Bayesian Belief Dynamics

---

Chapter 13: Bayesian Belief Dynamics

A Formal Account of Confabulation Compounding, Collective Convergence, and the Subcriticality Threshold

---

Chapter Opening

Position Statement: This chapter establishes a formal Bayesian account of how confabulated Belief compounds within an Individual and converges across a coupled population, and it defines the Subcriticality Threshold that separates a disturbance a population absorbs from a cascade that locks in. The updating that produces the lock-in is valid updating: the error is in the evidence, not in the inference performed on it.

Dependency Statement: This chapter requires prior understanding of:

• The three-source taxonomy (A-I-M) and why motivation requires disaggregation (Chapter 4)
• Source Opacity as a structural feature of neural integration: the loss of source-tags during common-currency processing (Chapter 9)
• Confabulation as the mechanism by which Individuals explain their opaque desires using plausible but unreliable narratives (Chapter 11)

Derivation Preview: The chapter's central claims (the Bayesian Mimetic Crisis Principle, the Floor-Breach Ignition Corollary, and the Subcriticality Threshold) are derived from four axiomatic foundations:

1. Axiom 3 (Source Opacity): After common-currency integration, Individuals cannot reliably introspect which source (A, I, or M) generated a given desire
2. Axiom 4 (Confabulation): The narrative system generates plausible explanations for desires regardless of actual source, with systematic bias toward authentic-sounding attributions
3. Standard Bayesian updating: Individuals rationally revise Beliefs in light of evidence, treating high-likelihood evidence as confirmatory
4. Social proof as evidence: Individuals treat others' expressed Beliefs as informative evidence for their own Belief-updating

Chapter Preview: The chapter demonstrates:

• How valid Bayesian reasoning, applied to Confabulated evidence, produces systematic convergence on false Beliefs (the Individual Confabulation Cascade)
• How individual cascades synchronise through social proof to produce population-level convergence (Collective Dynamics)
• Why M-coupled Beliefs lock in above the Subcriticality Threshold by an operation indifferent to their accuracy, which is what makes the Terminal Phase of the Crisis State Architecture an attractor region in the parameter space (Bayesian Mimetic Crisis Principle; the attractor result itself is the Unifying Crisis Principle, derived at Chapter 17, Section 17.2)
• Why the Bayesian machinery is content-agnostic, explaining both the infinite variation in historical crisis content and why comparative analysis failed to identify the common architecture (Content Independence)
• Why sustained deprivation at the worst-off segment raises amplification and lowers correction at once, so that a unique critical depth exists at which the Subcriticality Condition fails (Floor-Breach Ignition Corollary)
• Why population-level intervention is necessary and not sufficient, because a cascade inside one Decision Hub needs no population to complete it (the Leadership Compression Principle that states the condition is derived at Chapter 17, Section 17.7)
• How the same Bayesian machinery that produces crisis also permits prevention, through parameter shifts that maintain the system below the Subcriticality Threshold
• Why the crisis architecture is overdetermined by minimal conditions more general than AIM's specific axioms, making the threat robust across theoretical frameworks (Robustness Analysis)

---

13.1 Introduction: Why Bayesian Formalism?

Confabulation is not a failure of reasoning. An Individual who cannot introspect the source of a desire, and who accounts for it from the two channels that remain legible, is producing exactly the account the architecture makes available; the account then enters the Belief Register as evidence, and is weighted there by a process doing nothing wrong. This is what makes a formalism necessary. Ordinary description can say that people deceive themselves and that the deception spreads. It cannot say why the spread is convergent rather than dispersive, why it accelerates instead of settling, or where the line falls between a disturbance a population absorbs and one from which it does not return. Those are quantitative questions about an updating process, and Bayesian formalism is the instrument that answers them.

The formal specification sits in Appendix A: the slow Bayesian layer at A.5.6 and the network extension at A.5.7. Where this chapter states a threshold, an amplification rate, or a condition on how Evidence propagates between Individuals, the corresponding Appendix locus carries the dynamical-systems statement of it, and the two treatments are of one object. One result of the pairing governs what the architecture licenses as intervention: the conditions under which a coupled population crosses into cascade and the conditions under which it is held below that crossing are the same parameters running in opposite directions. Crisis and prevention are not two mechanisms. They are one mechanism at two signs.

The argument proceeds through six stages. First, the Individual Confabulation Cascade shows how a single person's Authenticity Belief, their confidence that their desires reflect genuine A or I sources rather than Mimetic transmission, updates toward certainty for the wrong Hypothesis through self-reinforcing feedback. Second, the Collective Convergence Mechanism demonstrates how Individual cascades synchronise through social proof, producing population-level Belief shifts that accelerate rather than correct the underlying error. This section introduces two structural parameters, Coupling Strength (κ) and Decoherence Time (τ), that govern whether Bayesian updating in a coupled system tracks Truth or merely tracks the system's own dynamics, and presents a Domain Matrix classifying epistemic regimes by their Coupling Strength-Decoherence Time profile.

Third, the Bayesian Mimetic Crisis Principle establishes that M-coupled Beliefs compound through the population by a lock-in process indifferent to their accuracy once Coupling Strength exceeds the Subcriticality Threshold, which is the condition under which the Terminal Phase of the Crisis State Architecture becomes an attractor region in the parameter space, and that the principle's Content-independence explains why comparative historical analysis has failed to identify the common crisis architecture. Fourth, the chapter marks the limit of its own account: crisis does not require population-level convergence, because concentrated authority allows M-Captured leadership to impose crisis dynamics regardless of where the collective posterior sits, and the Leadership Compression Principle stating that condition is derived at Chapter 17, Section 17.7. Fifth, the Subcriticality Threshold identifies the parameter space within which corrective mechanisms can outpace mimetic amplification, providing the formal foundation for intervention architecture. Sixth, the Robustness Analysis demonstrates that crisis dynamics depend on minimal sufficient conditions that are more general than AIM's specific axioms, strengthening rather than weakening the conclusion.

Not all evidence is epistemically suspect. Direct sensory observation, physical measurement, logical inference from established premises, and testimony from sources with no stake in the observer's Belief state remain reliable inputs to Bayesian updating. The problem identified here is specific: mimetically-transmitted information (Beliefs acquired through social channels where the transmitter's own Belief may be M-influenced, and where transmission is amplified by prestige cues, visibility, and social proof) deserves particular scrutiny before adoption. The Confabulation Cascade corrupts precisely this channel, leaving other evidential sources intact but insufficient to overcome socially-amplified false Belief.

---

13.2 The Bayesian Framework: Standard Machinery, Novel Application

13.2.1 Bayes' Theorem and Belief Updating

Bayesian inference provides the mathematical framework for rational Belief revision. For any Hypothesis H and evidence E:

where P(H) is the prior probability of H, P(E|H) is the likelihood of observing E given H, and P(H|E) is the posterior probability after incorporating E. P(E) is Evidence Probability: the overall probability of observing a particular piece of evidence regardless of which Source Hypothesis is true. This machinery is epistemically optimal when evidence is exogenous, generated independently of the Hypothesis being evaluated.

The AIM Framework applies this standard machinery to a novel domain: an Individual's Belief about the source of their own desires. Let H_authentic represent the Hypothesis "my desires authentically reflect A or I sources" and H_mimetic represent the alternative "my desires are primarily M-driven." Source Opacity ensures the Individual cannot directly observe which Hypothesis is true; the decomposition was discarded during common-currency integration. They must instead infer source from available evidence: the character of their desires, their behavioural patterns, and crucially, the explanations they generate for why they want what they want.

13.2.2 The Evidence Problem

Here the standard Bayesian framework encounters a structural difficulty. The primary evidence available to an Individual for evaluating their own motivational sources consists of:

1. Phenomenological experience: How the desire feels (urgent, authentic, chosen)
2. Behavioural observation: What they find themselves doing
3. Self-generated explanation: Why they believe they want what they want

Categories (1) and (2) are ambiguous; M-driven desires feel as urgent and authentic as A or I desires after integration, and a single act driven by M looks like one driven by A or I; the difference shows only in how the behaviour runs under conditions that move one contribution and not the others (deficit corrected, audience removed, task completed). Category (3) is the critical failure point: self-generated explanations are Confabulations: plausible narratives constructed post-hoc by the narrative system, which has no privileged access to actual motivational sources.

The Confabulation is not a lie. The Individual genuinely believes their explanation. But the explanation is generated to account for the integrated output ("I want X") using culturally available narrative templates ("because it aligns with my values," "because I genuinely need it," "because it reflects who I am"). These templates systematically favour authentic-sounding explanations because authentic motivation is more socially valued than mimetic motivation. The narrative system generates explanations that make the Individual look good, to themselves and others, regardless of actual source.

This produces a specific evidence structure: the primary evidence available for self-assessment is endogenously generated by the very system being assessed, and that generation process is biased toward confirming H_authentic regardless of Truth. That evidence structure is Endogenous Feedback: the Individual's Beliefs about Hypotheses systematically shape the distribution of evidence the Individual observes, so evidence is co-produced by the Prior Beliefs being tested. That is the Problem of the Priors: classical washing-out assumed independent evidence would dominate initial probability assignments, and the assumption fails under Source Opacity because the evidence is endogenously generated through Confabulation.

---

13.3 The Individual Confabulation Cascade

The fast valuation layer and the Bayesian layer this chapter develops run on different timescales and feed into one another. The Decision Hub's Common-Currency Scalar (the Scalar) reaches Conscious Cognition as Evidence in three stages. The Scalar-to-Likelihood Lift expands the single number back into a set of likelihoods: how expected that degree of wanting would be under each explanation the mind can represent. The Belief is updated on those likelihoods. The settled Belief is then emitted back to the fast layer, where it adjusts the weight each motivational source carries. Stability depends on all three gains together (the fast loop, the slow Belief loop, and the coupling between them) held below a joint threshold. The slow-layer gain is strictly positive wherever the Thinker is recruited, because Self-Coupling Strength never falls to zero, and on the inflated branch recall of the unresolved question keeps the Thinker recruited, which is why Bayesian Lock-In is structural rather than contingent. Three coupling parameters await empirical calibration: Self-Coupling Strength, the Restriction-Induced Concentration Factor, and the Evidence-Emission Coefficient. Appendix A.5.6 carries the formal specification.

13.3.1 The Self-Reinforcing Loop

Consider an Individual who begins in genuine uncertainty, giving even odds either way on whether their desires reflect authentic sources. The complementary residual on the mimetic Hypothesis is Mimetic Recognition: the probability an Individual assigns to the Hypothesis that their causal narrative about a want might be incomplete or structurally generated rather than reflecting the want's true origin. They experience a want, generate an explanation for that want, and treat that explanation as evidence for source-assessment.

If the explanation sounds authentic ("I want this promotion because I care about the work"), the Individual updates:

The explanation counts as evidence for authenticity because an authentic-sounding account is more expected if the desire really is authentic than if it is mimetic. The inference is sound on its face. What defeats it is that Confabulation generates authentic-sounding accounts at the same rate whichever the true source was, so the tilt toward authenticity is an artefact of the generating process rather than a fact about the desire.

The first update therefore moves the Authenticity Belief above even odds.

At t₁, the Individual experiences another want, generates another explanation. But now their prior has shifted; they are more confident in their own authenticity. This shifted prior affects both the new explanation generated (the narrative system draws on the "I am authentic" self-model) and the interpretation of that explanation (confirming evidence is weighted more heavily than disconfirming evidence through standard confirmation bias).

The second update moves it higher again, above where the first left it.

13.3.2 Cascade Dynamics

Iterating this process across many decisions and explanations produces systematic drift of the Authenticity Belief toward certainty. That iterative process is the Bayesian Confabulation Mechanism: Source Opacity produces stable false Beliefs about motivational source through narrative generation that becomes incorporated into the Self-Model as Bayesian priors for interpreting subsequent evidence. Each cycle:

1. The current self-model (increasingly confident in authenticity) shapes explanation generation
2. Generated explanations sound authentic (because the narrative system optimises for this)
3. Authentic-sounding explanations are treated as evidence for authenticity
4. The posterior shifts further toward H_authentic
5. The updated posterior becomes the prior for the next cycle

The interpreted wanting that enters each cycle is Evidence at Iteration: each Posterior becomes the Prior for interpreting the next iteration.

The cascade has a direction; it moves toward higher Authenticity Belief, not toward Truth. This directionality emerges from the asymmetry in Confabulation: the narrative system generates authentic-sounding explanations regardless of actual source, but these explanations are interpreted as evidence for authenticity. The process is not symmetric; there is no corresponding mechanism generating mimetic-sounding explanations that would update toward H_mimetic.

13.3.3 The Individual Fixed Point

The cascade terminates at a Bayesian Fixed Point: a stable Belief state where the Authenticity Belief has approached certainty and the Individual's self-model has become resistant to revision. At this point:

• The prior Authenticity Belief is so high that even genuinely disconfirming evidence produces minimal posterior shift
• The explanation-generation system, drawing on the "I am authentic" self-model, produces increasingly elaborate justifications
• Alternative interpretations (perhaps this is M-driven) are dismissed as implausible given the accumulated "evidence"

The Fixed Point is stable in the technical sense: small perturbations (occasional doubt, external challenge) are absorbed without shifting the equilibrium. The Individual has achieved certainty about a Hypothesis that may be false, and that certainty was achieved through valid Bayesian updating on invalid (endogenously generated, systematically biased) evidence.

The Individual Fixed Point is not a state in which the Thinker falls silent. Ground-Resolution is invoked only by an open question (Chapter 4, Section 4.4.3), and a Belief held at near-certainty would seem to leave none. Structural Restlessness supplies one. Structural Restlessness is the architecturally necessary condition whereby stable, final satisfaction is unavailable to any Individual: an Appetite closes when its deficit is repaired and then recurs, a want for an Intrinsic Object closes when the Object is sufficient for the task and the next task brings its own, and Mimetic Desire has no closure event at all (Chapter 9, Section 9.7). An A-framed or I-framed account of a want predicts how strongly the want is felt, and that the want closes on a definite event: the repair of the deficit, or the acquisition of an Object sufficient for the task. The first prediction is confirmed on every cycle, because the Scalar carries no source-tag and matches the felt level whichever source drove it. The second is refuted wherever the want is M-driven: the closure event arrives and the want does not close, because each acquisition moves the comparison against which the next want is computed (Chapter 8, Section 8.5.2), and the felt deficit regenerates. Each refuted closure is a departure from what the Belief Register predicts, and the departure recruits the Thinker.

Between events the Thinker is recruited by recall. An unresolved question is an open question whether it arrives from the field or from memory, and recall falls most often on the most urgent unresolved questions, which is rumination: the want that should have closed and did not is returned to the Thinker without any fresh observation.

Each recruitment resolves onto the incumbent Hypothesis again. Where the Hypothesis that the want is Mimetic is absent from the Hypothesis Space or is not recalled, the refuted closure is absorbed by the next account the incumbent frame supplies ("the better version will do it"), and at near-certainty that resolution is written with full weight. Recruitment at the Individual Fixed Point therefore confirms the Belief rather than relieving it. The slow-layer gain accordingly stays positive at exactly the point where the Belief is most certain: the refuted closures that Structural Restlessness guarantees, and the rumination that follows them, keep the Thinker recruited, and every recruitment is a further write of Bayesian Lock-In (Appendix A, Section A.5.6).

13.3.4 Empirical Signatures

The Individual Confabulation Cascade generates testable predictions:

1. Confidence-accuracy divergence: Self-reported confidence in motivational self-knowledge should increase with age and decision-making experience, while accuracy (measured behaviourally) should not show corresponding improvement. The cascade predicts growing certainty without growing accuracy.
2. Explanation elaboration: Individuals further along the cascade should generate more elaborate, detailed justifications for their choices, not because they have better self-knowledge, but because their narrative systems have accumulated more material from which to construct explanations.
3. Resistance asymmetry: Challenges to the "I am authentic" self-model should produce more defensive response than challenges to specific Beliefs, because the self-model is protected by accumulated Bayesian confidence.
4. Choice blindness persistence: In experimental paradigms where choices are covertly switched (Johansson et al., 2005), participants generate confident explanations for choices they did not make. The cascade predicts this effect should be stronger in participants with longer decision-making histories: more accumulated confidence in their own authenticity.
5. Deficit-irrelevant attribution shift: The only quantity Conscious Cognition receives about a focal Object is the Decision Hub's integrated Scalar, and that Scalar carries the Body Manager's contribution whether or not the Body Manager's current deficit has anything to do with the Object. The cascade therefore predicts that a bodily deficit irrelevant to an Object changes the source attribution an Individual gives for wanting it: matched Individuals valuing the same Status-laden Object under induced hunger, thirst, or pain, against a sated baseline, should return systematically different posteriors over why they want it. Which way the attribution moves is a measurement rather than a prediction, since it turns on the state-dependence of the likelihood lift, but the presence of a difference follows from tag-stripping alone.

The criteria that would disconfirm these predictions are stated at Section 13.12.5, Predictions 1.1 to 1.4; the deficit-irrelevant attribution shift bears on Scalar-Only Delivery, which is stated with the third commitment of the falsification surface at Appendix B, Section B.2.

The Individual Confabulation Cascade has a formal articulation in Appendix A.5.6 as three axiom restatements drawn from Common-Currency Integration, Source Opacity, Confabulation, and Bayesian Updating. Together they specify that conscious cognition receives only the Decision Hub's integrated Scalar, that conscious cognition converts that Scalar into an evidence proposition by selecting the highest-posterior Hypothesis in the representable Hypothesis Space, and that the Belief Register updates on that evidence by canonical Bayesian updating, regardless of whether the evidence accurately decomposes the Scalar. The three restatements are the formal correspondent of the prose treatment in this section, the architectural reason individually rational Bayesian updating produces convergence on false Hypotheses about motivational source.

---

13.4 Collective Convergence: From Individual Cascades to Population Dynamics

13.4.1 The Social Proof Mechanism

Individual Confabulation Cascades do not operate in isolation. Individuals observe others' expressed Beliefs, stated preferences, and public behaviours, treating these observations as evidence for updating their own Beliefs. This produces a collective dynamic: Individual A's expressed Confabulation becomes evidence that Individual B incorporates into their own cascade, and vice versa.

The population-level counterpart is the shared level of expressed Belief: the share of expressed Beliefs, public narratives and cultural messages affirming authentic motivation. Each Individual updates their own Authenticity Belief using that shared level as evidence:

A high shared level is itself strong evidence: if most people affirm authentic motivation, this is more expected under the Hypothesis that authentic motivation is common than under the Hypothesis that people are systematically self-deceived.

But that shared level is itself nothing more than the aggregate of the Individuals' own Beliefs. The system is reflexive: the shared level affects individual updating, and individual Beliefs constitute the shared level. What makes it more than an average is the coupling: each Individual's publicly expressed Confabulation functions as social proof that updates other Individuals' Prior Beliefs, producing synchronised convergence rather than independent averaging.

