Position Statement

This chapter establishes what the AIM Framework claims the human motivational system is. It does so in two stages preceded by orientation and followed by consolidation. The orientation, in Sections 4.1 and 4.2, establishes why AIM is presented axiomatically rather than narratively and how the nine axioms set out in Section 4.3 were arrived at through derivation from the convergence of three motivationally distinct neural systems on a single common-currency integration locus. The first foundational stage, in Section 4.3, states the nine core axioms in canonical form with their foundational citations. The second foundational stage, in Section 4.4, specifies the Minimal Architecture Specification: the four processing components (Decision Hub, Body Manager, Thinker, Observer), the Belief Register mechanism, the operational loop, and the architectural localisations and derived consequences that follow. Section 4.5 completes the account of integration by distinguishing its two joint outputs, the Common-Currency Scalar (the Scalar) and Phenomenal Affect. The consolidation, in Section 4.6, situates the foundational chapter within the treatise's larger arc, articulates the asymmetric stakes under which the treatise is offered, and bridges into the disaggregation argument of Chapter 5.

The chapter is foundational rather than argumentative: it states premises and specifies architecture. The argumentative case for disaggregating human motivation into Appetites, Intrinsic Motivation, and Mimetic Desire is the work of Chapter 5, which depends on the foundations laid here and which this chapter does not itself attempt to mount.

The chapter's central claim is that the human motivational system is a four-component architecture in which three motivationally distinct sources (the homeostatic-deficit signalling system, the dopaminergic intrinsic-engagement system, and the mirror neuron system) converge on a single integration locus that produces a scalar valuation for action selection, and that this architecture, together with the satiation and termination dynamics that distinguish the three sources, entails Source Opacity, Confabulation, Object Parasitism, and Bayesian Belief dynamics as structural consequences rather than as independent claims. The agent-based causal default is stated as a ninth axiom, on content, on which the architecture's formal results do not depend. These consequences are the load-bearing substrate on which every derivation in the chapters that follow depends.

Dependency Statement

This chapter depends on:

Chapter 2 for the canonical definitions of all capitalised terms (Appetites, Intrinsic Motivation, Mimetic Desire, Decision Hub, Body Manager, Thinker, Observer, Source Opacity, Confabulation, Object Parasitism, Belief Register, Behavioural Signatures, Dashboard Gauge Principle, Romantic Lie, Mimetic Premium, and the further capitalised terms used throughout);
Chapter 3 for the Output-Capacity Entailment Principle (OCEP), the Completeness Criterion for General Theory Status, the Replacement Requirement, and the methodological infrastructure within which the empirical premises of the axioms are evaluated;
Chapter 1 for the introductory framing of the three motivational sources, the integration architecture, and the asymmetric stakes under which the treatise is offered;
The Preface for the broader frame within which the methodological discipline of derivation is understood.

This chapter is depended on by every subsequent chapter in the treatise. Chapter 5 derives the disaggregation argument from the axioms and the Minimal Architecture Specification stated here. Chapter 9 (Neural Integration and Source Opacity) provides the formal proof of Source Opacity via the Data Processing Inequality, which presupposes Axiom 2's integration and Axiom 3's statement, in Section 4.3, of what that integration loses. The chapters on Appetites, Intrinsic Motivation, and Mimetic Desire develop each source at length using the axiomatic and architectural foundation laid here. The later chapters on Bayesian Belief dynamics, institutional capture, civilisational risk, and the Knowledge Patch (the Patch) each derive their claims from the axioms and the architecture stated in this chapter rather than from independent observation.

This chapter cites empirical studies as the best available accounts of the neural mechanisms underlying the architecture the AIM axioms describe. The explanatory obligation is imposed by OCEP (Chapter 3): the human motivational system demonstrably produces common-currency integration, source-tag loss, confabulation, preconscious mimetic transmission, and the further outputs the axioms enumerate. These outputs require neural explanation. Where a cited study faces replication challenges, the relevant question under the Replacement Requirement is not whether the study is beyond dispute, but whether a superior account of the same output exists. If it does, AIM adopts the superior account; if it does not, the cited study remains the best available discharge of the obligation. The architecture of AIM depends on the outputs, not on any particular mechanistic study.

Derivation Preview

This chapter derives:

The presentation of AIM as an axiomatic rather than narrative framework, justified by the requirements of examinability, falsifiability, parsimony, derivation-chain visibility, and the unreliability of introspective assent in a domain that includes the reader's own cognition, together with the reach of the derivation chain and the conservation discipline that fixes it: the body spends a finite energy budget, the architecture settles where that budget is spent rather than whether, and no step in the chain may introduce a cause absent from the step before it or allow an upstream signal to stop (Section 4.1);
The nine core axioms as the formal compression of a derivation that proceeded from Mimetic Desire as a triangular structure, through identification of the mirror neuron system as a candidate substrate, to the convergence of three motivationally distinct systems on the ventromedial prefrontal cortex and ventral striatum, to the satiation and termination dynamics that distinguish them, to the structural consequences entailed by the architecture (Section 4.2);
Axiom 1 (Three-Source Taxonomy), Axiom 2 (Common-Currency Integration), Axiom 3 (Source Opacity), Axiom 4 (Confabulation), Axiom 5 (Object Parasitism), Axiom 6 (Differential Satiation Dynamics), Axiom 7 (Preconscious Transmission), Axiom 8 (Bayesian Belief Dynamics), and Axiom 9 (Agent-Based Causal Default), together with their interlocking structure and the standing of Axiom 9 as the one axiom whose subject is content rather than structure, load-bearing for the form the architecture's failures take and not for the architecture or the results derived from it (Section 4.3);
The Minimal Architecture Specification, comprising the Signed Valuation Principle, the four processing components (Decision Hub as integrator and action-selector, Body Manager as homeostatic-deficit signaller, Thinker as prior-maintainer and Bayesian ground-resolver, Observer as conditional Object-field Salience assessor), the Belief Register mechanism, the operational loop, and the architectural localisations and derived consequences and the crisis geometry of the coupled loop comprising the Internal Loop Gain, the critical threshold and Saddle-Node Fold, the Parasitism Theorem, the five Knowledge Patch channels, and Subcriticality Maintenance, together with the statement (rather than derivation) of the Total Stability Condition and the Critical Coupling Strength, which Appendix A, Sections A.5.7.4 and A.5.7.5 define and bound (Section 4.4);
The two joint outputs of a single integration event, the Common-Currency Scalar as action-selection decision variable and Phenomenal Affect as the concurrently generated felt state, together with the application of the Source Opacity Theorem to both outputs, their differing durations, the analytic posteriority of the Thinker to both, and the conversion of the Scalar into action that the Affect does not undergo; and the four couplings by which the Scalar that leaves the body is carried to civilisational scale, namely integration at the Decision Hub, the Scalar's simultaneous conversion outward into action and inward into Evidence for the Thinker, the conversion of action into Evidence for whoever observes it, and the accumulation of inter-Individual Evidence into institutional structure (Section 4.5);
The consolidation of the foundational material into a principled basis for the disaggregation of human motivation, the articulation of the asymmetric stakes under which the treatise is offered, and the bridge into the disaggregation argument of Chapter 5 (Section 4.6).

Chapter Preview

4.1 Why AIM Requires Axiomatic Foundations Establishes the five reasons AIM is presented axiomatically rather than narratively: examinability of confronting conclusions, falsifiability of claims, parsimony of premises, visibility of the derivation chain, and the unreliability of introspective assent in a domain that includes the reader's own cognition. Closes with how far the derivation chain reaches and why: a motivational signal does not terminate at the boundary of the body, and the derivations are held to the conservation discipline of the physical sciences, under which no step may introduce a cause absent from the step before it or allow an upstream signal to stop.

4.2 How the Nine Axioms Were Arrived At Reconstructs the nine-step derivation by which the axioms were arrived at, beginning from Mimetic Desire as a triangular structure, through identification of the mirror neuron system as a candidate substrate, identification of the integration locus and the two further convergent systems, alignment with Self-Determination Theory's account of Intrinsic Motivation, derivation of the satiation and termination dynamics, derivation of Source Opacity via convergence-and-integration, derivation of the phenomenological diagnostics, and derivation of the remaining axiomatic structure. Closes with the role of the five quality-control constraints (parsimony, falsifiability, neural-architectural compatibility, dissociability, behavioural diagnosability) within the derivation, and a methodological-discipline acknowledgment.

4.3 The Nine Core Axioms States the nine axioms in canonical form: Three-Source Taxonomy, Common-Currency Integration, Source Opacity, Confabulation, Object Parasitism, Differential Satiation Dynamics, Preconscious Transmission, Bayesian Belief Dynamics, and Agent-Based Causal Default. Each axiom is stated with its canonical statement, explanation, supporting evidence where applicable, and foundational citations limited to no more than three per axiom. Establishes the standing of Axiom 9 as the one axiom whose subject is content rather than structure, on which the architecture and its formal results do not depend and on which the account of the form the architecture's failures take does depend.

4.4 The Minimal Architecture and Its AIM Instantiation Specifies the four processing components of the architecture, the Signed Valuation Principle that governs their inputs to the Decision Hub, the Belief Register mechanism, the operational loop that couples them, the architecture's characteristic failure mode, and the architectural localisations and derived consequences. States the dynamical structure of the coupled loop, the critical threshold $G_{cr i t}$ , the Saddle-Node Fold geometry, the Parasitism Theorem, the five Knowledge Patch channels, names (rather than derives) the Total Stability Condition and Critical Coupling Strength $κ_{c}$ , which Appendix A, Sections A.5.7.4 and A.5.7.5 define and bound, and identifies Subcriticality Maintenance as the normative target. Closes with a forward pointer to Appendix A for the formal mathematical treatment.

4.5 The Two Outputs of Integration — Scalar and Affect Completes the account of integration by establishing that a single integration event emits two distinct joint products: the Common-Currency Scalar (the action-selection decision variable) and Phenomenal Affect (the concurrently generated felt state). Shows that the Source Opacity Theorem applies to both, that they differ in duration (the Scalar resolves rapidly while the affect persists), that the Thinker, being the deliberative faculty, is analytically posterior to both and can act on an affective episode only after the fact and prospectively, and that the Scalar converts to action while the affect does not, which is what fixes where the AIM Framework's diagnostics attach. Closes by following what leaves the body through the four couplings that carry the account to civilisational scale, and notes that the Saddle-Node Fold established for the single Individual is preserved at the coupled-population scale by derivation rather than by analogy, at a decohered operating point under one-signed coupling.

4.6 From Foundation to Argument Consolidates the foundational chapter, summarises what Sections 4.1 through 4.5 have established, articulates the load-bearing role of the axioms and the Minimal Architecture Specification for every subsequent chapter in the treatise, distinguishes what the foundation has done from what the chapters that follow will do, and articulates the asymmetric stakes under which the treatise is offered. Bridges into Chapter 5 (Why Motivation Requires Disaggregation).

4.1 Why AIM Requires Axiomatic Foundations

A theory of human motivation can be presented in two ways. It can be presented as a narrative: a story about why people want what they want, organised around examples, illustrations, and the texture of lived experience, with theoretical claims emerging from the accumulation of cases. Or it can be presented axiomatically: a small set of precisely stated premises is fixed at the outset, and every subsequent claim is derived from those premises by explicit reasoning, with examples used to illustrate derivations rather than to substitute for them. The two presentations are not equivalent, and the choice between them is not stylistic. It determines what the theory can do.

AIM is presented axiomatically. The reasons are practical, and they bear on what the AIM Framework must do.

The first reason is that the conclusions AIM reaches are confronting. Many of them contradict intuitions that have been cultural inheritance for thousands of years, including the intuition that desires originate within the Individual who experiences them, the intuition that introspection is a reliable guide to one's own motives, and the intuition that institutions which appear to function to maintain social order and cohesion are in fact functioning. Confronting conclusions invite resistance, and resistance is appropriate; conclusions are not to be accepted on the author's word. An axiomatic presentation gives the reader an alternative to acceptance or rejection. It allows the reader to examine each axiom on its own terms, examine the derivations that follow from the axioms, and identify precisely where, if anywhere, the logic of the argument breaks down. The reader who rejects a conclusion can locate the axiom that produced it and contest the axiom directly. The reader who accepts the axioms is bound, on pain of inconsistency, to accept the derivations of they do in fact follow. This is what an axiomatic presentation gives the reader that a narrative presentation cannot: a precise specification of what is being claimed, what is being assumed, and what follows from what.

The second reason is that the conclusions AIM reaches must be falsifiable. A theory whose claims cannot be wrong is a theory that cannot be right either, in the sense that matters; it is a vocabulary rather than a science. Falsifiability requires that each claim specify the conditions under which it would be refuted. Axiomatic presentation makes those conditions explicit. Each axiom states what is the case in language sufficiently precise that a contrary empirical finding can be identified as contrary; each derivation states what follows from the axioms in language sufficiently precise that a contrary entailment can be identified as contradicting either the axioms or the derivation rule. Narrative presentation, by contrast, tends toward a register in which contrary findings can be absorbed by adjusting the narrative's emphasis rather than by surrendering its claims. AIM's commitment to falsifiability requires the more demanding presentation.

The third reason is parsimony. A theory of human motivation that posits more sources, components, or mechanisms than the explanatory work requires is a theory that has admitted excess machinery, and excess machinery is where errors hide. Excess machinery also tends to grow: each additional posit invites further posits to coordinate with it, and the theory expands without the expansion being earned. Axiomatic presentation forces parsimony at the foundation. The premises must be the minimum required to generate the explanatory and predictive work the theory does; any premise that can be removed without losing a derivation is a premise that should be removed. The discipline this imposes is not aesthetic; it is functional. A parsimonious foundation is easier to test, easier to refute, and easier to repair if a part of it fails.

The fourth reason is that the derivations matter. AIM is not, primarily, a catalogue of claims about motivation. It is a derivation chain: a sequence of entailments from a small set of axioms about what the human motivational architecture is, through statements about what that architecture necessarily does, to statements about what consequences follow at the scale of the Individual, the institution, the network, and the civilisation. The derivation chain is the theory. The axioms exist to be its starting points. Any presentation that obscures the derivation chain, by interleaving claims at different levels of inference or by treating distinct entailments as if they were independent observations, undermines what AIM is doing. Axiomatic presentation keeps the derivation chain visible at every step.

The fifth reason is more particular to this treatise. AIM's domain is the architecture of human cognition, and the claim AIM makes about that architecture is that it is largely invisible to the Individual whose cognition it is. Source Opacity, the Individual's structural inability to read, by introspection of the integrated signal, the source decomposition of any given desire, is one of the framework's central results. A theory of an architecture that operates beneath conscious access cannot be tested by appeal to what the reader feels is true about their own motivation, because what the reader feels is precisely the output of the architecture rather than a window into it. The theory must therefore stand on grounds independent of introspective endorsement. Axiomatic presentation supplies those grounds. It allows the reader to evaluate the theory by examining its premises and its derivations, rather than by checking whether the theory's claims match the reader's pre-existing self-understanding. In a domain where self-understanding is itself one of the things the theory is explaining, this is the only sound basis on which evaluation can proceed.

These five reasons taken together fix the presentation. AIM is presented axiomatically because the conclusions are confronting and must therefore be examinable, because the claims must be falsifiable and must therefore be precise, because the foundation must be parsimonious and must therefore be minimal, because the derivation chain is the theory and must therefore be visible, and because the domain includes the reader's own cognition and cannot therefore rely on the reader's introspective assent.

There is a tradition of axiomatic work in philosophy and the social sciences that AIM does not belong to. Axiomatic presentation does not mean that the axioms were derived from pure reason without reference to empirical findings. The nine axioms in Section 4.3 are empirical claims about how the human motivational system works, grounded in the published findings of neuroscience, psychology, and behavioural research; the foundational citations attached to each axiom identify the empirical basis on which that axiom rests. Axiomatic presentation does not mean that the framework is closed to revision. Each axiom specifies the conditions under which it would be refuted, and a definitive empirical refutation of any axiom would require the framework to be revised or rejected. Axiomatic presentation does not mean that the framework is more certain than the empirical findings from which it is constructed; it means that the framework is no more certain than those findings, and that its inferential structure is exposed to view rather than concealed in the connective tissue of narrative.

The chain reaches as far as it does because the physical process it describes reaches that far. A motivational signal does not terminate at the boundary of the body, it transduces. The body is a system with a finite energy budget, drawn in through the channels set out in Section 1.8, and at every moment that budget is being spent: on maintaining tissue, on holding attention on one thing rather than another, on moving the limbs. What the architecture specified in this chapter settles is not whether the budget is spent but where. The signal leaving the Decision Hub does not itself supply the energy of the act; it directs it, in the way that a valve does not supply the pressure in a pipe but determines what the pressure does. The physical act then follows as a consequence. A body moves, a thing is made or broken, a word is said, a resource changes hands, another Individual is helped or harmed. The architecture works within four constraints, and they are not of one kind. One budget runs down: the body's energy envelope, drawn in as fuel and oxygen and spent continuously, whose state the Body Manager reports as deficit (the drive to breathe, hunger, fatigue, sleep pressure) without itself deciding where it is spent. Two capacities limit what can be taken in at any moment without being depleted by use: the Observer's attention, which selects only so much of an unbounded field (Section 4.4.7), and the Thinker's deliberation, which holds one line of resolution at a time (Section 4.4.3). Neither capacity is used up by its own exercise, but both are bounded by the envelope: as fatigue and sleep pressure rise, the Observer selects less of the field and the Thinker holds a line of resolution less securely (Lim and Dinges, 2010; Appendix A, Section A.2.2). One channel spends all of it: the Decision Hub selects one action at a time, and because the direction of attention is among the actions it selects, the same selection fixes both what the Individual acts on and what the Individual attends to (Section 4.4.1). Competition between the components is competition at that channel. When the Observer's Salience dominates the integrated Scalar, the Decision Hub's selection moves to what the Observer marked, and the Scalar the Decision Hub returns to the Thinker is the one the Thinker must then resolve, so the Thinker's work is displaced without any draw on the Thinker's own capacity (Section 4.4.5).

The derivations are therefore held to a conservation discipline, which is the ordinary accounting rule of the physical sciences applied to an account of motivation. Energy is neither created nor destroyed but only converted from one form into another, and the same is required of every step in the chain: no step may introduce a cause that was not present in the step before it, and no step may allow an upstream signal simply to stop. Where an account says that a motivation "leads to" a social outcome without specifying the conversion, something has gone missing from the books, and what has gone missing is usually the mechanism on which the whole explanation depended. Holding the accounting closed is what obliges the argument to continue across the space beyond the Individual's skin. Having fixed what the sources are and how they integrate, the treatise cannot put the pen down at the body's edge, because the signal does not stop there.

The next section, Section 4.2, reconstructs the derivation by which the nine axioms were arrived at. The derivation history is itself part of what justifies the axioms taking the form they do. A reader who knows where the axioms came from, what constraints they had to satisfy, and what alternatives were considered and rejected can evaluate the axioms against those constraints. After Section 4.2, Sections 4.3 and 4.4 state the axioms and specify the architecture in their canonical form, and Section 4.5 completes the account of integration by distinguishing its two joint outputs. Section 4.6 closes the chapter by consolidating what has been established and bridging into Chapter 5.

4.2 How the Nine Axioms Were Arrived At

The nine axioms set out in Section 4.3 were not posited at the outset of this project and then defended. They were the formal output of a derivation that proceeded from a single concept, through a question about that concept's neural substrate, to an architectural account of human motivation, to the satiation and termination dynamics that distinguish the architecture's components, to the entailments that follow from those dynamics, and finally to the compressed statement of those entailments as a premise set. This section reconstructs that derivation. Following the derivation permits the axioms to be evaluated against its evidential and architectural constraints rather than against the author's authority.

The derivation has nine steps, which are set out in turn below.

The first step is the starting concept: Mimetic Desire as a triangular structure. The concept was acquired from the work of René Girard, whose literary analyses establish that desire is not a straight line from a desiring subject to a desired Object but a triangle with a Model, another Individual, at its apex: the subject desires the Object because the Model desires it, the Object's perceived value being raised, in the subject's valuation, by the Model's pursuit of it. Girard's account is presented in literary and anthropological terms rather than neural terms, and within his frame the question of what neural structure would support such desire-copying is not specifically addressed. But the question is natural once the concept is in hand. If desire can be transmitted from a Model to a subject, the transmission must occur through some pathway in the subject's brain, and that pathway must connect the subject's perception of the Model to the subject's own valuation of the Object. The first step in the derivation was to ask what pathway that might be. Girard himself framed his observations as a hypothesis, rather than a complete theory. AIM takes that hypothesis as a starting point for this theory.

The second step is the identification of the mirror neuron system as a candidate substrate. Mirror neurons are a class of neurons that fire both when an Individual performs an action and when the Individual observes another performing the same action. They were first identified in macaques, and a human homologue is evidenced at the systems level, and they constitute a neural mechanism by which observation of another's behaviour generates, in the observer, a partial neural representation of the observed behaviour as if the observer were performing it. The mapping from this mechanism to Girard's triangular desire is direct: if the subject observes the Model pursuing an Object, the subject's mirror neuron system represents the Model's pursuit, and that representation contributes to the subject's own valuation of the Object. The literature on mirror neurons had already entertained this possibility before AIM was formulated, and several authors had suggested that mirror neurons may provide a substrate for the acquisition of desire from observed others. The derivation took this suggestion as a working Hypothesis.

The third step is the question of how mirror neuron output becomes action. A mirror neuron system that represents observed behaviour without that representation contributing to the observer's own action selection is a system that does nothing. The representation must reach the part of the brain that selects what the observer does next. The relevant integration locus is well established in the literature: the ventromedial prefrontal cortex and the ventral striatum encode a common-currency subjective value signal for each Object the Individual considers, and action selection is performed on that signal. The third step in the derivation was therefore to ask what other systems converge on that integration locus, because the existence of one system feeding the integration locus implied the existence of others.

The fourth step is the identification of the two further convergent systems. Working through the literature on common-currency valuation, the systems that converge on the ventromedial prefrontal cortex and the ventral striatum, alongside the mirror neuron system, are the homeostatic-deficit signalling system and the dopaminergic intrinsic-engagement system. The homeostatic-deficit system is the one that signals when the body is out of balance: the drive to breathe, hunger, thirst, pain, thermoregulatory deficit, and the broader class of signals indicating that the Individual's physiological or basic-security state has departed from a set-point. The dopaminergic intrinsic-engagement system is the one that rewards engagement in activities that are rewarding in the doing rather than for their outcomes: curiosity, mastery, exploration, and the broader class of activities that the Individual continues without external reward. These two systems, together with the mirror neuron system, are the three motivational pathways that converge on the integration locus. No fourth pathway was identifiable in the literature, and parsimony, which holds that no more pathways should be posited than the explanatory work requires, fixed the count at three.

The fifth step is the alignment of the dopaminergic intrinsic-engagement system with Self-Determination Theory's account of Intrinsic Motivation. Self-Determination Theory, developed by Edward Deci and Richard Ryan and elaborated over decades of work, distinguishes intrinsically motivated activity, pursued for its own sake, from extrinsically motivated activity, pursued for external reward. The framework identifies three innate psychological needs, autonomy, competence, and relatedness, whose satisfaction supports intrinsic motivation, and it documents the conditions under which intrinsic motivation flourishes or is undermined. AIM's account of Intrinsic Motivation aligns closely with the Self-Determination Theory account at the level of phenomenology and behaviour, and the foundational citations for the I-component in Section 4.3 acknowledge that alignment. The two frameworks differ in their treatment of motivation outside the intrinsic category, in their account of what happens at the integration locus, and in the consequences they derive at the scale of the institution and the civilisation, but at the level at which Self-Determination Theory has done its most thorough work, AIM treats Self-Determination Theory's findings as direct and substantial support for its own I-claims. AIM does not, however, treat Self-Determination Theory as providing a complete account of the functions of the dopaminergic intrinsic-engagement system.

The sixth step is the derivation of the satiation and termination dynamics. The three convergent systems do not operate in the same way at the timescale of behaviour, and the differences in how they operate are what make them separately identifiable from observable behaviour. The homeostatic-deficit system is the Body Manager of the Minimal Architecture Specification set out at 4.4, the dopaminergic system is its Thinker, and the mirror neuron system is its Observer. The Body Manager operates on a deficit-and-restoration cycle: the deficit is detected, a signal is generated to motivate corrective action, the corrective action is taken, the deficit is corrected, and the signal subsides. Once the Individual has eaten, the hunger signal does not continue to compete for action selection; it is gone until the next deficit arises.

The Thinker registers the absence of whatever the process it is running requires, and a process of any extent requires a series of Objects rather than one. The Individual making a chair requires timber, then a saw, then nails, then sandpaper, each at its own stage and not before, and a deficit is registered at each stage for the Object that stage requires. Action directed at the Object corrects that deficit, and it corrects it by doing the work the stage required, which is what advances the process to the stage registering the next. Carving the wood is not obtaining the wood. Acquisition is the special case in which the state the process requires is possession, which is why the tools an I-process needs satiate on reaching functional adequacy while the greater part of a process closes its stages by doing rather than by getting. Each correction is a hand-off rather than a discharge: the standing deficit, which is for the finished chair, is undiminished by the nailing of the joint, and what the nailing removes is an obstruction to the next stage. The Decision Hub's indifference to how many Objects occupy the intermediary position in series (4.4.1) is what permits a process to run through them.

