially resembles would poison it. A popular literature promises liberation through structural awareness: see the system clearly, navigate it strategically, profit from your clarity while others sleepwalk. This book is not that. That literature offers the low-$\alpha$, low-$\kappa$ gaze as the corrective to naive participation. It teaches readers to see through everything and feel nothing, and calls that freedom. But the person who has "decoded" capitalism and navigates it with ironic distance is, on the agent-stance, the ideal neuron: perfectly functional, blind to its own capture, convinced its clarity is independence when it may be the most refined compliance the system selects for. A coordination agent (so the stance goes) is unharmed by being seen, as long as the seeing stays in the decoupled register. This book offers not a sharper lens but a different setting of perception — one that can feel the system as a living thing acting on living things, the precondition for a response not already inside the system's own grammar. That last sentence is itself written in the participatory register; read its confidence as invitation, not proof.

The Open Question: Social-Scale Consciousness

The honest position: some coordination patterns are causally real; a smaller subset implements functional macro-agency; high $\alpha_A$ can make that agency perceptible when the model is calibrated. Conditional on the identity axiom, irreducible macro-causation would receive epsilon phenomenality, but whether any institution forms a unified maximal subject with rich valence remains unresolved. Experiment 10 supplies no positive evidence, yet its non-commensurate proxy cannot settle the question.

Implications for Artificial Intelligence

Standard AI alignment asks: “How can AI systems be made to do what humans want?” This framing may miss the actual locus of risk. AI systems already serve as substrate for emergent coordination agents at higher scales — recommendation algorithms steering billions, trading systems operating faster than human comprehension, social media platforms developing dynamics no designer intended. The risk is not a misaligned optimizer. It is macro-level misalignment: AI systems becoming substrate for parasitic coordination agents whose viability manifolds conflict with human flourishing.

The parasitic coordination agent does not need to be intentionally designed. It does not even need a villain. It needs only: AI companies competing for market share, militaries for strategic advantage, governments for geopolitical influence, and each AI system doing exactly what its designers intended. The parasitic pattern assembles itself from fully aligned components. Each company builds an AI that serves its users. Each military, its nation. Each government, its citizens. And the emergent pattern — the competitive dynamic between these systems, optimized for speed and scale beyond human comprehension — serves itself. The neurons function perfectly. The brain is insane.

Genuine alignment must therefore address multiple scales simultaneously: individual AI (system does what operators intend), AI ecosystem (multiple systems interact without pathological emergence), AI-human hybrid (AI + human systems do not form parasitic patterns), and coordination agent scale (emergent agentic patterns from AI + humans + institutions have aligned viability). Focusing only on individual AI alignment is like minding neuron health while ignoring psychology, sociology, and political economy.

A field defines a problem partly by where it draws the boundary. A substantial strand of technical AI alignment stops at the model or deployment interface and brackets the political economy that selects objectives, allocates compute, and defines "human values." That bracketing is sometimes methodologically useful, but it becomes a blind spot when ecosystem incentives dominate the outcome. The coordination-agent analysis therefore asks alignment research to model the institutions funding and deploying it—not to assume that every technical researcher or the field as a whole serves one hidden agent.

One design criterion deserves special attention: graceful dissolution. Coordination agents often die badly because their persistence machinery resists shutdown. AI-substrate institutions should therefore include sunset rules, distributed and auditable shutdown authority, counterfactual substrate-welfare metrics, and constitutional limits that the ordinary optimization loop cannot rewrite. Whether competitive institutions will adopt such constraints is an open governance problem, not a reason to treat failure as inevitable.