13.4.2 Convergence Dynamics

This reflexive structure produces convergence dynamics. The distribution those updates begin from is the Collective Prior: the population-level starting assumption that shapes how new Individuals interpret evidence before they have formed independent judgments. If the Collective Prior starts above even odds, most expressed Beliefs affirming authenticity, then:

1. Individuals observe a high shared level of expressed Belief and update their own Authenticity Belief upward
2. These updated Beliefs are expressed, raising the shared level
3. The now higher shared level provides stronger evidence for the next round of individual updating
4. The process iterates, the population's mean Authenticity Belief climbing toward unanimity

The convergence has three important properties:

Self-reinforcement: Each round of updating strengthens the signal for the next round. Early convergence produces evidence that accelerates later convergence.

Correlation creation: Even if Individual cascades began independently with varied priors, social proof synchronises them. Individuals who started sceptical observe a high shared level and update toward the consensus. The population moves together toward the Fixed Point.

Acceleration: The rate of convergence increases as the population's mean Authenticity Belief rises. Where three in five expressed Beliefs affirm authenticity the social proof signal is moderate; where nine in ten do, it is overwhelming. The cascade accelerates as it progresses.

13.4.3 The Collective Fixed Point

The system converges on a Collective Fixed Point, the Collective Prior having approached certainty, with the vast majority of expressed Belief affirming authentic motivation. At this point:

• Social proof overwhelmingly supports H_authentic
• Individuals who privately doubt face a choice between their doubts and the unanimous evidence of social consensus
• Expressing doubt becomes socially costly (it challenges the collective self-model)
• The collective Belief state is stable, resistant to revision even by accumulating external evidence

The Collective Fixed Point is the Bayesian architecture of mimetic crisis. It is the state in which a population has achieved collective certainty about a false Hypothesis, and that certainty was achieved through individually rational updating on socially-transmitted evidence that was generated by the same Confabulation process in others.

Two features of that stability are worth separating, because only the first is usually noticed. The Collective Fixed Point resists incoming evidence, which the preceding points describe. It also does not depend on the conditions that produced it. The low-Belief state and the Collective Fixed Point are both stable states of the same population, separated by an unstable boundary, and a population rests at one or the other rather than between them. What carries a population across that boundary is Confabulated private evidence accumulating alongside social proof. What does not carry it back is the withdrawal of either. Remove the pressure that drove the cascade and the population does not drift toward the low state, because nothing about the Collective Fixed Point requires the original pressure to persist; the state is held by the mutual evidence the population now supplies itself. Returning a population requires pushing collective Belief back across the boundary, which is a different operation from removing a cause, and a more expensive one.

This is the collective form of the Ignition-Release Asymmetry (Chapter 4, Section 4.4.13.2, with the formal statement at Appendix A.5.8.3). Withdrawal of the driver is sufficient to prevent a cascade that has not begun and insufficient to end one that has.

13.4.4 Grievance Beliefs and Scapegoat Convergence

The same mechanism operates for any socially-transmitted Belief, not only Beliefs about motivational authenticity. Of particular importance for crisis dynamics are Grievance Beliefs: Beliefs attributing experienced suffering to the conduct of an identifiable agent or agent-class (Appendix A, Section A.11.8; prose derivation at Chapter 16, Section 16.3). A Grievance Belief is not a Source Belief. A Source Belief is the Individual's distribution over the source Hypotheses for their own want; a Grievance Belief is about who caused the deficit, and it carries a Source Belief about that deficit with it, which is what the Grievance Gateway's A-Check and I-Check test.

When a population experiences diffuse harm (economic precarity, status anxiety, A-insufficiency, loss of I-meaning), the Information Gap Principle (Chapter 2) predicts that explanations will converge on agent-based causes. Confabulation generates narratives attributing suffering to identifiable actors rather than systemic dynamics, because agent-based narratives satisfy the brain's causal-attribution requirements while systemic explanations do not.

If Individual A generates and expresses a scapegoat explanation ("Group X is responsible for our suffering"), this becomes evidence that Individual B incorporates:

Individual B treats A's testimony as informative: if Group X really were responsible, we would expect people to say so. B updates toward H_GroupX and expresses their own (now higher) Belief. This expression becomes evidence for Individual C, and so on.

The convergence mechanism is identical to the authenticity-Belief case: expressed Confabulations become social proof that amplifies further Confabulation expression, producing collective convergence on a scapegoat explanation that may have no correspondence to actual causation. Content-Independence holds within the type and not across it: any agent and any deficit can fill a Grievance Belief, and the cascade dynamics do not vary with what fills it, but the Belief type that carries a population across the threshold does not vary either.

13.4.5 The Evidential Asymmetry

The corruption of evidence identified here is specific to mimetically-transmitted information. Not all evidence is suspect.

Reliable evidence channels include:

• Direct sensory observation of physical states
• Measurement through calibrated instruments
• Logical inference from established premises
• Testimony from sources with no stake in the observer's Belief (and no M-amplification in transmission)
• Historical records created without knowledge of current debates

Suspect evidence channels include:

• Socially-transmitted Beliefs where transmission is amplified by prestige cues
• Testimony from sources whose own Beliefs may be M-influenced
• Information that has passed through visibility-amplification (viral spread, trending topics)
• Explanations for collective suffering generated during crisis conditions
• Self-reports about motivational sources

The Bayesian corruption operates specifically on the second category. An Individual can still trust their thermometer, their direct observations of physical reality, and their logical inferences. What they cannot trust, without additional scrutiny, is information that has passed through social transmission channels where M-amplification and Confabulation may have operated.

This asymmetry has practical implications: the Two-Question Diagnostic and the Presumption Against Scapegoating are not generalised scepticism about all evidence. They are targeted filters for the specific evidence channel that the Confabulation Cascade corrupts. Rigorous evaluation of mimetically-acquired information does not require abandoning epistemic trust in other domains.

---

13.4.6 Two Parameters Governing Coupled Dynamics: Coupling Strength and Decoherence Time

The preceding sections established that Bayesian updating in a coupled system (where expressed Beliefs constitute evidence for other Individuals) produces convergence dynamics qualitatively different from Bayesian updating on independently sampled evidence. Two structural parameters determine whether a given system's updating dynamics track Truth or merely track the system's own internal dynamics. These parameters are implicit in the analysis already presented; naming them provides the vocabulary this chapter requires. This section introduces them and does not derive them: their derivation requires the generalised-probability apparatus of Chapter 14, and it is given at Chapter 14, Section 14.7, with the formal treatment at Appendix 14A.3 and 14A.4.

Coupling Strength (κ) measures the degree to which a system's evidence base is endogenously generated. In a system with zero Coupling Strength, each Individual's evidence is sampled independently from external reality. No Individual's expressed Belief constitutes evidence for any other Individual's updating. This is the regime in which standard Bayesian inference operates as textbooks describe it: rational updating on independent evidence converges toward Truth. In a system with high‑coupling‑strength, expressed Beliefs circulate as evidence, and each Individual's posterior is partly constituted by other Individuals' posteriors (which are in turn partly constituted by theirs). The evidence base becomes reflexive. The Collective Convergence mechanism described in Sections 13.4.1 through 13.4.4 is the description of what happens when Coupling Strength is substantially above zero: Individual A's Confabulation enters Individual B's evidence stream, B's updated Belief enters C's evidence stream, and the system converges on a collectively held position that may bear no relationship to external reality.

Coupling Strength is not a property of Individuals. It is a property of domains, information environments, and institutional structures. The same Individual operates at near-zero Coupling Strength when reading a calibrated thermometer (the thermometer's output is independent of the reader's Beliefs about temperature) and at high‑coupling‑strength when evaluating a political candidate (the Individual's assessment is partly constituted by their social environment's expressed assessments, which are partly constituted by theirs). What shifts between these cases is not the Individual's rationality or intelligence but the evidential structure of the domain.

Science achieves its epistemic power not because scientists are smarter or more rational than other people, but because the institutional architecture of science (blinded peer review, replication requirements, methodological transparency, adversarial collaboration) systematically reduces Coupling Strength. The evidence that enters a scientist's updating process is, by institutional design, less contaminated by other scientists' expressed Beliefs than the evidence that enters a voter's, an investor's, or a social media user's updating process. Science is institutional engineering that forces the claims it processes toward decoherence against an accessible Referent, migrating them into the Truth-Tracked regime and holding Coupling Strength low for those claims, rather than a property that makes the field low-coupling as a whole. (On the tracked-state and Referent-attachment treatment of this regime, see Appendix A, A.12.4 and A.12.6.) This is why science works where individual rationality, operating without institutional support in high‑coupling‑strength environments, does not.

Decoherence Time (τ) measures the delay before external reality imposes costs that coupled dynamics cannot absorb. In a system with short‑decoherence‑time, reality intrudes quickly. An engineer's false Belief about load-bearing capacity is corrected by structural failure within hours or days. The bridge collapses, the evidence is unambiguous, and no amount of social consensus can sustain the false Belief against the physical evidence. In a system with long‑decoherence‑time, reality's corrective signal may be delayed by years, decades, or generations. A population's false Belief about the cause of its economic suffering may persist for decades because the connection between the Belief and its consequences is diffuse, delayed, and mediated by institutional complexity. During this delay, the coupled dynamics have time to drive the system toward the Collective Fixed Point before reality can interrupt the process.

Decoherence Time is determined by the relationship between a domain's buffering capacity (the resources available to absorb the costs of divergence from reality) and the rate at which those buffers are consumed. A well-capitalised financial institution has long‑decoherence‑time: it can sustain positions divergent from fundamental value for extended periods because its capital reserves absorb interim losses. A subsistence farmer has short‑decoherence‑time: false Beliefs about planting seasons are corrected by crop failure within a single growing cycle. Long Decoherence Time is not intrinsically dangerous (it can provide time for deliberation and self-correction) but it becomes dangerous in combination with high‑coupling‑strength, because it gives the coupled dynamics time to reach the Collective Fixed Point before external correction arrives.

At the architectural level (Appendix A.5.7.5), Coupling Strength and Decoherence Time govern different things. Coupling Strength sets where the threshold sits, in the sense that a population whose evidence base is sufficiently endogenous crosses into the regime where coupled dynamics outpace exogenous correction. Decoherence Time does not move the threshold. It governs how long a coupled system that has crossed the threshold has to run before reality intrudes. The same crossing in a short-Decoherence-Time environment produces a self-limiting episode (bank runs, market panics, brief moral panics) because reality-contact arrives before the collective fixed point stabilises. The same crossing in a long-Decoherence-Time environment produces a crisis because the coupled dynamics reach the fixed point first. The architectural reason long-Decoherence-Time environments are more dangerous, then, is not that they lower the threshold but that they extend the period over which a population that has already crossed the threshold can compound before correction reaches it. The formal correspondent of this distinction is given in Appendix A.5.7.5.

The conjunction of Coupling Strength and Decoherence Time defines four epistemic regimes, summarised in the following Domain Matrix (the target being an Optimal Coupling Strength above zero rather than zero, per the Positive Mimesis Preservation commitment of Appendix A.5.7.1):

The Crisis State Architecture's five Phases (Section 13.6.2) can be understood as a trajectory through this matrix. The Grounded Phase corresponds to the Classical or Slow Classical domain (low‑coupling‑strength, short to moderate τ). An exogenous shock increases Coupling Strength (A-precarity makes Individuals more susceptible to social evidence) and begins extending Decoherence Time (institutional complexity delays reality-contact). The Drift Phase marks the transition into the Coupled Domain. The Calcified Phase represents deep entrenchment within the Coupled Domain (high‑coupling‑strength, long‑decoherence‑time, with institutional capture actively extending Decoherence Time by destroying the channels through which reality would otherwise intrude). The Terminal Phase represents the point at which Decoherence Time has been extended beyond the system's capacity to self-correct before catastrophic reconfiguration.

Coupling Strength and Decoherence Time as introduced here are conceptual parameters describing structural properties of epistemic environments. What they measure is evidence endogeneity and reality-contact delay, and their combination determines whether updating tracks Truth. Chapter 14 develops the formal apparatus demonstrating that Coupling Strength and Decoherence Time are not merely useful labels but mathematically necessary parameters characterising the boundary between probability systems that can be modelled on a single classical (Kolmogorovian) sample space and those that cannot. The formal treatment shows that above a critical coupling strength (Critical Coupling Strength, formally specified in Appendix A.5.7.5), the probability space itself becomes non-Boolean, meaning that certain joint measurements (such as simultaneously assessing one's status satisfaction and one's welfare adequacy) are not merely difficult but algebraically undefined. That formal result strengthens every claim made in this chapter but is not required for it. The classical Bayesian analysis presented here is sufficient to establish the crisis architecture; the non-classical analysis of Chapter 14 shows the problem is deeper than the classical treatment alone reveals.

The social proof mechanism operates with differential weighting that the preceding analysis did not specify. Not all expressed Beliefs carry equal evidential weight in the observer's Bayesian updating. The observer's Decision Hub assigns greater credibility to testimony from Individuals who carry higher PM-valuation in the observer's motivational system. This is narrative authority yield (the third transfer channel identified in Chapter 8): the higher-Status party's claims are treated as more informative, producing larger likelihood ratios in the observer's updating equation. The mechanism is not deliberate deference to authority (which I-Override could in principle detect and correct) but preconscious weighting: the PM-component of the observer's valuation of the source has been integrated before the testimony is evaluated, inflating the perceived reliability of the testimony before conscious assessment begins. The consequence is that Bayesian cascade velocity is not uniform across the population. Testimony from high-PM Individuals accelerates convergence disproportionately, because each expression of their Belief produces larger posterior shifts in observers than equivalent expressions from low-PM Individuals. This is the architectural basis of what is colloquially called "influencer effects" and provides the formal mechanism through which the Leadership Compression Principle operates even below the threshold of formal authority: high-PM Individuals can shift collective priors through narrative authority alone, without commanding institutional levers.

The Build-Up Time Decomposition

Coupling Strength and Decoherence Time govern whether a system enters the regime in which Bayesian dynamics are coupled to within-population evidence and how long the coupled dynamics have to operate before reality intrudes. They do not, by themselves, decompose the integral that determines when cumulative Mimetic Premium crosses the supercriticality threshold. That decomposition is necessary because it is the integral, not the rate, that drives the phase trajectory of the Crisis State Architecture.

Cumulative Mimetic Premium accumulates by compounding small per-transaction extractions across very large numbers of transactions. The supercriticality build-up time is decomposable into four factors. The first is per-transaction Mimetic-Premium magnitude, the mean fraction of the integrated valuation in typical transactions that is Mimetically derived. By the Per-Transaction Minority Claim, this is small in well-functioning societies. The second is transaction frequency in the population, which rises as the share of essential goods and services markets in which Mimetic-Premium extraction occurs increases. The third is the absorption capacity of the A-I-Buffer, the rate at which a society's A-Floor and Intrinsic Infrastructure (I-infrastructure) can absorb cumulative Mimetic-Premium extraction before A-sufficiency or I-access begins to fail. The fourth is active Subcriticality Maintenance, any institutional reduction in the integration through measures that limit per-transaction extraction, distribute A-sufficiency universally, protect I-infrastructure, and preserve opt-out from Mimetically dominated networks.

The decomposition matters for trajectory deflection. Per-transaction Mimetic Premium cannot easily be reduced because Source Opacity is architectural. Transaction frequency, A-I-Buffer absorption capacity, and Subcriticality Maintenance can all be operated on. The intervention architecture appropriate to the framework is therefore not arbitrary. It targets the three factors in the decomposition that are tractable, leaving the architectural floor untouched and operating on the integrand to keep the integral below threshold.

This decomposition also explains why supercriticality typically takes very long periods of distributed mimesis to develop. The integral grows slowly when per-transaction Mimetic Premium is small (the typical case), transaction frequency is moderate, A-I-Buffer absorption is robust, and Subcriticality Maintenance is operating. It grows quickly when any of those factors degrades. It grows extremely quickly when an M-captured Individual gains executive authority, which is the subject of the next subsection.

The formal apparatus for this decomposition, including the four-factor parameterisation of the M-component drift coefficient and its connection to Coupling Strength and Decoherence Time, is set out in Appendix A.

13.4.7 Path-Dependence and the Order of Belief

When Evidence reaches an Individual, that Evidence does not arrive into a blank space. The Evidence arrives into a Belief Register that already holds a distribution of weight across competing Hypotheses. The weight the new Evidence receives depends on what is already in the register, and what is already in the register depends on what arrived before. The order in which Evidence reaches an Individual is therefore not a neutral administrative fact. It shapes the Belief the Individual ends up holding.

This much is familiar from everyday experience. What the framework adds is a more precise account of why order matters and why the problem is structural rather than correctable by good reasoning alone.

Think of the Belief Register as a folder. The folder contains every Hypothesis the Individual currently holds as a live candidate for what is true. Each Hypothesis in the folder carries a weight: the Individual's current assessment of how likely it is to be correct. When new Evidence arrives, the weights are updated. An Individual with good epistemic practices updates them in a principled way, giving more weight to Hypotheses that would have predicted the Evidence and less to those that would not.

Two distinct problems can arise, and both are architectural rather than matters of individual failing.

The first problem is that the right Hypothesis is not in the folder at all. A Hypothesis that is absent from the Belief Register cannot receive any weight, regardless of how compelling the Evidence for it would be if the Hypothesis were present. Evidence that points toward a Mimetic source arrives into a register that contains no Mimetic Hypothesis, and so it is absorbed by whichever A-framed or I-framed Hypothesis fits it best. This is the mechanism behind Bayesian Lock-In: the Belief Register does not resist the Evidence; it simply has nowhere to put it that would implicate the correct source. AIM Vocabulary Acquisition addresses this problem directly by adding the missing Hypothesis to the folder.

The second problem arises even when the right Hypothesis is present. If the Individual has already landed on near-certainty about another Hypothesis, new Evidence arriving in favour of the correct one finds almost no weight available to it. The existing dominant Hypothesis absorbs the Evidence before the correct one can take hold as the leading candidate. The order in which the Evidence arrived determined the outcome, not the balance of the Evidence itself. Had the Evidence pointing toward the Mimetic source arrived first, the register would have been in a different starting state, and the update would have gone differently. Epistemic Suspension addresses this problem by keeping the register from collapsing to near-certainty prematurely, maintaining enough distributed weight across Hypotheses that subsequent Evidence can still shift the outcome. The same dependence on order runs in the other direction: a True Belief that arrived first, on testimony, can be worn away by Evidence that arrives later in small items, none of which moves it far (Chapter 14, Section 14.8.3.2).