Not every stage recruits the Thinker. Where a stage's requirement is already resolved and to hand, action proceeds on cached value, and it is the stage that departs from what was predicted, the nails that are not where they were left, that returns the Thinker to the problem.

This structure is not peculiar to the pursuits ordinarily called intrinsic. It is what the Thinker does whenever a process is running, and which component's signal set the process is a separate question from how the process runs. A hungry Individual's Thinker takes the correction of that hunger as its problem and signals the absence of the money, the shop, the rice, the pot, or the heat, in the same form as the furniture maker's Thinker signals the absence of the sandpaper, and an Individual procuring medicine runs a process of the same shape as an Individual learning to cook. Cooking is a craft for one Individual and provisioning for another, and the stages are indistinguishable from outside.

The difference between the sources therefore sits in the generating condition rather than in the process, and it is a difference in what becomes of that condition once the process completes. A homeostatic set-point drifts back, so the Body Manager commissions the same process again, and the recurrence belongs to the set-point and not to the pursuit it commissions. A goal state the Thinker set for itself does not drift: the chair is finished and stays finished, the piece is learned, the problem is worked through, and the standing deficit closes because the state it was for has been reached. That closure is the satiation of an I-process, and it presents as accomplishment rather than as satiety. The next process the Thinker sets need not be that one.

Boredom belongs to neither axis. Boredom reports that the Object under attention holds no unresolved residue for Ground-Resolution to work on, which is a fact about what that Object still affords and not about what the running process requires. The two signals are independent, and every combination of them occurs: a process may be fully supplied and abandoned to boredom, which is the state of the fourth identical chair leg with every tool to hand, and a process may be blocked for want of an Object with engagement wholly undiminished, which is the state of the missing nails. Where the process is itself an inquiry, completion and the exhaustion of residue arrive together, because the state the process was for is a committed posterior and reaching it is what leaves nothing further to resolve. Coincidence there is not identity. The I-process accordingly terminates on completion, on the exhaustion of resolution, or on interruption, and only the second of the three is boredom.

The Observer runs a process of its own, and it is a process of a different kind. Its process is the ongoing identification of Salient Objects in the environment: a continuous conditional Salience assessment over the whole perceived Object field, tracking the movement and the Salience of any Object for the guidance of action, the branch moved by the wind, the vehicle through the corner, the cup on the table. Models are Objects in that field and are read by the same apparatus, their observed orientation tracked for the same action-guiding reason as any other Object's movement, and the Salience installed by observing another Individual's directional valence toward or away from an Object is therefore entailed by a general perceptual function rather than produced by a faculty dedicated to reading agents.

A process of identification over a field has no completion condition. There is always more field to identify over, and identifying is itself what generates the signal, so the process runs for as long as the Individual is conscious and perceiving, and the identifications succeed one another without accumulating toward a terminus. What the process registers is Salience rather than deficit, so no closure is available to it even in principle. It also does not terminate in boredom, because boredom is the exhaustion of the resolution the Thinker was running and the Observer's process does not depend on conscious engagement at all. An Object leaving the perceived field ends that Object's reading and not the process, which ceases only when the Individual stops perceiving

The Observer's process reaches the Thinker's by one route. The Salience contribution enters the Decision Hub as one signed term, is integrated with the other two, and returns to the Thinker as tag-stripped Evidence, and Ground-Resolution sets a process on whatever attribution it then returns. That process has the standing deficit, the series of staged Objects, and the hand-offs of any other. The Individual earns the money, finds the seller, pays and takes delivery, and each stage registers and closes its own deficit exactly as the stages of the chair do. Nothing in the conduct of the process discloses what set it.

What separates the Observer is again what becomes of the setting condition once the Thinker's process completes, and here the condition is being regenerated by a process that has not stopped. Salience tracks an Object as pursued rather than as possessed, so possession ends the movement the reading was tracking and the signal collapses at its source, having closed no deficit because none was ever recorded, the Salience of the pursued Object simply ends while the Observer, still identifying, forms the reading against the next Object. The condition relocates rather than drifting back or staying closed. The Thinker's process therefore completes without discharging any deficit of its own, which is why completion arrives here with at most a brief sense of success and without the durable accomplishment that closure on a Thinker-set goal state carries. Where the reading was comparative, the field the next identification runs over has itself been altered by the acquisition, and the demand re-forms one position further on; that upward recalibration is the downstream expression of the architecture rather than its cause, and the non-satiation is already entailed by the unboundedness of the field the identification runs over.

The processes performed by the Body Manager, Thinker and Observer are three systems running sub-processes of one system working in concert with each other, not three systems that happen to differ. The Body Manager monitors set-points, the Observer identifies Salient Objects, and the Thinker runs the goal-directed processes that any of the three conditions can set, so the stages of a pursuit look the same whichever condition set it and the Decision Hub integrates all three contributions before any of them reaches that pursuit. The separation is available only in what happens after a pursuit completes: whether the demand returns for the same Object on a cycle, holds closed, or re-forms one position further on. That is why the satiation dynamics are read from behaviour over an interval rather than from the pursuit itself.

The names of the three sources. Intrinsic Motivation and Mimetic Desire carry the names of the literatures that characterised them: Self-Determination Theory (Deci and Ryan, 1985) and Girard (1961). Appetites takes its name from a different place. Girard rests Mimetic Desire on a distinction between desire, which is copied from a Model and has no point of satiation, and the appetites and needs grounded in the body, which mimesis can attach to but does not constitute (Girard, 1972, 1978). Hunger and thirst are the plainest members of that second class, and in ordinary speech they are what an appetite for food or drink means. The ordinary word also carries satiation in its default sense: an appetite is the kind of wanting that a meal or a drink brings to an end, and speech marks the exception by adding a word, an insatiable appetite, when a wanting does not end. That default is the property by which the AIM Framework separates the source from Mimetic Desire, which carries no closure condition (Axiom 6). The ordinary word is wider than that, and the physiology of hunger draws the line elsewhere: Cannon and Washburn (1912) separated hunger, the sensation of the empty stomach's contractions, from appetite, the learned anticipation of particular foods. The capital marks the defined term. Appetites in the AIM Framework names the homeostatic deficit signal, which is Cannon's hunger rather than his appetite, and the system that signal belongs to was characterised by the physiology of homeostasis (Cannon, 1932). Which food or drink would repair the deficit is not part of the signal; identifying that is the Thinker's work (Section 4.4.3). None of the three names is coined, because the AIM Framework unifies those literatures and does not stand apart from them. Each describes one of the three convergent systems accurately within its own frame, and none supplies the integration that joins the three or the consequences that follow from it. A framework that renamed what those literatures had already named would present as their rival, when its claim is that each is correct about its own system and incomplete about the motivational system as a whole.

The letters carry an order as well as an abbreviation. A, I and M is the order in which the architecture's normative structure takes the sources up: A is secured first, because an elevated Appetite dominates the Decision Hub and displaces the I-processes (Chapter 6, Section 6.7). An Individual whose deficits remain uncorrected will die. I is enabled on that floor, and M is bounded, because it records no deficit and so carries no closure condition that could limit it (Axiom 6). An elevated Appetite does not displace M, which runs for as long as the Individual is conscious and perceiving (Section 4.4.2). The same order governs the duties to protect A, enable I and bound M (Chapter 17, Section 17.12.5).

A borrowed name brings its literature's frame with it, and in one case that frame is narrower than the component it names. Girard's term names what the Individual experiences, a wanting copied from a Model. The component the AIM Framework places under that name is the Observer, whose output is Salience: a reading of which Objects in the field are marked as valued, formed over the whole perceived field and installed by Models as its social special case. Girard's account is an account of that special case, which is also where the escalating pathologies of the source live (Section 4.4.7); the general case, Salience formed with no Model present, has no counterpart in it. Mimetic Desire in Girard's sense is what that reading becomes once it is integrated, tag-stripped and resolved by the Thinker into a ground (Section 4.4.7). The source keeps Girard's name because his is the literature that identified it; the specification of the Observer governs what the source does. The same holds for all three names: where a source's name and its component's function would be read differently, the function governs.

The seventh step is the derivation of Source Opacity from the convergence-and-integration structure. Three systems feed signals into a single integration locus that produces a single Scalar for action selection. The integration is a many-to-one map: three inputs, one from each system, are summed into one output, and every value of that output is produced by many combinations of the three. Once the summation has occurred, the source of any given contribution to the Scalar is no longer represented in the Scalar. The intuition is captured by an analogy of pipes: if two pipes join into one, and a quantity of water is taken from the joined pipe, the quantity cannot be traced back to which of the two source pipes it came from, because the joining operation did not preserve that information. The same is true of the integration at the ventromedial prefrontal cortex and the ventral striatum: the source-tag identifying which of the three systems contributed what is not preserved in the integrated Scalar. Downpipe processes, including the conscious-cognitive system that the Individual identifies with their own deliberation and self-knowledge, operate on the integrated Scalar and therefore have no access to the source decomposition. This is Source Opacity. Source Opacity is not a contingent failure of self-awareness; it is a structural consequence of the architecture. The formal proof of Source Opacity, via the Data Processing Inequality applied to the integration mapping, is presented in Chapter 9. The derivation in this chapter is the verbal form of that proof.

The eighth step is the derivation of the phenomenological diagnostics from how each system's signal presents in conscious experience. Source Opacity destroys the source-tag at the integration step. But the three systems differ in how their signals present in conscious experience upstream of integration, and those differences leave behavioural traces from which source can be inferred even though they cannot be introspected by disaggregating the Scalar itself. Appetites persist into conscious awareness as long as the deficit is uncorrected: the Individual feels hungry, feels the pain, feels the cold, and the feeling does not subside until the cause of the signal is addressed. The phenomenology is available because the underlying signal is sustained over a timescale long enough for conscious cognition to register it. Intrinsic Motivation is the Thinker's own operation presenting in conscious experience: the Individual is aware of the activity because the Thinker's engagement is what the activity consists of. Thinking about a problem one finds interesting, playing music one is mastering, working on a piece of writing one is invested in, are activities the Individual notices because the noticing is the activity. Mimetic Desire is different from both. The mirror neuron system operates on a timescale, characterised in the literature as roughly sixty to three hundred and forty milliseconds for the relevant integration steps, that is faster than conscious awareness can register. By the time the Individual is consciously aware of an Object, the mirror neuron system's contribution to that Object's valuation has already been delivered to the integration locus, and the system is already conditioning on the next Object in the Individual's perceptual field. The acquisition event leaves no phenomenological residue, and in the human substrate conscious awareness was not yet present when the event occurred. This is what makes M structurally invisible to introspection in a way A and I are not: M is not merely tag-stripped at integration, it leaves no residue, and in the human substrate it is also preconscious at acquisition.

The ninth step is the derivation of the remaining axiomatic structure. The four-component architecture (the three convergent systems plus the integration locus that becomes the Decision Hub in the Minimal Architecture Specification) and the three satiation-and-termination dynamics together entail the rest of the axiomatic structure. Confabulation, the generation of plausible but inaccurate narratives about the causes of choice, follows from Source Opacity combined with the Individual's need to give an account of their own behaviour. Object Parasitism, the operation of M by inflating the perceived value of A-Objects and I-Objects rather than by generating a separate Object-class, follows from the mirror neuron system's mode of operation: it assesses Salience of Objects in the perceptual field, and the Objects in the perceptual field are largely the same Objects that A and I systems are valuing. Bayesian Belief dynamics follow from the operation of the conscious-cognitive system on the evidence available to it, an evidence base that is itself partly produced by the architecture rather than independent of it. The agent-based causal default is grounded in what the brain developed to do: the mirror neuron system's tuning to agents (entities perceived as Models) rather than to abstract systemic causes. Each of these entailments was derived from the architectural and dynamic foundations rather than posited as an independent claim, and each is stated as an axiom in Section 4.3 because the formal compression of an entailment is itself a starting point for further derivation.

The five constraints that operate as quality controls on this derivation, mentioned in Section 4.1 and recurring throughout this treatise, did their work at specific points in the nine steps above. Parsimony did its work at step four, fixing the count of convergent systems at three by requiring that no fourth system be posited unless the explanatory work demanded it; and again throughout step nine, requiring that each entailment be derived rather than independently claimed. Falsifiability did its work at step six, requiring that the satiation and termination dynamics be specified in language sufficiently precise that contrary findings could be identified as contrary; and again at step seven, requiring that Source Opacity be specified as a structural consequence of integration rather than as a vague claim about introspective limits.

Neural-architectural compatibility did its work at every step, requiring that no axiom contradict established findings on common-currency valuation, dimensionality reduction, mirror neuron function, dopaminergic reward processing, or homeostatic regulation. Dissociability did its work at step eight, requiring that the three sources be distinguishable from observable behaviour and not merely from theoretical stipulation. Behavioural diagnosability did its work at the boundary between the foundational chapter and the disaggregation chapter that follows it, requiring that the architecture support practical inference about which source is dominant in a given Individual's behaviour at a given moment, since a theory of motivation that cannot be applied to behaviour is a theory that does no useful work. The constraints are not the source of the derivation. They are the discipline within which the derivation was conducted.

The conclusions arrived at along the path above were not anticipated at the outset. Many of them, including the conclusion that introspection is structurally unreliable for source identification, the conclusion that Mimetic Desire is preconsciously acquired and therefore inaccessible to self-correction by conscious deliberation alone, and the conclusions that follow at institutional and civilisational scale in later chapters, were as confronting to the author as they will be to the reader. The discipline of following the entailments wherever they led, rather than redirecting the derivation when it pointed toward uncomfortable conclusions in tension with prior intuition, is the discipline this treatise asks its readers to share. The Preface and Chapter 1 set out the broader frame within which that discipline is offered, and Section 4.6 of this chapter returns to it at the close of the foundational material. The axioms set out in Section 4.3 are the formal compression of where the derivation arrived, not the formal expression of where it began. The reader who reads them in that light is reading them as they were intended.

4.3 The Nine Core Axioms

AIM rests on nine axioms that together constitute the framework's irreducible premises. These axioms are not arbitrary stipulations but empirically grounded claims about human motivational architecture. They represent the minimal commitments required to generate AIM's explanatory and predictive power.

The axioms are stated below in canonical form with their foundational citations. No more than three canonical sources are cited for each axiom. This discipline is necessary because all deductions made throughout this book flow directly from these axioms. Other empirical evidence may be cited to support how these axioms play out; however, in each case the axiom is what is relied on, not the illustrative example. Further empirical validation is explicitly outside the scope of this text (no new empirical evidence is offered); however, further development, peer review, falsification, or refinement are all welcomed.

4.3.1 Axiom 1: Three-Source Taxonomy

Canonical Statement: Human motivation arises from exactly three neurologically distinct sources (Appetites, Intrinsic Motivation, and Mimetic Desire) and no additional source is required for a parsimonious, falsifiable account of human motivation.

The Three Sources Defined:

Appetites (A): Homeostatic drives that report the state of the body (oxygen and the drive to breathe, hunger, thirst, sleep, pain, thermoregulation), pushing behaviour toward returning the body to its set-points and subsiding rapidly once it has returned.

These are the motivations that come from the body being out of balance: needing air, food, water, rest, warmth, or bodily safety. They feel urgent while active, tightly drive behaviour toward relief, and fall away quickly once the need is properly met. This is why securing Appetites is a precondition for Intrinsic Motivation to show up reliably: when A-signals are elevated, they dominate the Decision Hub. Thinking does not stop, because working out what would fix the deficit is thinking, but the pursuits that are rewarding in the doing are pushed aside, and with them long-term, systemic and creative thought. Mimetic Desire is not pushed aside: the part of the mind that notices what stands out in the surroundings keeps running for as long as one is awake and perceiving, and under deficit what stands out first is whatever within reach would fix it, which is usually food, water or shelter rather than another person.

The A-category is bounded by the body. It comprises the biological drives in the narrow sense (oxygen and the drive to breathe, hunger, thirst, thermoregulation, sleep pressure) and the body's own condition: present exposure, present tissue damage, present pain. Physical safety enters A only as that condition. The classification of an Object in the field as a future threat is not an A-signal; it is Observer Salience conditioned by the Thinker's threat-prior (Section 4.4.2). A circumstance in the world, such as a lack of money or of shelter, does not enter A as a circumstance either. It reaches A only through what it does to the body: sustained stress, whether set off by the circumstance itself or by the Thinker's appraisal of it, raises the stress hormones, tightens the chest and shifts the body's set-points, the condition the physiology of stress calls allostasis and, when prolonged, allostatic load (Sterling and Eyer, 1988; McEwen, 1998). The Body Manager reports that bodily state as it reports any other deficit, and the Thinker, which registers the body's condition directly (Section 4.4.2), has it too. What the Body Manager reports is the state of the body, never the circumstance that produced it: knowing that the money has run out belongs to the Thinker, as does identifying the Objects that would repair the deficit (Section 4.4.3). Shelter, income, healthcare and safety are the means by which A-deficits are prevented and repaired, which is why securing them secures A, but they are not what A signals.

Intrinsic Motivation (I): Self-endorsed engagement in activities that are rewarding in the doing (curiosity, mastery, play, exploration, care, aesthetic and creative work) that Individuals continue in private without audience or external reward, typically strengthening as competence and meaning deepen.

The critical diagnostic feature of Intrinsic Motivation is Private Persistence (the continued pursuit of an activity without audience, external reward, or social reinforcement). If an activity continues when no one is watching and no external reward is contingent on performance, across an interval long enough for an established Mimetic valuation to release (Section 4.4.13.2), the motivation is not Mimetic Desire in its special case, and where the engagement also deepens with competence it is I-driven. This distinguishes Intrinsic Motivation from Mimetic Desire in its special case, which loses its input without audience and falls away once the loop holding it releases.

Importantly, the desire to be with other people is an Intrinsic Motivation (for example, loneliness, the desire to perform to an audience as expression rather than status-seeking), as these desires persist in private and drive the Individual to find ways to connect with others. Not all social motivation is mimetic; genuine connection and care are I-driven (including empathy).

A critical distinction governs I-satiation dynamics:

Intrinsic engagement itself (mastery, curiosity, flow) is non-satiating and deepens relative to the same mode, instrument, or tool. A guitarist pursuing mastery does not reach a point where mastery is "full" and further engagement becomes unwanted. Competence generates new frontiers; curiosity answered generates new questions; flow deepens as challenges increase (Csikszentmihalyi, 1990).
I-enabling Objects (tools, instruments, materials, access to locations for the process) do satiate. The guitarist wants a guitar. Once she possesses a guitar adequate for her current level and anticipated development, the Object-desire terminates. She does not want additional guitars beyond functional adequacy.

Further Object-acquisition beyond what the process requires signals likely M-contamination rather than genuine I-need. The diagnostic is what the upgrade does once acquired: if the new Object goes into the process and the process changes with it, the current tool was limiting the process and the upgrade met a genuine I-functional need; if the new Object is used while it is new and the process then continues on the old one, or does not continue at all, the upgrade was M-driven (the Fallow Signature, Chapter 7, Section 7.5.2.1).

Mimetic Desire (M): The Salience channel: in the general case M is a continuous, signed reading of which Objects in the perceived field are marked as valued and where each sits, conditioned on priors the Thinker maintains. In its distinctive special case, M's input is the observed pursuits and valuations of other Individuals, so that what Models (Individuals whose behaviour, preferences, or status serve as reference points for another Individual's valuation of Objects) pursue, display or signal value for is read as Salient, strongly shaped by observability and prestige cues, and prone to rivalry when several Individuals' readings converge on the same scarce Object. M encodes no deficit; the wanting the Individual experiences is what integration makes of the reading (Axiom 3). Section 4.2 sets out why the source carries Girard's name for this system.

In the most significant special case Mimetic Desire is characterised as the wanting that comes from watching others want. When Individual A observes Model B pursuing Object X with apparent enthusiasm, A's valuation of X increases, not because A has learned new information about X's properties, but because of B's pursuit of it (Girard, 1961).

What passes from B to A is not a desire but a reading: B's pursuit marks X as Salient in A's field. A comes to want X with the urgency of a need because integration renders that reading as a deficit (the Salience-Deficit Conflation Principle, Section 4.4.0), and A experiences the wanting as self-generated because the reading's source is not preserved (Axiom 3).

Mimetic Desire is strongly shaped by:

Observability. M responds to what can be seen or otherwise observed. Desires for Objects that cannot be displayed or compared are less subject to M-amplification; desires for highly visible Objects are more subject.
Prestige cues. M is amplified when the Model is high-status. Wanting what high-status others want is more intense than wanting what low-status others want (Zink, Tong, Chen, Bassett, Stein, and Meyer-Lindenberg, 2008).
Scarcity. M is amplified when the Object is scarce or contested. If many Individuals want the same limited Object, M intensifies through rivalry dynamics.

M is prone to rivalry because mimetic transmission causes multiple Individuals to converge on the same Objects. When A and B both want X because they have copied each other's desires (directly or through a shared Model C), they become rivals. The rivalry can escalate through positive feedback: A's intensified pursuit of X signals to B that X is even more valuable, intensifying B's desire, which signals back to A, and so on. This escalation dynamic is central to AIM's account of social pathology and is developed fully in Chapter 8 (Mimetic Desire).

Exhaustiveness and Mutual Exclusivity:

The taxonomy is exhaustive: every motivational signal entering the Decision Hub originates from one of these three sources. There is no fourth source required to account for any human motivation. This claim is defended in Sections 5.5 and 5.6.

The taxonomy is mutually exclusive: each motivational signal originates from exactly one source. However, any given Object can receive motivational weight from multiple sources simultaneously. A meal can address hunger (A), provide gustatory pleasure in the eating (I), and signal status through its expense or exclusivity (M). The sources are distinct; the Objects they target may overlap.

Foundational Citations:

For Appetites: Cannon, W.B. (1932). The Wisdom of the Body. W.W. Norton & Co. Cannon's work establishes homeostasis as the fundamental principle governing physiological drives. The body maintains set-points for temperature, blood chemistry, hydration, and nutrition; deviations from these set-points generate motivational signals that drive behaviour toward restoration. This provides the conceptual foundation for A as deficit-driven motivation with natural satiation upon restoration.
For Intrinsic Motivation: Deci, E.L. & Ryan, R.M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. Plenum. Deci and Ryan's Self-Determination Theory establishes that humans have innate psychological needs for autonomy, competence, and relatedness that generate motivation independent of external reward or homeostatic deficit. Activities satisfying these needs are pursued for their own sake and are undermined by contingent external incentives (the Overjustification Effect). This provides the conceptual foundation for I as process-rewarding motivation distinct from both A and M.
For Mimetic Desire: Girard, R. (1961). Deceit, Desire, and the Novel: Self and Other in Literary Structure. Johns Hopkins University Press. Girard's literary analysis establishes the triangular structure of desire: the subject desires the Object because the mediator (Model) desires it. Desire is not a straight line from subject to Object but a triangle with the Model at the apex. This provides the conceptual foundation for M's special case, the wanting transmitted through observed Models, and for the escalating pathologies that live there; the general case, Salience over the whole perceived field with no Model present, extends Girard's account (Section 4.4.7).

4.3.2 Axiom 2: Common-Currency Integration

Canonical Statement: The three motivational sources converge in a Common-Currency valuation system centred in the ventromedial prefrontal cortex (vmPFC) and ventral striatum that encodes a single scalar subjective value signal for each Object, enabling unified choice by converting unlike inputs into a comparable metric. That integration locus is the ventromedial prefrontal cortex and the ventral striatum / nucleus accumbens.

Explanation:

Although A, I, and M are distinct systems with different properties, the brain must eventually produce a single "do this next" decision. The Decision Hub integrates all three sources into one priority signal, like mixing three audio channels into a single speaker output.

This integration can be approximated as a weighted sum:

$U (x) = w_{A} \cdot U_{A} (x) + w_{I} \cdot U_{I} (x) + w_{M} \cdot U_{M} (x)$

where $w_{A} + w_{I} + w_{M} = 1$ . The weight $w_{M}$ is labelled $β$ in Appendix A (Section A.2, restatement M3); the two notations refer to the same parameter.

This is the AIM Integration Equation. The weights represent the proportional influence of each source on the current decision. These weights are not fixed traits but state-dependent variables that shift with context: hunger elevates the Appetites weight; a more salient Object field elevates the Mimetic Desire weight, which is non-decreasing in field salience (Appendix A, Section A.9.3); and any rise in one weight is taken from the others, because the three sum to one. The weights must be inferred from behaviour and context rather than read directly from self-reports (due to Source Opacity and Confabulation).

The Common-Currency architecture explains why:

Trade-offs are possible. An Individual can sacrifice A-satisfaction for M-gains (dieting to look attractive) or M-satisfaction for I-engagement (abandoning a prestigious career for meaningful work) because all three sources feed the same valuation system and can be compared.
Source Opacity is inevitable. The integration is a many-to-one map: three inputs (A, I, M) are summed into one output (scalar value). Source tags cannot be preserved in the integrated signal because the integration maps many source combinations onto each value of the output.
Prediction requires disaggregation. Although the integrated signal is one-dimensional, the satisfaction conditions of the three sources differ (Axiom 6, Differential Satiation Dynamics). Predicting whether satisfying a desire will produce durable welfare requires knowing which source led it, which is not readable from the integrated signal and must be inferred from the Individual's resulting behaviour (Axiom 6).

