Part IV: Social Reality

Coordination Agents

Existing Theory

Part III showed that ideological identification expands the self-model to include supra-individual patterns—nation, movement, religion, cause—and manages mortality terror by making that self-model partially immortal. The manifold framework asks what lives on the other end of the coupling. When many individuals expand their self-models to include a shared pattern, and the pattern begins regulating its substrate to ensure its own persistence—what is that pattern, exactly?

From Attractors to Agents

Begin with the weakest version. A coordination attractor is any persistent macro-scale pattern that stabilizes correlated behavior across many agents by shaping incentives, attention, and shared models. A language—it structures communication, resists individual modification, persists through speaker turnover. A dress code, a market convention, a social norm, a shared aesthetic. These patterns attract behavior without actively regulating their substrate. Stable because deviation is costly, not because the pattern acts to prevent it.

The campfire may be the oldest. For tens of thousands of years—plausibly hundreds of thousands—humans gathered around fire every night. Not occasionally. Every night. The fire was not merely warmth. It was the first coordination technology: a shared attentional anchor synchronizing perception across a group, a boundary between inside and outside, the temporal structure—the evening gathering—in which the day's events were collectively processed, the dead remembered, the absent imagined. Around the fire, certain stories stuck. Certain figures—the ancestor who found water, the elder who read the weather, the one who faced the predator—became narrative anchors around which other memories clustered. Not yet coordination agents. But coordination attractors with a distinctive property: they accumulated cultural information across generations, growing denser with each retelling, each figure pulling new material into its gravitational orbit.

Evolutionary coordination patterns sit between attractors and agents. They adapt because variants are differentially retained, even when no integrated macro-policy represents a threat and selects a response. Languages, ecosystems, and decentralized market conventions can evolve in this sense without yet qualifying as agents. Adaptation alone is therefore insufficient.

But some coordination attractors do more. They do not merely attract behavior—they act to preserve themselves through their substrate. A religion that modifies its doctrine to survive a new cultural environment is adapting, not passively attracting believers. A market that lobbies for deregulation is reshaping the conditions of its own persistence. A nation that educates children in its founding myths is manufacturing future substrate.

A portrait composed entirely of fruits, flowers, and vegetables forming a human face — the whole has expression that no individual part possesses — Giuseppe Arcimboldo, *Vertumnus (Portrait of Rudolf II)*, 1591The coordination agent: a social-scale pattern whose parts cannot perceive the whole.

A collective pattern $G$ qualifies as a coordination agent at scale $\sigma$ only when persistence machinery is joined by a causally load-bearing macro-policy loop:

Persistence through substrate turnover. $G$ survives replacement of members.
Boundary maintenance. Inclusion, exclusion, and identity criteria are actively maintained.
Resource regulation. Information, resources, norms, or attention are routed in ways that preserve the pattern.
Substrate modification. The pattern changes constituent behavior or belief in persistence-relevant ways.
Adaptive response. It reconfigures under perturbation rather than merely being replaced by a fitter variant.
Macro-representational state. A distributed variable tracks the pattern's own condition or environment and is available to control.
Macro-motivational state. The system maintains a sufficiently stable ordering over macro-outcomes.
Decision procedure. Representation and motivation are transformed into coordinated action through an identifiable mechanism.
Reasons-responsive control. Intervening on a represented macro-reason changes macro-action after individual incentives and network structure are held fixed.

This taxonomy keeps different phenomena apart. A dress code is an attractor. A language is often an evolutionary coordination pattern. A corporation or state with integrated reporting, deliberation, and actuation may be a coordination agent. A decentralized market—or "capitalism" at civilizational scale—may instead be an ecology of agents plus attractor dynamics unless a unified macro-policy loop is identified. The label is earned case by case.

Superorganism applies at a further threshold: when a coordination agent's self-maintaining dynamics are complex enough, its adaptive responses flexible enough, its substrate regulation comprehensive enough that the analogy to biological organisms becomes not metaphorical but structural. It has something analogous to metabolism (resource extraction and allocation), an immune system (memetic defense, dissent suppression), a reproductive strategy (conversion, cultural transmission), and a viability manifold (conditions for persistence). Whether it has anything analogous to experience remains the open question this chapter will not pretend to close.

Many mature organizations are reflexively self-representing: org charts, financial statements, censuses, constitutions, intelligence reports, and doctrines encode claims about what the system is. They count as a macro self-model only when recurrently updated, available to the decision procedure, and causally load-bearing in control. That is measurable operational self-representation, not phenomenal self-awareness. Whether anything is felt remains separate.

Four Distinct Claims

The analysis rests on four claims at decreasing confidence. Keep them separate—conflating them is the primary source of both overclaim and dismissal.