Both problems have a further consequence at the network scale. When an Individual expresses a conclusion to neighbours, that conclusion arrived at through a particular Evidence sequence in a particular starting register. Neighbours who receive it as Evidence are then in the position of updating their own registers on a conclusion that was itself shaped by sequence. Across a network of Individuals, each updating on conclusions that were themselves sequence-dependent, the population's collective Belief state reflects the order in which information flowed through the network's coupling structure, not only what information was available. Two populations with access to identical bodies of Evidence, receiving it through different institutional sequences and different coupling strengths, can arrive at dramatically different collective Beliefs.

This is the architectural basis for the framework's claim that the sequence and structure of Evidence flows is not a second-order concern for institutional design. It is a first-order determinant of whether a population navigates a mimetic crisis or is consumed by one. The formal treatment of this path-dependence, including the conditions under which Belief trajectories through a network become irreversible, is developed in Appendix A.5.7.2.

The intervention consequences follow directly. The network Evidence-coupling channel (Channel 4 of the Knowledge Patch footprint) is sensitive to sequence and coupling structure for exactly the reasons set out above. Patch-competent Evidence intake operates not only by attenuating the weight an Individual assigns to neighbours' expressed conclusions, but by maintaining a Belief Register that is still open to the influence of sequence: one that has not yet collapsed to the point where no incoming Evidence can shift the dominant Hypothesis. Epistemic Suspension, by keeping the register from fully collapsing, preserves this openness. Once the Individual Fixed Point is reached and the register has collapsed to near-certainty on a single Hypothesis, the architecture that would allow resequencing to matter is gone.

Appendix locus: Appendix A.5.7.2, N3 Non-Commutativity Propagation.

13.4.8 The Ambient Evidence Amplification Effect

Information received from other Individuals reaches the Thinker as Evidence through ordinary sensory intake, weighted by the coupling between the Individuals (Section 13.4.6), and enters Bayesian Updating on the same terms as any other Evidence. That Evidence moves the Belief Register on the question the Evidence bears on while the Thinker is resolving that question, and the Thinker is recruited to a question in three ways (Chapter 4, Section 4.4.3): no committed posterior exists, as with the first report of a distant event; an item departs from what the committed posterior predicts, as a doubt about a settled question does; or the question is recalled unresolved, which is rumination (Section 13.3.3). Three properties of the channel then fix what continuous exposure does to a population.

First, an item's Salience bears on whether the Thinker is recruited and not on the weight the item carries once the Thinker is. A strongly Salient Object, of either sign, can move the Scalar away from what the Belief Register predicts and so recruit the Thinker, but the network term of each Individual's likelihood depends only on the coupling to the source and on how firmly the source emitted the item, and contains no Salience (Appendix A, Section A.5.7.3). On a question the Thinker is already resolving, a routine item moves the posterior as far as an alarming item from the same source.

Second, the channel credits restatement as corroboration. An Individual holds no channel that distinguishes a restatement of one attribution from independent corroboration of that attribution, so arrivals are credited one by one, and the weight an attribution accumulates tracks how often the attribution has been repeated rather than the weight of what stands behind it (Appendix A, Section A.5.7.2).

Third, the race among Hypotheses cannot circle. Coupling weights are non-negative, so each arrival adds support to the attribution it carries, and by the Monotone Belief-Race Lemma the race sorts steadily toward the best-supported Hypothesis (Chapter 4, Section 4.4.13.2). Where coupled Evidence makes up most of what arrives, the best-supported Hypothesis is the one repeated most.

Together these entail the Ambient Evidence Amplification Effect: a continuous inflow of coupled Evidence through the routine channels a population attends to accumulates in the Belief Registers of that population item by item, toward the attribution the channels repeat most, with no single item large enough to register as a change of mind and no ignition event required. The Ambient Evidence Amplification Effect does its work on matters whose Referent the recipients never contact, which is most of what lies outside their own Appetitive and Intrinsic circumstances. On those matters each recipient's updating is Contactless Updating (Chapter 14, Section 14.8.3.2): no Evidence arrives from outside the coupled flow to correct the Belief, and the accumulation that installs a False Belief carries a True Belief away by the same operation. Near certainty the probability scale is compressed, so the accumulation is least visible where a Belief is held most firmly. The Ambient Evidence Amplification Effect runs one way until contact with the Referent arrives, through forced visibility or exogenous Evidence, and where the Ambient Evidence Amplification Effect has carried a population to an inflated state, that state ends at once rather than by degrees (Appendix A, Section A.5.7.6).

The Ambient Evidence Amplification Effect is the network counterpart of Bayesian Lock-In within one Individual. Within the Individual, Self-Coupling Strength returns the Individual's own recalled Beliefs as Evidence; across the network, the coupling between Individuals returns the population's restatements as Evidence, and the slow-channel network gain carries that coupling where the Individual's slow-layer gain carries Self-Coupling Strength (Appendix A, Section A.5.7.4).

The Ambient Evidence Amplification Effect is distinct from the Information-Cascade Amplification Effect (Chapter 20, Section 20.4.3). The Information-Cascade Amplification Effect acts on how many couplings reach an Individual at once: a denser structure of Evidence couplings carries a network toward the critical threshold, and a denser structure of observation lowers the threshold, while the attention Individuals can spend on others has room to spare. The Ambient Evidence Amplification Effect acts at a fixed structure, through the constancy of what that structure delivers, and is bounded by what the Thinker can take in, so that once the Thinker's capacity is spent the Ambient Evidence Amplification Effect grows only as routine channels take a larger share of that capacity. The rate at which items arrive moves the operating point at which the network sits and not the threshold, which is fixed by the topology of the coupled loop and not by any timescale (Appendix A, Section A.5.7.5).

Appendix locus: Appendix A.5.7.2, N2 Evidence as Coupling-Mediated Input.

13.5 The Two Accumulation Rates

Distributed mimesis builds cumulative Mimetic Premium slowly. No single Decision Hub controls more than its own integrated valuations and the small fraction of network influence its Status entitles it to, so the accumulation runs across very large numbers of interpersonal transactions and the Build-Up Time Decomposition of Section 13.4.6 governs how long it takes.

That decomposition should not be read as a civilisational constant. The build-up time it describes is the time distributed mimesis takes, and a second rate operates alongside it: an Individual holding executive institutional authority can raise both the magnitude and the frequency of Mimetic-Premium extraction across an entire governed population within a single executive term, which is a step-change rather than an increment in the rate. Distributed mimesis is the slow build and executive capture is the fast finish, and the second is why transitions into terminal-phase trajectories are observed unfolding across years even though the precondition for them required centuries to assemble.

Two consequences bear on the Bayesian account. The build-up time of Section 13.4.6 is an upper bound on how long a population has, not an estimate of it, because the four factors it decomposes can all be moved from a single institutional position. And the corrective rate that Section 13.8.3 sets against amplification is itself a target: the channels through which the framework's vocabulary propagates, the density of concealment within the population, and the recognition that would otherwise translate into exit are all reachable from the same position. The mechanism by which they are reached, and the five institutional channels through which it operates, is Executive M-Capture, derived at Chapter 17, Section 17.7.4.

13.6 The Bayesian Mimetic Crisis Principle

13.6.1 Statement of the Bayesian Mimetic Crisis Principle

The preceding analysis enables a formal statement of what the Bayesian architecture delivers at population scale, connecting the cascade dynamics to the Crisis State Architecture of Section 13.6.2:

Bayesian Mimetic Crisis Principle (derived here from the spectral threshold of Appendix A.5.7.5 and the failure-mode analysis of A.6.3): Given (1) Source Opacity, (2) Confabulation as the mechanism for explaining opaque desires, (3) content-agnostic Bayesian updating on Confabulated evidence (Axiom 8), and (4) social proof as evidence for individual updating, M-coupled Beliefs compound through the population by a lock-in process that operates identically on accurate and inaccurate Beliefs once Coupling Strength exceeds the Subcriticality Threshold, and the felt crisis is delivered through A-precarity, I-domain destruction, and Status displacement with no true deficit beneath it.

The Bayesian Mimetic Crisis Principle is a statement about updating rather than about outcomes. The machinery that consolidates an accurate Belief about the physical world consolidates an inaccurate Belief about a scapegoated group by the same operation, because it weights evidence by its evidential form and has no access to its accuracy. Once the coupling is high enough that the population is supplying most of its own evidence, the consolidation no longer depends on anything outside the population, and the Belief holds because it is circulating.

What a population in that condition then does is a separate question, answered in the domain where the architecture reaches ground: the Terminal Phase region of the Crisis State parameter space functions as an attractor, a configuration toward which the system evolves under its own dynamics absent exogenous intervention. That result is the Unifying Crisis Principle, derived at Chapter 17, Section 17.2. The two are one result at two levels of description. This chapter establishes that the updating locks in; Chapter 17 establishes what follows from a population in which it has.

The Crisis State Architecture's phases read as a trajectory in the cascade parameters, which is what makes the attractor claim available at all.

13.6.2 Crisis State Phases as Cascade Progression

The Four Crisis State Variables

A population's position among the five phases of the Crisis State Architecture is read from four quantities measurable in the population. Each is the population-scale trace of a result derived elsewhere, and none has dynamics of its own: each measures a quantity whose dynamics the architecture already fixes. The four are the Crisis State Variables, and where the boundaries between phases lie in the Crisis State Variables is a matter of measurement.

Status Integrity is the degree to which an Individual's position in a field is predicted by the two Referent-backed components of the Status the Individual receives, Status-A and Status-I, rather than by the Mimetic component, Status-PM (Chapter 8, Section 8.12.3): the Referent-backed share of that prediction, taken across the fields of a population. Where Status Integrity is high, position tracks contribution to other Individuals' Appetitive needs and to their engagement and mastery; as Status Integrity falls, position tracks position already held. At a share of one half the Mimetic component predicts position as well as the two Referent-backed components together, which is the Calcification Threshold (Chapter 8, Section 8.12.9). The Calcification Threshold is crossed field by field, so a population can hold fields on both sides of the Calcification Threshold at once. No observer can read which source composes the valuation the observer directs at another Individual (Chapter 4, Section 4.3.3), and contribution goes on being valued after the ceiling hardens, so the legitimacy a population perceives in its hierarchies is a Belief about Status Integrity, and that Belief falls later than Status Integrity does, latest where Concealment Density is high.

Price Inflation Ratio is the ratio of the Mimetic Premium in the prices of essential goods to the Appetitive and Intrinsic components of the same prices, under the Price Decomposition (Chapter 8, Section 8.13.1; Appendix A, Section A.12.5). The Price Inflation Ratio is taken over transactions and aggregated across transactions, because a purchase has a decomposition only relative to its purchaser and a good has none of its own (Chapter 8, Section 8.13.2). In the markets an Individual cannot forgo, the Price Inflation Ratio reads the first two of the four factors of the Build-Up Time Decomposition (Section 13.4.6): the Mimetic Premium each transaction carries, and the share of essential markets in which transactions carry a Mimetic Premium. Mimetic Premium financed by debt enters the claims that Appendix 12A sets against generative capacity, whose service limit is the Interest-Only Threshold (Appendix 12A, Section 12A.5).

Exit Capacity is the degree to which the Individuals of a population can refuse the markets and arenas that carry the Mimetic Premium without the refusal taking them below the A-Floor or out of their I-domains. Refusal is the weaker half of exit: an Individual who refuses the arena but stays among its Models goes on observing them, so Exit Capacity also counts the capacity to leave a Reference Set's observation for an interval long enough for the Reference Set's priors to release, with provisioning maintained (Chapter 20, Section 20.8). Exit Capacity measures how far the Opt-Out Paradox has run (Chapter 17, Section 17.2): where access to Appetitive sufficiency or to an I-domain has become contingent on continued participation, recognising the dynamic opens no exit. Exit Capacity is lowest where an Appetitive need can be met only in a market the Mimetic Premium occupies, as shelter is met in housing, where renting moves the Mimetic Premium into the rent rather than escaping it (Chapter 8, Section 8.13.2). At the worst-off margin low Exit Capacity closes both routes: an Individual who exits falls below the A-Floor, and one who stays pays the Mimetic Premium out of the provision the A-Floor protects, deepening the deficit whose critical depth Section 13.8.4 derives. Exit Capacity is the quantity Opt-Out Preservation maintains (Chapter 17, Section 17.22).

Concealment Density is the degree to which the Mimetic Hypothesis is unavailable to a population's Individuals at the point where a want or a price is resolved. The Mimetic Hypothesis can be unavailable in two ways (Appendix A, Section A.8.4). The Mimetic Hypothesis can be absent from the Hypothesis Space, which is the Vocabulary Gap of the Camouflage Architecture (Chapter 5, Section 5.8.4) and is what suppressing the channels that would carry the Hypothesis brings about (Chapter 17, Section 17.7.4). Or the Mimetic Hypothesis can be held and not applied, because the decomposition that would apply the Mimetic Hypothesis exceeds what the Thinker can carry and the decision settles on Social Proof instead. That second way is Complexity as Concealment: the structures that persist where the Mimetic Premium is large are the ones that leave the Mimetic Premium illegible, and their complexity scales with the Mimetic Premium they protect (Chapter 12, Section 12.26). Where the Mimetic Hypothesis is unavailable, Evidence pointing to a Mimetic source is absorbed by the Hypotheses that remain (Section 13.4.7), so correction does not take hold and Decoherence Time lengthens. Concealment Density acts on the attribution and not on the signal: the Macro-Residues of Mimetic Premium extraction accumulate and remain measurable at population scale however high Concealment Density rises (Appendix A, Section A.8.4). Concealment Density falls discontinuously when forced visibility admits the missing Hypothesis, and that fall belongs to the Calcified-to-Terminal transition, at which the Concealment Infrastructure collapses to the A-I functional baseline (Appendix A, Section A.5.7.6).

The Crisis State Variables move together through derived mechanisms. Status-elevation of extraction pathways lowers Status Integrity and accelerates Mimetic transmission, which raises the Mimetic Premium carried in prices (Chapter 17, Section 17.7.4). A Mimetic Premium in the markets for essential goods lowers Exit Capacity, as the Housing Exception Principle shows for shelter (Chapter 8, Section 8.13.2). The complexity that conceals the Mimetic Premium scales with the Mimetic Premium, which is Complexity as Concealment (Chapter 12, Section 12.26), so a rising Price Inflation Ratio raises Concealment Density. And a high Concealment Density delays the reading of a fall in Status Integrity, since the hardening of a hierarchy is then accounted for by the Hypotheses that remain. A population's phase is therefore located by all four Crisis State Variables at once.

The Crisis State Architecture's five phases map onto cascade dynamics as follows:

Grounded Phase: Baseline M-Activity. Individual Confabulation Cascades operate at low intensity. Authenticity Belief drifts upward but slowly in each Individual, and the population's mean Authenticity Belief remains moderate. The system is in a subcritical regime where corrective mechanisms (direct experience contradicting Confabulations, I-domain engagement providing alternative reference points, A-floor security enabling deliberative assessment) outpace mimetic amplification. Crisis State Variables remain within safe bounds: Status Integrity is high, Price Inflation Ratio is low, Exit Capacity is adequate, and Concealment Density is manageable. In Coupling Strength-Decoherence Time terms, the system operates in the Classical Domain: Coupling Strength is low (evidence is predominantly exogenous), Decoherence Time is short (A-floor security and I-domain engagement provide rapid reality-contact), and corrective mechanisms outpace mimetic amplification.

Drift Phase: Trigger Amplification and Scapegoat Convergence. An exogenous shock (economic crisis, military defeat, pandemic, technological disruption) produces widespread A-insufficiency or I-disruption. The shock increases M-weight in population motivation (A-urgency makes Individuals more susceptible to M-following; I-disruption removes alternative value sources). Cascade velocity increases. As Status Integrity declines toward the Drift/Calcified boundary, the Information Gap Principle activates. Populations experiencing diffuse harm generate agent-based explanations (Axiom 9). Grievance Belief in the guilt of the named group begins climbing through social proof dynamics. Early adopters of the scapegoat narrative provide evidence for later adopters. The narrative gains momentum. The exogenous shock has shifted the system's Coupling Strength-Decoherence Time profile: Coupling Strength rises as A-precarity increases susceptibility to social evidence, and Decoherence Time begins extending as institutional complexity delays the corrective signal. The system is transitioning from the Classical Domain into the Coupled Domain.

Calcified Phase: Institutional Capture and Sacrificial Preparation. Grievance Belief in the guilt of the named group approaches the threshold where collective action becomes possible. Institutional capture begins, with leadership positions occupied by Individuals whose personal cascades align with the collective dynamics. Legal, economic, and social restrictions on the scapegoated group are implemented, experienced as legitimate response to confirmed threat. Concealment Density reaches levels where structural analysis becomes impossible from within the system. Exit Capacity collapses as participation in M-escalation becomes the precondition for A-sufficiency. Institutional capture actively manages both parameters: Coupling Strength is maximised through state media saturation, loyalty-signalling requirements, and the elimination of independent information channels, while Decoherence Time is extended through concealment infrastructure that prevents reality-contact from reaching the population before the Collective Fixed Point stabilises.

Terminal Phase: Sacrificial Violence and Catastrophic Reconfiguration. The restrictions escalate to violence. The Collective Fixed Point has been reached, Grievance Belief in the group's guilt having approached certainty, and the violence is experienced not as aggression but as necessary defence against a confirmed enemy. The Confabulation Cascade has converted M-driven targeting into felt moral necessity. The system is now deep within the Coupled Domain, Coupling Strength at its maximum and Decoherence Time effectively infinite within the crisis timescale, and the Collective Fixed Point has been reached before any corrective signal could arrive. The violence either succeeds (destroying the scapegoated group, but leaving underlying M-dynamics unresolved) or fails catastrophically (war, civilisational collapse, mutual destruction). The Terminal Phase is the attractor region in the parameter space, the configuration toward which the system evolves through its own dynamics.

Fast-Layer Correspondent of the Calcified-to-Terminal Transition

The Calcified-to-Terminal transition is a collapse rather than a decline, and the difference decides what can be done about it. The Concealment Infrastructure is the network-scale structure that holds valuation at its inflated level until visibility is forced. While the three couplings of the mimetic loop stay positive, and while more realised valuation always produces more environmental Salience up to the Salience Ceiling at which it saturates, the loss of stability at the critical threshold takes one particular form, which the mathematics calls a fold. The inflated state does not wobble its way down. Sitting just below it is a second, precarious state, and as conditions tighten the two move toward one another until they meet and both cease to exist. With nothing left to hold it, the system drops all at once to what Appetites and Intrinsic Motivation alone would have sustained: the functional baseline, with the mimetic contribution gone. It does not settle into a sustained oscillation instead, the alternative the mathematics would otherwise allow and calls a Hopf cycle, in which the system rises and falls around its resting state and can be corrected on each downswing, where a fold leaves no way back down through intermediate states (Appendix A, §A.5.7.6).