Supporting Evidence: The Wanting-Liking Dissociation

Additional support for the Common-Currency architecture comes from Kent Berridge's research on the Wanting-Liking Dissociation (the neurobiological separation between "wanting," or incentive Salience mediated by dopaminergic circuits, and "liking," or hedonic pleasure mediated by opioid circuits) (Berridge, 2004). "Wanting" and "liking" are neurobiologically distinct systems that can be independently manipulated.

M amplifies wanting without proportionately amplifying liking. The Individual pursuing an Object for relative position experiences intense wanting, the urgency of the pursuit and the willingness to expend effort, but often experiences muted liking. The pleasure when the Object is reached is disappointing, and the victory feels hollow. This explains the chronic dissatisfaction characteristic of M-dominance: wanting intensifies while liking remains flat or declines.

The Wanting-Liking Dissociation also explains M's lack of a natural satiation signal, whose ground is the Salience-Deficit Asymmetry of Axiom 6. Liking is, in one degree or another, the phenomenology of satisfaction, and each source's liking follows the deficits that source closes. In the bodily case the dependence is measured: the same stimulus is pleasant to a body short of its set-point and unpleasant to one already restored, the alliesthesia that ties bodily pleasure to the correction of a deficit (Cabanac, 1971), so the liking of Appetites accompanies a deficit as it closes and fades with it, as with eating. Intrinsic Motivation closes deficits at two grains, and the literature of intrinsic motivation already separates the two likings that result. Each stage of a process registers and closes its own deficit as the work proceeds (Section 4.2), and the liking that accompanies those closures, one after another, is the joy in the doing: the "inherent satisfactions" by which Self-Determination Theory defines intrinsic motivation (Ryan and Deci, 2000), and the activity enjoyment that flow intensifies (Csikszentmihalyi, 1990). The standing deficit of the goal state closes when the state is reached, and the liking that accompanies that closure is accomplishment, the joy and pride that the study of achievement emotions places on the outcome side of the same division (Pekrun, 2006). In I, wanting and liking therefore track together, because every stage of the pursuit closes something.

M records no deficit, so there is nothing for it to satiate against, and a reading that closes no deficit has no liking of its own to deliver; whatever pleasure an M-led acquisition brings is the A- or I-liking of the Object the reading fell on (Section 4.4.11). An M-led success still brings the joy and pride of an outcome, the promotion achieved or the contest won (Pekrun, 2006), but that liking is fleeting and does not sustain the Individual. What closes is a pursuit that cannot be continued: possession ends the reading that set it (Section 4.2), and the Object, once obtained, has lost the Salience that made it wanted. Wanting is what the reading feeds once integration renders it as lack, so in M the two come apart: when wanting escalates on a reading that closes nothing, no satiation signal is generated, and the Individual continues wanting more even though acquiring more produces no additional pleasure.

Foundational Citations:

Levy, D.J. & Glimcher, P.W. (2012). The root of all value: a neural common currency for choice. Current Opinion in Neurobiology, 22(6), 1027-1038. The definitive meta-analysis confirming that the vmPFC and ventral striatum integrate diverse reward types (monetary, primary, social, aesthetic) into a common value signal that predicts choice.
Chib, V.S., Rangel, A., Shimojo, S. & O'Doherty, J.P. (2009). Evidence for a common representation of decision values for dissimilar goods in human ventromedial prefrontal cortex. Journal of Neuroscience, 29(39), 12315-12320. Demonstrates that the vmPFC encodes the value of qualitatively different goods (food, money, consumer items) on a common scale, enabling direct comparison and trade-off.
Bartra, O., McGuire, J.T. & Kable, J.W. (2013). The valuation system: A coordinate-based meta-analysis of BOLD fMRI experiments examining neural correlates of subjective value. NeuroImage, 76, 412-427. A comprehensive meta-analysis localising the valuation system and confirming that subjective value across domains is represented in overlapping neural regions.

4.3.3 Axiom 3: Source Opacity

Canonical Statement: After Common-Currency Integration, the source decomposition is not preserved in the output signal. Downstream circuits, including those supporting conscious introspection, cannot reliably identify which source generated a given motivational input from the integrated signal alone.

The loss is not confined to identity. The three inputs are not of one kind, and integration changes what one of them reports as well as erasing which one reported it; what each component encodes, and what integration makes of the one that encodes no deficit, is specified with the components at Section 4.4.0.

Explanation:

Once the brain has collapsed A, I, and M into a single "I want this" feeling, it has lost the tag indicating which channel contributed what. This is not a failure of self-awareness but an architectural fact: three inputs (A, I, M) have been summed into one output (scalar value) by a many-to-one map, necessarily discarding the source decomposition.

The loss is structural rather than incidental. There is no separate "source memory" system that preserves the breakdown; the information is simply not represented in the integrated signal. Downstream processes, including conscious awareness, verbal report, and deliberative reasoning, operate on the integrated signal and therefore cannot recover the source decomposition from it.

Source Opacity has four implications:

A Salience reading is received as a demand for correction. Integration confers the felt character of a correction on the Mimetic contribution, a lack where the reading is positive and a threat where it is negative, which is the Salience-Deficit Conflation Principle derived with the components at Section 4.4.0, and it is the principal cost of Source Opacity: not that the Individual cannot say where a want came from, but that a reading of where value sits arrives as a lack, and a reading of what is to be kept from arrives as a threat.
Introspection is unreliable for source identification. When asked "why do you want this?", the Individual cannot directly access the answer because the source tags have been discarded. They can only generate post-hoc explanations based on plausible-sounding narratives (Confabulation).
Wisdom traditions' prohibitions fail. The prohibition "thou shalt not covet" assumes the Individual can identify which desires are covetous (M-driven) and which are legitimate (A-driven or I-driven). Source Opacity means precisely that this identification is architecturally impossible through introspection. The wanting arrives already unlabelled, and it arrives in the form of a need: the prohibition asks the Individual to renounce what presents to them as a lack requiring correction, which is a categorically harder demand than renouncing a preference, and it is the demand the architecture guarantees will be experienced.
Behavioural diagnostics are required. Since introspection cannot reliably identify source, Behavioural Signatures (the observable patterns in an Individual's behaviour across contexts that carry information about which motivational source is dominant) must be used instead: Does the desire persist in private? (I) Does it satiate when the obtained Object is applied to, or consumed by, the body? (A) Does it track observability and prestige cues? (M) These diagnostics operate on observable behaviour rather than subjective report.

Scope of the claim. The Axiom is a narrow claim about one signal: the integrated motivational value on which the Decision Hub selects an action, and what can be recovered from that value. It is not a claim about neural representation elsewhere in the system. The brain represents a great deal of other information about the Object being valued, what kind of thing it is, where it is, what it looks like, and that information reaches the Thinker by the ordinary channels of perception and inference, untouched by integration (Section 4.4.3). What cannot be recovered is how much each source contributed to the value that moved the Individual to act. Nor does the Axiom deny structure among the motivational inputs before they are integrated. The valuation system keeps some: alongside a value signal that is the same whatever kind of good is being valued, distinct regions register the value of particular kinds of good, such as food, money or trinkets (McNamee, Rangel, & O'Doherty, 2013). Those categories are kinds of Object, not sources of motivation. And once three quantities have been added together, the total cannot be taken apart again. Many different combinations of the three give the same total, so the total alone cannot show which combination produced it, and that holds however the three were organised before they were added. The Individual acts on, and reads, the sum.

Source Opacity identifies the constraint surface, the precise locus where self-knowledge breaks down. The constraint is permanent. The integration is a many-to-one map: every value of the Scalar is produced by a whole family of source combinations, so no function of the Scalar can recover which combination produced it, and that loss is strict. The Data Processing Inequality then guarantees that no processing downstream of the Scalar recovers what the integration discarded. The one architectural premise is that introspection is such downstream processing, reading the Scalar and not the channels that fed it (Scalar-Only Delivery). The Source Opacity Theorem is proved formally in Chapter 9 (Neural Integration and Source Opacity) and stated at Appendix A, Section A.2 (restatement M5a). Given that premise, it follows from the mathematics of many-to-one compression and not from any further commitment of the AIM Framework.

But the integrated Scalar is not the only information channel available. An Individual's behavioural output across contexts (public versus private, sated versus hungry, with audience versus alone) carries superimposed signatures of each source: state-dependence and satiation for A, Private Persistence and deepening for I, non-satiation for M, and audience-dependence for M in its special case. These signatures are not read off any single value of the Scalar. They are the course the Scalar, and the behaviour it drives, takes across contexts, which no instant of it contains. The Two-Question Diagnostic (introduced fully in Chapter 18, The Knowledge Patch) operates on this second channel. It functions as a prism, disaggregating superimposed Behavioural Signatures into distinguishable bands, not by reversing Source Opacity, but by routing around Source Opacity entirely, reading the Scalar's course across contexts and the behaviour it drives rather than interrogating any single value of the integrated signal. Source tags are permanently destroyed for introspection. The observable Behavioural Signatures remain available for inferential reconstruction, post hoc. That reconstructive, post-integration inference is Source-Tagging: the Individual re-attributes a source to an integrated wanting by applying conceptual instruments, contextual cues, and I-Override, because Common-Currency Integration discarded the native source-labels.

I-Override Information Architecture:

Source Opacity destroys the source decomposition of the integrated signal: which channels (A, I, or M) contributed, and in what proportions. All other information processed prior to integration (contextual cues, temporal sequence, behavioural history, environmental conditions, bodily states, and the Content of the desire itself) remains accessible to conscious awareness.

This means that while the Individual cannot directly perceive "this desire is M-driven," they can access the contextual information that would allow inference: "I started wanting this right after seeing my colleague acquire one"; "this desire intensifies when I imagine others seeing me with it"; "this desire evaporates when I imagine having it in complete privacy." The AIM Framework provides the conceptual vocabulary to make such inferences explicit, enabling what is termed I-Override: the Thinker's inference of the source-attribution that Source Opacity leaves unreadable, which routes around Source Opacity rather than restoring the source-tags. Because I-Override is an operation of the Thinker, it is available only once the Thinker has been recruited. Until then, action-selection proceeds automatically, on the cached values of habit and on the priors the Individual already holds (Section 4.4.3), and nothing is in place to perform an override. The Thinker is recruited by an open question, a novelty or a prediction-mismatch (Section 4.4.3), so I-Override can occur only at those points. Where the recruitment comes before the action is selected, the inference can redirect the action; where it comes after, the inference can only revise the Beliefs on which the next selection runs.

Why the Mimetic contribution in particular cannot be recovered by this route is derived at Section 4.4.12, where three conditions that hold together (the Legibility Gradient, the Salience character of the signal, and the destruction of its tag) are shown to entail non-satiation, Confabulation and the escalation dynamics that follow from them.

Foundational Citations:

Nisbett, R.E. & Wilson, T.D. (1977). Telling more than we know: Verbal reports on mental processes. Psychological Review, 84(3), 231-259. The landmark paper establishing that humans lack introspective access to the cognitive processes driving their choices. Subjects confidently reported reasons for their choices that could not have been the actual causes (experimentally controlled factors that subjects were unaware of).
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. Demonstrates Choice Blindness (the failure to detect when the outcome of a choice has been surreptitiously swapped). Subjects who chose photograph A, but were handed photograph B, often failed to notice the switch and proceeded to confabulate reasons for preferring B.

4.3.4 Axiom 4: Confabulation

Canonical Statement: The brain routinely generates sincere but inaccurate narratives about the causes of choice, filling gaps in introspection with plausible reasons that do not match the actual motivational sources. Mimetic Desire is especially vulnerable to misattribution because its signal is a Salience reading received under the category of deficit, and because it leaves no residue against which that reception could be checked.

Explanation:

When asked to explain their choices, Individuals construct stories that sound reasonable but are often wrong. These are not lies or in any way dishonest; the Individual genuinely believes their explanation. They are post-hoc rationalisations created after the decision has already been made.

Confabulation is not an occasional error but a structural feature of how human cognition handles the gap between experienced wanting and absent source-information. The pattern is consistent:

The Individual experiences wanting (phenomenologically vivid).
The Individual lacks access to the wanting's origin (neurally obscured by Source Opacity).
The conscious mind generates a plausible narrative ("I want this because...").
The narrative draws on available cultural materials (authentic choice, genuine preference, personal values, rational assessment).
The narrative feels subjectively true, indistinguishable from genuine recollection.

Confabulation intensifies under scrutiny. When the narrative is challenged, the Individual does not reconsider the source of the desire; instead, they generate increasingly elaborate defences of the confabulated narrative (Haidt, 2001). The challenge activates motivated reasoning, which produces deeper entrenchment rather than correction. This is not stubbornness but Bayesian updating on the Self-Model (the Individual's internal representation of their own motivational architecture): each time the Individual articulates the confabulated explanation, it becomes evidence for the Self-Model, increasing the posterior probability that the explanation is correct. Challenge triggers additional articulation, generating additional Self-Model evidence, further increasing commitment.

Confabulation also co-opts the introspective search the Individual substitutes for I-Override. The Individual who suspects their desire might not be authentic will introspect, seeking the "real" reason. But introspection operates on the integrated signal (from which source tags have been discarded), so the "real" reason it discovers is simply another confabulated narrative, experienced as a deeper Truth. The attempt to detect Mimetic Desire through self-examination becomes the occasion for more sophisticated Confabulation.

The Romantic Lie:

The Romantic Lie (Girard's term for the culturally transmitted Belief that desires originate autonomously within the Individual, that authentic selfhood is the wellspring of wanting) is not a modern Western phenomenon but an architectural constant (Girard, 1961). Every culture generates confabulated narratives about desires' origins. The specific Content varies: Western Romanticism emphasises authentic individual choice; honour cultures emphasise righteous defence of dignity; ascetic traditions emphasise disciplined transcendence. But the structure is constant: M is misattributed to other sources because no accurate attribution is introspectively available. The Romantic Lie is what Confabulation looks like at civilisational scale: the collective version of the narrative that each Individual generates to fill the gap left by Source Opacity.

Foundational Citations:

Gazzaniga, M.S. (2000). Cerebral specialization and interhemispheric communication: Does the corpus callosum enable the human condition? Brain, 123(7), 1293-1326. Gazzaniga's split-brain research reveals the left-hemisphere Interpreter, a neural system that generates confident, coherent, and entirely incorrect explanations for actions initiated by the right hemisphere. In intact brains, the same Interpreter system operates on Source Opacity: when the actual motivational source is inaccessible, the Interpreter generates a plausible narrative and the Individual believes it.
Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review, 108(4), 814-834. Establishes that moral judgments are produced by rapid intuitive processes, with conscious reasoning serving primarily as post-hoc justification rather than causal driver. When reasons are experimentally undermined, the judgment persists ("moral dumbfounding"), confirming that the stated reasons are confabulated rather than causal.
Wegner, D.M. (2002). The Illusion of Conscious Will. MIT Press. Demonstrates that the experience of consciously willing an action is constructed after the fact rather than causing the action. The sense of agency is itself a confabulated narrative, generated by the same Interpreter system that generates confabulated motivational explanations.

4.3.5 Axiom 5: M as Amplifier (Object Parasitism)

Canonical Statement: Mimetic Desire does not generate its own independent Object-category. Its Salience reading falls on Objects that already carry A- or I-functional roles, and integration renders that reading as additional value attached to them, a Mimetic Premium experienced as intrinsic to the Object rather than as socially added.

Explanation:

Mimetic Desire is a parasite, not a creator. Mimetic Desire does not make Individuals want things that have no A-function or I-function; it makes Individuals want more of things that already serve A or I purposes, and to pay more for them than their A-functional or I-adequate value warrants.

Consider a house. Its A-function is shelter: protection from the elements, physical safety, a base for daily activities. Its I-function might include satisfaction in maintaining a garden, comfort in a space that supports creative work, or the work of studying and adapting its design. Its M-function is the Salience it carries: the pleasure its design gives as perceived (Chapter 8, Section 8.1.9), and the status it signals within a reference group through its size, location, architectural prestige, and price. Object Parasitism means the M-component does not create a separate "status house" Object; it inflates the valuation of the same house that also provides shelter and intrinsic engagement.

This parasitic structure is what makes Mimetic Desire so difficult to detect. When the Individual pursues a larger house, they can point to genuine A-considerations (more space for family members, neighbourhood safety) and genuine I-considerations (garage for craft, large garden for gardening). The M-component (that the house is larger than the neighbour's, that it signals position within the reference group) need never be consciously acknowledged because the A and I justifications are genuinely present. This is not hypocrisy. The Individual is not lying when they cite A and I reasons. Those reasons are real. The M-inflation is additional: it operates by amplifying the motivational weight attached to A-Objects and I-Objects without generating any distinctive phenomenology.

The Mimetic Premium (PM, the portion of an Object's perceived value attributable to M-amplification rather than to A-function or I-adequacy) represents the socially transmitted valuation component. PM is not a price set by a single Model; it is the internalised network average of confabulated valuations across the Individual's reference group. Because each Individual in the network confabulates their own M-valuation as authentic (Source Opacity, Confabulation), the network average is itself a confabulated quantity, a collective fiction experienced as objective market value.

The Mimetic Premium has a counterpart on the side of the Individual who holds the Object. Individuals are themselves Objects in one another's fields, valued by the same architecture that values a house, and the aggregate valuation that other Individuals direct at a given Individual, averaged over time, is that Individual's Status. The Mimetic Premium and Status are one Salience contribution seen from its two ends: the Mimetic Premium is the component of an observer's valuation of an Object attributable to the Salience the Object carries, and the aggregate of those premiums across the observers who value the Individual holding it is the Status the Object confers (the Status-Premium Duality, derived at Chapter 8, Section 8.12). The house's M-function is this relation. Its size, location and price draw the reference group's valuations toward the Individual who holds it, and the premium paid for the house is paid for that draw.

A critical consequence: Status cannot be purchased, although the Mimetic Premium is paid as though it could. Three features establish this. First, Status is constituted within other Individuals' valuation architectures, and nothing an Individual spends writes into them. An Individual can act to become Salient but cannot set the valuation another computes (the Salience-Emission Asymmetry, Section 4.4.12), so the premium buys the Object and not the valuation the Object was bought to draw.

Second, Status gained through the Mimetic Premium is zero-sum across the reference group. It is a relative position, so an expenditure that raises one Individual's Status lowers the Status of the other members of the reference group, each of whom now stands lower relative to the Individual who gained. The loss is spread across the group rather than borne by any one member, and it falls most sharply on those whom the Individual's holding of the Object excludes from it. The field pays for a redistribution of rank and gains none in total (the Zero-Sum Status Competition, derived at Chapter 8, Section 8.12.4). Third, the Mimetic Premium an Individual pays reflects their own M-valuation, not an objective property of the Object; the same Object carries a different Mimetic Premium for different Individuals depending on their reference groups and Model structures, so the premium measures the payer's valuation of the signal and not the signal's effect on anyone else.

Object Parasitism generates a double parasitism: M parasitises A-Objects and I-Objects at the motivational level (inflating their perceived value), and the Mimetic Premium parasitises the A-Functional Price and I-Adequate Price at the economic level (inflating what Individuals are willing to pay). This produces Price Decomposition, the analytical separation of any Object's total price into its three components: the A-Functional Price (what the Object would cost if purchased solely for its ability to correct deficits, or maintain the baseline), the I-Adequate Price (the additional cost of quality sufficient to support genuine process-engagement), and the Mimetic Premium (the residual price attributable to status signalling, positional competition, and socially transmitted valuation).

Foundational Citations:

Lebreton, M., Kawa, S., Forgeot d'Arc, B., Daunizeau, J. & Pessiglione, M. (2012). Your goal is mine: Unraveling mimetic desires in the human brain. Journal of Neuroscience, 32(21), 7146-7157. Demonstrates that Objects are rated more desirable once perceived as the goal of another agent's action, that observing the action engages the mirror-neuron system while desirability is represented in the brain valuation system (ventral striatum and ventromedial prefrontal cortex), and that mimetic desire is induced by the first modulating the second, with the strength of that coupling predicting individual susceptibility.
Zink, C.F., Tong, Y., Chen, Q., Bassett, D.S., Stein, J.L. & Meyer-Lindenberg, A. (2008). Know your place: Neural processing of social hierarchy in humans. Neuron, 58(2), 273-283. Establishes that the brain tracks social hierarchy through dedicated circuits that modulate valuation in the ventral striatum, demonstrating the neural pathway through which M-signals amplify Object valuation.
Campbell-Meiklejohn, D.K., Bach, D.R., Roepstorff, A., Dolan, R.J. & Frith, C.D. (2010). How the opinion of others affects our valuation of objects. Current Biology, 20(13), 1165-1170. Directly demonstrates that knowledge of others' valuations modulates the Individual's own vmPFC valuation of the same Object, confirming the neural mechanism of mimetic valuation amplification.

4.3.6 Axiom 6: Differential Satiation Dynamics

Canonical Statement: The three motivational sources exhibit fundamentally different satiation profiles. Appetites satiate episodically and recur cyclically. Intrinsic Motivation does not satiate at the level of the Thinker's function (though each process the function sets closes when the state it was for is reached, and I-enabling Objects satiate at functional adequacy). Mimetic Desire does not satiate at all, because the Observer's function is continuous conditional Salience assessment over the whole perceived Object field, which cannot complete while perception continues; the reset of its reference points with each achievement is the expression of that non-satiation, not its cause.

Explanation:

A, I, and M do not merely differ in what they make Individuals want; they differ in what happens after the want is satisfied. This difference is the key to predicting welfare outcomes and designing effective interventions.

Appetites: Cyclic satiation. A-desires terminate upon deficit-correction and return when the deficit recurs. The hungry Individual eats, hunger subsides, the Individual is genuinely satisfied (not suppressing ongoing desire but experiencing its absence), and some hours later hunger returns. This cyclic pattern means A-needs can be genuinely met: providing food to the hungry actually terminates their hunger until the next cycle. A-satisfaction produces durable welfare gains within each cycle.

Intrinsic Motivation: Deepening without satiation. I-engagement does not terminate upon success; competence generates new frontiers. The guitarist who masters one technique discovers new techniques to pursue. The researcher who answers one question discovers new questions to ask. I-engagement produces sustainable wellbeing precisely because it deepens rather than resets. Each achievement opens new horizons rather than closing the book (Csikszentmihalyi, 1990). However, I-enabling Objects (tools, instruments, access) do satiate at functional adequacy: once the guitar is good enough to support the current level of engagement, desire for a better guitar terminates (until skill advances enough to require better equipment).

Mimetic Desire: No satiation. Mimetic Desire records no deficit to close, so there is no quantity against which it could satiate, and acquisition removes the condition that produced the signal without closing anything. The upward recalibration of the comparison set after each gain (the Reference-Set Reset, whose trajectory is the Mimetic Treadmill developed in Chapter 8) is the lived expression of that non-satiation, not its cause. Why the Observer's function admits no closure condition is derived with the components at Section 4.4.12.

This explains why societies that achieve material abundance (A-sufficiency) often fail to achieve widespread happiness. Surplus resources flow into M-channels (status competition) rather than I-channels (intrinsic engagement), producing no welfare gain despite rising expenditure. The Easterlin Plateau (Easterlin, 1974) is the macroeconomic signature of Differential Satiation: aggregate spending increases, but because the marginal spending is M-driven, aggregate welfare stagnates.

The Salience-Deficit Asymmetry. A and I encode deficits; M encodes Salience. This is the load-bearing asymmetry beneath the surface symmetry of the Integration Equation. A deficit signal carries its own termination condition, since closing the deficit silences the signal. A Salience signal carries none, since the quantity it reads (an unbounded environment, as Section 4.4.7 specifies) has no terminal value. The non-satiation of M is therefore not a contingent behavioural observation but an entailment of what M encodes together with the unboundedness of the Observer's domain, and the felt-deficit character of M is an entailment of Source Opacity operating on that Salience signal at integration.

A complementary and deeper statement of the same non-satiation holds on the emission side. Where the Salience signal is the Individual's own drive to be Salient to others, its terminating quantity, sufficient valuation by others, is doubly unavailable: it is exogenous (computed in the other Individual's architecture and not operable by the first) and event-shaped (each acknowledgement passes rather than persists). No endogenous, closable termination signal is therefore available even in principle, which is why the emission drive recurs one position higher each time it is met (the Salience Ratchet Principle) rather than resolving. This emission-side account supplements, and does not displace, the domain-unboundedness account above: the reception side explains why the detection signal has no terminal value, the emission side explains why the drive to be Salient cannot close its loop.

The asymmetry also fixes how mimetic crises terminate. With self-regulating feedback on the A-channel and the I-channel and none on the M-channel, the loop-gain threshold is crossed by a Saddle-Node Fold rather than by sustained oscillation, which is derived at Section 4.4.13.3 and formally at Appendix A.5.3.

Foundational Citations:

Berridge, K.C. (2004). Motivation concepts in behavioral neuroscience. Physiology & Behavior, 81(2), 179-209. Berridge's work on incentive Salience and the Wanting-Liking Dissociation provides the neural framework for understanding appetitive motivation and the divergence between wanting and liking that characterises M-dominant states.
Csikszentmihalyi, M. (1990). Flow: The Psychology of Optimal Experience. Harper & Row. Csikszentmihalyi's research on flow states establishes that optimal experience feeds back into higher complexity and skill, preventing simple satiation. The flow channel requires increasing challenge matched to increasing skill; the process is self-perpetuating rather than self-terminating.
Veblen, T. (1899). The Theory of the Leisure Class. Macmillan. Veblen's foundational analysis of pecuniary emulation and conspicuous consumption establishes that status-seeking is inherently comparative and therefore cannot reach a natural endpoint. Each achievement shifts the comparison set; relative position is zero-sum across the population.