Social ontology. Some macro-patterns are causally real at their scale: interventions on them alter outcomes, and the macro-variables can provide stable causal compression. Strongest claim.
Functional agency. Some of those patterns implement a macro-policy loop—representation, outcome ordering, decision, and coordinated actuation—not merely adaptation by selection. Strong where the mechanism is visible; unearned for civilization-sized abstractions lacking one.
Perceptibility. Holding $\alpha_A(G)$ high makes the pattern perceptible through an agent model. That stance is calibrated only when it improves complexity-penalized prediction or intervention. It says nothing by itself about $\alpha_P(G)$ or the entity's actual interior.
Macro-subjecthood. Conditional on the identity axiom, any irreducible macro causal structure receives at least epsilon phenomenality. The open question is whether a coordination agent forms a unified maximal complex at its own scale, with enough integration for rich valence or moral patiency, rather than remaining a collection of individual subjects linked through low-bandwidth channels.

Claims 1–2 are defended where their intervention and control conditions are met. Claim 4 is unresolved at the level that matters—unified macro-subjecthood. Claim 3 is a model-selection question. Define calibration evidence $B_G=\log P(D_G\mid M_{\text{agent}})-\log P(D_G\mid M_{\text{mechanistic}})$ : high $\alpha_A$ with high $B_G$ is calibrated agency perception; high $\alpha_A$ with low $B_G$ is projection; low $\alpha_A$ with high $B_G$ is macro-agent blindness.

The Debt the Agent-Stance Owes

The danger in this chapter is a specific slide, and naming it is the only protection. Claims 1 and 2 commit only to dynamics: a market is a real causal pattern (claim 1) with viability conditions and self-preserving tendencies (claim 2). Claim 3 says those dynamics are perceptible as agency when $\alpha$ is held high. The slide uses the perception claim 3 licenses to smuggle back the phenomenology only claim 4 could grant — from "the market exhibits a viability gradient" to "the market wants growth," from "the furnace metaphor tracks a competitive dynamic" to "the furnace burns," from "demon was the low- $\alpha$ -era name for a parasitic dynamic" to "demons are malevolent agents." Each re-imports interiority the formal commitments never bought. The prose that follows uses such language; read it as stance-talk, not ontology, until the debt below is paid.

The agent-stance is not tested against a caricature in which individuals are independent. Its competitor must be the best available micro/network/institutional model, including repeated games, information asymmetries, selection, and explicit organizations. Add a macro-agent state only if it produces a complexity-penalized gain on held-out perturbations, stable across substrate turnover, and identifies interventions that the lower-level model misses.

A candidate test compares $M_\mu$ , the strongest micro/network model, with $M_G$ , the same model augmented by a macro-state, macro-policy, and viability margin. The stance earns its keep when $\Delta_{\text{agent}}=\operatorname{NLL}(M_\mu)-\operatorname{NLL}(M_G)-\lambda\Delta\operatorname{Complexity}>0$ on out-of-distribution threats and when interventions on the macro-state transfer across member turnover. Boundary-locked coordination is one candidate signal, but individual incentives must be modeled rather than assumed absent.

What a Coordination Agent Needs

Like any self-maintaining system, a coordination agent has conditions it needs to persist. Its viability manifold $\viable_G$ includes: belief propagation rate (recruitment must exceed attrition—the pattern starves if it stops converting), practice maintenance (rituals performed with sufficient frequency and fidelity), resource adequacy (material support for infrastructure), memetic defense (resistance to competing patterns—the pattern's immune system), and adaptive capacity (ability to update to environmental change).

A religion losing members is approaching its viability boundary. A growing ideology is expanding its viable region. A corporation restructuring after a market shift is performing exactly the same operation as an organism adjusting to environmental change: reconfiguring internal dynamics to remain within the viable region of state space.

Coordination agent lifecycle — individual agents cycle in and out while the central pattern persists across substrate turnover.

Part III introduced ideology as the individual-level mechanism: expand the self-model to include a supra-individual pattern, gain a longer viability horizon, manage mortality terror. The coordination agent lives on the other side of that coupling. When many individuals identify with the same pattern, it acquires an aggregate viability manifold reducible to no individual's. Its persistence requires the substrate to maintain certain beliefs, perform certain practices, allocate certain resources, suppress certain competitors. The individual's identification is the mechanism; the coordination agent is the beneficiary. Part III's warning about parasitic ideology was not premature—it described, at the individual level, a dynamic that operates at the collective level: the expanded self-model can be exploited by the pattern it includes.

Ritual as Metabolism

Part III examined how religious practices serve human affect regulation. From the coordination agent’s perspective, rituals are not worship but metabolism — the rhythmic process by which the pattern feeds, repairs, and replicates: substrate maintenance (keeping humans in states conducive to pattern persistence), belief reinforcement (repeated practice strengthening commitments), social bonding (cohesion, raised barriers to exit), resource extraction (offerings, tithes, labor), signal propagation (public ritual advertising presence), dissent suppression (participation marking deviants for correction), and — most fundamentally — attention direction (where the substrate looks, what enters the collective stream, what gets broken down and absorbed). Attention direction is less a metabolic function than the digestive medium itself. What the pattern attends to through its substrate, it metabolizes. What falls outside collective attention starves. The sermon, the feed, the curriculum, the news cycle — each a digestive organ, converting raw world into forms the pattern can absorb. The distinction: a ritual is aligned if it serves both human flourishing and pattern persistence, exploitative if it serves persistence at human cost. Meditation is aligned — it benefits humans AND maintains the pattern. Extreme fasting, self-harm, warfare are exploitative.