Three architectural properties force this geometry: the absence of self-terminating negative feedback in the M-channel, Source Opacity suppression of error-correction before collapse, and the monotone accumulation of Concealment Infrastructure up to the point of forced visibility. The phenomenology follows: bubble collapses without pre-crash oscillation, scapegoating cascades that do not taper, and institutional consensus that does not negotiate its way back to Truth.

The same geometry runs in the other direction, and this matters for how the phases are read. A population does not ease into the Calcified Phase and it will not ease out of the Terminal Phase. Pressure accumulates while very little appears to change, and then the whole configuration moves at once, which is why the phases above describe states rather than points on a slope.

What sits at the inflated state is the population's Belief about the scapegoated group, and it holds there because it is circulating rather than because it is supported: each person's conviction arrives at everyone else as Evidence, so the Belief is supplied from inside the population rather than from anything outside it. Above a threshold of collective Belief, that circulation outpaces whatever doubt the population still generates, and the Collective Fixed Point is reached from anywhere above it. Below the threshold, doubt outpaces circulation and the descent completes itself without further help. The threshold is therefore a point to be crossed downward rather than a level to be approached: Belief reduced but left above it returns, because circulation is still supplying more than doubt removes, which is why measures that visibly moderate a population's expressed convictions can be followed by their full return.

And the conditions that carried the population over the threshold sit higher than the conditions it must be brought back below. The state media saturation, the loyalty-signalling requirements, the elimination of independent information channels and the concealment infrastructure that lengthened Decoherence Time do not have to be returned to their pre-crisis levels. They have to be taken below them, and how far below grows with the coupling and concealment the Calcified Phase has accumulated. Section 13.10 takes up what this means for the cost of prevention set against the cost of release. The derivation is given in Appendix A §A.5.3, where the fold is the Saddle-Node Default and the sustained oscillation, which arises only under externally supplied damping, is the non-generic Hopf Sub-Case.

Subcriticality Maintenance, the normative goal of this chapter, operates by reducing all five gain components sufficiently to bring their total below the critical threshold, keeping the system in the subcritical (decaying) regime. Adding damping to the existing mimetic loop cannot produce the oscillation; converting the collapse into a cycle requires the introduction of a new dynamical variable with intrinsic negative feedback, and Appendix A.5.3 and A.5.7.6 establish that such a variable converts the crossing at moderate gain and lag. What remains open is which institutional structures instantiate one, and the spectral conditions a specific augmented system must satisfy. Where the system cannot be brought subcritical fast enough by gain reduction alone, a second correction is available alongside conversion rather than instead of it: the structural distribution of authority specified by the Leadership Compression Principle (Chapter 17, Section 17.7) and its Time-Buying Principle, which lengthens the interval between leadership M-shock and crisis imposition and buys the time within which the other two can complete their work. The three responses operate on different objects and none substitutes for another, which Section 13.10 develops.

Reset Phase: Post-Crisis Return to Grounded. Catastrophic reconfiguration destroys accumulated Mimetic Premium. The system's Crisis State Variables return to the Grounded region of the parameter space, but the underlying architecture (Source Opacity, Confabulation, Mimetic dynamics) remains intact, and drift toward Terminal recommences absent intervention. The reductions in inequality that follow catastrophe are therefore resets in a recurring cycle rather than stable departures from the stratifying dynamic, because the Reset Phase destroys the accumulated Premium and leaves intact the architecture that accumulated it. That is the Stratification Inevitability Principle: in a high-M system, stratification is the standing output of Mimetic escalation and Mimetic Premium extraction, and a post-crisis levelling is the Grounded Phase of the next cycle rather than an equilibrium of its own.

In the monetary domain the recommencement is a derived result. The Jubilee Theorem (Appendix 12A, Section 12A.8) establishes that a clearance of accumulated claims executed with the Mimetic Premium unbounded re-crosses the Interest-Only Threshold in finite time, on a schedule that shortens as the clearance is less complete, which is why the historical clearances were periodic, and that the drift does not recommence where the clearance executes under a standing M-Tax. The Reset Phase's recommencement is therefore not a fate but a parameter condition.

The Reset Phase is not a passive return to equilibrium. It is produced by the specific forms that catastrophic reconfiguration takes: war that physically destroys infrastructure, displaces populations, and eliminates the institutional channels through which endogenous evidence circulated; economic collapse that exhausts the financial buffers sustaining divergence from fundamental value, forcing direct confrontation with material scarcity; environmental catastrophe (famine, resource exhaustion, ecological breakdown) that imposes biological reality with a force no social consensus can override; or institutional disintegration that dismantles the media, bureaucratic, and coercive apparatus through which high‑coupling‑strength conditions were maintained. Each of these is a form of Decoherence Time reaching zero: reality arriving with such violence that the system's remaining buffers are overwhelmed in a compressed timescale. The endogenous evidence structure collapses because the infrastructure sustaining it (functioning media, stable institutions, material surplus, physical safety) has been destroyed. Coupling Strength does not gradually decline but is broken by the catastrophe itself. In Coupling Strength-Decoherence Time terms, the system is forced back into the Classical Domain not by corrective intervention but by the sheer material weight of the catastrophe eliminating the conditions under which coupled dynamics could operate. Populations amid rubble, famine, or economic ruin update on exogenous evidence (hunger, displacement, physical danger) because endogenous evidence channels no longer exist.

But the architectural capacity to regenerate high‑coupling‑strength, long‑decoherence‑time conditions remains entirely intact: Source Opacity persists, Confabulation persists, Preconscious Transmission persists. The moment material conditions stabilise sufficiently for endogenous evidence channels to reconstitute (functioning media, surplus production, institutional regrowth), Coupling Strength begins rising again and the cycle recommences. This is why the historical record shows not single crises but recurring cycles of crisis separated by periods of relative stability: each catastrophe resets the parameters, and each recovery regenerates the conditions for the next escalation.

13.6.3 The Cycle and the Termination Condition

The Crisis State Architecture describes a cyclical dynamic: Reset returns the system to the Grounded region, from which drift toward Terminal recommences absent intervention. Successful scapegoating does not resolve the underlying architecture, as Source Opacity persists, Confabulation continues, and M-dynamics remain operative. The "resolution" of the Terminal Phase merely resets conditions for the next cycle, often with intensified parameters (accumulated grievances, depleted institutional buffers, normalised violence).

This cyclical structure has characterised human civilisation. The historical record, from Rome's cycles of crisis and recovery through European history's recurring pogroms and wars to the twentieth century's unprecedented destructions, represents iterations of the same architectural dynamic operating at varying scales.

At the civilisational scale, and specifically in relation to the risk of nuclear war between states in the grip of supercritical Mimetic rivalry, the framework's claim rests on entailment from the dynamics validated at survivable scales, extended under the Precautionary Asymmetry. It is stronger than speculative extrapolation and weaker than a directly tested prediction: an architectural entailment under a declared asymmetry between non-equivalent error modes. The Asymmetry is derived at Chapter 17, Section 17.18, its formal statement is at Appendix A.6.3, and the falsification section of this chapter states what would disconfirm the claims it covers.

However, a termination condition now exists that did not exist historically: nuclear weapons. Previous cycles, however catastrophic, permitted recovery. Populations survived. Civilisations could be rebuilt. The nuclear termination condition removes this assumption. A sufficiently severe nuclear exchange could produce civilisational collapse from which recovery is impossible, not merely another cycle, but the end of cycles.

In Coupling Strength-Decoherence Time terms, the nuclear termination condition is the scenario in which catastrophic reconfiguration is so total that the preconditions for Coupling Strength-regeneration are permanently eliminated, not because the architecture has been repaired, but because there is no surviving population, infrastructure, or institutional substrate from which to regenerate it. Every previous cycle in recorded history has been survivable in the specific sense that the catastrophe, however devastating, left enough material and social substrate intact for endogenous evidence channels to eventually reconstitute: cities were rebuilt, trade networks reformed, media institutions re-emerged, and Coupling Strength began climbing again. The nuclear termination condition is unique because it threatens to destroy the substrate itself. A sufficient nuclear exchange does not merely reset Coupling Strength to zero; it eliminates the possibility of Coupling Strength ever rising again, because it eliminates the civilisation within which Coupling Strength operates. The cycle does not restart. The architecture is not repaired. The species-level experiment in cumulative cultural evolution, sustained by the same coupling infrastructure that generates crisis, terminates.

This transforms the crisis architecture from an iterating pattern that humanity has survived (barely, repeatedly, at enormous cost) into an existential risk. The cycle continues until interrupted or until a nuclear-armed iteration reaches the Terminal Phase with sufficient intensity to trigger the termination condition.

13.6.4 Content Independence: Why Crisis Architecture Permits Infinite Variation

The Bayesian machinery formalised in this chapter is Content-independent. The cascade dynamics (Individual Confabulation producing self-reinforcing Belief, collective convergence through social proof, progression toward the Fixed Point) operate identically regardless of what specific Content fills the Hypothesis slot (Appendix A, Section A.5.8.6).

Let H represent any scapegoat Hypothesis: "Group X is responsible for our suffering." The Bayesian updating proceeds:

Nothing in this equation constrains what H contains. The architecture processes:

• H = "Jews are responsible" (Nazi Germany)
• H = "Kulaks are responsible" (Soviet Union)
• H = "Tutsis are responsible" (Rwanda)
• H = "Intellectuals are responsible" (Cambodia)
• H = "Immigrants are responsible" (various contemporary cases)

with identical dynamics. The cascade velocity, convergence pattern, and Fixed Point stability are determined by the structural parameters (observability, A-precarity, I-domain vitality, social proof intensity), not by the specific group designated or the specific grievance articulated.

This Content-independence means the Bayesian architecture permits infinite variation in crisis Content while maintaining identical crisis dynamics. Any group can fill the scapegoat slot. Any grievance can serve as trigger. Any narrative can provide the Confabulated explanation. The surface features of crises vary enormously; the underlying architecture is invariant.

13.6.5 Why Comparative Historical Analysis Failed

Content-independence explains why traditional comparative analysis of historical crises has failed to identify the common causal architecture.

Comparative methodology proceeds by identifying similarities and differences across cases, seeking patterns that distinguish crisis from non-crisis outcomes. The analyst asks: What did Nazi Germany, Stalinist Russia, and Rwandan genocide have in common that peaceful societies lack?

The search for Content similarities produces limited results:

• Different target groups (Jews, kulaks, Tutsis: no common demographic feature)
• Different triggering events (economic collapse, revolution, assassination: no common trigger type)
• Different ideological frameworks (racial theory, class theory, ethnic nationalism: no common ideology)
• Different political systems (democracy collapsing, post-revolutionary consolidation, post-colonial fragility: no common regime type)
• Different economic conditions (industrial depression, agricultural collectivisation, post-colonial poverty: no common economic profile)

The comparative method, focused on Content, concludes that crises are heterogeneous, each explained by its particular historical circumstances, with no general theory possible. This conclusion is methodologically sound given Content-focused premises, but the premises are wrong.

The architectural analysis reveals what Content-comparison cannot see: identical Bayesian dynamics operating on different Content. The structure is invariant; only the Content varies. Comparative analysis focused on Content is analogous to comparing the words in different books while ignoring that all books share the same grammatical architecture. The grammar is invisible to word-comparison; the crisis architecture is invisible to Content-comparison.

13.6.6 Methodological and Predictive Implications

This analysis explains why the crisis architecture was not identified through historical scholarship despite centuries of crisis documentation. Historians, trained in contextual specificity, correctly identified the unique features of each crisis. But the methodology that enabled this contextual richness (attention to specific Content, particular actors, contingent sequences) systematically obscured the architectural commonality.

The architectural perspective requires a different methodological orientation:

• Treating specific Content as instantiation rather than explanation
• Asking what structure different crises share rather than what Content they share
• Recognising that surface dissimilarity can mask deep structural identity

Content-independence generates specific predictions:

Novel Content, identical dynamics: Future crises will designate scapegoats that cannot be predicted from past content (new groups, new grievances, new narratives) but will exhibit the same cascade dynamics, convergence patterns, and Phase progression. Content-prediction is impossible; structural prediction is possible.

Content-focused intervention failure: Interventions targeting specific Content ("never again" focused on anti-Semitism specifically; anti-racism focused on particular racial categories) will fail to prevent crises with different Content. The architecture will route around Content-specific barriers, finding new H to fill the scapegoat slot. Only architecture-targeted intervention (Subcriticality maintenance) can prevent the general class of crises.

Cross-cultural generality: The architecture should produce crisis dynamics across all cultures and historical periods where the minimal conditions obtain, not because cultures share Content, but because the Bayesian machinery is species-general. Apparent cultural variation in crisis susceptibility should trace to structural parameters (A-floor security, I-domain vitality, authority distribution), not to cultural Content.

13.6.7 The Infinite Variation Problem for Prevention

Content-independence poses a specific challenge for prevention: you cannot enumerate and prohibit all possible scapegoat contents. Any list of "protected groups" is finite; the space of possible H is infinite. A society that successfully prohibits persecution of Groups A, B, and C has done nothing to prevent persecution of Group D, E, or any group not yet conceived.

This is why AIM's intervention architecture targets the mechanism rather than the Content:

• The Knowledge Patch (the Patch) teaches recognition of any scapegoat dynamic, not protection of specific groups
• Observability Governance reduces amplification of any false evidence, not specific false claims
• A-floor provision reduces susceptibility to any Confabulated explanation by restoring deliberative capacity
• Authority distribution prevents any M-captured leadership from imposing crisis, regardless of their specific ideology

The Presumption Against Scapegoating is explicitly Content-neutral: it applies whenever a population converges on any designated enemy, asking whether the attribution passes the Three-Test Battery regardless of who the target is. This Content-neutrality is not a limitation but a feature; it addresses the infinite variation problem by targeting the invariant architecture rather than the variable Content.

---

13.7 Why Population-Level Subcriticality Is Not Sufficient

The account to this point locates the crisis transition in the population. Individual cascades synchronise, the collective posterior climbs, and Section 13.8 gives the condition under which amplification outruns correction. Read on its own, that account carries an assumption: that a crisis trajectory is settled by where the collective posterior sits, with leadership as the instrument through which a population's Belief is executed rather than as a determinant in its own right.

The assumption does not hold, and saying so is a limit on what this chapter's own machinery delivers rather than a qualification of it. Where authority is concentrated, the Individual Confabulation Cascade of Section 13.3 runs to certainty inside one Decision Hub on self-generated evidence, and an Individual who has crossed the threshold for action and holds command over crisis-relevant resources, control of the dominant information channels, appointment power, and Appetitive leverage over dependent populations can impose crisis dynamics whatever the collective posterior is doing. The route bypasses the collective layer entirely: nothing in it requires the population's Belief to have moved at all.

Two rates decide whether that matters, and both are already in hand. Imposition through concentrated authority operates on timescales of days to months. The interventions this chapter's Subcriticality analysis specifies, distribution of the Knowledge Patch, Appetitive floors, protected Intrinsic domains, visibility governance, operate on timescales of months to decades. So population-level Subcriticality is necessary and is not sufficient, and the insufficiency is structural rather than a matter of execution.

The Leadership Compression Principle that states the condition, the four capacities that define its threshold, and the architectural requirements that follow from it are derived at Chapter 17, Section 17.7, where the apex route stands beside the floor route as the second of the two ignition loci. What this chapter establishes is the Bayesian half: why a cascade inside one Decision Hub needs no population to complete it, and why the corrective mechanisms that operate at population scale cannot reach a decision taken on the apex timescale.

---

13.8 The Subcriticality Threshold

13.8.1 Defining the Threshold

Subcriticality is the parameter space within which corrective mechanisms outpace mimetic amplification. Below the threshold, Individual Confabulation Cascades are interrupted before reaching the Fixed Point, collective convergence is slowed or reversed by counterevidence, and the system remains in the Grounded or early Drift region of the Crisis State parameter space.

Formally, the chapter works with two rates: the mimetic amplification rate (cascade velocity) and the corrective intervention rate, the combined speed of Knowledge Patch penetration, I-domain engagement, A-floor security, and reality-testing through direct experience. The system is subcritical when the corrective rate exceeds the amplification rate. The Subcriticality Threshold is the parameter configuration where the two rates balance: the boundary between regimes where correction dominates and regimes where amplification dominates. The rates are defined formally in Appendix A.5.8, and the condition they stand in at Appendix A.5.8.4. The Subcriticality Maintenance Principle that governs operation below this threshold is formally specified in Appendix A.5.7.5 (three-lever formulation) and A.5.7.7 (five-channel Knowledge Patch footprint).

The Subcriticality Threshold can be expressed in Coupling Strength-Decoherence Time terms. The mimetic amplification rate rises with Coupling Strength, which enters it through the weight expressed Beliefs carry as evidence in other Individuals' updating (Appendix A.5.8.4): at high Coupling Strength, expressed Beliefs constitute strong evidence for other Individuals' updating, which accelerates cascade velocity. The corrective rate is an inverse function of Decoherence Time: short‑decoherence‑time means reality-contact arrives quickly, providing exogenous evidence that competes with the endogenous evidence circulating in the coupled system. The subcriticality condition therefore translates to: reality-contact must arrive faster than the coupled dynamics can drive the system toward the Collective Fixed Point. Equivalently, the institutional environment must maintain Coupling Strength below the critical coupling strength (Critical Coupling Strength) at which cascade velocity exceeds corrective capacity, and Decoherence Time short enough that exogenous evidence interrupts the cascade before convergence stabilises.

13.8.2 Parameters Affecting the Amplification Rate

Mimetic amplification rate increases with:

Observability infrastructure: Social media, continuous news cycles, real-time information feeds accelerate social proof dynamics. Each expression of Belief reaches larger audiences faster, providing more evidence for more Individuals' updating cycles. Delivered without pause, the same channels accumulate an attribution by repetition even at a fixed audience, which is the Ambient Evidence Amplification Effect (Section 13.4.8).

A-precarity: Populations experiencing A-insufficiency have reduced deliberative capacity. Cognitive resources consumed by survival computation are unavailable for cascade-interruption. A-urgency also increases M-susceptibility; desperate Individuals follow models more readily.

I-domain collapse: When Intrinsic engagement is unavailable (education reduced to credentialing, work reduced to compensation-seeking, community reduced to status display), Individuals lack the alternative reference points that would reveal M-capture through contrast.

Reference group homogeneity: When Individuals' information environments contain only confirming voices, social proof is unanimous. Heterogeneous reference groups provide counterevidence that slows convergence. The operative quantity is homogeneity, together with how live the coupling is, and not the number of Individuals: once attention is fully spent, observing fewer Individuals redistributes it rather than reducing it (Appendix A, Section A.5.7.4). The restructuring that answers homogeneity is set out at Chapter 18, Section 18.2.2.

Crisis conditions: Exogenous shocks (economic collapse, pandemic, military defeat) simultaneously increase A-precarity, disrupt I-domains, and provide the diffuse harm that triggers scapegoat-seeking. Crises are amplification-rate accelerators.