4.3.7 Axiom 7: Preconscious Transmission

Canonical Statement: The Observer's reading of other Individuals' valuations is formed and delivered to the Common-Currency system faster than conscious report, through the mirror-neuron and social-cognition systems by which a general perception channel reads the directional valence of agents; in the human substrate the best-current estimate of that latency is 60–340 milliseconds. M-signals therefore reach the Common-Currency system with no accompanying phenomenology the Individual could consciously catch. That their source cannot be recovered afterwards is Source Opacity (Axiom 3), which holds whatever the latency.

Explanation:

By the time an Individual becomes consciously aware of wanting something in a social context, the mimetic signal has already been processed and fed into the Decision Hub. Mirror neurons (networks of neurons in premotor, parietal, and prefrontal cortex that activate both when an Individual performs a goal-directed action and when they observe another Individual performing that action) and social-reward circuits operate faster than conscious thought, automatically reading the goals observed in others and the value at the Objects those goals are directed to (Rizzolatti and Craighero, 2004; Caggiano et al., 2012).

The Timing Architecture:

T = 0-100ms: Visual processing of observed action.
T = 60-340ms (the Transmission Window): Mirror-neuron activation reads the Model's goal and the value at its Object; valuation circuits in the ventral striatum begin modulating the Object's subjective value.
T = 300-500ms: Attentional gating and conscious awareness become possible.
T = 500ms+: The Individual experiences wanting Object X as a present state.

By the time the Individual becomes consciously aware of wanting X (T = 300-500ms), the mimetic transmission is already complete and integrated (T = 60-340ms, the Transmission Window). The acquisition event (the moment when the Model's desire became the Individual's desire) occurred in a preconscious window that is structurally inaccessible to introspection.

The Libet Window:

The Libet Window (the 350-800 millisecond interval between neural initiation of action and conscious awareness of intention, established by Libet, Gleason, Wright, and Pearl, 1983) provides additional context. The brain begins preparing for voluntary action 350-800ms before the Individual consciously "decides" to act. This creates a temporal window during which mimetic inputs (which activate at 60-340ms) can influence the decision architecture before consciousness arrives. By the time the Individual reports "I decided to do X," the neural systems have already integrated the mimetic signal, and consciousness confabulates a narrative ("I chose this because it aligns with my values") that misattributes the M-driven choice to authentic A or I motivation.

Mirror Neurons and Predictive Activation:

Mirror neurons fire predictively, up to 480 milliseconds before an observed action occurs, indicating they represent the goal or intention behind the action, not merely the visible movement (Rizzolatti and Craighero, 2004). This predictive activation allows the brain to simulate others' goals internally, which becomes one pathway through which the Observer reads a Model's valuation: seeing someone pursue an Object activates the same neural circuits that would fire if the observer were pursuing it themselves, marking the Model's goal Object as Salient in the observer's field.

Social-reward circuits (including mirror neuron systems and related social-cognition networks) are why humans are such powerful social learners; they support language acquisition, skill transmission, and coordination. But they also make humans vulnerable to mimetic contagion when combined with Source Opacity. The same architecture that enables cultural learning enables status-driven desire transmission, and the system cannot distinguish between them.

Preconscious Transmission ensures that mimetic influence is experienced as a fait accompli of desire rather than a deliberative choice. The Individual does not experience "Person B wants X, should I also want X?" They experience "I want X" with no accompanying phenomenology indicating social origin. This timing architecture is why the acquisition event cannot be caught by conscious reflection; that its source cannot be recovered afterwards is Source Opacity (Axiom 3), which holds whatever the latency. The mimetic acquisition event is not merely forgotten or suppressed; it was never consciously registered in the first place.

This is why wisdom traditions' prescriptions to resist covetousness or "examine your desires" have limited efficacy against M. The prescriptions assume the Individual can catch the desire at its inception and evaluate it. Preconscious Transmission means the desire has already been integrated before any catching is possible.

Foundational Citations:

Libet, B., Gleason, C.A., Wright, E.W. & Pearl, D.K. (1983). Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). Brain, 106(3), 623-642. Libet's landmark experiments establish the temporal gap between neural preparation for action and conscious awareness of intention, demonstrating that neural processes initiate actions before conscious decisions.
Caggiano, V., Fogassi, L., Rizzolatti, G., Casile, A., Giese, M.A. & Thier, P. (2012). Mirror neurons encode the subjective value of an observed action. Proceedings of the National Academy of Sciences, 109(29), 11848-11853. Demonstrates that the observation responses of mirror neurons in macaque premotor area F5 are modulated by the value of the Object the observed action is directed to, so that the mirror system carries a reading of where value sits in an observed act and not only a representation of the movement.
Chartrand, T.L. & Bargh, J.A. (1999). The chameleon effect: The perception-behavior link and social interaction. Journal of Personality and Social Psychology, 76(6), 893-910. Demonstrates automatic, non-conscious mimicry of social partners' behaviours, postures, and mannerisms, establishing that social transmission can bypass conscious processing entirely.

4.3.8 Axiom 8: Bayesian Belief Dynamics

Canonical Statement: Human Belief revision follows Bayesian updating, where posteriors are computed from priors and evidence, posteriors become priors for subsequent updating cycles, and the updating process exhibits Content-Agnosticism: it operates identically regardless of the accuracy, quality, content, or crisis-potential of the Beliefs being processed.

Explanation:

The brain's Belief-revision machinery does not care what it is updating about. It treats a Belief about breakfast preferences and a Belief about whether a neighbour is a threat with the same mechanical process. This Content-Agnosticism means that once a false prior enters the system (for example, via Confabulation), Bayesian updating will faithfully propagate and entrench it. The machinery cannot flag "this prior was confabulated" because it has no content-evaluation function.

In compact form, the updating rule is Bayes' theorem:

$P (H ∣ E) = \frac{P ( E ∣ H ) \cdot P ( H )}{P ( E )}$

For AIM's Confabulation dynamics, this becomes:

$P_{t} (B) = \frac{P ( E _{t} ∣ B ) \cdot P _{t - 1} ( B )}{P _{t - 1} ( E _{t} )}$

with each posterior $P_{t} (B)$ serving as the prior for the next update, so that yesterday's posterior enters today's update as the next prior, regardless of whether the original prior entered the system through veridical inference or confabulated attribution.

Three properties of this architecture matter for AIM:

Self-reinforcement. Each time an Individual articulates a confabulated explanation ("I want this because I genuinely love it"), the articulation becomes evidence for the Self-Model. The posterior probability that the explanation is correct increases. Repeated articulation (especially under challenge, which triggers defensive elaboration) drives the posterior toward certainty, not because the explanation is true, but because the updating machinery treats repeated articulation as repeated evidence.
Content-Agnosticism. The updating process cannot distinguish accurate priors from confabulated priors. A prior that entered through veridical perception and a prior that entered through Confabulation are processed identically, and that identical handling is Content-Agnostic Processing. This means Confabulation is not a one-time error that might self-correct but a seed that the updating machinery will cultivate into a stable false Belief.
Cascade potential. In social environments where Individuals observe and respond to one another's expressed Beliefs, one Individual's confabulated Belief becomes evidence for others' updating. Their updated Beliefs become evidence for still others. Above a threshold the cascade is self-reinforcing, and at every level it is content-agnostic, while below the threshold perturbations self-correct (Appendix A, Section A.5.8.3): the machinery faithfully propagates the confabulated Belief through the network regardless of its truth-value. The formal parameters governing cascade dynamics (the conditions under which cascades become self-sustaining, and the mechanisms through which they might be interrupted) are developed in Chapter 13 (Bayesian Belief Dynamics).

This axiom becomes critical in Chapter 13, where Content-Agnostic Processing is shown to produce Confabulation Cascades at collective scale and those cascades are mapped onto the Crisis State Architecture (the framework describing how four continuous Crisis State Variables, measured across five Crisis State Phases from Grounded through Terminal and Reset, capture a social system's trajectory toward or away from mimetic crisis) at Chapter 13, Section 13.6.2, and in Chapter 17 (Violence as Architectural Output), where the Crisis State Architecture's trajectory is shown to issue in violence.

Foundational Citations:

Helmholtz, H. von (1867). Handbuch der physiologischen Optik. Voss. Establishes perceptual inference as unconscious Bayesian-style updating under uncertainty, the foundational insight that the brain is a prediction machine that revises its models through evidence accumulation.
Tenenbaum, J.B., Kemp, C., Griffiths, T.L. & Goodman, N.D. (2011). How to grow a mind: Statistics, structure, and abstraction. Science, 331(6022), 1279-1285. Develops the Bayesian framework as a unifying account of cognitive inference and learning, demonstrating that human learning across domains follows Bayesian updating norms.
Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127-138. Formalises predictive processing and Bayesian Belief updating as a core organising principle of cognition, providing the computational architecture within which content-agnostic updating operates.

4.3.9 Axiom 9: Agent-Based Causal Default (Cognitive Closure through Agent Attribution)

Canonical Statement: Human causal cognition is architecturally optimised for detecting and modelling agents (predators, prey, rivals, collaborators) and defaults to agent-based explanations when causal demands arise, because the brain's threat-detection systems, social-reasoning circuits, and theory-of-mind architecture converge on agentive causation as the computationally tractable mode of explanation. Systemic, emergent, or distributed causes exceed the computational capacity of these systems and are therefore structurally unavailable as default explanatory completions, ensuring that Confabulation (Axiom 4) characteristically takes agent-based form.

Explanation:

When something goes wrong, the brain's first question is "who did this?" not "what systemic dynamics produced this?" Agent-detection was survival-critical in ancestral environments: the cost of mistaking a predator for a shadow (false negative) vastly exceeded the cost of mistaking a shadow for a predator (false positive). The resulting bias, what Barrett (2004) terms the Hyperactive Agency Detection Device (HADD), produces a systematic tendency to perceive agency where none exists.

The same architecture now provides the default template for Confabulation. When Source Opacity creates an information gap about why an Individual wants something, and Confabulation fills that gap, the gap is filled with an agent-shaped explanation because agentive causation is the computationally cheapest mode of explanation the brain's causal-inference machinery can generate.

"Structurally unavailable as default explanatory completions" means systemic explanations are not spontaneously generated as the default automatic output when a causal gap arises. This does not mean systemic explanations are impossible to formulate with deliberate effort; the existence of AIM itself is an existence proof against that claim. The architecture is biased against generating systemic explanations, and biased toward generating agent-based alternatives. The Hypothesis that would have revealed Mimetic Desire (that wanting arises from a systemic, three-source interaction rather than from properties internal to a single Individual) is precisely the kind of explanation the architecture does not readily produce.

Where the Object being valued is itself another Individual, the classifier's output is not a single label but a weighting across the agent kinds, and the valuation the Decision Hub integrates is correspondingly a weighted combination of what each kind implies. The formal treatment is at Chapter 11, Section 11.13.1.

This axiom explains why:

Confabulation takes agent-based form. When Individuals explain their desires, they attribute them to themselves as agents ("I want this because I chose it, because I value it, because it reflects who I am") rather than to systemic dynamics ("I want this because social-reward circuits copied a Model's desire preconsciously and the Decision Hub integrated it before I could evaluate it"). The agent-based explanation is not more accurate; it is more computationally available.
Scapegoating targets agents rather than systems. When collective distress accumulates (as it does during mimetic crisis), the brain's default is to seek an agent responsible. The scapegoat mechanism (developed in Chapter 11, Confabulation, Projection, and the Scapegoat Mechanism) targets Individuals or groups rather than systemic dynamics because the causal-inference architecture cannot spontaneously generate systemic explanations under crisis conditions.
The synthesis question was invisible. The Hypothesis that would have unified motivational science ("motivation arises from three structurally distinct sources whose interaction produces emergent phenomena invisible to introspection") is a systemic Hypothesis. Axiom 9 predicts that this Hypothesis-type would not be spontaneously generated within any discipline whose causal-inference framework defaults to agent-based explanation, which is all of them.

The standing of this axiom.

Axioms 1 to 8 specify structure: how many sources there are, what the integrator does to them, what the integration destroys, what the components do with what survives it, and how the resulting Beliefs update. Every formal result in the AIM Framework derives from that structure. Axiom 9 specifies none of it, and enters the formalism only as a constraint on the classifier upstream of the fast loop (Appendix A, Section A.2.1). It is a claim about content, about which explanatory completions the architecture reaches for when a causal demand arises and the structure has left an information gap, and it is the one axiom whose subject is what fills the architecture rather than what the architecture is. The Structure-Content Distinction set out in 4.3.11 therefore bears on it as it bears on no other axiom.

Axiom 9 is not load-bearing for the architecture. Remove it and the four components remain, the integration remains, Source Opacity remains, Confabulation remains as the characteristic output of a Thinker resolving grounds from a tag-stripped Scalar, and the satiation dynamics, the transmission window and the updating rule are untouched. The mathematics is undisturbed for the same reason: the formal treatment takes Axiom 9 as a constraint on the classifier that feeds the fast loop rather than as a variable within it (Appendix A.2.1), so the loop-gain condition, the critical threshold, the fold geometry and the parasitism result are derived without it. An Individual whose causal cognition defaulted to systemic rather than agentive explanation would still be running this architecture, would still confabulate, and would still be subject to every result the loop dynamics establish. What would change is the shape of what was confabulated.

That shape is what Axiom 9 supplies. The architecture entails that a gap will be filled; it does not, on its own, say with what. Axiom 9 fixes the completion, and the completion converts a general result about Confabulation into the specific predictions the treatise makes: that the Individual attributes a mimetically installed desire to their own authorship rather than to a process that has no author; that a population under accumulating distress converges on a person or a group rather than on a dynamic; and that the systemic Hypothesis, the one that would identify the actual cause, is the Hypothesis the architecture is least likely to generate. The Scapegoat Mechanism does not follow from Axioms 1 to 8 alone.

It is carried as an axiom rather than as an empirical observation because it is presupposed at every scale the AIM Framework treats: in the Individual's account of their own wanting, in the dyad's reading of a rival, in the network's convergence on a target, and in the institution's search for a party to hold responsible. Its evidential standing is the standing of the findings cited below, and it is falsifiable on the same terms as the rest: a demonstration that causal cognition under closure-pressure does not default to agentive completion would refute it. Such a refutation would take with it the AIM Framework's account of the form the failures take, and would leave the architecture that generates them standing.

Foundational Citations:

Barrett, J.L. (2004). Why Would Anyone Believe in God? AltaMira Press. Develops the Hyperactive Agency Detection Device (HADD) account of agent-attribution bias under uncertainty, establishing that humans systematically over-attribute events to intentional agents.
Heider, F. & Simmel, M. (1944). An experimental study of apparent behavior. American Journal of Psychology, 57(2), 243-259. Demonstrates spontaneous attribution of agency and intention to geometric shapes moving on a screen. Participants described triangles and circles as "chasing," "bullying," and "hiding," revealing the depth of the agent-detection default.
Rosset, E. (2008). It's no accident: Our bias for intentional explanations. Cognition, 108(3), 771-780. Establishes the default preference for intentional-agentive causal explanations over non-agentive alternatives, even when non-agentive explanations are more accurate.

4.3.10 Three Layers Beneath the Agent-Based Causal Default: Update-Rate, Formation, and Scaling

Section 4.3.9 established that Confabulation characteristically takes agent-based form because agent-based explanations are the computationally cheapest completions available to the causal-inference architecture. Does Axiom 9 entail that systemic Hypotheses cannot be formed at all? It does not. Axiom 9 states that systemic explanations are structurally unavailable as default explanatory completions, not that they are unavailable as achievable completions under non-default conditions. The existence of AIM itself is proof that the systemic-Hypothesis type can be formed. What Axiom 9 explains is why formation is rare, not why it is impossible. This section supplies the proximate mechanism beneath that claim, traces its consequence for how systemic Hypotheses are formed, and identifies a further, independent consequence for how correct Hypotheses scale once formed.

The Configural Update-Rate Asymmetry. The Observer, as established earlier in this chapter, is the general perception channel of the architecture, not a dedicated social or agent-detection faculty. It samples the surrounding field continuously and attaches Salience to whatever it detects, whether or not that field contains another agent. The question this section answers is why a general-purpose channel of this kind ends up so heavily saturated with agent-based content in practice.

The answer lies in a simple asymmetry of information density. An apple hanging on a branch presents almost no state-change signal from one moment to the next, at most swaying slightly in the wind, and generates correspondingly little fresh data for a continuously-sampling perception channel to register. An animal in the same field can transition from static rest to pursuit, to attack, to flight within a fraction of a second, and each such transition constitutes a new, spatially localised, behaviourally significant configuration change. Heider and Simmel's classic 1944 demonstration that observers spontaneously attribute intention and agency to moving geometric shapes, purely from their motion pattern and with no static-object cues present whatsoever, illustrates precisely this point: agency-attribution tracks configural change, not object-identity. A general-purpose perception channel sampling a field containing both an apple and an animal will therefore accumulate vastly more update events attributable to the animal, not because the channel has any built-in preference for agentive content, but because the animal is the only entity in the field generating a comparable rate of discrete, trackable state changes.

Diffuse systemic causes fare worse still on this measure than even a static object. A housing market, an institutional process, or a distributed regulatory failure has no single spatial location whose configuration can be tracked, no discrete transition from one observable state to another that a perception channel could register as an event. Systemic causes are not merely slow relative to agents, they are non-configural — they do not present the kind of unified, trackable entity-state that the Observer's channel is built to accumulate evidence about at all. This is why systemic Hypotheses cannot simply be perceived and must instead be constructed.

The Transmission-Deliberation Asymmetry. This update-rate differential is the proximate cause of the timing asymmetry between how agent-based and systemic Hypotheses reach the Decision Hub. The Observer's continuous sampling, dominated by agent-configuration data for the reasons just given, supplies the Thinker, through the Salience it delivers to the Decision Hub and the Scalar that returns, with Evidence already shaped by agent classification against the Agent-Type Vector, so that agent-based candidates require only classification against an already information-rich, continuously-updating stream. Constructing a systemic Hypothesis, by contrast, requires the Thinker to deploy Imagination, the generative operation that assembles candidate Hypotheses through novel connections between elements the Belief Register has not already flagged as jointly relevant, precisely because no equivalent continuous, configural evidence stream exists for systemic causes to be read off directly. This construction is necessarily slower, not because deliberation is inherently sluggish, but because it must build its evidentiary basis from a comparatively sparse and disconnected source rather than harvesting a dense, ready-made stream the way agent-detection does.

Under default closure-pressure conditions, this produces a race the fast channel wins almost by construction. The agent-based candidate, read off that dense stream, typically reaches the point of Belief-fixation before the systemic candidate has finished being built, and once the closure-demand that drove the search has been satisfied by the fast candidate, further search terminates before the slow candidate is completed. Critically, this is not an argument that systemic Hypotheses are architecturally excluded from the Hypothesis Space. It is an argument that a Bayesian system correctly assigns lower prior weight, in the first instance, to a Hypothesis-type built from a comparatively starved evidence stream, and that this starvation is itself a direct entailment of the Configural Update-Rate Asymmetry rather than an independent computational ceiling.

The asymmetry is conditional, not absolute, in exactly the way Chapter 11's treatment of I-Override establishes. Where Epistemic Suspension holds the Interference Term above zero, the closure-demand does not terminate the search the moment the fast candidate arrives, giving the Thinker's slower, evidence-starved construction the additional time it needs to complete despite its inherent disadvantage. Where AIM Vocabulary Acquisition has expanded the Hypothesis Space to admit systemic candidates directly, the Thinker's construction task is shortened because the connecting concepts are already available rather than needing to be assembled from more primitive elements under time pressure. Both conditions are trainable capacities and available vocabulary, not fixed ceilings. The rarity of systemic-Hypothesis formation under ordinary conditions is therefore a statement about how infrequently these enabling conditions co-occur with sufficient causal demand to trigger the search in the first place, compounded by the update-rate deficit systemic causes inherently suffer, not a statement about an architectural impossibility.

The Scaling Asymmetry of Correct Hypotheses. A further and independent consequence follows once the first has been resolved. Suppose an Individual has, against both the update-rate deficit and the resulting Transmission-Deliberation Asymmetry, succeeded in constructing the correct systemic Hypothesis. Transmitting this Hypothesis to others is not simply a matter of clear articulation, because Axiom 8's content-agnostic Belief-revision machinery processes the correct signal against listeners whose own priors have, independently and for the identical architectural reasons just described, already converged on an agent-based explanation supplied by their own update-rate-dominated perception channels. The correct Hypothesis therefore arrives at an already-decohered prior with a correspondingly high likelihood-ratio requirement, and this resistance compounds across a population through the exchange of Evidence as each listener's continued expression of the dominant agent-based explanation becomes further endogenous evidence reinforcing that explanation in everyone else.

This is a distinct layer from the two above, not a restatement. The Configural Update-Rate Asymmetry and the Transmission-Deliberation Asymmetry describe why one Individual struggles to construct the correct Hypothesis. The Scaling Asymmetry describes why, even once constructed, that Hypothesis struggles to propagate — a problem resolved only by changing the listening population's priors through repeated exposure or by reducing the population's Coupling Strength, not by any further effort on the Hypothesis-holder's part.

Together, these three layers complete the non-defeatist reading of Axiom 9 opened at Section 4.3.9. Agent-based attribution dominates because agents generate more perceptible evidence per unit time than objects or systems; this update-rate deficit is what makes systemic-Hypothesis formation rare rather than impossible; and a second, independent update-rate-driven deficit at the population level is what makes propagation of a correctly-formed Hypothesis rare rather than impossible. Each layer is named, each has an identified intervention point, and all three are consistent with the AIM Framework's Non-Utopian Framing: vast improvement over the default trajectory is available without requiring that any of the three difficulties be eliminated outright.

4.3.11 Content and the Structure-Content Distinction

One further distinction remains: the relationship between AIM's structural claims and the Content those structures process.

The Structure-Content Distinction:

AIM is a structural theory. It specifies the architecture through which motivation is generated, integrated, and experienced (three sources, Common-Currency Integration, Source Opacity, Confabulation, Object Parasitism, Differential Satiation, Preconscious Transmission, Bayesian Belief Dynamics). It does not specify the particular Content that fills these structures. Axiom 9 is the single exception, and a partial one: it constrains not which Objects fill the architecture but which explanatory completions the architecture generates when a gap arises, which is a claim about the form Content takes rather than about its substance. Its standing is set out at 4.3.9.

Content (the specific Objects, Beliefs, narratives, cultural materials, and situational details that fill AIM's structural architecture in any particular case) is determined by the Individual's environment, culture, personal history, and current circumstances. AIM predicts the dynamics of Content processing (what happens when Content enters the system), not the Content itself.

Content varies; structure does not. The Objects that fill M-driven desire differ across cultures, historical periods, and Individuals: land in feudal societies, credentials in meritocratic ones, followers in digital ones. But the dynamics of M-driven desire in its special case (audience-dependence, non-satiation, rivalry escalation, reference-point reset) are structurally constant because they arise from the architecture, not from the Content.
Structure determines Content-trajectory. Although AIM does not predict which specific Object an Individual will pursue, it predicts what will happen after they pursue it based on which source is dominant. A-driven pursuit will terminate upon acquisition (until deficit recurs). I-driven pursuit will deepen with engagement. M-driven pursuit will reset upon acquisition and recreate felt deficit. These predictions hold regardless of the specific Content.
Content-determination is structurally governed. Which Content becomes Salient to an Individual is itself partly determined by structure: A renders deficit-relevant Content Salient (food when hungry, shelter when insecure); I renders engagement-relevant Content Salient (the next challenge, the next question, the next creative possibility); M renders Model-observed Content Salient (what the reference group is pursuing, displaying, or competing for).

The Undifferentiated Content Trap:

A common error in evaluating motivational theories is to focus on Content rather than structure. Two Individuals pursuing the same Object (the same house, the same career, the same relationship) appear to have the same motivation when examined at the Content level. The structural question (which source is dominant?) reveals that they may have entirely different motivations, entirely different satisfaction trajectories, and entirely different intervention needs.

The Undifferentiated Desire Fallacy (Chapter 5, Section 5.1) is, at root, a Content-level analysis mistaken for a structural one: because the Content of desire appears similar across Individuals, the structure of desire is assumed to be uniform. AIM's contribution is to demonstrate that structurally different sources can produce Content-identical desires with fundamentally different properties.