At low $\alpha_A(\text{market})$ , the market is modeled as decomposable dynamics. At high $\alpha_A(\text{market})$ , it is modeled through goals, constraints, and policy-like responses. Both can expose real structure; neither receives parity by declaration. The agent model must win the calibration test above. "Wants," "punishes," and "rewards" remain stance-talk for measurable directional tendencies unless macro-subjecthood is independently supported.

This separates two questions often fused in philosophy of religion. A person can genuinely perceive a represented deity or religious institution through a high- $\alpha_A$ agent model; that is a fact about the perceiver and the coordination dynamics. Whether the represented deity has a referent outside the social system, and whether any relevant macro-pattern is a phenomenal subject, are separate claims. Taking the experience seriously therefore requires neither naive realism nor eliminativism.

A rationalist who assigns $\alpha_A(\text{market})\approx0$ may miss a useful macro-policy model; a participatory perceiver who assigns it near 1 may anthropomorphize an attractor or ecology. Rigidity in either direction is the error. Calibration is adjudicated by differential prediction, causal intervention, and willingness to lower or raise ascription when the evidence changes.

Historical vocabularies should be translated without collapsing distinct referents. For any named deity $g$ , distinguish $D_g$ , the culturally represented deity-model; $C_g$ , the religious coordination pattern composed of adherents, rituals, texts, and institutions; and $E_g$ , any alleged extra-social referent. This framework can strongly analyze $D_g$ and $C_g$ while remaining agnostic about $E_g$ .

Gods: represented large-scale agents $D_g$ , often coupled to religious coordination agents $C_g$ , perceived at high $\alpha_A$ .
Spirits: represented localized agency associated with place, lineage, practice, or event; the corresponding social pattern may be a local attractor without being the represented spirit itself.
Demons: historical phenomenology of threatening or substrate-opposed agency. In the analytical vocabulary, use structurally parasitic only when persistence requires avoidable substrate harm under the formal test below.
Angels: historical phenomenology of protective or aligned agency. In the analytical vocabulary, use mutualistic or symbiotic when the counterfactual viability conditions are met.
Rituals: synchronization and retrieval protocols that may regulate both participants and the coordination pattern they instantiate.

This is not an argument that religions were secretly right or secretly wrong. An animist and an ecologist may foreground different structures in the same forest. Where the models make compatible predictions, they can be complementary registers; where they conflict, causal and predictive adequacy adjudicate. Neither perceptual richness nor mechanistic austerity is exempt from calibration.

Founding figures can function as founder-indices or narrative attractors inside a coordination agent: compressed addresses under which laws, stories, practices, and interpretations accumulate. Whether Abraham, Moses, the Buddha, or Muhammad is represented historically, compositely, or archetypally matters for historical claims; the structural claim is narrower—that a salient figure can index and retrieve a distributed cultural state across generations.

One major function of religion is cultural memory given institutional form. Scriptures compress selected histories and interpretations; rituals retrieve and re-enact them; founder-indices organize access. Religion is not exhausted by this function, and the represented deity is not identical to the institution that transmits it. The distinction among representation, coordination system, and alleged external referent remains in force.

Perception Across Registers

Charisma as multi-manifold coherence. Charismatic people produce the impression of simultaneous alignment across multiple manifolds—your friend, your ally, your source of meaning—all at once, without the gradient conflicts that normally arise. Whether this reflects genuine alignment or sophisticated mimicry is exactly what distinguishes the aligned leader from the cult leader. The affect system registers both as positive, which is why charisma is dangerous: it disarms the detection system.

Being "seen" as manifold recognition. A specific affect signature—warmth, relief, sometimes tears—fires when another person accurately perceives the manifold you are on. Not the one you perform, not the one you wish you were on, but the one you actually inhabit. The relief is the detection system registering: someone is tracking reality here. This is why good therapy works, why genuine friendship heals, why one moment of real recognition from a stranger can stay with you for years.

Ritual as Measurement Synchronization

In the trajectory-selection framework (Part I), collective patterns become observable not because something new enters existence but because the observer's attention expands to sample at the scale where the pattern operates. Ritual works partly by synchronizing the collective's measurement distribution—coordinating where participants direct attention, what temporal markers they share, what affective states they enter together. A synchronized collective measures at the collective scale, and what it measures, it becomes correlated with. When ritual attention weakens, the coordination agent does not cease to exist; the distributed attention pattern that constituted its observability has dissolved.

This logic extends to communication between observers. When observer $A$ reports an observation to observer $B$ , $B$ 's future trajectory becomes constrained by that report—weighted by trust:

P_B(x\mid r_A)\propto P_B(x)\,P(r_A\mid x,\tau_{AB},\iota_A)

A report constrains another observer in proportion to the reporter's reliability, access, and incentives. Collective belief therefore depends not only on trust but on evidence dependence: one observation repeated through a network is not many independent observations. Religious testimony, scientific consensus, news, and rumor differ in provenance tracking, correction, and correlation structure. Control of testimony matters because it shapes the collective posterior, but the update is Bayesian only when those dependencies are represented.