The Knowledge Patch as Access Pathway, Not as Cure

The Knowledge Patch is the first and most architecturally consequential of the five intervention channels enumerated in Appendix A.5.7.7, not because it reduces any channel parameter directly, but because it opens the access pathway through which all five reductions become available to the Individual and therefore society as a whole. In the same way that an intravenous line cures no condition of its own but opens the route by which medicine can reach a patient who cannot receive it otherwise, the Patch does not correct Source Opacity, does not eliminate Confabulation, and does not dissolve Bayesian Lock-In. What the Patch does is route around these architectural conditions, providing the behavioural proxies (the Two-Question Diagnostic, the Presumption Against Scapegoating, Epistemic Suspension, and AIM Vocabulary Acquisition) through which the Individual can direct the other four channels against dynamics that the Individual cannot introspect on directly.

The architectural force of this framing is that each of the five channels of Appendix A.5.7.7 requires the Patch to be operative at the level of the Individual. Channel 1, the I-Override gate reducing the fast-layer loop gain by intercepting some fraction of M-driven actions, requires the Individual to recognise that M-weighted action is occurring, and Source Opacity denies the Individual introspective access to that recognition unless a behavioural proxy is in place. Channel 2, Hypothesis-Space enlargement reducing the Restriction-Induced Concentration Factor, the distortion a missing explanation forces onto the explanations that remain, is operationalised through AIM Vocabulary Acquisition, which is itself a Patch component. Channel 3, Interference-Term maintenance reducing the Evidence-Emission Coefficient, how firmly a settled Belief is sent back to drive fast reactions, is operationalised through Epistemic Suspension, which is itself a Patch component. Channel 4, attenuating the Evidence coupling between Individuals, is operationalised through Patch-competent Evidence intake that recognises coupling-mediated Evidence as coupling-mediated rather than as observation. Channel 5, reducing the Observability coupling across the network, is nominally institutional but requires Patch-competent institutional designers to recognise why the intervention matters at all. The Patch is the common prerequisite of all five.

The Patch is therefore necessary but not sufficient. Patch adoption without the five-channel treatment behind it is an intravenous line with no medicine flowing through it: the access is open, but nothing is administered. Attempting the five-channel treatment without Patch adoption is an attempt to administer medicine by a route the condition does not admit (like squirting the medicine in a syringe onto skin): the treatment cannot reach its target because the architectural conditions that would permit it to are not in place. The AIM Framework's interventionism is accordingly ordered. Patch adoption is the first move. The four channels the Patch makes available are the treatment. Institutional design at the fifth channel is the collective-action complement that closes the gap between individual Patch efficacy and population-scale stability.

This ordering refines the non-utopian commitment recorded in Appendix A.6 and the Positive Mimesis Preservation Principle (Chapter 8, Section 8.15.2). The Patch does not cure any architectural condition. It cannot cure Source Opacity, which is an information-theoretic entailment of Common-Currency Integration and is architecturally irreducible. It cannot cure the Bayesian Lock-In that follows from Self-Coupling Strength being structurally positive. It cannot cure the path-dependence of cross-Individual Belief dynamics that follows from N3 Non-Commutativity Propagation. What the Patch does, and what no other intervention in the AIM Framework's architecture can do, is open the pathway through which the reductions of all five gain components (the fast-layer loop, the slow Belief loop, the coupling between them, and the two network channels) become operations that an individual or an institution can actually perform. Everything downstream in the prevention programme depends on this opening.

13.8.3 Parameters Affecting the Corrective Rate

Corrective mechanism rate increases with:

Knowledge Patch distribution: The Two-Question Diagnostic and the Presumption Against Scapegoating, once acquired, enable manual Source-tagging by inference. Individuals with the Patch can recognise Confabulation and decline to incorporate Confabulated evidence into their updating.

A-floor security: When basic needs are reliably met, Individuals retain deliberative capacity. They can pause before updating, consider alternative explanations, and resist social proof when it conflicts with direct experience.

I-domain vitality: Active Intrinsic engagement provides experiential contrast; the Individual knows what non-M-driven motivation feels like and can use that knowledge to diagnose M-capture.

Institutional buffering: Independent judiciary, free press, distributed authority, and protected civil society provide structural resistance to cascade progression. They create friction that slows amplification even when individual-level correction is insufficient. At the population scale, formal democratic and civic design often institutionalises these forces as constitutional checks, independent media norms, low-theatre dispute resolution, and social insurance.

The framework's architectural analysis assigns each of these arrangements a formal role that extends beyond the normative case for their value. That role is visible in what happens to a system that is already in crisis.

Earlier in this chapter the Crisis State Architecture was described in terms of a trajectory: escalation builds, the Belief Register locks in, Concealment Infrastructure holds the inflated state together, and eventually the whole structure collapses. The collapse, when it comes, is not gradual. It is a jump: the system moves discontinuously from the inflated state to a post-cascade baseline. Think of it as a bridge that holds under increasing load until it does not, and then fails all at once rather than bending slowly into a safe position. The architecture predicts this as the default outcome for a supercritical system, because nothing in the individual-level architecture supplies a brake. The Mimetic channel has no self-terminating feedback. Source Opacity withholds the compounding from the Individual's introspection, so it goes uncorrected unless inferred from behaviour. The Concealment Infrastructure that sustains the inflated state does not oscillate; it accumulates until a forcing event breaks it. The result is a one-way collapse rather than a recoverable correction.

This is where institutional design enters the architecture in a formal rather than merely normative sense. Constitutional checks, independent media norms, low-theatre dispute resolution, and social insurance each operate, at the level of the population's coupled Belief dynamics, as braking forces that the individual-level architecture does not supply on its own. They push back against deviation. They slow the approach to the point of no return. And where they are strong enough and correctly structured, they can convert the system's trajectory from a one-way collapse into something that oscillates: a cycle of escalation and correction from which within-episode recovery is possible, rather than a jump from which recovery must wait until after the collapse has run its course.

The difference between these two outcomes is not a matter of degree but of type. A system that oscillates can be corrected while the episode is still in progress. A system that collapses can only be reconstructed afterward, under the conditions that follow a Terminal Phase. The formal conditions under which institutional buffering is sufficient to convert a collapsing trajectory into a recoverable one rest on the bifurcation-type derivation at Appendix A.5.7.6, and the specification of what sufficient buffering requires structurally is in Appendix A.5.7.8.

The practical implication is direct. The erosion of any one of the four buffering mechanisms named above is not merely a reduction in the network's normative quality. It is a reduction in the braking force that keeps the system in the recoverable regime. This matters at every scale the framework addresses, from organisational rivalry to geopolitical competition between nuclear-armed states. At that last scale, the difference between a system that can correct within an episode and a system that collapses without recovery is the difference between a crisis that ends and a war that does not.

Appendix locus: Appendix A.5.7.8.

Observability Governance: Reducing unnecessary observability of status markers, comparison data, and prestige cues reduces the information that amplifies M-dynamics.

Each intervention dimension operates on Coupling Strength, Decoherence Time, or both. The Knowledge Patch reduces Coupling Strength at the Individual level by introducing a conceptual vocabulary that enables Individuals to discount endogenous evidence (recognising that socially transmitted Beliefs may be M-contaminated rather than independently informative). A-Floor Provision shortens Decoherence Time by restoring direct reality-contact through material security (an Individual whose basic needs are met has less reason to accept social consensus uncritically and more cognitive capacity for deliberative assessment). I-domain protection reduces Coupling Strength by providing alternative evidence sources (engagement in I-domains generates evidence about one's own capacities and satisfactions that is independent of social consensus). Observability Governance reduces the network Observability gain by limiting the channels through which realised valuations are amplified into observed Model signals (reducing the rate at which one Individual's choices become inputs to other Individuals' comparison).

The preceding analysis treats "institutional buffering" and "social capital" as quantities that resist cascade dynamics. The Social Capital Decomposition Principle, which follows from the Unidirectional PM Flow Principle (Chapter 8, Section 8.14.2), refines this treatment by identifying that what is conventionally called social capital decomposes into two structurally distinct components. The first is A-I Buffer: genuine trust, cooperation, mutual aid, shared I-engagement, and the institutional arrangements that sustain these. A-I Buffer is the component that actually resists cascade dynamics, because it provides alternative value-sources (I-domain engagement) and material security (A-floor provision) that reduce susceptibility to M-amplification. The second component is PM: the extraction mechanism that operates through Status hierarchy, producing the deference patterns, credibility differentials, and attention asymmetries analysed throughout this chapter. PM does not buffer against cascade dynamics. It accelerates them, because PM-differential is the mechanism through which high-Status testimony carries disproportionate evidential weight (as established at Section 13.4.1).

When institutional analysis treats "social capital" as uniformly protective, it conflates a genuine buffer (A-I component) with an extraction mechanism (PM component) that is itself part of the crisis architecture. The Subcriticality Threshold should therefore be understood as sensitive to A-I Buffer depth specifically, not to "social capital" as a composite measure. Societies with high aggregate social capital but extreme PM-concentration may have less crisis-resistance than societies with lower aggregate social capital but deeper A-I Buffer, because the PM-component of the former's social capital is accelerating the very dynamics the A-I component is attempting to resist.

13.8.4 The Critical Depth of Worst-Off Deprivation

Sections 13.8.2 and 13.8.3 set an amplification rate against a corrective rate, and the Subcriticality Condition holds where the second outpaces the first. One parameter appears in both, and its appearance in both is what makes it decisive rather than merely important.

The corrective mechanisms of Section 13.8.3 presuppose that exogenous evidence retains enough weight to displace Mimetically-sourced priors. Sustained Appetite deficit erodes exactly that weight. An Individual whose bodily provision is chronically insecure downweights incoming evidence, and a reduction in the weight carried by exogenous evidence is observationally indistinguishable from a rise in Effective Coupling Strength, the coupling an Individual actually experiences in a given context once the raw parameter has passed through the filters of content-distance and recency: in both cases the posterior tracks the population's own output rather than the world's. The same erosion drains the Belief Heterogeneity Buffer, since reality-contact that no longer displaces priors no longer sustains the cross-Individual Belief variance from which the Buffer's damping derives. So deprivation raises the amplification rate and lowers the corrective rate at once, and both movements are monotone in the depth of the deficit.

The deficit that matters is the worst-off segment's, not the population mean's. The mean can sit comfortably above the A-Floor while a segment sits below it, and the segment below it is where the rate condition first fails, because the Subcriticality Condition has to hold across the guarded range of collective Belief rather than on average across Individuals.

It follows that there is a unique critical depth of sustained worst-off deprivation at which amplification first outruns correction across that whole range. Below it the population self-corrects. At it the Subcriticality Condition fails. This is the Floor-Breach Ignition Corollary, stated formally at Appendix A, A.5.8.7, on the collective mean-field dynamics of A.5.8.3 and the rate condition of A.5.8.4.

Two consequences follow, and the second is why the result is stated here rather than left implicit in Section 13.8.3. Universal Appetite-floor security is an entailment of Subcriticality Maintenance rather than a commitment adopted alongside it: the A-Floor Requirement is referenced to the worst-off Individual for the reason just given, and a floor that holds on average does not hold at all. And nothing about leadership appears anywhere in the condition, which means the population route into crisis is complete without an apex. The Floor-Breach Ignition Corollary is therefore one locus of a paired result whose other locus sits at the opposite extreme of the distribution of provision and authority. The pairing is the Dual-Locus Ignition Theorem, derived at Chapter 17, Section 17.7.8, where the apex condition it pairs with is established, and the discharge the floor condition produces is the mass-mobilised rupture whose economic face is developed at Chapter 17, Section 17.5.

13.8.5 The Buying-Time Strategy

Subcriticality maintenance is not a one-time achievement but a continuous process. The strategy is not to eliminate M-dynamics (impossible given human architecture) or to achieve permanent safety (there is no "mission accomplished" moment), but to buy time: maintaining the system in the subcritical regime indefinitely through sustained intervention.

The mathematics that permits crisis also permits prevention. The same Bayesian machinery that produces convergence on false Beliefs can be interrupted by:

1. Changing the evidence: Distribute the Knowledge Patch so that Confabulated explanations are recognised as such rather than incorporated as evidence
2. Changing the priors: Build strong I-self-models through I-domain engagement, so that alternative Hypotheses ("this is M-driven") have higher initial probability
3. Changing the updating: Provide A-security so that deliberative assessment replaces desperate M-following
4. Changing the transmission: Implement visibility governance to reduce social proof intensity

Each intervention shifts the balance of the two rates toward correction dominance. Cumulative intervention compounds; a higher corrective rate now produces more Individuals with the Patch, who are less susceptible to future amplification and who can transmit correction to others.

In Coupling Strength-Decoherence Time terms, the engineering objective is to maintain the system in the Classical or Slow Classical quadrant of the Domain Matrix, where updating tracks Truth (even if slowly), rather than allowing drift into the Coupled Domain where updating tracks system dynamics. This does not require eliminating Coupling Strength (which would destroy the social coupling infrastructure on which positive mimesis, cultural transmission, and I-driven relatedness depend) but maintaining Coupling Strength below‑coupling‑strength_c while keeping Decoherence Time short enough that exogenous evidence remains competitive with endogenous evidence. The goal is parametric management, not parametric elimination.

The goal is trajectory deflection, not cure. Deferral maintained over centuries is civilisation-scale success. Demanding ultimate solutions is itself a mimetic trap, often a status display; vast improvement over the status quo is achievable even if perfection is not.

13.8.6 Subcriticality and the Content of Belief

Convergence to Truth and convergence to consensus are different events with different mechanisms, and they are told apart by what does the converging.

Where a Referent is available and a population's Beliefs are in contact with it, the evidence reaching each Individual is sampled from the world. The washing-out result holds: posteriors move toward the true Hypothesis regardless of where the priors started, and the agreement that results is held in place by the Referent rather than by the agreeing. Agreement of that kind is stable and cheap to maintain, because what corrected the population into it goes on correcting the population within it. This is the short-Decoherence-Time regime of the Decoherence Trichotomy, and a population that has come to agree that water carries cholera has been corrected by cholera.

Where the Referent is absent, or present but not tracked, the evidence reaching each Individual is not sampled from the world. It is expressed Belief, generated by the Beliefs under test, and the Washing-Out Principle records this as the exact boundary at which classical convergence fails. Convergence still occurs, and occurs faster, because endogenous evidence is unanimous by construction. What it converges on is consensus, and the destination is the Collective Fixed Point of §13.4. Nothing outside the population holds that agreement in place, so it must be maintained by the mechanisms that produced it, and it discharges rather than corrects when a Referent finally arrives.

Crisis lives in the second case. The Beliefs that carry a population across the threshold are Grievance Beliefs, attributions of diffuse harm to agents: who is responsible for the decline, which group is taking what is owed, whose presence explains the loss. The Agent-Based Causal Default supplies the agentic reading before any evidence is in, and the harm is diffuse precisely because no Referent is in contact to settle the attribution. A population cannot be corrected into agreement about who is to blame for a distributed cause. Whatever agreement it reaches on such a question, it has reached by coupling.

Heterogeneity of Belief is a corrective resource on those terms, and §13.8.3 has already counted it as one: the counterevidence a heterogeneous reference group supplies is a term on the corrective side of the threshold. Where no Referent supplies exogenous evidence, competing Hypotheses are the only exogenous evidence available. Each faction's account of what has gone wrong contradicts every other faction's account, so contrary evidence arrives as a routine event rather than an exceptional one, and each cascade draws probability mass from the others before any reaches the action threshold. This is the Belief Heterogeneity Buffer, and its damping comes from the count of live Hypotheses because in this class of Belief the count is what is available: accuracy cannot supply the damping where accuracy cannot be established.

What makes heterogeneity rather than volume the damping term is the way this population differs from a merely noisy one. In a system of independent evaluators, a faulty member can corrupt the reports it sends but not the procedure the others run to assess them, so faulty members are outvoted and error stays bounded. The Confabulation Cascade has no such separation. Each Confabulation becomes a prior shaping how the next evidence is read, so what is corrupted is not the report but the procedure that evaluates it, and the population is not a set of sound evaluators receiving some bad inputs. Adding Individuals therefore does not help, because more Individuals running a corrupted evaluation produce more confident agreement rather than correction. What damps the cascade is the number of live Hypotheses among them, not the number of them, which is why the Buffer is a property of the Belief distribution rather than of population size.

The configurations order by count. Many competing accounts is stabilising, and stabilising even where every account is substantially Confabulated. Two accounts locked against each other is destabilising, and two is the specific number at which the damping inverts, because mutual targeting manufactures its own confirming evidence: each side's hostility toward the other is real, observable, and exactly what the other side's Grievance Belief predicts, so the interference that damped the multipolar case becomes amplification (Buffer Inversion). One account is the completed cascade. What carries the system toward the threshold is the consolidation itself, from many to two, and not the arrival of any particular Belief.

Two consequences follow.

The first is diagnostic. Rapid convergence of a population's attributions in a long-Decoherence-Time domain is a warning indicator, and the accuracy of the attribution converged upon does not enter the indicator, because in that domain accuracy could not have caused the convergence. A society whose competing explanations of its own condition are collapsing into two is approaching the threshold while each side experiences the collapse as its own side finally being heard. Consolidation presents to the participants as progress and multipolarity as dysfunction, so the friction doing the protective work is felt as a failure to agree.

The second is architectural. The interventions this chapter has enumerated (Knowledge Patch distribution, A-floor security, I-domain vitality, institutional buffering, visibility governance) are specified on rates and thresholds and not on the content of anyone's Beliefs. A-floor security operates through the deliberative capacity it restores, and restores it in a person of any political persuasion. Independent courts and an independent press slow cascade progression regardless of which cascade. An intervention conditional on shared Belief would have to obtain that agreement in a domain where agreement is available only by coupling, and would therefore manufacture the convergence the threshold exists to prevent.

Belief Multipolarity Principle: Where Decoherence Time is long relative to cascade velocity, so that expressed Belief rather than a Referent is the dominant evidence, collective subcriticality is sustained by the multiplicity of live Hypotheses in a population and not by their correctness. Convergence of a population's attributions in such a domain is a movement toward the Subcriticality Threshold irrespective of the accuracy of the attribution converged upon, and the destabilising transition is the consolidation of many competing accounts into two, at which point mutual targeting supplies each faction with genuine confirming evidence and the damping the multiplicity provided inverts into amplification. Subcriticality Maintenance therefore requires institutional architecture whose operation is independent of the content of Belief. Multiplicity is necessary and not sufficient, because a spreading cascade consumes it. Where Decoherence Time is short, a live Referent re-imposes the Truth continuously and convergence upon the Truth is truth-tracking; the Belief Multipolarity Principle makes no claim against agreement of that kind.