The following table summarises the key Content-related principles:

Principle	Statement	Implication
Content-Agnosticism	Belief updating operates identically across subject matters, with no internal variable for the Truth, moral value, or crisis-potential of the Belief	The same updating dynamics apply to a Belief about a product, a party, or a scapegoat
Content-Trajectory Independence	The trajectory of Belief change is set by the structure of updating and the pattern of inputs, not by Content	Structurally identical input patterns produce structurally identical trajectories whatever the Beliefs concern
Content-Channel Asymmetry	What spreads through mimetic channels can enable resistance to mimetic dominance where the Content is I-enabling rather than M-amplifying	Diagnostic vocabulary travels by the same channels as status signals and can work against them

Falsification Condition:

If it could be demonstrated that content, rather than source, determines satisfaction trajectories (that pursuing a given Object produces the same welfare outcome regardless of whether the pursuit is A-driven, I-driven, or M-driven), the Structure-Content Distinction would be undermined and AIM's predictive architecture would require fundamental revision. Consider black pepper. Its A-function (flavouring food, aiding digestion) and its I-function (its role in skilled culinary practice) have not changed in five centuries. What has changed is its scarcity, and with it, its M-Salience. In medieval and early modern Europe, black pepper was rare, expensive, and visibly displayed at feasts in locked spice boxes. It functioned as a Status marker precisely because scarcity made it a reliable signal of wealth. Acquisition did not satiate at culinary sufficiency; accumulation beyond any cooking requirement signalled rank, and reference groups constantly reset expectations upward. The same substance that today costs a few dollars per kilogram and satiates at household sufficiency (purchased, restocked when depleted, generating no further desire) once drove non-satiating positional competition among European elites. The pepper is the same. The flavour is the same. The culinary purpose is the same. What collapsed was the scarcity that made pepper M-Salient, and with it, the entire non-satiating trajectory.

No one today accumulates pepper beyond functional need, because pepper no longer signals Status. The A-trajectory (episodic satiation at sufficiency) and the I-trajectory (deepening engagement for the dedicated cook) remain available and unchanged. Only the M-trajectory disappeared, and the reason is specific. What collapsed was pepper's capacity to discriminate positions. Observing another Individual value an Object installs a Mimetic signal toward it, and that remains as true of pepper as of anything else; scarcity is not a condition on installation (4.4.7.1). What abundance destroyed is the comparative-position content of the observation. Where everyone can have the thing, having it places no one relative to anyone, so the Observer's assessment of it no longer carries the positional loading that drove accumulation past culinary sufficiency. Scarcity is not what M needs in order to install. It is what an Object needs in order to register rank. Three trajectories were available for one Object. Content did not change. Scarcity changed, M-Salience collapsed, and the satisfaction trajectory shifted from non-satiating accumulation to episodic sufficiency. AIM predicts this; undifferentiated models cannot.

4.3.12 Summary: The Axiomatic Foundation

These nine axioms form the non-negotiable core of the AIM Framework. They are treated as foundational premises that define the theory's structure, generate its predictions, and establish its falsification criteria. Any claim, application, or extension within AIM must be consistent with these axioms, and any empirical finding that definitively contradicts one or more of these axioms would constitute serious evidence for revising or rejecting the framework.

The axioms are not independent; they form an interlocking system:

Axioms 1 and 2 establish the three-source structure and its integration mechanism.
Axioms 3 and 4 explain why the three-source structure is invisible to introspection.
Axiom 5 explains why M operates parasitically on A and I rather than independently.
Axiom 6 explains why satisfying different sources produces different welfare outcomes.
Axiom 7 explains why Preconscious Transmission precedes conscious awareness, grounding Axioms 3 and 4 in neural timing.
Axiom 8 explains how Content-Agnostic Processing propagates and entrenches Confabulation across update cycles.
Axiom 9 explains why Confabulation defaults to agent-based causal attribution rather than systemic explanation.

The interlock is not uniform in what it bears. Axioms 1 to 8 are load-bearing for the architecture and for every formal result derived from it; Axiom 9 is load-bearing for the account of the form the architecture's failures take. A refutation of Axiom 9 would cost the AIM Framework its explanation of why Confabulation is agent-shaped and why collective distress converges on scapegoats, and would leave the architecture, and the mathematics derived from it, standing (4.3.9).

Together, the axioms generate AIM's explanatory power: the ability to predict when preference-satisfaction will produce welfare gains (A), when it will produce deepening engagement (I), and when it will produce only treadmill effects (M), which are predictions that undifferentiated models cannot make.

4.4 The Minimal Architecture and Its AIM Instantiation

AIM's three-way decomposition of motivation into Appetites, Intrinsic Motivation, and Mimetic Desire operates at a specific level of abstraction, and the claims the AIM Framework makes about that decomposition are calibrated to that level. A, I, and M are reading surfaces over a continuous and entangled neural substrate, in the same sense that the gauges on a vehicle's dashboard are reading surfaces over the engine's fuel, cooling, and lubrication systems. A fuel gauge resolves the state of a complex supply system along a single behaviourally actionable dimension. It is not the fuel system, and it does not claim the fuel system is simple. Its usefulness lies precisely in the abstraction: it tells the operator what can be acted on, at the level at which action is possible.

A, I, and M work the same way. Each summarises its substrate along a behaviourally actionable dimension: homeostatic deficit for A, conscious-cognitive engagement and Bayesian resolution for I, and continuous conditional Object-field Salience assessment for M. Each is effective for disaggregation not because the underlying neural systems are themselves cleanly modular (they are not), but because the behavioural patterns the three components pick out are consistent and reliable enough to support diagnosis and intervention. This is the Dashboard Gauge Principle, and it fixes the level at which AIM's architectural claims are pitched. The AIM Framework does not claim neural-level granularity. It claims reading-surface granularity, which is the level at which its derivations and interventions operate.

A component specified at that level, rather than at the level of its neural substrate, is a Disaggregation Primitive. The separate identifiability of the Primitives is a property of observable behaviour and associational regularity, not a property of neural-module separability. The neural systems that produce A-behaviour, I-behaviour, and M-behaviour overlap and interact constantly, and nothing in the AIM Framework denies this. What the AIM Framework claims is that, when behaviour is watched closely enough, three distinct patterns and association dynamics emerge that are stable, recurrent, and diagnostically actionable. That is the level at which A, I, and M are separately identifiable. It is sufficient for everything the AIM Framework asks of the Primitives, and it leaves the underlying neural complexity unclaimed.

The minimal architecture set out below — the Minimal Architecture Specification (MAS), as it is referred to throughout the treatise — is specified at this level. Its functional statements about what A, I, M, and the Decision Hub each do are statements about what each component does at the reading-surface level, not about how the supporting neural systems are organised. The neural-substrate detail is developed in Chapter 9.

The nine axioms established in Section 4.3 fix what the system is. Axiom 1 establishes the three-source taxonomy. Axiom 2 establishes Common-Currency Integration and supplies the integrator. Axiom 3 establishes Source Opacity, which is a mathematical entailment of Axiom 2 via the many-to-one integration and the Data Processing Inequality. Axiom 4 establishes Confabulation as the characteristic output of the narrative-construction function operating on tag-stripped input. Axiom 5 establishes Object Parasitism as M's operating mode. Axiom 6 establishes the differential satiation dynamics across A, I, and M. Axiom 7 establishes the Preconscious Transmission window within which M-signals are integrated before conscious awareness. Axiom 8 establishes Bayesian Belief dynamics as the updating rule. Axiom 9 establishes the agent-based causal default under which Confabulation characteristically produces agent-shaped attributions.

The axioms do not yet specify, at the component-function level, what each source does operationally, how the components interact over time, which component maintains the priors the other components rely on, or what the architecture's characteristic failure mode looks like when the interaction is corrupted. This section supplies that specification. The four components are set out in §§4.4.1–4.4.3 and 4.4.7; §4.4.6 supplies how the order of experience shapes the Thinker's multi-Hypothesis ground-resolution; the Minimum Installation Condition on Observer-side Mimetic installation is stated in 4.4.7.1; the Belief Register mechanism that mediates downstream access to the Thinker's posteriors at preconscious timescales is set out in 4.4.8, the operational loop that couples the components in 4.4.9, the failure mode the architecture entails in 4.4.11, and the consolidation of localisations and derived consequences in 4.4.12, and the crisis geometry that the coupled loop entails in 4.4.13, with 4.4.10 providing a summary table. Each component sub-section specifies the component's function abstractly, in terms any system instantiating the architecture would satisfy, and then identifies what the AIM Framework claims instantiates that function in the human motivational system.

The minimal architecture and its AIM instantiation are separately falsifiable. The human system might fail to instantiate a minimal architecture that remains a valid specification for some other system. This separability is not an escape from the record: if the human instantiation is falsified, the AIM Framework has failed as a theory of human motivation, whatever residual validity the abstract specification retains for some other system. The minimal architecture might require revision in ways that leave the AIM instantiation claim untouched.

For ease of reference, the four components specified in this section are referred to throughout by reader-facing names. The integrator and action-selector is the Decision Hub. The homeostatic-deficit signaller is the Body Manager. The prior-maintainer and Bayesian ground-resolver, the component that performs deliberative inference, is the Thinker. The continuous conditional Object-field Salience assessor is the Observer. These names are chosen to identify each component by what it does at the reading-surface level, in the spirit of the Dashboard Gauge Principle, and to avoid the unwieldy technical compounds that the function-specification work in this section would otherwise propagate through the rest of the treatise.

4.4.0 The Signed Valuation Principle

Each of the three input-bearing components (the Body Manager, the Thinker, and the Observer) supplies the Decision Hub with a signed valuation signal, not an unsigned magnitude. The signal can take positive values (informing attraction toward the Object), negative values (informing avoidance of the Object), or zero (informing indifference). The sign is intrinsic to the signal: it is not a separate avoidance channel running alongside an attraction channel, but a single signed Scalar per component per Object per moment. The Decision Hub's integration is consequently an integration over signed terms, and the integrated value V can take any real value, positive (net attraction), negative (net avoidance), or zero (net indifference). Action-selection on V therefore selects approach actions when V is positive, avoidance actions when V is negative, and disengagement when V is at or near zero. This is the Signed Valuation Principle.

Disengagement is a selection outcome and not the absence of one. The loop has run, the integration has occurred, and no motor output has issued, so the interval is behaviour without action. Intervals of that kind are not gaps in the record; they are where several of the framework's diagnostics do their work, and 4.4.1 sets out why.

The architectural consequences of the Signed Valuation Principle are three. First, each component is independently valence-sensitive: each can independently contribute attraction, avoidance, or indifference toward the same Object simultaneously, and the Decision Hub integrates the three signed contributions into a single net signed Scalar. The Body Manager can signal a current internal deficit (hunger) and a current internal aversive state (nausea after ingestion), and these are the same signed channel taking different values under different conditions, not two separate channels. Binding either signal to an Object that has not contacted the body is the Observer's locative job, conditioned by the Thinker's priors. The Thinker can resolve an engagement as deepening (positive I-valuation) or as cognitively threatening, boring, or value-violating (negative I-valuation). The Observer can flag an Object as Model-pursued and comparative-position-elevating (positive M-valuation) or as Model-avoided, low-status, or comparative-position-degrading (negative M-valuation).

Second, the source-opacity result holds with full force at the integration step: V carries no information about which component supplied which sign or which magnitude, and an Individual experiencing a strongly negative V toward an Object cannot introspectively distinguish A-aversion from I-aversion from M-aversion any more than they can distinguish A-attraction from I-attraction from M-attraction. The Knowledge Patch's Two-Question Diagnostic and Presumption Against Scapegoating apply symmetrically to the avoidance side of the architecture. Third, the four derived consequences set out in 4.4.12 (the I-Supremacy Result, the A-Disadvantage, the I-Template Override Result, and the I-Corrective Asymmetry) all hold on the avoidance side, though a net-negative reading cannot compound through the loop, so the amplification that makes template configuration most consequential is an approach-side property: cultural configuration of the Thinker's templates can produce systematic Object-aversion (the heretic, the unclean, the contaminated, the scapegoat) by the same channel-count arithmetic that produces systematic Object-attraction, with the Belief Register mechanism carrying the cultural aversion-templates at action-execution timescales just as it carries the attraction-templates. The scapegoat appears in that list under one of two aspects the Observer's channel carries toward the same figure, and 4.4.7 separates them once the Observer has been specified.

The Signed Valuation Principle is not an additional axiom; it is a feature of the minimal architecture that follows from the requirement that the architecture be capable of producing avoidance behaviour, which Axiom-1 and Axiom-2 already implicitly require. The Signed Valuation Principle simply makes that feature explicit, so that downstream derivations can refer to it directly.

The crisis-geometry results stated later in this chapter and derived in Appendix A (the threshold at which the Internal Loop Gain reaches one, the Saddle-Node Default, and the Hopf exclusion) hold for exogenous drives of either sign, provided the couplings that close the fast loop are one-signed (Appendix A, Section A.5, the Signed-Input Stability Corollary). The case in which one of those couplings itself changes sign is a named specialist task (Appendix A, Section A.7, Task 12b), and the chapter's crisis-geometry claims should be read as covering one-signed couplings.

Because the three inputs are not of one kind, the loss that Axiom 3 states (Section 4.3.3) is not confined to identity. All three are directed at Objects, which is what makes them commensurable enough to sum. The Body Manager encodes a deficit, a departure from a homeostatic set-point that some Object would close, which may be a condition of the Individual's own body and then needs no third thing to be corrected through. The Thinker encodes a deficit of the same form, the absence of an Object that would support the engagement it seeks, whether an instrument, a problem, a material, a collaborator, or a condition of the Individual's own body such as fitness, which is an Object in the same sense as any of them and needs no third thing to be pursued through.

The Observer encodes no deficit. It encodes Salience, a signed reading of which Objects in the environment are marked as valued and which as to be kept from, and its function is to point the other two channels at targets, since a deficit signal with no Salience reading is a drive with nowhere to go. The integrated Scalar admits one interpretive category, a demand for correction, and its sign states which correction is demanded: a positive V demands that the Individual obtain or approach the Object, a negative V that the Individual escape, expel or keep away from it. Integration therefore does not merely strip the Observer's tag. It converts a signal that reports where value is located into a signal that reports that something must be corrected, and it leaves the Individual no operation by which the reading could be received as a reading.

A further feature of the minimal architecture follows immediately from the Signed Valuation Principle. Because the Observer's Salience contribution is a signed input to the same Common-Currency Scalar that carries the Body Manager's and the Thinker's signals, and because that Scalar is acted on solely as a signed demand for correction, the Salience contribution cannot be integrated and yet experienced as Salience: it is necessarily felt as the correction its sign demands. A positive reading is felt as a lack, a pull toward the Object. A negative reading is felt as a threat, a recoil from the Object, which under Phenomenological Collapse is experienced as fear, disgust or contempt, and as a state emanating from the body itself. The felt lack and the felt threat each carry the magnitude of V and nothing of its sources. This is the Salience-Deficit Conflation Principle (Chapter 2). On the positive pole a pointer to where value sits arrives wearing the form of a lack: acquiring the Object closes nothing, there having been nothing open, and no introspective operation will disclose that. Where the dominant Salience contribution is mimetically installed — set because a Model was observed valuing the Object, with no A-deficit or I-opportunity Referent beneath it — the Salience-Deficit Conflation Principle entails a felt deficit for which no corrective mechanism exists, which is the architectural root of M's non-satiation (Axiom 6). The Salience-Deficit Conflation Principle is not an additional axiom; it makes explicit a consequence already entailed by Axioms 2 and 3 operating on the Observer's signed input, so that downstream derivations can refer to it directly.

The negative pole is installed and revised by the same means as the positive. Where the dominant negative Salience contribution is mimetically installed — set because a Model was observed recoiling from, shunning or condemning the Object, with no A-aversion or I-aversion Referent beneath it — the Individual feels a threat with no Referent beneath it. Avoidance discharges the felt threat: once the Object is out of the field or out of reach, the Observer's reading of it falls, V returns toward zero, and the felt state ends (Chapter 8, Section 8.11.3). The template that set the sign is a Belief like any other and is revised as any other is, by Evidence (Section 4.4.8). Three routes carry that Evidence. A Model observed approaching the Object without harm enters through the same observation channel by which the avoidance was installed. Testimony enters the Belief Register as Evidence weighted by Coupling Strength (Section 4.5; Appendix A, Section A.5.7.3). And the Individual's own contact with the Object, followed by no harm, is Evidence of the most direct kind. Avoidance closes the third route and leaves the other two open. While the Individual keeps the Object at a distance, the Individual's own conduct supplies no Evidence against the template, so a mimetically installed aversion is revised chiefly from outside the Individual who holds it: by Models who approach what that Individual avoids, and by what that Individual is shown and taught. That avoidance prevents the disconfirmation of the threat it is aimed at is the established account of how avoidance maintains anxiety (Salkovskis, 1991), and that aversion toward a group falls with contact, including the indirect contact of knowing an in-group member who has it, is the finding of the contact literature (Allport, 1954; Wright et al., 1997). In the formal system the internal cycle can raise V above its exogenous forcing and cannot carry it below it, so a negative reading does not compound through the fast loop. What carries it between encounters is the Belief Register, and the Belief Register moves under Evidence (Appendix A, Section A.5, the Signed-Input Stability Corollary). A felt aversion that sustains engagement with its Object, rather than distance from it, is not the negative pole: what is selected there is approach, and Section 4.4.7 separates the two aspects under which the Observer's channel carries the same figure.

4.4.1 The Decision Hub (Integrator and Action-Selector)

Axiom grounding. The Decision Hub is the direct architectural rendering of Axiom 2 (Common-Currency Integration). Axiom 3 (Source Opacity) is an immediate consequence of the Decision Hub's operation under the Data Processing Inequality and is not an independent architectural stipulation.

Function-only specification. The Decision Hub is the component whose primary function is action-selection. It receives valuation inputs from the other three components, integrates them into a single Common-Currency Scalar, and emits a selected action for execution. It also returns the integrated Scalar to the Thinker as the feedback signal on which Bayesian ground-resolution operates. The valuation of a candidate the Decision Hub does not select is not discarded: its Scalar returns to the Thinker on the same path, and an unselected claim the Belief Register cannot already resolve is held as a prior about what will be required next, where it competes at later selections (Chapter 12, Section 12.3.1). Action-selection is single-channel: the Decision Hub emits one selected action at a time, and the direction of attention is among the actions it selects. The Decision Hub is therefore the bottleneck through which the body's energy budget is spent and the point at which the components compete (Section 4.1).

The Decision Hub integrates unlike motivational signals into one Common-Currency Scalar. The integration is tag-stripping: the integrated Scalar does not carry information about which source contributed what weight. This is the Source Opacity Theorem, proved via the Data Processing Inequality.

The result is opacity of the source decomposition, not opacity of value, and it does not depend on the integrated representation being one-dimensional. What the summation destroys is the correspondence between contribution and source; whatever further structure the integrated representation carries, it does not carry that. A finding that valuation is represented with more structure than a single number therefore leaves Source Opacity where it was, provided the integrated representation still maps many source combinations onto each of its states, because the result rests on the map being many-to-one and not on the count of dimensions on the far side of it. Appendix A states the result for a scalar output (Sections A.2 and A.6.2).

Source Opacity and the two barriers at the Decision Hub.

Source Opacity, as the term is used throughout, names the fact that the Individual has no introspective access to the source-decomposition of an integrated valuation, and it is the product of integration alone (Axiom 3; Appendix A, Section A.2, restatement M5a). Self-knowledge of the M-contribution is barred by more than Source Opacity: Chapter 9 sets out three independent barriers to it, of which Source Opacity is one. Two of those barriers can be stated here, because they bear directly on the Decision Hub, and because an objection to Source Opacity sometimes turns on which of the two is doing the work.

The first barrier is Preconscious Transmission (Axiom 7). The Observer's Salience signal is generated and integrated within the Preconscious Transmission Window (on the order of 60 to 340 ms in the human instantiation), before the Thinker can subject it to deliberative scrutiny. The claim is about timing, and it holds independently of how the components are wired. It makes no difference whether the M-contribution reaches the Thinker only through the Decision Hub or also through some hypothetical direct Observer-to-Thinker channel, because on either routing the signal arrives faster than the Thinker can evaluate it. Timing therefore forecloses conscious interception of the M-contribution at the moment of action-selection, because the M-contribution arrives with no phenomenology the Thinker could catch; reading its source decomposition afterwards is foreclosed by the integration itself, whatever the timing. This barrier is not Source Opacity and does not produce it: relaxing the timing changes how fast action-selection runs, not what the integration destroys.

The second barrier is integration itself, and it is Source Opacity proper (Axiom 3, an entailment of Axiom 2). Common-Currency Integration maps the three signed source-inputs onto a single Scalar V many-to-one, and the Data Processing Inequality establishes that no downstream operation can recover the source decomposition the integration step has destroyed. The claim is about information, not about timing, and it would hold even if the Thinker had unlimited time to deliberate over V.

The independence of the two barriers is clearest in the counterfactual that the second forecloses. Suppose, contrary to the architecture, that integration were lossless, so that no many-to-one summation occurred and V retained a source-tag recording how much of the valuation the M-channel had supplied. On that counterfactual the timing barrier would still hold (the signal would still arrive preconsciously), but a retained M-tag would constitute an in-principle backup, a residual record available to later deliberation that some quantified portion of the valuation driving the decision had entered through a preconscious mechanism. The Thinker could in principle read that tag after the fact and weight its ground-resolution accordingly, attributing the corresponding portion of its wanting to M rather than confabulating an A-ground or I-ground. The Data Processing Inequality is precisely what forecloses this counterfactual. The many-to-one summation at integration destroys the tag irreversibly, so the in-principle backup that lossless integration would have preserved does not exist downstream of the Decision Hub.

The two barriers therefore do distinct work. Preconscious Transmission is decisive on its own at the moment of action, denying real-time conscious interception. Integration is decisive on its own across all time, denying even retrospective disaggregation by destroying the source-record that retrospective disaggregation would require. Source Opacity is the second of the two and not their conjunction, which is why the relaxation of the specific timing window leaves Source Opacity intact. The timing window is a relaxable empirical calibration of the first barrier, whereas the destruction of the decomposition is an essential entailment of Common-Currency Integration and is not relaxable without abandoning Axiom 2.

What integration destroys is the decomposition, not the contributions. The Scalar carries all three contributions as their sum: it rises and falls with each of them, and none is recoverable from it. Because the three obey different termination dynamics (Axiom 6), the Scalar's course across conditions that move one contribution and not the others, a deficit corrected, an audience removed, a task completed, differs according to which contribution led, so the leading channel can be inferred from that course and from the behaviour it drives, though never read off any single value (Appendix A, Section A.2, restatement M5a). The inference returns the leading channel and not the weights. It is available to an observer and, through the same contrasts run on oneself, to the Individual, and because it is inference rather than introspective reading it routes around Source Opacity and does not breach it. An Individual who confabulates an A-ground or an I-ground is therefore short not of information but of the Hypothesis that could use it, which is the Confabulation Rationality Lemma's account (Chapter 11, Section 11.2).

The integration maps many source combinations onto each value of its output, so the representation that would carry source-tags does not exist downstream of it. The Decision Hub emits action selections based on the integrated Scalar, not on the component signals directly. The signals received from the Body Manager, the Thinker, and the Observer are each signed scalars (positive, negative, or zero), per the Signed Valuation Principle (4.4.0), and the integrated Scalar V is therefore itself signed, with action-selection consequently selecting approach when V is positive, avoidance when V is negative, and disengagement when V is at or near zero.

The Decision Hub does not interpret, ground, or explain its outputs. It does not maintain priors. It does not resolve Salience into the grounds that the other components supply signals from. Interpretation is the Thinker's function, not the Decision Hub's. The Decision Hub's feedback to the Thinker is the tag-stripped Scalar itself, not three separately-tagged source-contributions. This feedback-architecture feature is what makes ground-resolution a Bayesian inference problem rather than a lookup: the Thinker receives a Scalar from which the information required for perfect ground-attribution has already been destroyed, and must infer the source-composition using its own priors.

The Scalar converts to an action, and that conversion is the transduction. The Scalar is not replaced at the boundary by a different quantity; it is carried across into a medium in which it can move something, which as a signal inside the body it cannot. What the Individual does about a valuation, and what anyone else can see of it, is that same Scalar continued in the medium that admits of motion. The action is not a sign standing at some remove from the valuation that produced it. It is that valuation, carried across.

The action the Decision Hub selects is directed at an Object, and that action is the transduction: it puts the same Scalar into a medium in which the Scalar can occupy an Object. The Object then carries the signal for as long as action is directed at it. Because the Decision Hub does not interpret its outputs, it does not distinguish an Object that satisfies the deficit from one that only unlocks a further Object, and is therefore indifferent to which Object occupies the intermediary position and to how many occupy it in series.

The minimum the transduction requires is an effector and a deficit that the effector's engagement closes. It does not require that anything outside the Individual be reached. Where the deficit is a condition of the Individual's own body, the body is the Object, and the action corrects the deficit directly: the Individual scratches the itch, shifts off the numbed limb, withdraws the hand from the heat. Nothing is acquired, nothing is obtained, and no third thing stands between the effector and the correction. The chain of actions just described is what the architecture does where the correcting condition is not already to hand, where the Object that closes the deficit lies at a distance, is held by another, or is itself gated, and the length of that chain is a fact about the deficit's circumstances rather than a condition on action. The limiting case, in which the effector reaches the correcting condition with no intermediary at all, is not a degenerate instance of the architecture but its shortest complete run.

The word is exact rather than figurative. Transduction is the conversion of a signal from one medium into another with its pattern preserved, and that is what the selected action performs. The integrated Scalar is a signal inside the body, and nothing moves because of it until the action converts it into force applied to an Object. A loudspeaker performs the same operation on an audio signal. The signal arriving at it is a varying voltage, and a voltage moves nothing in a room. The coil converts that voltage into motion of the cone, the cone vibrates the air, and the travelling pressure waves are the sound. The voltage stays in the device. What is now in the room is the same signal, in a medium that can propagate. Each parcel of air displaces the next, and the pattern is what travels, until it reaches a listener or dissipates.