The penultimate clause bounds the resource. Multiplicity is not a stock that stays where it is put: a cascade converts adherents off their prior Hypotheses as it spreads, so the heterogeneity that was damping it is drained by its own progress, and the drain accelerates as the count of live accounts falls. This is the ratchet of the Susceptibility-Dispersion Result, which pairs the protective content axis with an anti-protective dispersion of adoption thresholds. Disagreement buys the interval in which the other four interventions operate. It does not substitute for them.

The Belief Multipolarity Principle exempts its own transmission, and by its scope clause rather than as a special case. The framework's claims about the architecture (the three sources, the destruction of source information at integration, the threshold structure of a coupled population) are governed by a live Referent, and the evidence for them is sampled from the world rather than generated by holding them. On the Hardness–Referent Matrix they sit on the truth side, in the regime where washing-out holds and convergence is the Referent's work. Widespread agreement about the architecture is therefore agreement of the safe kind, which is what makes the distribution of the Knowledge Patch the first move of the intervention set rather than a contradiction of it. What the framework does not seek is agreement about the attributions the architecture is used to examine, or about the institutional route by which a derived condition is met. Its own derived requirements (that the floor be held, that authority be gated, that the vocabulary be supplied) sit with the substrate claims, and the boundary between what is derived and what is chosen is the one the Hardness–Referent Matrix draws. Uptake sufficient to hold the system subcritical is the standard in either case.

The Belief Multipolarity Principle constrains what institutions may be built on, not what Individuals may believe. The configuration of Belief across a population (how many accounts are live, how strongly they are coupled, how fast exogenous evidence arrives, how dispersed the adoption thresholds are) determines whether the system sits above or below the threshold, and that configuration is a governable quantity. The contents of any Individual's conviction are not. The framework does not secure stability by securing agreement. It builds an environment in which disagreement is survivable indefinitely, and leaves the domains that have Referents free to converge on them.

---

13.9 Robustness: Minimal Sufficient Conditions for Crisis Architecture

13.9.1 The Independence of Crisis Dynamics from Specific Axioms

The preceding analysis derived crisis dynamics from the specific AIM architecture: Source Opacity prevents reliable introspection, Confabulation generates plausible but false explanations, and Bayesian updating on Confabulated evidence produces convergent false Belief. A natural objection asks: what if Source Opacity is less complete than claimed? What if Confabulation operates differently? What if the three-source taxonomy requires revision?

This section establishes that the crisis architecture is robust to axiom failure. The Bayesian mechanism requires only the following minimal sufficient conditions:

Condition 1: False Evidence Generation Some mechanism produces evidence that is treated as informative but does not reliably track Truth. This evidence may be self-generated (Confabulation), other-generated (propaganda, manipulation), or system-generated (algorithmic amplification of engagement-maximising content regardless of accuracy).

Condition 2: Social Transmission with Amplification False evidence spreads through social channels, with transmission probability positively correlated with features other than truth-value: prestige of source, emotional Salience, narrative coherence, visibility, and social proof.

Condition 3: Bayesian Incorporation Individuals treat socially-transmitted information as evidence, updating their Beliefs in accordance with (approximately) Bayesian principles. High-consensus social Belief is interpreted as evidence for the believed proposition.

Condition 4: Reflexive Constitution Individual Beliefs, once updated, are expressed and thereby become part of the social evidence that others incorporate. The system is reflexive: individual Belief → expression → social evidence → individual Belief.

These four conditions are jointly sufficient for crisis dynamics. If they obtain, the Collective Fixed Point is reachable regardless of the specific mechanism generating false evidence.

13.9.2 AIM as One Instantiation

The AIM Framework provides a specific explanation for how Condition 1 is satisfied: Source Opacity ensures that the primary evidence for self-assessment (phenomenological experience of desire, self-generated explanations) does not reliably track motivational source. Confabulation generates authentic-sounding explanations regardless of actual source.

But Source Opacity and Confabulation are not the only mechanisms that could satisfy Condition 1. Alternative mechanisms include:

Motivated reasoning: Individuals generate evidence that supports desired conclusions regardless of Truth, driven by emotional investment, identity protection, or cognitive dissonance reduction, without requiring Source Opacity specifically.

Information environment corruption: Deliberate disinformation, algorithmic content curation optimised for engagement rather than accuracy, and filter bubbles that systematically exclude disconfirming evidence: external mechanisms that produce false evidence without requiring any failure of individual cognition.

Cognitive limitations: Bounded rationality, heuristic reasoning under uncertainty, and systematic biases identified by behavioural economics: mechanisms that produce false evidence through processing limitations rather than architectural opacity.

Social construction of evidence: Standards of evidence that privilege consensus over accuracy, institutional capture of knowledge-production, and prestige-based rather than accuracy-based credibility attribution: mechanisms operating at the social rather than individual level.

Each of these mechanisms, alone or in combination, could satisfy Condition 1. The crisis architecture does not depend on AIM being the correct explanation for false evidence generation; it depends only on some mechanism satisfying the condition.

13.9.3 The Robustness Implication

This robustness has a counterintuitive implication: the crisis architecture is more concerning, not less, if AIM's specific axioms are uncertain.

If Source Opacity and Confabulation were the only mechanisms capable of generating socially-amplified false Belief, then falsifying these specific claims would dissolve the crisis architecture. The threat would be contingent on AIM being correct.

But the minimal conditions analysis reveals that falsifying Source Opacity or Confabulation would not dissolve the crisis architecture; it would merely require identifying which alternative mechanism satisfies Condition 1. The crisis dynamics follow from any mechanism producing socially-amplified false evidence, and multiple such mechanisms plausibly exist.

This means:

• The crisis architecture is overdetermined; multiple independent mechanisms could produce it
• Falsifying AIM-specific claims shifts the explanatory burden but does not eliminate the threat
• The historical regularity of crisis trajectories (which the architecture explains) remains to be explained even if AIM is wrong about the specific mechanism

13.9.4 What Would Actually Dissolve the Crisis Architecture

Given the minimal conditions, genuine dissolution of the crisis architecture would require demonstrating that none of the following obtain:

1. No false evidence generation: All mechanisms by which individuals form Beliefs about causes of suffering reliably track Truth. Self-generated explanations accurately reflect actual causes. Socially-transmitted information is systematically filtered for accuracy before transmission.
2. No amplification of false evidence: Social transmission probability is uncorrelated with truth-irrelevant features. Prestige, emotional Salience, narrative coherence, and visibility do not affect transmission independent of accuracy.
3. No Bayesian incorporation of social evidence: Individuals do not treat others' expressed Beliefs as evidence. Social consensus does not update individual priors.
4. No reflexive constitution: Individual Belief expression does not become evidence for others. The feedback loop between individual and collective Belief is broken.

Each of these conditions is empirically implausible. False evidence generation is documented across psychology, behavioural economics, and information science. Amplification of false evidence is the defining feature of misinformation dynamics. Bayesian incorporation of social evidence is a well-established feature of human inference (and often rational: others' Beliefs are evidence, absent systematic corruption). Reflexive constitution is inherent in social communication.

The crisis architecture is therefore robust across a wide range of possible worlds, including worlds where AIM's specific axioms are false but alternative mechanisms satisfy the minimal conditions.

13.9.5 Implications for Intervention

The robustness analysis clarifies intervention targeting. Effective intervention need not depend on AIM being the correct explanation for Condition 1; it must target the minimal conditions themselves:

Targeting Condition 1 (False Evidence Generation)

• If the mechanism is Source Opacity + Confabulation: Knowledge Patch enables manual Source-tagging by inference
• If the mechanism is motivated reasoning: debiasing training, adversarial collaboration, pre-commitment to evidence standards
• If the mechanism is information environment corruption: platform accountability, algorithmic transparency, institutional integrity
• If the mechanism is cognitive limitations: decision aids, structured deliberation, expertise systems

The specific intervention depends on the operative mechanism, but the category of intervention, reducing false evidence generation, is robust across mechanisms.

Targeting Condition 2 (Amplification) Observability Governance addresses Condition 2 regardless of the specific Condition 1 mechanism. Reducing prestige-based transmission, dampening emotional Salience in information design, and breaking engagement-maximising algorithmic amplification, all reduce transmission probability for false evidence whatever its source.

Targeting Condition 3 (Bayesian Incorporation) The evidential asymmetry analysis (Section 13.4.5) addresses Condition 3: not all evidence is suspect, but mimetically-transmitted information deserves particular scrutiny. This intervention is robust; it targets the specific evidence channel vulnerable to Conditions 1 and 2, whatever the mechanism producing false evidence.

Targeting Condition 4 (Reflexive Constitution) Low-visibility deliberation contexts, delayed expression requirements, and anonymised contribution to collective decision-making all reduce reflexive constitution by breaking the direct link between individual Belief and social evidence.

13.9.6 The Residual AIM Contribution

If the crisis architecture is robust across mechanisms, what distinctive contribution does AIM provide?

Explanatory specificity: AIM identifies why false evidence is generated in the domain of motivational self-assessment specifically; Source Opacity and Confabulation explain the particular phenomenology of sincere but inaccurate self-attribution. Alternative mechanisms explain false evidence in other domains; AIM explains it in this one.

Intervention precision: If Source Opacity is the operative mechanism, the Two-Question Diagnostic is precisely targeted; it reconstructs source-discrimination through deliberative inference, since Common-Currency Integration leaves no source-tag to read. If motivated reasoning or information environment corruption are operative instead, different interventions may be more effective. AIM provides the diagnosis that enables intervention selection.

Architectural understanding: AIM explains not merely that false evidence is generated but how it is generated through the integration of motivational signals in common-currency valuation with subsequent source-tag loss. This architectural understanding enables prediction about which contexts will produce more or less false evidence (high-visibility, high-stakes, A-urgent contexts should produce more Confabulation).

Connection to Mimetic Desire: AIM's distinctive contribution is identifying Mimetic Desire as a third motivational source that is systematically disguised through Confabulation, mistaken for A-necessity or I-value. Alternative mechanisms might explain false evidence generally; AIM explains false evidence about M specifically, which is crucial for understanding crisis dynamics that hinge on M-driven scapegoating being experienced as genuine threat-response.

The crisis architecture is robust; AIM provides the specific instantiation that explains the particular domain where false evidence is most consequential for civilisational outcomes.

13.10 The Symmetry of Crisis and Prevention

Appendix locus: A.5.3 (Saddle-Node Default and Hopf Sub-Case architectural inaccessibility at the fast layer, and the Ignition-Release Asymmetry); A.5.6 (the Monotone Belief-Race Lemma, the slow-layer exclusion); A.5.7.6 (bifurcation-type preservation at the network scale, derived at decohered operating points). The Symmetry of Crisis and Prevention holds within the same bifurcation type, the Fold, in both directions of approach; conversion to a Hopf cycle is not available by attenuation of the existing couplings and requires introducing a new dynamical variable with intrinsic negative feedback, as set out at A.5.3 and A.5.7.6.

The architecture of crisis and the architecture of prevention are the same architecture. They draw on the same five channels, they are governed by the same stability threshold, and they move in opposite directions across it. The threshold is one value; the channel configuration at which it is met is not. A system ignites at one configuration and releases at a lower one, so the reduction that ends a crisis is larger than the reduction that would have prevented it. The claim is structural, not rhetorical.

The Total Stability Condition is the joint threshold below which mimetic compounding decays rather than escalates. Read as a working estimate, the condition is the sum of five contributions, one from each channel, and the estimate holds when the channels are weakly coupled, so that adding their separate contributions is a good reading of the whole. Each contribution measures how strongly that channel amplifies mimetic signal. When the sum sits above the threshold, compounding takes hold: the Belief Register locks in, and the trajectory runs toward the Terminal Phase. When the sum sits below the threshold, disturbances die away, and the Individual and the network return toward a functional baseline. The sum is that working estimate. Whether the coupled system as a whole is still settling is the exact test, and Appendix A, Section A.5.7.4 states the test. The five channels, and the Knowledge Patch interventions that reach them, are as follows. The interventions do not pair one to one with the channels: two of them act on the second channel, the third channel has none of its own, and the intervention on the first channel reaches the fifth as well (Appendix A, Section A.5.7.7).

The first channel is the fast-layer effective loop gain, the gain of the within-Individual mimetic loop before conscious access. This is the rate at which the Decision Hub converts observed Model behaviour into compounding Mimetic drive. The Knowledge Patch intervenes here through the I-Override gate, which inserts a step between the felt valuation and the motor output. The gate does not prevent Mimetic activation. It reduces the probability that activation converts directly into the behaviour that would sustain the loop. Because the fifth channel's gain carries the same Decision Hub weight, the same interception attenuates it as well.

The second channel is the slow-layer loop gain, the gain of the Bayesian Belief-updating cycle. This is the rate at which Confabulation concentrates the Belief Register onto A-framed or I-framed Hypotheses, reinforcing the fast loop at the level of sincere Belief. The Knowledge Patch intervenes here through AIM Vocabulary Acquisition, which enlarges the Hypothesis Space the Individual holds available to include the Mimetic source as a live candidate. When the Hypothesis that a valuation is Mimetically driven is representable, the Belief Register can update toward it. When that Hypothesis is absent from the space entirely, no Evidence can place weight on it, and Bayesian Lock-In proceeds by architectural default.

The third channel is the cross-coupling remainder, the higher-order amplification the fast and slow layers generate between them beyond the leading loop that joins them. That leading loop belongs to the second channel. A locked Belief Register conditions the Observer's templates, the template-coded Salience the Observer delivers raises the Mimetic contribution to the Decision Hub's valuation, which raises the fast-layer gain, which generates further Evidence for the locked conclusion; the slow-layer loop gain is the gain of that cycle, bounded at the Belief Register although the cycle runs through the fast layer (Appendix A, Section A.5.6). No Knowledge Patch intervention acts on the remainder directly (Appendix A, Section A.5.7.7), but the remainder is built from the same couplings those interventions reduce, and falls with them. Epistemic Suspension, which maintains a non-zero Interference Term in the Belief Register, keeping competing Hypotheses in play rather than allowing the register to collapse to a single conclusion, acts on the second channel alongside AIM Vocabulary Acquisition: it lowers the rate at which the slow layer emits Evidence for a conclusion that has not been earned. Suspension does not require uncertainty about the facts. It requires the Individual to hold the source diagnosis open while the facts are assembled.

The fourth channel is the network Evidence-coupling gain, the gain with which neighbours' emitted conclusions enter an Individual's Belief Register. This is the gain of the social Belief-propagation pathway: each Individual's locked Belief Register becomes Evidence for neighbours, and if coupling is strong enough the network cascades. The Knowledge Patch intervenes here through Patch-competent Evidence intake, which reduces the effective weight an Individual assigns to neighbours' conclusions when those conclusions were formed under Source Opacity. Patch-competent intake is not scepticism in the colloquial sense. It is a specific operation: checking whether an emitted conclusion is consistent with a Mimetic source before treating it as diagnostic.

The fifth channel is the network Observability gain, the gain with which observed Model behaviour reaches Individuals across the network. This is the channel through which one Individual's realised valuation becomes another's Model signal, scaled by the institutional and technological structures that govern visibility. Observability Governance intervenes here by reducing the density and amplitude of Model signals flowing across the network, through disclosure norms, platform design, sumptuary arrangements, and analogous institutional structures.

The symmetry is this: the five channels that must be elevated for crisis to occur are exactly the five channels that must be reduced for prevention to take hold. There is no sixth channel that crisis uses and prevention does not, and no channel that prevention requires that crisis leaves untouched.

Class 1 interventions, as defined in Appendix A.5.1, target the ignition pathway: the stream of observed Model signals reaching the Individual. At the network scale, Observability Governance (Channel 5) is the institutional instantiation of that ignition-pathway reduction, operating on the network Observability gain that governs how one Individual's realised valuation becomes another's Model signal. Class 2 interventions target the internal loop gains: the within-Individual and cross-Individual coupling strengths that determine whether the cycle compounds. At the network scale, Channels 1 through 4 together constitute the Class 2 targets. The two framings are the same architecture described at individual and network scale respectively. See Appendix A.5.1 for the individual-scale definition and Appendix A.5.7.4 and A.5.7.7 for the network-scale extension.

The symmetry can be misread in three ways the architecture does not support.

The symmetry is directional rather than temporal. It does not entail that prevention and crisis take the same time, proceed at the same pace, or are equally easy to initiate. The Saddle-Node Default (specified in Appendix A.5.3) means that crisis, once supercritical, tends toward a discontinuous collapse: the system jumps. Prevention, operating through the Knowledge Patch and institutional design, works by gradual gain reduction across channels that are already coupled and already partially locked. Nor are the costs. The Ignition-Release Asymmetry fixes the difference: because the release configuration sits below the ignition configuration, prevention undertaken before a system ignites is cheaper than the reduction required to release it afterwards, and the gap widens the further the system has travelled. The direction is symmetric; the dynamics and the price are not.

The symmetry is architectural rather than moral. It does not imply that the causes of crisis and the causes of prevention are morally equivalent, that Individuals who sustain supercritical loops are culpable in the same way as those who undermine preventive institutions, or that any normative equivalence holds between escalation and de-escalation. The architecture describes what is coupled to what. Moral evaluation is a further question that the architecture informs but does not resolve.

The symmetry is cumulative rather than substitutive. Reducing gain in one channel does not substitute for reducing gain in the others. The Total Stability Condition adds the five contributions on that same working estimate. A partial intervention that drives one channel well below threshold while leaving the others near or above it may move the total downward without crossing it. Sustained subcriticality requires the sum to fall below threshold, which typically requires coordinated gain reduction across multiple channels rather than a single large intervention in one.

The symmetry is operable, in the current state of the architecture's intervention tools, only through the Knowledge Patch. Its interventions reach every channel but the third directly, and the third is the higher-order remainder of the coupling between the two layers the first two interventions act on. No other specified intervention reaches as many at once. The relationship between the Knowledge Patch and the framework's intervention architecture is developed in the Access Pathway subsection within §13.8, which establishes the Patch as the access pathway to the architecture rather than as a cure for Mimetic Desire.

The Symmetry of Crisis and Prevention is a property of the updating rather than a recommendation drawn from it. What follows from the Symmetry of Crisis and Prevention once a population has passed the point at which it can supply its own correction is derived at Chapter 18, Section 18.16.3.

The Three-Regime Resolution of Collapse and Oscillation

If the same five channels that drive a society toward crisis are the channels that prevent it, what does the trajectory look like in between? Does a society under mimetic pressure crash, or does it cycle, rising and falling without reaching the cliff? The answer is that it does both, in different conditions, and which one happens depends entirely on how many of the five channels are being managed and how deeply.

Three regimes cover every possibility.

The first is the unmanaged regime. Here, none of the five channels is meaningfully attenuated. People have not learned the vocabulary that would let them recognise mimetic pull as mimetic, so the corrective Hypothesis is missing from the list their minds can entertain. Institutions broadcast status comparisons rather than dampening them. The Salience feedback that turns one person's wanting into the next person's wanting runs unobstructed. Evidence flows through the population unchecked by anything resembling epistemic suspension.