So here: the neural Scalar stays in the body and is re-emitted for as long as the deficit stands. What occupies the Object is that same signal, now in a medium that admits of motion, which is why the Object thereafter carries the signal rather than standing for it. The Objects that follow are not further speakers. They are the air: each carries the signal, once transduced, and hands it on where it is itself gated, until the Object that satisfies the deficit is reached.

The same image carries the integration and its cost. An orchestra recorded to a single channel reaches the loudspeaker as one voltage. The violin and the timpani are both in it, summed, and neither is labelled: probe the cable and there is one quantity varying over time, with no part of it marked as belonging to either instrument. That is the shape of what the Decision Hub does to A, I and M, and it is why the Individual reads a magnitude and never a provenance. Then the cone vibrates the air and the listener in the room hears the violin. Nothing was restored to the cable, and nothing needed to be. What changed is that the pattern is now in a medium that can be attended to, and attending to a medium is a different operation from reading a signal.

Behaviour stands in the same relation to the sources, and it is wider than action. The action is where the signal enters a medium that can be observed, and once it is there it takes on a shape across time and context that no single value of the Scalar it came from contains. That shape is not read off the action alone. It is read off the interval the action sits in, and the informative part of the interval is frequently the part in which nothing was done: whether the pursuit stops when the Object is obtained, whether it survives the removal of an audience, whether it resets at the same remove once the Object is held. Each of those is a comparison between an action taken and an action not taken, so the channel they occupy is a channel of behaviour rather than of action, and an Individual asleep, or declining, or waiting out a stretch with no audience is supplying it. Those are the Behavioural Signatures, and reading them is not the reduction run in reverse. The summands remain unrecoverable from the sum, as 4.3.3 establishes and the Data Processing Inequality guarantees; the second channel the Signatures occupy is available because transduction opened it, and it is wide because behaviour extends past the moments transduction occupies.

Mapping onto the AIM instantiation. In the AIM Framework, the Decision Hub is localised in the human substrate to the ventromedial prefrontal cortex and the ventral striatum, per the foundational citations to Axiom 2. The weighted-sum approximation of the Decision Hub's integration, in which the integrated value is the sum of the Appetitive, Intrinsic and Mimetic valuations of the Object each multiplied by its weight, with the weights state-dependent and summing to one, is the AIM-level rendering of the minimal architecture's Common-Currency Scalar. The weighted-sum approximation is load-bearing for prediction under ordinary conditions and is sufficient for the derivations the treatise requires. Finer specifications (non-linear interactions, context-dependent weight adjustment, temporal dynamics within a single choice episode) are compatible with the weighted-sum approximation as first-order descriptions and are developed where downstream chapters require them.

The Integration Equation's surface symmetry (three sources contributing on equal terms to a weighted sum) is surface only. The three weighted terms reflect three channels of direct input to the Decision Hub, not three symmetric sources of influence over the integrated value. Why that matters, and what the resulting asymmetry implies, is developed after the remaining components have been specified, in 4.4.12.

A note on individual variation. The weights an Individual assigns to A, I, and M inputs at the Decision Hub vary across the population, and this variation accommodates what is commonly termed neurodivergence without requiring the weights to be explained by reference to any partition between innate and acquired contributions. The neural firing sequences that produce the weights are the joint product of all prior genetic expression and all prior input, processed through a connectome that is itself that same joint product. The partition the nature–nurture dichotomy presupposes is not well-defined at the firing-sequence level and therefore not well-defined at any level that supervenes on it. Chapter 9 develops this point in the context of the architectural consequences of Common-Currency Integration.

4.4.2 The Body Manager (Homeostatic-Deficit Signaller)

Axiom grounding. The Body Manager is the direct architectural rendering of the A-component of Axiom 1 (Three-Source Taxonomy). Axiom 6 (Differential Satiation Dynamics) establishes the Body Manager's termination condition: consumption or repair satiates, and the signal cycles with the underlying deficit.

Function-only specification. The Body Manager is the component whose primary function is homeostatic deficit signalling and deficit-repair termination. It produces valuation signals scaled to the deficit states of the body or body-analogous system it monitors, and it terminates those signals on consumption or repair.

The Body Manager generates deficit signals indexed to internal physiological state. Signal magnitude tracks deficit magnitude. The Body Manager terminates on consumption or repair. Among the deficits the Body Manager signals is the state of the body's energy budget itself. The drive to breathe reports the oxygen supply, and with it the clearance of carbon dioxide, on which every expenditure runs, and it falls due within minutes; hunger, fatigue and sleep pressure report the envelope running down over hours and days. The Body Manager reports the supply and does not allocate it. Allocation is the Decision Hub's selection, and the dominance of elevated A-signals at the Decision Hub described below is the budget protecting its own supply. This termination is the canonical closed-loop termination condition in the minimal architecture, the reference case against which the other components' termination dynamics are specified. The Body Manager carries high legibility phenomenology: its deficits are registered in the system as sensations that the system's conscious-cognition component can detect directly. This legibility constrains confabulation about the presence of a deficit, though not about the source or cause of the deficit.

Body Manager signals are signed per the Signed Valuation Principle (4.4.0): they take positive values when a current internal deficit is in force (hunger, thirst, cold) and negative values when a state already inside the body is itself aversive (nausea after ingestion, heat already on the skin, pain from tissue already damaged). The same channel carries both. The Body Manager does not index those signals to Objects that have not contacted the body. Anticipatory avoidance of an Object still in the field, spoiled food, a predator still outside contact, a tree falling toward the Individual, is Observer-signed Salience conditioned by the Thinker's priors as read from the Belief Register, including priors about what will damage the body if it is not avoided. That signal is fear under Phenomenological Collapse (Chapter 2). The Body Manager does not issue it.

Consumption and repair are two terminations rather than one. Consumption closes a deficit by taking in what the Individual did not have: food, water, warmth drawn from a source beyond the body. Repair closes a deficit by restoring a bodily condition, and the effectors can frequently do that unaided, the itch scratched, the numbed limb shifted, the deeper breath drawn, the eyes closed against the glare. The deficit is registered, the action is selected, the effector engages, the signal terminates, and nothing is acquired at any point in the cycle. Repair is therefore the case against which the requirements of action are most economically read. An action requires a valuation, a threshold crossing on it, and engagement of an effector, and it requires nothing further. That an Object beyond the Individual must be reached is a contingent fact about which deficits the body is carrying, not a fourth requirement. Where the correcting condition is not already to hand, the chain of actions lengthens through intermediaries as 4.4.1 sets out; where it is, the correction is a single action.

The Body Manager requires no priors to perform its function. Deficit signalling is generated by the body's physiological state, not by inference, and therefore has no Hypothesis Space that priors could inform. Its signalling can be modulated by learned information already inside the body; what it lacks is the deliberative cognitive apparatus, which is the Thinker's. What the architecture does not assign to the Body Manager is the classification of an Object in the field as a future threat. That classification is a Thinker prior, delivered as signed Salience by the Observer. The abstraction claims the component's functional role, not that the underlying substrate is a single anatomical module or that it carries no learned or anticipatory signal. The Body Manager accordingly has no upstream channel: no component sends templates, priors, or instructions to it, because the Body Manager is not performing the kind of operation that templates or priors could assist. The Body Manager also has no outward template-supply function: it does not supply priors to the Thinker, because it does not maintain priors, and it does not supply priors to the Observer, because prior-supply is the Thinker's function. The Body Manager's sole channel of influence over the integrated value is its direct output to the Decision Hub.

The consequence for the system as a whole is that the Body Manager is structurally incapable of correcting drift in the Thinker or the Observer. When the Thinker's priors have been configured such that engagement with the Body Manager's deficits is discounted (the fast, the vigil, the ordeal, the Glory-template's reframing of bodily sacrifice as honour), the Body Manager continues to fire at full magnitude, but its signal can be outvoted at the Decision Hub by the Thinker's direct input and the Thinker's upstream shaping of the Observer's input. The Body Manager cannot feed back into the Thinker to update the priors that are overriding it.

Mapping onto the AIM instantiation. In the AIM Framework, the Body Manager is A, Appetites, as defined in Axiom 1. The deficit states A signals include oxygen and the drive to breathe, hunger, thirst, thermoregulation, sleep pressure, tissue damage and reproductive drive, together with the bodily state that sustained stress produces when it shifts those set-points (Section 4.3.1). A's boundary is the body: material circumstances such as income, shelter, safety and healthcare reach A only through their bodily consequences, and what defines the category within that boundary is the satiation dynamic (once the deficit is corrected, the demand for deficit correction ends). Stress acts on the body the Body Manager reads, whatever set it off, the circumstance or the Thinker's appraisal of it. The route runs through the body and is not a channel; it is usually set off by circumstances the Individual does not wholly control, and it acts over hours and days rather than within a pass of the loop, which is why the formal treatment takes A as an exogenous input on the fast timescale (Appendix A, Section A.3).

When A-signals are elevated, A characteristically dominates the Decision Hub, and what it displaces is the I-process, not the Thinker. Identifying which Object would repair the deficit, and how to reach it, remains the Thinker's work (Section 4.4.3); what the Thinker stops doing is setting goal states for itself and pursuing them, the activities that are rewarding in the doing. Under deficit, finding food or medicine is provisioning, a process the Body Manager commissions, and it runs on the priors the Belief Register already holds about A-repair. The Thinker's serial deliberation is occupied by it, and the work that needs that line free is the work that goes: long-horizon planning, the systemic Hypotheses that only Imagination can assemble (Section 4.3.10), and creative search, which is recruited, if at all, to the deficit. A does not displace M. The Observer's process runs for as long as the Individual is conscious and perceiving (Section 4.4.7), a running-down envelope narrows the field it selects from without stopping it (Section 4.1), and under deficit the Salience it forms falls first on whatever within reach would repair the deficit, which is usually an inanimate Object rather than another Individual; its inflation of those Objects is Object Parasitism operating on A-Objects (Section 4.4.11). Imagination is closed from both sides: A occupies the Thinker's serial line, and the Observer, still active, withholds the quiescence Imagination requires (Section 4.4.4). This is not a failure of the Decision Hub but the minimal-architecture specification operating correctly, treating deficit-repair as the action-priority the deficit-scaled signal says it is. The dominance is conditional, however, on I's priors permitting A to dominate. Where I's priors have been culturally configured to discount A-signals, A's dominance can be overridden at the Decision Hub even when A-magnitude is large. The architectural basis for that override, and its consequences for Glory-culture, kamikaze, seppuku, martyrdom-for-name, and heroic last-stand mythology, is set out in 4.4.12 as the I-Template Override Result, once the Thinker and the Observer have been specified and the channel count on which the override depends has been established.

4.4.3 The Thinker (Prior-Maintainer and Bayesian Ground-Resolver)

Axiom grounding. The Thinker is the direct architectural rendering of the I-component of Axiom 1. Its ground-resolution function operates under Axiom 8 (Bayesian Belief Dynamics), which supplies the updating rule. Axiom 4 (Confabulation) is the characteristic output of the Thinker when its Hypothesis Space is impoverished. Axiom 9 (Agent-Based Causal Default) constrains the form the confabulated output takes when the ground-resolution reaches for a substitute attribution.

Function-only specification. The Thinker is the component whose two architecturally inseparable primary functions are, first, continuous maintenance of the priors that make assessment, recognition, and resolution possible for the whole system, and, second, Ground-Resolution, the evaluative operation that resolves whatever the Thinker is attending to into its best-supported explanation by redistributing probability mass across the Hypothesis Space, of which the resolution of the Decision Hub's tag-stripped Scalar into a ground-attribution is the source-attribution special case. The Thinker is the locus of deliberative inference in the architecture, and its operation is one contribution to Conscious Cognition rather than the whole of it: the Thinker is what we mean when we speak of thinking, and the deliberative work of inference, Hypothesis-formation, vocabulary acquisition, and revision of Belief is what the Thinker does. The Thinker's capacity is serial: the Thinker holds one line of Ground-Resolution at a time, within a working-memory capacity distinct from the Observer's attention (Section 4.4.7). That capacity limits what the Thinker can resolve at once. It is not a reservoir that effort exhausts (Hagger et al., 2016), and the fatigue that degrades deliberation is the body's energy budget running down, signalled by the Body Manager (Section 4.4.2).

Ordinary sensory input, including information received from other Individuals, reaches the Thinker at its deliberative timescale and is not routed through the Decision Hub. Speech, reading, inspection, and the rest of realtime intake remain the channels any account of perception and inference already grants. The architecture assumes those channels and does not contest them. What it specifies in addition is the Decision Hub path: the channel by which integrated valuation returns to the Thinker as source-opaque feedback. That path interacts with the ordinary architecture. The AIM Framework's work is to bring the additional channel into the same accounting as the channels already granted, not to replace them.

The Thinker's ground-resolution function is not invoked on every action-selection cycle. The Thinker is associated with two stores, and they do different work. The Memory Store is the inclusive record of encoded content, retaining episodic traces and semantic content without commitment, holding contradictory fragments, dormant facts and propositions never endorsed alongside those currently believed. The Belief Register holds what the Thinker currently commits to, and is the only one of the two that downstream components read. The Hypothesis Space over which Ground-Resolution runs is constituted from the Memory Store, which fixes a limit the architecture cannot circumvent: a Hypothesis whose constituent content is absent from the Memory Store cannot be formed at all, and is therefore not improbable but unavailable. Neither store is read by the Decision Hub.

Where the Decision Hub's cached-value (habit) store already holds a value adequate to the current context, action-selection proceeds at habit timescales on that store without recruiting the Thinker. The Thinker is recruited only by an open question: an Object the Belief Register cannot already resolve, either because no committed posterior exists for the Object or because the context diverges from what the stored posterior predicted. The Object need not arrive from the perceptual field. An Object the Thinker has assembled from the Memory Store and holds unresolved is an open question on the same terms as one perception delivers, so the activation condition can be met from within the Thinker as well as from outside it, and no problem posed by the world is required. A problem is the special case of an open question whose resolution bears on action. Novelty, prediction-mismatch and the self-set question are therefore the Thinker's activation condition rather than incidental inputs; Curiosity is the felt signature of that recruitment, and Boredom is the felt registration of its exhaustion.

The prior-maintenance function stores and continuously updates four classes of prior. Priors about the Body Manager are the phenomenological recognition templates for its signals (what each deficit feels like, when each is to be honoured versus endured or suppressed), the Object-classes that repair each deficit, and the causal pathways by which acquisition and consumption produce repair. These priors are held by the Thinker for its own use in ground-resolution and in shaping the Observer's input, not exported to the Body Manager, which does not require them. Priors about itself are the record of which engagements deepen capacity, which Objects the Thinker has previously resolved as engagement-rewarding, which domains the system has developed engagement in, and which activities produce flow. Priors about the Observer are the identification of which Individuals count as Models, Reference Set composition, comparative-position templates, Model-valuation-tracking templates, and the Hypothesis that a given Salience may be Observer-tracking of comparative position rather than a Body Manager deficit or a deficit of what the Thinker's process requires. Priors about the Object field are the categorical vocabulary through which perceptual input is parsed into candidate Objects in the first place.

Three of these four prior-classes remain internal to the Thinker's own operations. The priors about the Observer, however, are supplied outward to the Observer as the templates that condition the Observer's computation. This outward template-supply is the Thinker's sole channel of upstream influence over another component's operation, and it is load-bearing for the whole architecture. The Observer is not an independent Object-field evaluator; its computation is conditional on the Thinker-supplied priors, and the signal it produces therefore reflects both the Observer's distinctive input stream (observed pursuits, observed valuations, comparative-position data) and the Thinker's templates applied to that stream. When the Thinker's priors about Models and Reference Sets and Object-Salience are culturally configured in specific ways, the Observer's output reflects those configurations upstream of the Decision Hub, not merely the raw observation data.

The Bayesian ground-resolution function receives the tag-stripped integrated Scalar from the Decision Hub as an item of Evidence and performs Bayesian updating over the Thinker's Hypothesis Space to return a ground-attribution, which is fed back to the Decision Hub for action-initiation. The Scalar enters that updating on the same footing as any other Evidence reaching the Thinker, because the rule stated at Axiom 8 indexes on neither the origin of an item nor its subject matter, so the architecture holds no operation reserved for motivational feedback. The Object the Scalar concerns reaches the Thinker through its own perceptual intake, which is not routed through the Decision Hub. What does not reach it is the Observer's reading of that Object, the Observer's only output going to the Decision Hub and being tag-stripped there. The ground-resolution is performed necessarily on incomplete data, because the tag-stripping at the Decision Hub has already destroyed the source-identity information that perfect ground-resolution would require.

The quality of the resolution depends on the quality of the Hypothesis Space the Thinker holds. When the Hypothesis Space includes an adequate Hypothesis about Observer-driven Salience, Salience of that kind is available to be recognised as such. When the Hypothesis Space does not include that Hypothesis, Observer-driven Salience is resolved to the nearest available ground-attribution, characteristically a Body Manager deficit or a deficit of what its own process requires, and the system acts on an attribution that is high-confidence and wrong. This is not a cognitive failure, it is Bayesian reasoning operating correctly on an impoverished Hypothesis Space.

When resolution returns an I ground, the Thinker sustains engagement with the Object as a self-sustaining process that deepens with engagement and terminates only through endogenous termination (when novelty and capacity-deepening are exhausted, per Axiom 6) or through interruption (typically by an intruding deficit signal or by another system). Thinker valuations are signed per the Signed Valuation Principle (4.4.0): they take positive values toward engagements the Thinker resolves as deepening, capacity-building, or value-affirming, and negative values toward engagements the Thinker resolves as boring (Axiom 6 termination signal), value-violating, epistemically threatening, or otherwise aversive on Thinker-relevant grounds. The same channel carries both, with sign determined by the resolution the Thinker returns.

That the I-process does not satiate is entailed by the prior-maintenance and ground-resolution function itself, the function operates continuously and in private, and the deficits it signals are for the Objects whatever process it is running requires rather than for the function itself, so correcting one closes that deficit and leaves the function running, and non-satiation is architecturally guaranteed by the function specification, not an additional empirical claim, and the satiation of I-enabling Objects is the complementary entailment, the Object is contingent on the task and terminates with it while the process is not and does not. The consolidation of this point with the Body Manager's and the Observer's termination dynamics is the Satiation-Function Entailment of 4.4.12.

Three features of the Thinker jointly determine the whole system's failure mode and its intervention target. These features are stated here; their joint consequences are drawn out in 4.4.12.

First, the Thinker is the sole prior-maintaining component in the architecture. The Body Manager does not bear priors. The Observer does not bear priors of its own; it draws on the Thinker's priors as cached for downstream access (per 4.4.8). The Decision Hub does not bear priors. All cognitive content of any kind (priors, templates, Hypothesis Space, categorical vocabulary, causal knowledge) is maintained by the Thinker, and what the Belief Register holds is what the Thinker has committed there. No other component maintains priors that could be modified by external input or by the system's own experience.

Second, the motivational feedback on which the Thinker's ground-resolution function operates arrives only through the Decision Hub's tag-stripped Scalar. Source-identity information has already been destroyed at the moment this feedback enters the Thinker's ground-resolution function. The rule the feedback enters is content-agnostic in the sense Axiom 8 specifies, and the feedback itself is source-opaque: it reports that the integrated value was high or low, not which source drove it. The Thinker therefore cannot, from the feedback alone, distinguish a high Scalar driven by A-signal from a high Scalar driven by M-signal-shaped-by-I-priors, or distinguish a correct ground-attribution from a confabulated one that merely produces a consistent Scalar on the next iteration. The Confabulation architecture is, at this level, Bayesian updating on feedback from which the decomposition has been structurally removed, so that detection must run through a Hypothesis Space holding the M-Hypothesis.

Third, no feedback channel exists from the Body Manager into the Thinker's prior-maintenance function, and no feedback channel exists from the Observer into that function either. The Body Manager cannot update the Thinker's priors about itself, because the Body Manager does not bear priors of its own from which to send corrections. The Observer cannot update the Thinker's priors about itself, because the Observer is downstream of those priors and its output is therefore not independent evidence against them. The only input channel the Thinker has from its own architecture for updating its priors about which source drove an integrated valuation is the Decision Hub's Scalar, alongside the recall of its own prior state, and that feedback is source-opaque. Further Evidence continues to reach the Thinker through sensory input, including information received from other Individuals, and none of it carries the destroyed decomposition either.

The Thinker does not detect Salience in the perceptual field. Salience detection is the Observer's function. It does not select actions. Action-selection is the Decision Hub's function. The Thinker maintains the priors the Observer uses and updates on the Evidence that reaches it, whether through sensory input, including information received from other Individuals, or through the Decision Hub's Scalar, the resolution of which into a ground-attribution is that same updating applied to the source question.

What the Thinker does signal is deficit. Axiom 3 establishes that the Thinker encodes a deficit of the same form as the Body Manager's, the absence of an Object its process requires, and the Signed Valuation Principle at 4.4.0 has that deficit entering the Common-Currency Scalar alongside the Body Manager's. The two components differ in what closes the deficit and in whether it recurs, not in whether they signal one. Deficit-signalling is not a third Thinker function: Ground-Resolution, applied to what the process the Thinker is running requires, returns the absence of a required Object as its result, and that result is the signal.

Mapping onto the AIM instantiation. In the AIM Framework, the Thinker is I, Intrinsic Motivation, as defined in Axiom 1 and elaborated in subsequent chapters. The engagement-reward phenomenon that Self-Determination Theory and flow research have measured and described is the observable output of I's ground-resolution function operating in cases where the ground returns as I, with the accompanying deepening-through-engagement dynamic and termination-by-boredom-or-interruption pattern. I's Private Persistence, the hallmark diagnostic for I-driven engagement being the activity's continuation when no audience is present and no external reward is contingent on performance, follows from the ground-resolution specification: once the ground is resolved as I, engagement is self-sustaining and requires no external reinforcement. I-enabling Objects (tools, instruments, access that the process requires) satiate upon functional adequacy, because the deficit the Thinker signals for them closes when they are obtained, not because the I-process itself satiates. The Body Manager's deficit recurs on a physiological cycle; the Thinker's does not recur once the process has what it requires. That difference in recurrence, and not any difference in whether a deficit is signalled at all, is what separates the two satiation profiles.

Because the Thinker's activation condition is novelty and prediction-mismatch, and because M-driven Salience presents a prediction-mismatch that no consumption or repair discharges, the recruitment does not end with a single resolution: the ground-resolution function is run repeatedly against an explanandum it cannot close, and the candidate set expands rather than converging. That consequence is the Confabulation Proliferation Corollary, derived at Chapter 11, Section 11.2.1.

Confabulation in AIM's sense is the ground-attribution output of I when the Hypothesis Space about M is impoverished: the confident-but-wrong attribution of M-driven Salience to A or I. Mimetic Desire (the felt output) and Confabulation (the ground-attribution output) are the same emergent phenomenon viewed from different angles, both arising when I's Bayesian ground-resolution operates on the tag-stripped integrated Scalar with a Hypothesis Space that does not adequately include the M-Hypothesis. The Knowledge Patch in AIM operates on I's priors about M because I is the only component whose priors are architecturally modifiable. The architectural reasons the Patch must operate on I and must be externally installed are developed in 4.4.12.

4.4.4 The Observer-Quiescence Corollary and the Architectural Basis of Thinker Privacy

Axiom grounding. The Observer-Quiescence Corollary is jointly entailed by Axiom 8, Bayesian Belief Dynamics, as it applies specifically to Imagination, the generative operation of the Thinker, and by the Magnitude Clause of the Minimum Installation Condition set out in 4.4.7.1, which conditions the Observer's Salience weighting on the Thinker's cached priors as delivered through the Belief Register.

Function-only specification. The Thinker's ground-resolution function, specified in 4.4.3, presupposes a prior generative step: before probability mass can be redistributed across candidate Hypotheses, those candidates must first be constructed. This generative step is Imagination, and its output determines the boundary of what Bayesian ground-resolution can subsequently operate on. The Observer's Salience output, conditioned on the Belief Register's current state per 4.4.7 and 4.4.8, does not only weight Objects for action-selection at the Decision Hub. Because the Thinker's own Hypothesis-construction draws on the same cached priors that condition the Observer, elevated Observer-channel activity at a given moment reflects a prior landscape, and through the Scalar supplies Evidence that tends to reinforce it, weighted toward Model-relevant, comparative-position-relevant candidates. Under these conditions the generative operation's search is disproportionately populated by status-legible Hypotheses, and Hypotheses with no comparative-position payoff are not constructed at all, so Ground-Resolution never has them to act on. Under low Observer activity, this weighting relaxes, and the generative operation is free to assign non-trivial amplitude to Hypotheses that carry no Model precedent and no status payoff, including Hypotheses that describe the Individual's own social field systemically rather than through a single legible target. This is the Observer-Quiescence Corollary: low Observer activity is the architectural precondition for the construction of systemic Hypotheses, not a correlate of it.

Section 4.4.3 specifies the Thinker's activation condition as an open question, whether presented by the perceptual field or set from the Memory Store, and the generative operation inherits that condition. An Object the generative operation has itself assembled from the Memory Store occupies the Hypothesis Space on the same footing as an Object perception delivers. Where the Belief Register holds no posterior adequate to a question the Individual has set, the search for one is what Imagination performs, so recruitment of Imagination may be endogenous as well as exogenous.