In this regime, the system has no internal brake. The mimetic channel has no built-in stop, the way hunger has a built-in stop in eating. So when the loop is strong enough to compound on itself, there is no oscillation, no soft landing, no gradual return to fundamentals. The structure simply gives way all at once, and the system snaps to whatever functional baseline would have existed if the mimetic premium had never been there. Bubbles do not deflate, they pop. Calcified consensus does not negotiate itself back to Truth, it shatters when forced visibility arrives. And at civilisational scale, the fold is nuclear annihilation, the terminal one-way snap of an unmanaged supercritical system whose stakes have been escalated to species-level and whose architecture supplies no internal brake to stop the escalation short of detonation. This is the architecturally entailed outcome of running the system unmanaged. The technical literature names the geometry of this kind of one-way snap a fold, and the AIM Framework's prediction that this is the default shape of unmanaged mimetic crisis is the Saddle-Node Default treated formally in Appendix A.

One escape from this verdict suggests itself. The fast loop may lack a brake, but a society also runs a slower machinery: Belief revision. Beliefs are supposed to correct. Evidence arrives, the Belief Register updates, and the updating itself might be expected to supply the swing the fast loop cannot, conviction overshooting, correction arriving, the system wobbling its way back toward Truth.

The Monotone Belief-Race Lemma (formal statement at Appendix A, Section A.5.6) establishes that this expectation fails, and that it fails for the same reason the updating works at all. Hypotheses inside a Belief Register are in a race: whichever best explains the delivered Evidence gains ground, and the others lose it. But the race has two properties that together fix its shape. The referee is blind: under content-agnostic updating (Axiom 8), how strongly the Evidence supports a Hypothesis does not depend on how popular that Hypothesis currently is, so no runner gains ground because a different runner is ahead. And the track tilts toward the front-runner: a Hypothesis's gain is proportional to the mass it already holds, and the Self-Coupling channel feeds the leading Hypothesis's own dominance back to it as endogenous Evidence, so the leader is biased toward growing its lead. The first property means the race can only sort; the second means the sorting compounds. A blind-refereed race tilted toward its leader converges faster and entrenches harder than a fair one, which is Bayesian Lock-In stated as a race, and it is why a front-runner that leads only because the true Hypothesis is missing from the field becomes harder to overtake the longer it leads.

What neither property can produce is circulation. For Beliefs to cycle, the tilt would have to run cross-wise, each Hypothesis gaining because a different one is currently ahead, around a closed ring, and that structure is a violation of content-agnostic updating rather than a setting of it. The same two properties fix what Belief reform requires and guarantee that it remains available. Because the referee is blind, a Hypothesis that enters the field with genuinely better support gains ground from the moment it enters, however unpopular it starts and however long the incumbent has led; this is the property on which AIM Vocabulary Acquisition depends, since a newly representable Hypothesis could never overtake if support were graded by popularity. The front-runner is a front-runner at a given moment, not by standing appointment. What the tilt sets is the price: overtaking an entrenched leader takes longer and costs more Evidence the longer the lead has compounded, which is why Belief reform is difficult in proportion to the delay in beginning it, and never impossible while the race is still being run.

And the tilt itself has a point of intervention, because it operates through emission: the leading Hypothesis's dominance feeds back as endogenous Evidence at the strength the Interference Term permits, so Epistemic Suspension, by holding the Interference Term open, throttles the leader's self-reinforcement at its source while leaving the blind grading untouched. The consequence is that collective Belief has no thermostat, and the correction must therefore be worked rather than waited for: what the architecture withholds is self-correction, not correctability. The Bayesian layer contributes gain to the loop and never a swing, and a population that does not act on the race's open properties receives its return to reality as discontinuity.

The same result carries the Fold across scales. Because no Individual's Belief layer can rotate, and the couplings between Individuals run one-signed, the network-scale bifurcation is a derived result rather than an inference from the fast layer (Appendix A, Section A.5.7.6): at the decohered operating points characteristic of the Calcified Phase, the first loss of stability of the whole coupled population is the same non-oscillatory crossing, and the three routes to genuine oscillation are each the installation of a named structure rather than a drift of parameters, a restoring node carrying intrinsic negative feedback, a prevalence-responsive Evidence ring, or a self-coupling memory driven past its threshold.

The second is the partially-managed regime. Here, some of the five channels have been engaged but not enough of them, or not deeply enough, to bring the system below the critical threshold. An institutional buffer absorbs the worst of the Salience feedback. A fraction of the population carries the Knowledge Patch and exercises I-Override at decisive moments. Some exogenous appetitive or intrinsic signal arrives periodically and pulls the system back toward reality. None of this is sufficient to remove the underlying instability, but it is sufficient to supply, from outside, the restoring force the architecture does not generate from inside.

The geometry of this regime is not an oscillation. The bifurcation type is unchanged from the unmanaged case (Saddle-Node Default, per Appendix A.5.3), and what the external buffering achieves is an extension of the time-to-fold rather than a conversion into a different attractor. A naive linear reading of the trajectory can mistake the resulting ratcheting-then-jump pattern for a Hopf cycle, but Appendix A.5.3 establishes that a Hopf cycle cannot be produced by attenuating existing positive couplings or by adding additive damping; it requires introducing a new dynamical variable with intrinsic negative feedback (Appendix A), which is an architectural augmentation rather than a parameter adjustment. What the partially-managed regime actually exhibits is a Fold whose time-to-fold has been lengthened by external work that the architecture itself cannot do.

The clearest historical illustration is the run of nuclear close calls during the Cold War. The species avoided terminal collapse not because the underlying competitive dynamic between the executive leaderships of the two superpowers had been brought below threshold, but because individuals positioned downstream of executive authority, at single points in the operational chain, declined to execute orders or forward signals whose execution would have completed the trajectory to detonation. The Soviet officer who declined to forward a false satellite warning of an American first strike in 1983, and the Soviet submarine officer who refused to authorise a nuclear torpedo launch during the Cuban Missile Crisis in 1962, were not exercising executive authority. They were exercising a refusal at a node through which the executive's decision had to pass before becoming consequence.

The architectural lesson from these cases is not that the partial-management regime is safe; it is that under conditions of acute Leadership Compression at the executive level, the system can come to depend on downstream backstops at single operational nodes. Downstream backstops are not part of the Leadership Compression Principle's intervention architecture (Chapter 17, Section 17.7). They are an ad hoc property of the operational chain, not a designed feature of governance. The restoring force they supply is not distributed and not architecturally guaranteed. The historical record over the second half of the twentieth century reflects the closeness of the calls, not the robustness of the configuration.

The partial-management regime is therefore the regime in which a society can persuade itself that the danger has passed when in fact the threshold has not been crossed and the architecture has merely been masked. Two architectural responses are available, and they are not substitutes for one another. The first is to continue reducing the gain components until the total gain falls below the critical threshold, which moves the system into the third (subcritical) regime and is the only response that removes the underlying instability. The second is to lengthen the time-to-fold by distributing the authority over crisis-relevant resources (the structural constraints specified by the Leadership Compression Principle, Chapter 17, Section 17.7, and its Time-Buying Principle, Chapter 17, Section 17.7.6), which does not convert a supercritical system into a subcritical one but does buy the interval within which the first response can complete its work. Under nuclear conditions, both responses are required concurrently, because the first response operates on a generational timescale and the second response operates on the timescale of a single decision cycle.

The third is the subcritical regime. Here, the combined effect of the five channels of the Knowledge Patch footprint, vocabulary, epistemic suspension, I-Override, governance of who-can-see-whom, and institutional buffering, has been deep enough to bring the loop below the threshold at which it could compound on itself in the first place. The mimetic signal still arrives. People still notice what other people want. But the signal does not feed back on itself fast enough or strongly enough to escalate. It decays, the way a sound decays in a well-damped room. There is no bifurcation, no fold, no oscillation. There is just ordinary social life, with mimetic activation present but not running away with the system. This is the only regime in which the question of collapse-versus-oscillation does not arise, because neither is occurring.

The everyday meaning of critical can mislead. Critical in the technical sense names the condition at which the loop gain exactly equals the threshold, and a real system passes through that condition rather than inhabiting it. The condition is met twice over a full excursion, at two different configurations of the five channels: once on the way up, where the low state gives way, and once further down on the way back, where the inflated state gives way. The two regimes either side of the boundary are subcritical (below the threshold, third regime above) and supercritical (above the threshold). Both the unmanaged regime and the partially-managed regime are supercritical. They differ only in whether external damping is supplying the restoring force the architecture cannot supply for itself. Oscillation is therefore not a milder, intermediate condition. It is the same condition as collapse, with a different shape imposed on it from outside.

The trichotomy clarifies what the AIM Framework is actually proposing as the normative target. The aim is not to convert a collapsing society into an oscillating one. Oscillation is not the goal. Oscillation is the signature of a system that is still above the threshold, still in the same supercritical regime as the system that folds, and depending on external damping that can fail, and the nuclear near-misses make vivid how thin that thread is. The aim is to bring the system below the threshold entirely, into the third regime, by acting on enough of the five channels to make the underlying instability go quiet rather than merely cycle. That is what Subcriticality Maintenance names, and it is the only regime in which the dynamics genuinely settle.

The choice between collapse, oscillation, and a settled society is therefore the choice of how many of the five channels are operative, and at what depth. There is no fourth option.

The trichotomy is exhaustive over the architecture as it stands. Three regimes are available to a system whose loop runs on the couplings the architecture supplies, because a fourth would require a state variable the architecture does not contain. Building that variable is not a fourth option within the choice. It is the engineering work of changing which choice a society faces, and it is the design goal the AIM Framework sets for institutional construction. A structure that introduces a variable of its own, acts on one of the couplings that form the loop gain, and acts against rather than with, holds the system in a Governed Band: bounded by the structure's own feedback rather than by the Salience Ceiling, with margin preserved and the way back intact.

The structures a durable society accumulates have come closest to this without being built for it. Some introduce a variable of their own and some hold a setting; some act on the couplings and some act on coordinates the loop does not contain; and where all three conditions are met, they are met at strengths chosen for other purposes and engaged after the event they would have damped rather than before it. None of them converts. Qualification without conversion is what the partially-managed regime describes. That regime leaves two things to do, and they are not alternatives. The first is to reduce the five channels far enough that the loop can no longer compound at all, which removes the fold rather than deferring it. The second is to build a structure that converts, which does not remove the fold but holds the system away from it. The first is the more complete repair and the work of a generation. The second can be started on the timescale of a decision. A society with only the first has nothing holding it while the work proceeds. A society with only the second is depending on a structure that has to go on working indefinitely.

13.11 The Formal Specification of Cascade Dynamics

The cascade dynamics this chapter has developed narratively are formalised in Appendix A, Section A.5.8, which carries the individual Confabulation cascade, the cascade velocity profile, the collective mean-field dynamics, the rate form of the Subcriticality Condition, the Leadership Compression formalisation, and the Content-Independence result. Stated in prose, the formalisation establishes six results.

First, an Individual updating their Authenticity Belief on self-generated explanations converges to full false confidence from any starting point, because Confabulation manufactures authentic-sounding explanations regardless of source while the Individual reads them as diagnostic; the only stable resting point of the individual cascade is the Fixed Point. Second, cascade velocity follows an S-curve (slow initial movement, rapid mid-cascade acceleration, asymptotic approach) because there is little probability mass to shift early and little left to shift late.

Third, at the population scale the dynamics are bistable below a critical combination of private-evidence strength and social-proof strength: an interior threshold separates a self-correcting low-Belief basin from a cascading one, and at the critical combination the threshold collides with the low-Belief state, after which cascade is global. Fourth, Subcriticality is the condition that the corrective rate exceed the amplification rate everywhere above the safe Belief level: the phenomenological form of the same architectural condition that appears at the fast layer as amplification-per-cycle below unity and at the full coupled system as the Total Stability Condition. Fifth, population-level Subcriticality does not prevent crisis when leadership Belief and leadership authority both cross their thresholds (the Leadership Compression result, whose prose derivation is at Chapter 17, Section 17.7) so Subcriticality must be maintained at the leadership level as well, by selection constraints, authority distribution, or dismissal mechanisms. Sixth, no term of the updating dynamics indexes on the Content of the scapegoat Hypothesis: the architecture processes any candidate Hypothesis with identical dynamics, which is why comparative history organised around Content finds no common thread: the commonality lives in the parameters, not the Content.

13.11.1 Coupling Strength and Decoherence Time: Conceptual Definitions and Forward Reference

This subsection provides precise conceptual definitions for the two parameters introduced in Section 13.4.6, sufficient to support the analysis in this chapter. The full formal definitions, which require the non-Boolean probability framework, are derived at Chapter 14, Section 14.7, and are stated here in abbreviated form only.

Definition (Coupling Strength, κ). For a given epistemic domain, Coupling Strength measures the degree to which the evidence available to Individuals within that domain is generated by other Individuals' Belief-states within the same domain, rather than sampled independently from external reality. Coupling Strength is zero when all evidence is exogenous (no Individual's expressed Belief constitutes evidence for any other Individual's updating). Coupling Strength approaches its maximum when evidence is predominantly endogenous (each Individual's evidence base is substantially constituted by other Individuals' expressed Beliefs, which are in turn constituted by theirs).

At the formal level developed at Chapter 14, Section 14.7.1, Coupling Strength is defined as the supremum of the non-commutativity of measurement operators across the Belief state space. This means Coupling Strength measures the maximum degree to which the order in which an Individual evaluates different aspects of a situation changes the outcome of the evaluation. When Coupling Strength is above zero, there is no single classical probability space that can simultaneously model all evaluation contexts within the domain. This is a structural property of the domain's evidence architecture, not a cognitive limitation of the Individuals operating within it.

Definition (Decoherence Time, τ). For a given epistemic domain, Decoherence Time measures the expected time until external reality imposes costs that the coupled dynamics of the system cannot absorb. Decoherence Time is determined by the ratio of a system's buffering capacity (financial reserves, institutional resilience, relationship capital, geographic isolation from consequences) to the rate at which that buffer is consumed by divergence between Belief-states and reality.

At the formal level, Decoherence Time is defined as the timescale on which a decoherence channel (representing reality-contact) erases the order-dependence that Coupling Strength measures. Short Decoherence Time means reality rapidly collapses Belief-states back to correspondence with external conditions. Long Decoherence Time means coupled dynamics persist, allowing the system to evolve toward the Collective Fixed Point before correction arrives.

Domain Matrix (Summary Table). (On the tracked-state and Referent-attachment treatment of this regime, see Appendix A, A.12.4 and A.12.6.)

Forward Reference. The analysis in this chapter demonstrates that high‑coupling‑strength combined with long‑decoherence‑time produces the crisis dynamics formalised in the Bayesian Mimetic Crisis Principle, using classical Bayesian reasoning throughout. Chapter 14 develops the formal apparatus (Generalised Probability Theory, non-Boolean event algebras, interference terms) demonstrating that the boundary between the Classical and Coupled regimes is not merely a quantitative shift (more coupling, slower correction) but a qualitative phase transition: above Critical Coupling Strength, the probability space itself changes structure, and certain operations available in classical probability (such as simultaneous joint measurement of status satisfaction and welfare adequacy) become algebraically undefined. In that apparatus doubt is the non-zero Interference Term, and Epistemic Suspension, an operation of the Knowledge Patch, holds the Interference Term above zero against the decoherence that would collapse the Belief-state onto a single Hypothesis (Chapter 14, Section 14.9.3; Appendix A, Section A.5.6). This result provides the formal foundation for why the Knowledge Patch works: it is not merely a psychological intervention but a mathematical operation on the structure of the Belief space itself. These formal results strengthen every conclusion reached in this chapter. They are not required for it.

---

13.12 Chapter Closing

13.12.1 Establishment Summary

This chapter has established:

Individual Confabulation Cascade: Valid Bayesian updating on endogenously-generated evidence produces directional drift of the Authenticity Belief toward certainty, regardless of actual motivational source. The cascade terminates at a stable Fixed Point resistant to revision.

Collective Convergence: Individual cascades synchronise through social proof, producing population-level Belief shifts that accelerate as they progress. The Collective Fixed Point represents shared certainty about a potentially false Hypothesis.

Bayesian Mimetic Crisis Principle: M-coupled Beliefs compound through the population by a lock-in process that operates identically on accurate and inaccurate Beliefs once Coupling Strength exceeds the Subcriticality Threshold, which is the condition under which the Terminal Phase of the Crisis State Architecture functions as an attractor region in the parameter space. The dynamic is cyclical, with Reset returning the system to the Grounded region, from which drift toward Terminal recommences, until interrupted or until a nuclear-armed iteration triggers the termination condition. The attractor result in the domain where the architecture reaches ground is the Unifying Crisis Principle, derived at Chapter 17, Section 17.2.

Content Independence: The Bayesian architecture is content-agnostic; it processes any scapegoat Hypothesis with identical dynamics. This explains why comparative historical analysis, focused on content similarities across crises, failed to identify the common architecture. Surface variation in crisis content masks deep structural identity in crisis dynamics. The infinite variation in possible scapegoat designations means content-specific prevention is impossible; only architecture-targeted intervention can address the general class of crises.

Coupling Strength (κ) and Decoherence Time (τ) as the two structural parameters governing whether Bayesian updating in a coupled system tracks Truth or system dynamics, with the Domain Matrix classifying epistemic regimes by their Coupling Strength-Decoherence Time profile (Classical, Slow Classical, Unstable Coupled, and Coupled)

• The Coupling Strength-Decoherence Time characterisation of Crisis State Phases: the Grounded Phase as low‑coupling‑strength, short‑decoherence‑time (Classical Domain); progressive drift through rising Coupling Strength and extending Decoherence Time into the Coupled Domain; the Terminal Phase as Coupling Strength-maximal, Decoherence Time-effectively-infinite; catastrophic Reset as the exogenous shock (war, economic collapse, environmental catastrophe, institutional disintegration) that forces Coupling Strength-collapse by physically destroying the endogenous evidence infrastructure; and the nuclear termination condition as the scenario in which catastrophic reconfiguration eliminates the substrate for Coupling Strength-regeneration permanently, which the framework asserts as entailment from the dynamics validated at survivable scales rather than as an observed result.
• The insufficiency of population-level Subcriticality: Population-level Subcriticality is necessary and is not sufficient. Concentrated authority allows M-Captured leadership to impose crisis trajectories faster than population-level correction can operate, and the route requires no movement in the collective posterior at all. The Leadership Compression Principle stating the condition, and the architectural requirements that follow from the Leadership Compression Principle, are derived at Chapter 17, Section 17.7.
• Floor-Breach Ignition Corollary: A unique critical depth of sustained deprivation at the worst-off segment exists at which amplification first outruns correction across the guarded range of collective Belief, so that universal Appetite-floor security is an entailment of Subcriticality Maintenance rather than a commitment adopted alongside it (Section 13.8.4).