Section 4.4.3 further specifies Ground-Resolution as returning the most probable Hypothesis given the available Hypothesis Space and the evidence, under an updating rule that does not index on what a Hypothesis is about. Content-agnosticism at the evaluative half is what produces Confabulation, because the operation cannot detect that the true Hypothesis is absent from the Hypothesis Space. The generative half inherits the same property: Hypothesis construction draws on the Knowledge the Memory Store holds and proceeds under one rule, whether the assembled Hypothesis describes a state of affairs that can obtain or one that cannot. Constructing a candidate answer to a question the world poses and constructing a candidate state of affairs that could never obtain are therefore one operation running one rule, and the architecture holds no second generative capacity for the second case.

The two cases diverge downstream of the generative operation and at no point within it. They diverge in reach, because Imagination is bounded by the Memory Store and a Hypothesis whose constituent Knowledge the Memory Store does not hold is unavailable rather than improbable, so two explananda drawing on different Knowledge meet that bound at different points. They diverge in how available the evidence bearing on them is, since Ground-Resolution scores on what reaches the Thinker, and a constructed Hypothesis whose disconfirming evidence is out of circulation is scored on whatever else reaches it, which is typically what other Individuals appear to hold. They diverge in termination, since an explanandum the available evidence can settle terminates by resolution, while one it cannot terminates by Boredom or by interruption, which 4.4.3 specifies as the terminating conditions of the I-process.

Why Thinker privacy is architecturally required. Section 4.4.3 established that I's Private Persistence is the hallmark diagnostic for I-driven engagement, because once a ground resolves as I, engagement is self-sustaining and requires no external reinforcement. The Observer-Quiescence Corollary supplies the missing mechanism for why the Thinker's highest-value generative function characteristically operates behind that same privacy boundary. Audience presence does not merely fail to add anything to Imagination — it actively narrows the space Imagination can search, because an observed Individual carries elevated live Observer Salience conditioned on Model-tracking and comparative-position templates, and that elevated Salience, through the Scalar it loads, feeds back into which Hypotheses the generative operation constructs with meaningful amplitude in the first place. Privacy is therefore not a contingent condition under which the Thinker happens to prefer to work. It is the state in which the Observer-conditioned pre-filter on Hypothesis construction is at its weakest, and it is for this reason a structural precondition for the Thinker's capacity to generate genuinely novel or systemic Hypotheses, rather than an incidental feature of how creative work is customarily performed. This result is additive to, and independent of, the Reafference Pathway specified in 4.4.9.1: the Pathway explains why self-observation continues to reinforce wanting even in solitude, while the Observer-Quiescence Corollary explains why solitude nonetheless remains the precondition for the generative operation's widest search, because the Pathway's residual self-observation carries lower Model-density and weaker comparative-position content than externally audienced observation, and therefore imposes a correspondingly weaker pre-filter on Hypothesis construction.

Mapping onto the AIM instantiation. In the AIM Framework, this Corollary explains why imagination, creativity, and problem-solving present as a single private function of I rather than as three separate capacities. Each is an instance of the Thinker's generative operation constructing candidate Hypotheses under conditions of low M-conditioned pre-filtering, followed by Bayesian Updating resolving among them. The novelist drafting alone, the scientist working through a problem with no one watching, and the individual quietly reconsidering a settled Belief are all instances of the same architectural event: Observer-conditioned Salience weighting is quiescent, the generative operation's search is unconstrained by status-legibility, and the resulting candidate Hypotheses are available for Bayesian Updating to test against evidence on their merits rather than on their comparative-position payoff. The novelist and the scientist are not two capacities sharing one condition of privacy, and neither is the Individual assembling a state of affairs that cannot obtain: each is the generative operation constructing candidate Hypotheses from the Memory Store against an explanandum the Individual has set, and what separates them is the closure available downstream rather than the operation upstream.

4.4.5 Flow as the Low-Crowding Limit of Observer-Driven Crowding-Out

Section 4.4.4 established the Observer-Quiescence Corollary: low Observer activity is the architectural precondition for the Thinker's generative operation, Imagination, to conduct its goal-directed search of the Hypothesis Space without status-conditioned pre-filtering. This section shows that the quiescence in question is not a distinct architectural event but the low-intensity end of a single continuous process — the crowding-out of the Thinker's generative search by the Observer's read of observable valuing agents — and that flow is simply the name for the state at that end of the continuum.

The one process, stated as a continuum. On any operating cycle the Observer continuously samples the field for valuing agents and, on each perception of another Individual exhibiting directional valence toward an Object, installs a Mimetic signal (the Minimum Installation Condition, 4.4.7.1) whose Salience is conditioned by the priors the Thinker holds in the Belief Register. Under the Configural Update-Rate Asymmetry (4.3.10), agents generate a categorically higher rate of discrete, Salient configuration-change events than static objects or diffuse systems, so the Observer's sampling load scales with the number and Salience of observable valuing agents in the field. This sampling-and-updating load competes with the Thinker's generative operation at the Decision Hub, the single channel through which both are spent (Section 4.1): where the Observer's Salience dominates the Scalar, the Decision Hub's selection follows it and the Thinker is returned that Scalar to resolve. The load is therefore a dial, not a switch: as the density and Salience of observable valuing agents rises, the Observer's read progressively crowds out the Thinker's goal-directed search; as it falls, the search is progressively freed. There are not two states here and not two mechanisms — there is one process running at varying intensity.

Flow is the low-crowding limit. Flow is the state at the low-intensity end of this continuum: the condition in which few observable valuing agents are present in the field, or the Observer is rendered low-Salience with respect to those present, so that Observer-driven crowding-out of the Thinker's generative search approaches its minimum. Being in flow is, precisely, the Thinker not being crowded out by the Observer's sampling of observed agents. This is why flow is reliably associated with clear goals and a skill-challenge match rather than with aimlessness (Csikszentmihalyi, 1990): Imagination remains the Thinker's goal-directed generative operation in service of ground-resolution throughout, so the task's own structure continues to target the search; what the low-crowding condition removes is only the competing Observer load, not the search's aim.

The two signature features are one de-crowding, described twice. Flow is distinguished by two of its most robust reported features: a loss of self-consciousness, and a marked compression or apparent disappearance of subjective time (Csikszentmihalyi, 1990). AIM's architecture identifies these not as two independently-caused symptoms but as two descriptions of the single low-crowding condition. Loss of self-consciousness is the crowding-out process at its minimum on the self-directed channel: the Observer's comparative-position, status-conditioned read — including its read of the self as an observed, socially-rankable agent — falls quiet, which is what the recent active-inference literature independently characterises as the inhibition of higher-order self-conceptualisation with pre-reflective bodily awareness preserved (Parvizi-Wayne et al., 2024). Time compression is the same minimum, described on the duration-tracking channel: subjective duration tracks the density of discrete, attention-worthy events registered per interval rather than clock-time (Block and Grondin, 2014), and Event Segmentation Theory establishes that the prototypical event-boundary is a change in an agent's goal, action, or configuration (Zacks and Swallow, 2007) — the very event-stream the Configural Update-Rate Asymmetry identifies as the Observer's dominant input. When the Observer's read of observable agents falls to its minimum, the event-stream that duration-perception counts falls with it, and felt time compresses. The loss of self-consciousness and the loss of felt time are therefore the same event — the Observer's read of observable valuing agents dropping toward zero — registered on two different downstream channels.

Relation to crowding-out at other scales. This is the individual-architectural, real-time form of the crowding-out already named at other scales in the framework: Motivation Crowding-Out, the empirical suppression of intrinsic-reward activity by controlling extrinsic reward, and the I-Crowding-Out Dynamic, the sectoral displacement of I-engagement by M-intensive activity. In each case the Mimetic channel's operation displaces intrinsic engagement; flow is simply the limiting case in which that displacement, at the level of the single Operational Loop, approaches zero. Naming flow this way removes a redundant construct: it is not a separate phenomenon requiring its own corollary but the low-crowding boundary condition of the Observer-Quiescence Corollary itself.

4.4.6 How the Order of Experience Shapes the Thinker

The Thinker's ground-resolution has a further property, and it follows directly from what has already been said.

At any given moment, the Thinker is usually holding more than one Hypothesis about the same subject matter. It may be weighing whether a pull toward a new car is about a genuine transport need, about the enjoyment of driving, or about how the car will be seen by others. It may be weighing whether a colleague's coolness reflects something the Individual has done, something going on in the colleague's own life, or nothing at all. The Hypotheses are not always in strict competition (they often describe different sides of the same situation), but they are acting on the same subject matter, and each carries some probability that the Thinker updates as new evidence comes in.

The property is this. When two or more Hypotheses are acting on the same subject matter, the order in which evidence arrives changes where the Thinker ends up. Bayesian updating moves probability from one Hypothesis to another with each new piece of information. Each move reshapes the landscape for the next piece. An observation that would have strongly supported one Hypothesis if it arrived first can have a much smaller effect, or even the opposite effect, if it arrives after other observations have already shifted the distribution. Two people can start from the same priors, receive exactly the same evidence over their lives, and end up with different Beliefs simply because the evidence reached them in different sequences.

Four consequences follow.

The order of experience matters, not just the content. What the Individual encounters first, second, and third changes the Belief they end up with, even when the total body of evidence is identical.
Differences between Individuals are architecturally expected. Each person moves through the world in a personal order, meeting particular Models, Objects, and social responses in a sequence no one else shares. Divergence of Belief across a population is the normal output of the Thinker's operation, not a sign that some people are reasoning badly.
Competing Hypotheses interact rather than update in isolation. The weight the Thinker ends up placing on any one Hypothesis depends on what has been happening to the others along the way.
The Knowledge Patch works with this property rather than against it. Epistemic Suspension (Chapter 18) is the practice of holding more than one Hypothesis open for longer than the Thinker would naturally tolerate, so that the Individual's Beliefs are less dominated by the accident of early experience. The Patch does not remove order-sensitivity (that is architecturally given); it makes room for more evidence to bear on the outcome before the Thinker collapses onto a single answer.

The consequences for Confabulation are direct. Under an impoverished Hypothesis Space, Confabulation resolves Salience onto a wrong ground with high confidence. Order-sensitivity is the reason that two Individuals sharing the same impoverished space will confabulate onto different specific grounds, and the reason that a single Individual's confabulated ground can shift across time without any new evidence arriving. Where unpatched Individuals do land on the same confabulated ground, the convergence reflects shared sequences of experience (shared Models, shared Reference Sets, shared cached priors carried by the Observer's reafference pathway and the Belief Register's slow updating), rather than accurate resolution.

None of this claims the Thinker is departing from Bayesian updating. Strict conditionalisation on a fixed evidence set with fixed likelihoods is order-invariant: the same total evidence yields the same posterior in either order. Order-dependence enters because the evidence an observation yields is state-dependent: the Scalar on which the Thinker updates is itself shaped by the Belief-state the observation meets, which is the accumulated product of what came before, through the recall that re-enters prior belief as Evidence and through the templates under which the Observer computes (Appendix A, Section A.5.6). The likelihood the Thinker assigns stays content-agnostic, which is what keeps the Belief race monotone. That state-dependence is architectural (Source Opacity and Common-Currency Integration), not a failure of Bayesian discipline, and it is developed in Chapters 13 and 14, with the formal statement at Chapter 14, Section 14.9.

4.4.7 The Observer (Continuous Conditional Object-Field Salience Assessor)

The Perceptual Primacy of the Observer. The Observer is the general perception channel of the architecture, not a dedicated social faculty. Its continuous conditional Salience assessment runs over the entire parsed Object field, tracking the directional movement and Salience of any Object for action-guidance, the branch moved by the wind, the vehicle through the corner, the cup on the table, the tree falling toward the Individual. Mimetic installation is the special case entailed when the perceived field contains an Individual whose conduct toward another Object carries valence, because a channel that tracks directional movement in general necessarily tracks the directional valence of other Individuals by the same apparatus, other Individuals being Objects in the field whose movement is tracked for the same action-guiding reason as any other Object. Adopted desire is therefore architecturally entailed and requires no separate channel: a system that perceived directional movement in general but selectively failed to perceive the directional valence of agents would require a second, agent-blind perceptual channel the architecture does not contain. The general perceptual function includes threat-priors about Objects that have no agency: a falling tree is read against the Thinker's prior that unmanaged mass will damage the body, and the Observer emits negative Salience without a Model in the field. The Minimum Installation Condition specified in 4.4.7.1 is the social special case of this general perceptual function.

Axiom grounding. The Observer is the direct architectural rendering of the M-component of Axiom 1. Axiom 5 (Object Parasitism) describes the Observer's operating mode: it does not generate its own Objects but operates on the Object-field parsed through the Thinker's categorical priors, attaching Salience to Objects that already carry A- or I-functional roles. Axiom 7 (Preconscious Transmission) specifies the timing within which the Observer operates in the human substrate, the 60 to 340 millisecond window before conscious awareness.

Function-only specification. The Observer is the component whose primary function is continuous conditional Salience assessment over the perceived Object field, conditioned on priors maintained by the Thinker. The reading is locative as well as scalar: it marks which Objects in the field are valued and where in the field each one sits, which is what allows a deficit signal to be pointed at a target. The Observer's distinctive contribution is the input stream that no other component receives, and its only output goes to the Decision Hub. The reading reaches neither the Body Manager, which has no upstream channel from the other components of the MAS, nor the Thinker, which receives the integrated Scalar and not the reading that was one of its inputs. That is what produces the error modes: a report of the specific magnitude of the Salience signal is destroyed at integration rather than delivered onward, and nothing downstream can recover it from what survives.

Rendered at the level of the Decision Hub's inputs, the Observer's output is the Salience signal: it enters common-currency integration as one signed Salience contribution alongside the Body Manager's deficit signal (A) and the Thinker's deficit signal (I). Because that Salience contribution is integrated into a Scalar whose function is the selection of corrective action, it is necessarily rendered as, and experienced as, the correction its sign demands rather than as Salience — the Salience-Deficit Conflation Principle — which is why M-driven wanting presents phenomenologically as felt lack indistinguishable from a genuine need, and M-driven aversion as felt threat indistinguishable from a genuine danger.

The Observer operates continuously whenever the system is conscious and perceiving. It takes five input streams: the perceived Object field parsed through the Thinker's categorical priors; the Thinker's priors about the Body Manager's current deficit state (which permit weighting by current deficit relevance); the Thinker's priors about its own current deficit state, the Objects whatever process it is running requires; the Thinker's priors about the Observer itself (Models, Reference Sets, comparative-position templates); and the Observer's distinctive input stream, which comprises three kinds of observation, first, what other Individuals are pursuing, what they appear to value, and where they sit relative to one another, second, the Individual's own action once it has been executed, as re-perceived after the fact, and third, how other Individuals react to the Individual's own executed action, including whether they now appear to value the same Object more or less and whether the Individual's comparative position has shifted as a result.

These three are architecturally distinct and must not be collapsed. The first is Model observation proper, the ignition stream carrying what others pursue and value. The second is Reafference: the Individual's own executed action re-perceived through the Observer, on the fast valuation timescale. The third is the Observability coupling: that same executed action, having been observed and valued by others, returning as an altered reading of the environment. The second and third both begin with the Individual's own executed action, which is why they are readily taken for one stream; they differ in whether the return path passes through another Individual. Reafference does not, and the Observability coupling does, which is why the latter is the pathway by which one Individual's realised valuation becomes another Individual's Salience. Neither is the memory self-coupling of the Belief Register, which is a slow-layer Evidence channel treated at 4.4.9.1: all three of these streams are Observer inputs, and the self-coupling is not.

The Observer outputs Salience and value signals over Objects, weighted by the joint contribution of the Thinker's priors and the Observer's distinctive input. Observer Salience signals are signed per the Signed Valuation Principle (4.4.0): they take positive values toward Objects flagged as Model-pursued, comparative-position-elevating, or otherwise positively M-charged in the cached templates, and negative values toward Objects flagged as Model-avoided, low-status, comparative-position-degrading, contaminated, or scapegoat-coded. The same channel carries both, with sign determined by the Object's position in the cached M-template field. Negative-valenced Observer signals are the architectural source of socially transmitted aversion (the heretic, the unclean, the scapegoat) and of the avoidance side of the derivations in 4.4.12. The same negative pole carries the general case in which the Thinker's prior is that the Object will damage the body if unmanaged, whether or not a Model is present.

Where the Object is another Individual, the flag on which the sign depends is the Agent-Type Vector, the classifier's four-coordinate weighting of that Individual-Object across Collaborator, Rival, Predator and Prey (Axiom 9; Chapter 11, Section 11.13.1). A Predator-weighted reading is the case in which the valuation returned is aversive, and a Prey-weighted reading the case in which it is extractive. The classifier operates on the Object field the Observer already reads and modulates the salience weight specified here (Appendix A, Section A.2.1). Where the Thinker already holds an Agent-Type hypothesis for that Individual-Object, the hypothesis reaches the Observer through the Belief Register on the channel Section 4.4.8 specifies. Where it holds none, the read runs on the inherited template the Belief Register supplies for Individuals of that presentation, and the Thinker's resolution of the Scalar that read contributed to commits the first hypothesis for that Individual-Object.

The scapegoat carries both signs. The same Individual-Object is read negatively as contaminated and positively as a target for elimination, and the two readings are the same signed channel run under two aspects rather than a contradiction between them. The negative reading is the one the cached template carries: it is why scapegoat-coding sits among the negative flags enumerated above, and what it governs is contact, association, and the untouchability of the figure. The positive reading is the one presentation supplies: a Prey-weighted vector returns an extractive valuation, and what the Observer emits toward a vulnerable, visible and reciprocity-exempt Individual-Object is approach. Pursuit and elimination run on the second aspect; avoidance of contact runs on the first.

Which aspect is in force on a given cycle is decided by where each coordinate of the Agent-Type Vector is computed, and the two loci run at different speeds. The Predator and Prey coordinates are the ones an Individual-Object's presentation supplies, so they move in the Observer's limb at the architecture's shortest time constant, read under templates the Thinker writes (Appendix A, Section A.5.9); the Collaborator and Rival coordinates require behavioural evidence accumulated across encounters, so they move in the Thinker's limb at its longest (Chapter 11, Section 11.13.1). Vulnerability is legible in a single exposure. Standing as a collaborator or a rival is not, and a figure whose Prey-weight has moved on presentation faster than any Collaborator-weight could accumulate is selected before the evidence that would have told against the selection could arrive.

Neither aspect leaves a trace the Individual can read, because the classifier's four-coordinate weighting reaches the Decision Hub as nothing richer than a modulation of the signed Salience weight (Appendix A, Section A.2.1). The Observer's output carries the sign and not the coordinate that produced it. The Thinker's ground-resolution therefore has no evidence by which to separate an elevation that arose from a Prey-weighted reading from one that arose from an aversive template, and the most probable available ground for approach toward an Individual-Object the Individual classifies as an enemy is a threat. The resolution returns Predator-framing for a selection that ran on Prey-weight. Dual-Role Confabulation is that output: selection on prey-criteria, framing on predator-criteria, held sincerely in one attribution. The diagnostic Dual-Role Confabulation supplies is developed at Chapter 11, Section 11.19.

The Observer has no termination condition. Conditional assessment over a perceived field cannot complete while perception continues and while the conditioning inputs (the Body Manager's deficit state, the Thinker's deficit state, social-world observations) continue to exist. The Observer therefore does not satiate; non-satiation is architecturally guaranteed by the function specification, not an additional empirical claim on top of it. This is the minimal-architecture rendering of Axiom 6's claim that M does not satiate.

The Observer's field is unbounded and its operation does not close, but the attention it spends on the field is finite. The body's energy budget is finite and part of it is spent holding attention on one thing rather than another (Section 4.1), so at any moment the Observer selects only so much of the field for assessment, and over any period it spends only so much. The two properties are distinct and must not be collapsed. Non-closure belongs to the Observer's operating condition: while the system is conscious and perceiving there is always further field to assess, and no reading of it is terminal. Finitude belongs to the Observer's supply: what is assessed is drawn from a fixed budget, so Salience placed on one Object is attention not placed on another. The budget is the Observer's own and is distinct from the Thinker's working-memory capacity, which bounds deliberative intake. The Thinker does not draw on the Observer's budget, but it sets how that budget is allocated through the templates it writes, so that where the Thinker is loaded the allocation falls back on raw Salience. Where the architecture describes M, or anything M drives, as unbounded, the term carries the first property and not a claim about magnitude: M has no terminus of its own, while the magnitude it reaches at any time is limited by the saturation of the valuation response (Appendix A, Section A.4) and by the attention the Observers involved can supply (Appendix A, Section A.5.7.4). The population-scale consequence, that the Status a population can confer is bounded by the attention its members can spend on one another, is derived at Chapter 8, Section 8.12.7.

Where the escalating pathologies live. Non-closure belongs to the general case, and in the general case it relocates without escalating: Salience over a field of Objects moves from one Object to the next as each is reached, and nothing in the field reads the Individual's reading back. The escalating pathologies of M require the special case. Rivalry needs a second Individual whose reading converges on the same scarce Object; contagion and Coupling need readings that pass between Individuals; the inflation of the Mimetic Premium needs a market of observed valuations; the scapegoat needs a figure several Individuals read together; and the fold is the loop-gain condition met on exactly those inputs (Appendix A, Section A.5). Each needs one Individual's reading to feed back into another's, and only the special case supplies that return path. The general case is not free of pathology. Source Opacity and the Salience-Deficit Conflation operate in it as everywhere, so a brightly coloured package can be wanted as a lack and given a confabulated reason; but a choice no one observes feeds no one else's reading, and it ends where it is made. Most of M's pathology, and all of its escalation, lives in the special case.

The Observer does not maintain priors. It operates on the priors supplied by the Thinker through the Belief Register mechanism set out in 4.4.8, conditioning its live computation on the Belief Register's current state at each cycle. Storage, updating, and revision of those priors are Thinker functions and are not performed by the Observer. The Observer is therefore a prior-consumer rather than a prior-maintainer, and nothing in its own operation can alter the priors it consumes; alteration occurs only when the Thinker updates the Belief Register, on the Decision Hub's Scalar, on the recall of its own prior state, and on Evidence received from other Individuals. The Observer does not pursue Objects. It does not produce desire-as-wanting-to-pursue. It does not ground Salience in the Body Manager's deficits or in the Thinker's. It does not initiate acquisition. These are downstream functions of integration at the Decision Hub, ground-resolution at the Thinker, and action-selection at the Decision Hub.

The Observer has no upstream channel into the Thinker for its Salience output. That output is the Salience signal it delivers to the Decision Hub, and that signal is subsequently compressed with the other direct inputs into the tag-stripped Scalar. The Observer's output therefore cannot serve as independent evidence that the Thinker could use to correct its priors about the Observer, because by the time the Observer's output returns to the Thinker (as a component of the Decision Hub's Scalar), it has been tag-stripped and combined with the other direct inputs. This is the architectural reason the Observer cannot self-correct errors in the Thinker's priors about the Observer: the Observer's output is structurally indistinguishable from the other direct inputs at the point where it could have served as corrective evidence. The restriction is on the Salience path, not on perception as such. The Thinker continues to receive ordinary sensory input, including information received from other Individuals, at its deliberative timescale, as Diagram 4.1 records.

The Observer's distinctiveness as a component lies in the input stream that no other component receives: the observed pursuits and observed valuations of other Individuals, and comparative-position data derived from those observations. The Observer's conditional dependence on priors maintained elsewhere does not collapse it into the Thinker, because the distinctive input stream is neither produced nor stored by the Thinker. The Thinker supplies the templates by which the entities in the distinctive input stream are parsed, and the Observer does the continuous assessment work using those templates against live observation. Without the Observer, the other components' priors would still modulate Salience, but the Salience function would be blind to the distinctive input stream.

Mapping onto the AIM instantiation. In the AIM Framework, the Observer is the M input channel, and Mimetic Desire in Girard's sense is the phenomenological output that the Individual experiences when the M-driven component of a Salience signal is integrated at the Decision Hub, tag-stripped there, and then resolved by I's Bayesian ground-resolution under a Hypothesis Space that does not adequately include the M-Hypothesis. At the architectural level the source and the wanting it produces are not the same thing. The source, M, is the continuous salience-assessment function that operates on the perceived Object field with the Observer's distinctive input stream. Mimetic Desire in Girard's sense, the wanting the Individual experiences, is the emergent property that arises when the M-driven Salience is integrated and tag-stripped at the Decision Hub and then mis-resolved by I into a confabulated ground-attribution (a Body Manager deficit, a deficit of what a process requires, or both), at which point the Individual experiences the wanting as their own in a way that is phenomenologically indistinguishable from A-driven or I-driven wanting.

The Preconscious Transmission window identified in Axiom 7 (60 to 340ms) is the temporal signature of the Observer's operation in the human substrate, placing the Observer's output within the Decision Hub's input stream before conscious awareness of the signal is possible. In the human substrate this timing secures the absence of a phenomenological residue from the Observer's contribution, and it is what ensures that the Observer's output enters the Decision Hub through the same preconscious pathway as the Body Manager's output, denying the Individual any phenomenological cue that would support source-discrimination at the point of input. In the human substrate that absence is secured by the speed of transmission (Axiom 7); the source decomposition is unrecoverable afterwards whatever the speed, because integration discards it (Axiom 3). The inaccessibility rests on the residue being absent rather than on the transmission being quick.