Subcriticality Threshold: The parameter space where corrective mechanisms outpace mimetic amplification. Maintaining Subcriticality requires sustained intervention across multiple dimensions: Knowledge Patch distribution, A-floor provision, I-domain protection, visibility governance, and authority distribution.

Robustness: The crisis architecture depends on minimal sufficient conditions (false evidence generation, social amplification, Bayesian incorporation, reflexive constitution) that are more general than AIM's specific axioms. Falsifying Source Opacity or Confabulation would not dissolve the crisis architecture; it would merely shift the explanatory burden to alternative mechanisms satisfying the same conditions. The architecture is overdetermined, making the threat more robust than any single theoretical framework.

Evidential Asymmetry: The corruption identified here is specific to mimetically-transmitted information. Direct observation, physical measurement, logical inference, and testimony from non-M-influenced sources remain reliable. Rigorous scrutiny applies specifically to socially-transmitted Beliefs amplified by prestige cues, visibility, and social proof.

13.12.2 Axiomatic Recap

This chapter's conclusions are grounded in:

• Axiom 3 (Source Opacity): Established in Chapter 9: Individuals cannot reliably introspect motivational source after common-currency integration
• Axiom 4 (Confabulation): Established in Chapter 11: the narrative system generates plausible explanations biased toward authentic-sounding attributions
• Standard Bayesian inference: The mathematical framework for rational Belief revision, applied to evidence that is endogenously generated rather than exogenous
• Social proof dynamics: The treatment of others' expressed Beliefs as evidence, creating reflexive feedback between individual and collective Belief

The principles derived (the Bayesian Mimetic Crisis Principle, the Floor-Breach Ignition Corollary, Content Independence, and Subcriticality) follow deductively from these foundations. The chapter does not introduce new axioms; it demonstrates what the established axioms entail when combined with standard probability theory.

13.12.3 Implications for Intervention

The mathematics that permits crisis also permits prevention. Cascade interruption is achievable through:

• Distributing the Knowledge Patch: Enabling manual Source-tagging by inference so that Confabulated evidence is recognised rather than incorporated
• Securing A-floors: Restoring deliberative capacity by removing survival-level urgency from decision-making
• Protecting I-domains: Providing experiential contrast that enables M-recognition through comparison with genuine I-engagement
• Implementing visibility governance: Reducing social proof intensity by dampening prestige-based transmission and M-amplification infrastructure
• Distributing authority: Preventing single-point M-capture from producing civilisational harm through structural constraints on concentrated power.
• Targeting architecture, not Content: Recognising that Content-specific protection leaves the infinite space of alternative scapegoat designations unaddressed

A structural asymmetry runs through this intervention list. Some of the components target problems that can be genuinely solved; others target a condition that can only be managed. A-floor provision and I-domain protection belong to the first category. A problem solved today is not a problem that exists tomorrow: the population whose A-floors are secured does not face A-driven susceptibility to Confabulated evidence on an ongoing basis, because the structural condition generating that susceptibility has been removed. The population whose I-domains are protected accumulates experiential contrast against M-capture, and that accumulated capacity compounds rather than resets. These are interventions whose returns persist, subject to maintenance of the conditions that produced them.

By contrast, visibility governance, authority distribution, and Knowledge Patch dissemination operate on M-dynamics directly, and M has no natural satiation signal. Their returns are real and necessary, but they must be continuously maintained, because M-driven drift toward the Terminal Phase recommences the moment structural constraint relaxes. The buying-time strategy that follows from this analysis is therefore not uniform across its components: some components solve the contributing problem and hold the solution in place; others provide ongoing deflection against a condition that cannot, by architectural design, be solved. The normative and institutional implications of this asymmetry, including the resource-allocation principles it entails and the priority ordering it generates, are not derived here.

The buying-time strategy accepts that Source Opacity cannot be eliminated, M-dynamics will persist, and drift toward the Terminal Phase remains mathematically possible. But practical Subcriticality is maintainable through continuous effort. The goal is deflection from the Terminal region, not cure, and deflection maintained indefinitely is civilisation-scale success.

13.12.4 Forward Connection

The results established here are taken up as follows. Chapter 8 supplies the mechanisms behind the amplification-rate parameters, and its Section 8.14 derives the Mimetic Premium and the Unidirectional PM Flow Principle, the extraction dynamics that the Value Attribution Asymmetry of this chapter anticipates. Chapter 12 applies those dynamics to money, interest and financial complexity, the domains in which the Bayesian architecture operates on economic Belief. Chapter 14 derives the non-Boolean substrate on which the updating treated here actually runs, and the order effects that follow from it. Chapter 17 develops what the architecture produces in the domain where it reaches ground: the Unifying Crisis Principle at Section 17.2, the apex route and the four architectural requirements that follow from it at Section 17.7, and the race condition at Section 17.17. Chapter 18 develops the intervention this chapter shows to be architecturally possible, including the five-channel Knowledge Patch footprint and what its distribution requires at population scale.

13.12.5 What Would Falsify This?

The claims of this chapter would require significant revision under the conditions set out below, organised by the claim each would disconfirm. Each criterion is Tier 2 or Tier 3: its failure revises the claim it names and leaves the architecture intact. These criteria are not already established by the phenomena this chapter explains; they remain open empirical questions. The architectural falsification surface, including the self-correction of systems below the Subcriticality Threshold, is consolidated at Appendix B, Section B.2.

Falsification of the Individual Confabulation Cascade

Prediction 1.1: Self-reported confidence in motivational self-knowledge should increase with decision-making experience while accuracy (measured behaviourally through revealed preference consistency, choice blindness susceptibility, and audience-removal testing) should not show corresponding improvement.

Falsifying evidence: Systematic positive correlation between decision-making experience and behavioural accuracy in source-identification tasks. If experienced decision-makers reliably distinguish M-driven from A/I-driven choices at rates exceeding chance, the cascade mechanism is disconfirmed. Where the distinction is made by introspection alone, without the framework's vocabulary, the observation also bears on Source Opacity itself, which Appendix B, Section B.2 carries as its third commitment.

Prediction 1.2: Explanation elaboration should increase along the cascade trajectory: Individuals with longer histories of confident self-explanation should generate more detailed justifications without corresponding accuracy improvement.

Falsifying evidence: No relationship between justification elaboration and cascade progression, or positive correlation between elaboration and accuracy.

Prediction 1.3: Challenges to the "I am authentic" self-model should produce stronger defensive response than challenges to comparably-held specific Beliefs.

Falsifying evidence: Equivalent defensive response to self-model challenges and specific-Belief challenges, or self-model challenges producing less defensiveness (suggesting the self-model is not protected by accumulated Bayesian confidence).

Prediction 1.4: In choice-blindness paradigms (Johansson et al., 2005), confident explanation of choices the participant did not make should be stronger in participants with longer decision-making histories.

Falsifying evidence: Choice-blindness explanation equally confident, or less confident, in participants with longer decision-making histories than in matched participants with shorter ones.

The deficit-irrelevant attribution shift of Section 13.3.4 tests Scalar-Only Delivery, which is Tier 1 and is stated with the third commitment at Appendix B, Section B.2.

Falsification of Collective Convergence

Prediction 2.1: Population-level Belief in scapegoat explanations should show acceleration dynamics: slow initial growth followed by rapid convergence as social proof accumulates.

Falsifying evidence: Linear growth in scapegoat Belief, or deceleration as Belief levels rise. If social proof does not amplify convergence, the collective mechanism is disconfirmed.

Prediction 2.2: Belief homogeneity should increase during crisis progression: the spread of Authenticity Belief across Individuals should narrow as the population approaches unanimity.

Falsifying evidence: Maintained or increasing variance during crisis progression. If Individual Beliefs do not synchronise through social proof, the convergence mechanism requires revision.

Prediction 2.3: Introducing heterogeneous information sources should slow convergence by providing counterevidence that reduces social proof unanimity.

Falsifying evidence: No effect of information heterogeneity on convergence rate, or heterogeneity accelerating convergence.

Falsification of the Bayesian Mimetic Crisis Principle

Prediction 3.1: Historical crises should exhibit parameter trajectories consistent with the predicted Crisis State Phase transitions.

Falsifying evidence: Systematic deviation from the predicted Phase pattern, with crises that skip Phases, reverse Phase transitions, or progress through an unrecognised sequence. Isolated exceptions are expected (the model is probabilistic), but systematic deviation would disconfirm the Phase account of the Crisis State Architecture and leave the architecture intact.

Prediction 3.2: Societies with stronger institutional buffering (A-floors, I-domains, distributed authority, visibility governance) should show lower probability of Terminal Phase progression and slower cascade velocity when crises occur.

Falsifying evidence: No correlation between institutional buffering and crisis progression, or buffered societies showing faster cascade dynamics.

Prediction 3.3: Intervention effectiveness should decrease monotonically with Phase progression, with early interventions producing larger effect sizes than late interventions.

Falsifying evidence: Phase-independent intervention effectiveness, or late interventions outperforming early interventions.

Falsification of Content Independence

Prediction 4.1: Crisis dynamics should be structurally identical across different scapegoat Contents: cascade velocity, convergence patterns, and Phase progression should not vary systematically with the identity of the designated group.

Falsifying evidence: Systematic variation in crisis dynamics correlated with scapegoat Content rather than structural parameters. If crises targeting Group A consistently progress faster than crises targeting Group B, controlling for structural parameters, Content-independence is disconfirmed.

Prediction 4.2: Content-specific interventions should fail to prevent crises with different content. "Never again" campaigns focused on specific historical atrocities should not reduce crisis probability for novel scapegoat designations.

Falsifying evidence: Content-specific interventions providing general protection against all crisis variants. If anti-Semitism education reduces probability of crises targeting immigrants, Muslims, or other groups, the content-independence claim requires revision.

Prediction 4.3: Novel scapegoat designations should emerge unpredictably from past Content but predictably from structural conditions. Societies with high amplification rates and low corrective rates should generate crises regardless of which specific group is designated.

Falsifying evidence: Scapegoat Content predictable from prior Content history, or structural conditions failing to predict crisis emergence regardless of Content.

Falsification of the Insufficiency of Population-Level Subcriticality

Prediction 5.1: Threshold-crossing in concentrated-authority systems should occur through leadership capture rather than population consensus: crisis imposition should precede majority population endorsement.

Falsifying evidence: Systematic pattern of crisis imposition following (rather than preceding) population-level consensus. If populations must reach near-unanimity before leadership can impose crisis, the compression mechanism is disconfirmed.

Prediction 5.2: Crisis velocity should correlate positively with authority concentration: more concentrated systems should show faster progression from trigger event to catastrophic outcome.

Falsifying evidence: No correlation between authority concentration and crisis velocity, or distributed systems showing faster crisis progression.

Falsification of the Subcriticality Framework

Self-correction below the Subcriticality Threshold is Tier 1 and is stated with the fifth commitment at Appendix B, Section B.2.

Prediction 6.1: The Knowledge Patch (Two-Question Diagnostic + Presumption Against Scapegoating) should reduce individual susceptibility to cascade dynamics, measurable through reduced confidence updating on Confabulated evidence.

Falsifying evidence: No effect of Knowledge Patch acquisition on updating behaviour, or Patch acquisition increasing susceptibility.

Prediction 6.2: A-floor provision should reduce cascade velocity by restoring deliberative capacity.

Falsifying evidence: No relationship between A-security and cascade velocity, or A-security accelerating cascades.

Falsification of the Robustness Claim

The criteria below are Tier 3: each concerns which mechanisms satisfy the minimal conditions of Section 13.9 in actual populations, and its failure revises the Robustness claim and leaves the architecture intact. The dependence of crisis dynamics on the coupling conditions is Tier 1 and is stated with the fifth commitment at Appendix B, Section B.2.

Prediction 7.1: Crisis dynamics should be observable across diverse cultural and historical contexts, consistent with the claim that the architecture is overdetermined by multiple mechanisms satisfying the minimal conditions.

Falsifying evidence: Crisis dynamics systematically absent in contexts where AIM-specific mechanisms (Source Opacity, Confabulation) are inoperative but alternative mechanisms (motivated reasoning, information corruption) remain present. This would suggest the crisis architecture depends specifically on AIM axioms rather than on the minimal conditions.

Prediction 7.2: Interventions targeting different Condition 1 mechanisms should show effectiveness proportionate to the operative mechanism in each context: Knowledge Patch effective where Source Opacity dominates, platform accountability effective where information environment corruption dominates.

Falsifying evidence: Intervention effectiveness uncorrelated with mechanism diagnosis, suggesting either the mechanism diagnosis is incorrect or the intervention theory is flawed.

Prediction 7.3: The four minimal conditions should be empirically identifiable across crisis cases: false evidence generation, amplification, Bayesian incorporation, and reflexive constitution should all be present in documented crisis trajectories.

Falsifying evidence: Crisis trajectories where false evidence generation, Bayesian incorporation or reflexive constitution is demonstrably absent. If crises occur without reflexive constitution, the minimal conditions framework requires revision. A crisis arising without social transmission amplification would instead disconfirm the coupling condition, which is Tier 1 and stated with the fifth commitment at Appendix B, Section B.2.

Comparative Framework Falsification

Beyond specific predictions, the framework should be evaluated comparatively. The relevant question is not "does AIM perfectly predict all cases?" but "does AIM outperform competing frameworks in explanatory and predictive power?"

Comparative prediction: AIM should outperform undifferentiated preference models in explaining:

• Apex dysfunction (persistent dissatisfaction at maximum preference-satisfaction)
• Confidence-accuracy divergence (high confidence with low accuracy in self-assessment)
• Crisis trajectory regularity (repeated historical pattern of scapegoat convergence and sacrificial violence)
• Intervention effectiveness patterns (early intervention outperforming late intervention)
• Content independence (identical dynamics across varied scapegoat designations)

Falsifying evidence: Undifferentiated preference models providing equivalent or superior explanatory coverage with fewer theoretical commitments. If preference-collapse economics can explain these phenomena without the three-source taxonomy, Source Opacity, and the Confabulation mechanism, parsimony favours the simpler account. The criterion is Tier 3: its failure falsifies the claim that the AIM Framework is the more parsimonious account of these phenomena, and leaves the architecture intact.

Coupling Strength-Decoherence Time Predictions

The introduction of Coupling Strength and Decoherence Time as named parameters generates additional falsifiable predictions beyond those stated above for the Individual Confabulation Cascade, Collective Convergence, the Bayesian Mimetic Crisis Principle and Content Independence:

1. coupling‑strength gradient prediction: Domains with higher measured Coupling Strength (operationalised through order-effect experiments, as specified at Chapter 14, Section 14.1) should exhibit faster cascade velocity, higher Collective Fixed Point stability, and greater resistance to corrective intervention than domains with lower Coupling Strength, controlling for Decoherence Time. Systematic failure of this prediction would disconfirm the claim that Coupling Strength governs cascade dynamics.

2. Decoherence Time-gradient prediction: Within domains of comparable Coupling Strength, systems with longer Decoherence Time (greater buffering capacity relative to extraction rate) should exhibit longer persistence of false consensus and more severe eventual correction (the Seneca Asymmetry) than systems with shorter Decoherence Time. Bank runs (high‑coupling‑strength, short‑decoherence‑time) should produce intense but self-limiting crises; ideological polarisation (high‑coupling‑strength, long‑decoherence‑time) should produce slower-building but more deeply entrenched crises. Systematic failure of this pattern would disconfirm the Decoherence Time parameter's role.

3. Domain Matrix classification: It should be possible to classify real-world epistemic domains into the four quadrants of the Domain Matrix using independent measurements of Coupling Strength and Decoherence Time, and the classified domains should exhibit the dynamics predicted for their quadrant. A domain classified as Coupled (high‑coupling‑strength, long‑decoherence‑time) that consistently exhibits Classical dynamics (rapid self-correction, truth-tracking convergence) would disconfirm the classification framework.

4. Intervention-parameter mapping: Interventions that demonstrably reduce Coupling Strength (such as blinded evaluation procedures, audience removal, or Knowledge Patch distribution) should produce measurable reductions in cascade velocity and Fixed Point stability. Interventions that demonstrably shorten Decoherence Time (such as A-floor provision, transparency reforms, or buffer-capacity limits) should produce earlier and less severe correction cycles. Interventions that reduce Coupling Strength without shortening Decoherence Time, or shorten Decoherence Time without reducing Coupling Strength, should produce the partial effects predicted by their parameter-specific mechanism, not the full effects of simultaneous coupling‑strength reduction and Decoherence Time-shortening.

13.12.6 The Stakes

The nuclear termination condition makes this analysis practically urgent. Previous iterations of the crisis dynamic, however catastrophic, permitted recovery and learning. The current iteration, with nuclear-armed states exhibiting Crisis State indicators consistent with drift toward Terminal, may not permit recovery if it reaches the Terminal Phase with sufficient intensity, and because that failure is unrecoverable the framework treats the entailment as action-guiding at a confidence it would not warrant for a recoverable one. The buying-time strategy is not abstract future planning; it is immediate civilisational necessity.

This analysis is offered because it appears correct, not because its implications are comfortable. It is submitted to the reader for verification: the predictions are testable and the historical record is available to test them against. And it is submitted under the Precautionary Asymmetry: because understating a terminal-stakes risk is the error that cannot be recovered from, an analysis of the Terminal Phase that appears correct is to be acted on, not deferred pending a civilisational-scale test that the stakes themselves make unavailable. The mathematics that permits crisis is identical to the mathematics that permits prevention. Subcriticality is maintainable. Hopelessness is warranted only for a catastrophe already completed, and only because no Individual remains who can still limit the damage or rebuild. Short of that, the trajectory can still be deflected. Nuclear weapons are the case in which the survival of those Individuals cannot be assumed.

---

The chapter treats the fast loop, the slow Bayesian loop and the network coupling as three quantities that add together. Adding them is a good enough guide for institutional design while the three stay weakly connected, and it is what makes the five channels separately actionable. It is an approximation. Where the loops are strongly connected they have to be assessed as one system rather than as three, because a combination can be unstable in ways none of the three predicts on its own, and Appendix A.5.7.4 carries that exact treatment. How large the connecting terms actually are is not yet known, and finding out means measuring two things across a real population: who treats whom as a source of evidence, and who can see whose behaviour. The Subcriticality Threshold introduced at Section 13.8 is the same threshold the appendix reaches as the Total Stability Condition, and the Critical Coupling Strength is a property of that whole network rather than of any one person or any single link within it.

---

References

Johansson, P., Hall, L., Sikström, S., & Olsson, A. (2005). Failure to detect mismatches between intention and outcome in a simple decision task. Science, 310(5745), 116–119.