With all four components now specified, the instantiation claim can be stated in a form the reader can assess. The AIM Framework claims that the human motivational system instantiates the minimal architecture, with A instantiating the Body Manager, I instantiating the Thinker, M instantiating the Observer, and the Decision Hub instantiating the integrator-and-action-selector role. The instantiation claim is separately falsifiable from the minimal architecture itself. If the human motivational system were shown not to instantiate the architecture, for example if Salience computation were found not to be conditional on priors, or if prior-maintenance were found to be distributed across the A and M components, the instantiation mapping would be refuted while the minimal architecture might still stand as a valid specification for some other system, though a falsified human instantiation is a failure of the AIM Framework as a theory of human motivation whatever residual validity the abstract specification retains.

4.4.7.1 The Minimum Installation Condition

The Observer's function specification in 4.4.7, read jointly with the Signed Valuation Principle of 4.4.0, entails a two-clause condition on what is required for the Observer to install a Mimetic signal toward an Object from another Individual's conduct on any given operating cycle. This is M's social special case; the general case, the Observer's reading of the Object field against the Thinker's priors, requires no Model (4.4.7). The condition is stated here as the Minimum Installation Condition (MIC). MIC is valence-signed: installation occurs in the positive-valence and the negative-valence direction alike, the valence that signs it is the perceived Individual's own, and the downstream dynamics differ by sign.

Gate clause. The Observer installs a Mimetic signal toward an Object on any cycle in which its distinctive input stream contains perception of another Individual exhibiting directional valence toward that Object. Directional valence is either positive (approach, pursuit, reaching, acquisition, lingering gaze, positive display directed at the Object, expressed valuation) or negative (avoidance, recoil, withdrawal, contamination-coding, scapegoat-coding, expressed revulsion directed at the Object), and any conduct toward the Object discriminable from indifference carries it. The sign of the observed valence is the sign of the installed signal. Nothing further is required at the gate. The gate does not require that the Object carry A-functional or I-functional relevance for the observing Individual, that the observing Individual possess prior familiarity with the Object, that the Object be scarce, or that any particular slow-loop state obtain. Two conditions do not open it: another Individual merely co-present with the Object, and another Individual oriented toward the Object, by a passing glance or by position, with no valence readable in the orientation.

The gate is keyed to valence and not to co-presence because the elevation that co-presence produces has a target of its own. An Individual in the perceived field generates a categorically higher rate of trackable state-change than any static Object near it (the Configural Update-Rate Asymmetry, 4.3.10), so a continuously sampling Observer registers a Salience elevation on Individual-detection as such. The Observer's only output is signed Salience over an Object, and the perceived Individual is itself an Object in the field (Individual-as-Object), so the elevation binds to the perceived Individual. It is the Observer's reading of that Individual: it draws attention to the Individual and along the Individual's orientation, and it installs no valuation of anything else in the field. What passes from the perceived Individual to another Object is the Individual's relation to that Object, and the relation carries a sign only where the Individual's conduct toward the Object is discriminable from indifference. Attention follows orientation; Mimetic installation from another Individual follows valence.

The gate has no reverse. An observation event that has opened the gate has already delivered its installation to the Decision Hub, within the Preconscious Transmission window of Axiom 7, and no later act of the Individual withdraws it. What has been seen cannot be unseen, and what has been observed cannot be un-observed. Conscious refusal, correction and recantation all operate downstream, on the weight the Thinker assigns to what was installed, and never on the installation event itself.

Magnitude clause. The Salience weight applied to the installed signal is conditioned by the Thinker's priors about the Body Manager's current deficit state, its own current deficit state, and the Object field, as supplied read-only to the Observer through the Belief Register (specified in 4.4.8; Appendix A, Section A.5.9), and by the strength of the perceived Individual's directional valence toward the Object. An observation event that carries directional valence therefore opens the gate whatever the slow loop's state, and produces an installed M-signal whose sign is set by what the perceived Individual was doing and whose magnitude is set by the slow loop's current posteriors at the moment of observation and by how strongly the valence was expressed. An Individual whose deficit-state currently flags food as a high-priority deficit-repair category receives, on observing a Model reach for food, an installed M-signal of greater weight than the same Individual would receive in a sated state.

The Minimum Installation Condition is the social special case of the Observer's general perceptual function (the Perceptual Primacy of the Observer), the Observer tracks the directional movement of any Object in the field for action-guidance, and Mimetic installation is entailed when the tracked Object is itself an Individual whose conduct toward another Object carries valence, requiring no separate channel. The Minimum Installation Condition is accordingly subject to Installation Carrier Neutrality, the gate is keyed to the decoded observation event and not to the carrier or sensory format in which that event reaches the Observer, so the physical medium and the modality-specific form are incidental to installation. This is the input-side counterpart of Source Opacity (Section 4.4.1), carrier and format are immaterial upstream of the Decision Hub, just as the source decomposition is destroyed at and downstream of integration. The emergent crisis dynamics the treatise develops are properties of the coupling between many such channels, not of the perception channel in isolation.

4.4.8 The Belief Register

Axiom grounding. The Belief Register is the architectural form in which the Thinker's posteriors are made accessible to downstream operations that must run at timescales shorter than the Thinker's deliberative ground-resolution cycle permits. The Belief Register mechanism follows from Axiom 7 (which places the Observer's operation within the Preconscious Transmission window of 60 to 340 milliseconds in the human substrate, faster than any process that requires conscious deliberation) operating jointly with Axiom 8 (Bayesian Belief dynamics, which produce the posteriors that are cached).

Function-only specification. The Belief Register ( $b_{t}$ in Appendix A.5.6) is the architectural form in which the Thinker's current posteriors and salience-conditioning templates are made available for downstream access at timescales shorter than the Thinker's deliberative cycle permits. The Belief Register is the architecture's state component: the persistent store over the Hypothesis Space, written only by the Thinker, read by the Observer, and computing nothing. The Belief Register is not a fifth processing component, and the minimality of the specification is a claim about the components that compute, which remain four. The Belief Register is also the architectural channel through which the Thinker's updating propagates to the Observer. The Belief Register stores the Thinker's Beliefs and Salience templates, not a decomposition of valuation into A, I, and M contributions, because the Scalar on which the Thinker updates is source-opaque. The Belief Register is also the architectural channel through which the magnitude clause of the Minimum Installation Condition (4.4.7.1) is realised, because the slow-loop posteriors that set the Salience weight of each installed M-signal, in both the positive-valence and negative-valence direction, reach the Observer only through the Belief Register.

The two functions of one store. The Belief Register discharges two functions. The first is storage: the Belief Register holds the posteriors the Thinker commits through Ground-Resolution, so that a conclusion survives the cycle in which the Thinker reached it and stands as the prior for the next cycle. The second is salience-conditioning: on every Observer cycle the Belief Register furnishes the configuration under which the perceived field is parsed and weighted, which Models count, which Objects are M-charged or aversion-coded, which deficit categories are currently live, and the vocabulary through which the field is divided into Objects at all. An Individual whose Thinker has resolved that a given Object repairs a given deficit has, by that commitment, altered the Salience that Object carries when the Object is next in the field.

The second function is not a second store. What the Thinker supplies is the stored posterior itself: one distribution over the Hypothesis Space, put to two uses. The architecture is committed to the asymmetry between the two components and not to the mechanism that spans them. The Thinker maintains the posteriors; the Observer operates on the posteriors and cannot alter them; whether the configuration is fetched by the Observer or projected onto the Observer is immaterial to the architecture's derivations. What the Observer never does is hold priors of its own. Nor does salience-conditioning exhaust the Observer's output, because the Observer also marks Objects for which the Belief Register carries no prior, which is how content on which the Thinker has not yet performed Ground-Resolution enters the loop. The weight the Decision Hub places on the Observer's channel is meanwhile the Decision Hub's own, applied at integration: the Thinker's control of the Observer's channel is control of what is salient, not of how heavily the channel counts. The Belief Register reaches the integrated value only through the Observer (Appendix A, Section A.5.6).

Both functions are discharged by content that Ground-Resolution writes and that nothing else writes. A ground-attribution resolving onto the wrong ground therefore changes what the Individual subsequently notices, and the insight-behaviour lag runs at the rate at which Belief-Register Updates accumulate rather than at the rate the Thinker reaches conclusions.

Belief and desire. Belief issues in action. The Decision Hub selects on valuations formed under the Thinker's posteriors, which reach it on two routes: as the Thinker's own evaluative input, and, through the Observer's read of the Belief Register, as the configuration under which the field is made Salient. On either route what a Belief supplies is direction: which Object, reached by which action, would close which deficit. It does not supply the magnitude the Decision Hub acts on. That magnitude is always a deficit to be corrected, because the integrated Scalar admits no other interpretive category (Section 4.4.0), and every action the Decision Hub selects is therefore driven toward closing one. Three things can stand open to be closed. The Body Manager's deficit is a departure from a homeostatic set-point (Section 4.4.2). The Thinker's deficit is the absence of an Object that would support the engagement it seeks (Section 4.4.3). The third is a felt lack: the Observer's Salience, received under the only category integration allows, presents as something missing although nothing was open (the Salience-Deficit Conflation Principle, Section 4.4.0). Desire, in the architecture, is the Common-Currency Scalar directed at an Object, the felt urge to act (Section 4.5), and Belief is what fixes the Object. Neither issues in action without the other. A Belief with no deficit or felt lack beneath it supplies direction and no drive, and a deficit with no Belief to locate its correction is a drive the Observer must point unaided. This is the architectural form of the familiar explanation of action by a belief paired with a desire (Davidson, 1963), with the pairing specified: the two are two functions of one loop rather than two independent causes. The deficit makes an action necessary, the Belief makes it this action, and the Scalar that follows returns to the Thinker as the evidence on which the Belief is next revised. That Scalar falls where the deficit closed and stands where it did not. Where the lack was felt rather than open, acquiring the Object closes nothing and the Scalar stands, which is the architectural root of M's non-satiation (Axiom 6).

Two such timescales are architecturally important. The first is the Observer's live operating timescale, within which the Observer conditions its computation on the Belief Register's current state at each cycle without awaiting fresh ground-resolution. The second is the habit timescale, within which action-selection at the Decision Hub is driven directly by the Decision Hub's intrinsic cached-value (habit) store, executing familiar action without requiring a fresh ground-resolution cycle at all; the Observer's read of the Belief Register conditions live Salience but is not the route through which habitual selection runs.

The Belief Register is read-only from the downstream side. The Observer reads the Belief Register at each operating cycle to condition its Salience computation, and action-selection at habit timescales is driven by the Decision Hub's intrinsic cached-value (habit) store without recourse to the Thinker, but neither the Observer nor habit-routed action-selection can write to the Belief Register or alter its state. Belief-Register Updates are a Thinker function, performed via Bayesian updating on the Decision Hub's Scalar, and the revised posterior is the Belief Register's state, which moves on the slow timescale of repeated updating rather than in a single pass. That write is Ground-Resolution's output rather than a further operation performed on it: the posterior a cycle returns is the posterior the Belief Register holds, and there is no step between resolving and storing. The Belief Register therefore preserves the architectural commitment that the Thinker is the sole prior-maintaining component: the Thinker is the only writer, so the priors the Belief Register holds are the Thinker's however long the priors persist there, and no other component maintains priors of its own.

The cached habit store accounts for several otherwise puzzling features of the architecture's behavioural output. The hand's automatic reach for the door keys on returning home, the foot's adjustment to a familiar uneven step, the fingers' navigation of a familiar keyboard, and the entire repertoire of skilled practice are driven by the Decision Hub's intrinsic cached-value (habit) store; action-selection at habit timescales runs on that store, which is why these actions proceed without conscious deliberation and why they persist even when the Thinker's deliberative posteriors have begun to drift away from the values the store currently holds. A distinct and faster protective layer sits beneath this one: blink reflexes, flinching or ducking from a looming object, and jerking the hand back from a hot surface are executed within milliseconds by subcortical circuitry (spinal reflex arcs, the brainstem, and the superior colliculus), well before any deliberative or value-based cycle could complete. These protective responses are not value-based action-selection and do not run through the Belief Register or the habit store; the specification does not model them, noting only that they occupy a separate fast layer that the value architecture presupposes rather than implements.

The individual-scale basis of the Double Opacity Problem developed in Chapter 9 is the Decision Hub's cached-value store and not the Belief Register: habitual behaviour runs on the store, which bypasses the real-time integration cycle entirely, so the source composition of the integrations that formed the habit was destroyed when they were integrated, and the current instance involves no deliberation to inspect. The Individual whose deliberation has revised their posteriors will continue to act, at habit timescales, on the values the cached store still holds, because revising the Belief Register does not retune the store. This is the architecturally explained version of the familiar phenomenon in which an Individual knows better but continues to do otherwise.

Mapping onto the AIM instantiation. In the AIM Framework, the Belief Register is the store through which I's current posteriors reach M's preconscious salience-assessment function, and through M's read of the Belief Register they reach the Decision Hub's live integration. Habit-routed action-selection is not on that path: it runs on the Decision Hub's own cached-value store, which holds a value and no record of what produced it, and which I does not write. Repetition is what conditions that store. Each repeated cycle of fresh integration in a stable context lays down the value the store comes to hold, so the store is written by what the Individual has done and not by what I currently believes. The Belief Register is what makes M's contribution to the integration stream operate at preconscious speeds; M does not deliberate, because M reads the Belief Register and applies it. The Belief Register is also what makes culturally installed I-templates (the Glory-template, the martyrdom-template, the kamikaze-template, the seppuku-template, the heroic-last-stand-template) reach the Decision Hub at action-execution speeds rather than at deliberation speeds, through M's read of the Belief Register within a single operating cycle, which is why a soldier going over the top or a kamikaze pilot in final approach need not reconsider in the moment of action; the Belief Register has already been written, and M's reading of it is already in the integration when the action proceeds. The Belief Register also accounts for the characteristic challenge the Knowledge Patch faces in operation. Even when I's deliberative posteriors are updated by Patch-supplied content (AIM vocabulary, the Two-Question Diagnostic, behavioural signatures, presumption against scapegoating), the Belief Register must also be revised, through repeated Belief-Register Updates, before M's live computation tracks the updated posteriors. Habitual action-selection does not track them at all: the cached-value store is not read against current posteriors, so it is reconditioned only as repeated cycles of fresh integration lay down new values. The period between deliberative update and behavioural change is the period in which the Individual knows better but continues to act on outdated priors, and Patch-supplied insight does not translate into Patch-conforming behaviour without repetition. Both lags are architecturally entailed features of the routes that carry them rather than a contingent failure of will or attention.

4.4.9 The Operational Loop

Axiom grounding. The operational loop is the dynamic arrangement of Axioms 2, 3, 4, 7, and 8 operating together over time. Axiom 2 supplies the integration. Axiom 3 supplies the tag-stripping. Axiom 4 supplies the confabulated ground-resolution output. Axiom 7 places the M-signal's entry into the loop prior to conscious awareness. Axiom 8 supplies the updating rule by which ground-resolution outputs become posteriors that are written to the Belief Register and become priors for the next cycle.

Function-only specification. The four components operate in a continuous loop with the following structure. At each cycle: the Body Manager reports its current deficit state; the Observer, conditioned on the Belief Register's current state, performs continuous assessment over the perceived Object field using its distinctive input stream; the Thinker's ground-resolution output from the previous cycle contributes its direct input; these three streams enter the Decision Hub; the Decision Hub compresses them into a tag-stripped Scalar and selects an action; the Scalar is fed back to the Thinker for Bayesian ground-resolution on the next cycle; the resulting ground-attribution becomes an updated posterior that is both used directly by the Thinker for the next ground-resolution and written to the Belief Register, where it becomes the basis on which the Observer's next cycle conditions its computation. Habit-routed action-selection proceeds on the Decision Hub's own cached-value store, which repetition conditions and the Thinker does not write. Within each cycle, the Observer's installation of Mimetic signal onto Objects in the perceived field is governed by the Minimum Installation Condition specified in 4.4.7.1, with the gate clause determined by the valence of the Model-directed observation in the distinctive input stream for that cycle, the magnitude clause determined by the Belief Register's current state, and the post-acquisition re-ranking of the Object-field producing the next cycle's Salience distribution.

Why this is the minimal architecture. Remove any component and the system cannot produce the observed outputs. Remove the Observer and the social world cannot enter Salience computation. Remove the Thinker and priors collapse, leaving the Observer with nothing to condition its assessment on and leaving integrated valuations with no resolver. Remove the Body Manager and deficit-repair termination is unavailable. Remove the Decision Hub and the signals cannot be integrated for action-selection. Remove the Belief Register and the Thinker's posteriors cannot condition the Observer's live computation, severing the slow path from the fast path.

Diagram 4.1 — The minimal architecture. The three signed source signals (A, I, M) converge on the Decision Hub for integration and tag-stripping (Source Opacity) and action-selection; V returns to the Thinker for Bayesian ground-resolution over the Hypothesis Space, which is constituted from the Memory Store; the Thinker writes the Belief Register, which conditions the Observer by top-down projection and not by any read from the Decision Hub; the Decision Hub holds its own cached habit store. Action-selection issues in an action where V crosses the threshold and in behaviour without action where it does not, and both are observed and both return through the reafference pathway. Sensory input reaches all three source components, the Body Manager interoceptively and the Observer as the external perceptual and social field, both preconsciously, and the Thinker only as far as it can process at its deliberative timescale. That deliberative-timescale path is the ordinary sensory and communicative channel. It is assumed by the architecture and is not routed through the Decision Hub. Fast protective reflexes are a separate subcortical layer the architecture presupposes and does not model.

Loading diagram...

The propagation of the Thinker's updated posteriors to the Observer operates through Belief-Register revision: updated posteriors are written to the Belief Register through a Belief-Register Update, and the Observer's next cycle reads the revised Belief Register. At habit timescales, action-selection at the Decision Hub proceeds on the Decision Hub's intrinsic cached-value store without a fresh ground-resolution cycle, which is why routine action proceeds without conscious deliberation and why habits persist even when the Thinker's deliberative posteriors have begun to drift away from the values the store currently holds.

Where a signal goes when it subsides. The loop does not lose a signal when the signal subsides, as the conservation discipline of Section 4.1 requires. While the Observer's Salience for an Object is raised, the Salience passes into the integrated Scalar at every cycle. The Scalar is spent in the action the Decision Hub selects, which draws on the energy envelope, and returns to the Thinker. Where the Scalar poses an open question, Ground-Resolution writes the resolution to the Belief Register, which returns it to the Observer as the templates under which the Object is next read. Where the same selection is repeated, the repetition conditions the Decision Hub's cached-value store. When the Salience subsides, the Observer's reading is returning to rest after its content has been carried into the action, the Belief Register and the cached-value store. An Object that draws no Salience sends no signal, and nothing upstream has stopped. A habituated Object belongs with these: habituation lowers its Salience, action-selection runs on the cached-value store, and the Thinker is recruited onto the habit only where the habit no longer serves the deficit it was conditioned on, or fires in a context it was not conditioned for. Because the cached-value store does not register outcomes, that mismatch reaches the Thinker through the Scalar of a competing valuation or through ordinary deliberative intake (Appendix A, Section A.5.6).

This loop is the architectural basis for iterated Confabulation. The ground-attribution the Thinker returns on cycle t becomes an updated posterior for cycle t+1. If the Thinker's Hypothesis Space does not include an adequate Hypothesis about the Observer, the cycle t ground-attribution is confabulated, and the cycle t+1 posterior is updated toward greater confidence in the confabulation. The Scalar from the Decision Hub on cycle t+1 is consistent with the prediction the updated posterior produces, because the Scalar is source-opaque and matches the prediction regardless of which source drove it. Confabulation is therefore self-reinforcing across cycles, not because the Thinker is failing to update, but because it is updating correctly on source-opaque evidence using priors that include no Hypothesis capable of generating the correct ground-attribution.

This is the individual-scale locus of the Bayesian Lock-In that Chapter 13 develops at population-scale. The individual-scale version operates through the Thinker's updating cycle alone, and produces entrenched confabulated priors within a single Individual. The population-scale version operates when each Individual's confabulated output becomes evidence for other Individuals' Thinker prior-maintenance, producing cascading lock-in across a social network.

Mapping onto the AIM instantiation. In the AIM Framework, this loop is the AIM Operational Loop. At each cycle in an Individual's ongoing operation, A-signals from the body, M-signals from the Observer conditioned on the cached I-supplied priors, and I's prior ground-resolution output converge at the Decision Hub, are compressed into a tag-stripped Scalar, and the Scalar is returned to I for Bayesian ground-resolution that both triggers the next action-initiation and updates I's posteriors (including its posteriors about M) in the Belief Register. The cycle repeats continuously, and the individual-scale Bayesian Lock-In that characteristically develops when I's Hypothesis Space about M is impoverished is the AIM instantiation of the minimal architecture's iterated-Confabulation dynamic.

4.4.9.1 The Reafference Pathway, A Second Fast Loop Through the Observer

The operational loop described in 4.4.9 closes within a single cycle. The Decision Hub integrates the inputs, passes the combined signal to the Thinker for resolution, and the Thinker's resolution is returned to the Decision Hub for action-initiation. A second loop closes across cycles, and it runs through the Observer rather than through the Thinker.

When the Individual executes an action, the action itself becomes something the Observer can watch. Three kinds of self-observation follow. First, the Individual registers what they have just done at the moment of doing it, before any external evidence about the outcome has arrived. Second, the Individual perceives the action's outcome, whether the Object was acquired, whether the engagement began, whether the pursuit succeeded. Third, other Individuals respond to the Individual's executed action, and their responses, their renewed attention, their valuation shifts, their own comparative repositioning, become new observations in the Observer's distinctive input stream.

The Pathway runs on behaviour and not on action alone. An Individual who declines an Object observes the declining, and an Individual whose pursuit lapses across a stretch observes the lapse; both re-enter the Observer on the following cycle exactly as an executed action does. This is what makes an interception consequential beyond the moment it occurs. The interval the interception produces is itself observed, is itself written to the Belief Register through the ordinary cycle, and therefore conditions the Observer's next assessment in the way the action it displaced would have.

All three re-enter the Observer on the cycle following execution. The Individual's own completed action, and the social world's reaction to it, are for the architecture observations in exactly the same sense that a Model's pursuit of an Object is an observation. The Observer does not distinguish between "someone else pursued this" and "I pursued this," because at the level the Observer operates the source-identity information is not available. The consequence is that the executed action loads the next Observer cycle with elevated Salience for the Object just pursued, which loads the Decision Hub on that next cycle through the Observer's channel, directly and, once the Thinker has written the resulting attribution, through the templates the Observer reads. The loop can therefore compound, a pursued Object carries higher Salience on the next round, and higher Salience makes further pursuit more likely.

This second loop is the architectural reason that actions taken tend to reinforce the wanting that produced them, that public performance of a preference intensifies the preference over time, and that privacy attenuates but does not eliminate the reinforcement. Self-observation continues even when no one else is watching, because the Individual is always an observer of their own behaviour, which is why Private Persistence is a diagnostic for the presence of genuine engagement but not a guarantee that no observation-driven reinforcement is occurring.

The localisations of 4.4.12 do not change under the Pathway. The crisis geometry does: a diagonal self-observation term would move the threshold below 1, which is why the term is held outside the fast-layer formalism as a named specialist task (Appendix A, Section A.7, Task 12a). It is specified here because the diagnostic tools the AIM Framework relies on, audience-removal, Private Persistence, the Revelation Effect, the Care-Substitution Test, and the Gap Behaviour Test, all presuppose that the Individual's own executed action and the reactions of others to it re-enter the architecture through the Observer, and the intervention channels that reduce observability operate on this Pathway specifically. The Pathway is specified at the Disaggregation Primitive level, what any system instantiating the minimal architecture would satisfy; the question of what in the human neural substrate instantiates it is a separate question, developed in Chapter 9.

The Pathway is not the Self-Coupling Permanence channel formalised in Appendix A as $κ_{se l f}$ . That parameter is the strength with which the Belief Register's own prior state is recalled from memory and re-enters as Evidence. Reafference is self-observation of executed action through the Observer. The two are related but distinct, and the appendix's Observability channel sums over other Individuals only, so a diagonal self-observation term is not carried in the fast-layer formalism. Opening that term is a named specialist task in Appendix A: it would add a second fast cycle alongside the one that runs from the M signal through the integrated valuation and the Salience field back to the M signal, and move the Critical Loop Gain below one by the gain of that second cycle, while leaving the Hopf locus at negative loop gain for every value of the self-observation gain, so that the Saddle-Node Default survives the Pathway. Until that task is discharged, the Pathway remains functionally specified here and is not identified with $κ_{se l f}$ .

4.4.10 Summary Table

The four processing components and the Belief Register, with their inputs, their outputs, their termination conditions, and their AIM-instantiation identities, are summarised below. The rows supply the minimal-architecture specification. The final column identifies the AIM-instantiation component. The Observer row of the table is to be read subject to the Minimum Installation Condition of 4.4.7.1, which specifies that the Observer's output of Salience and value signals over Objects is gated on observation of Model-directed valence (gate clause, signed positive or negative), weighted by the Belief Register's current state (magnitude clause), and subject to the asymmetric termination conditions of 4.4.7.1, positive-valence installation is interminable in the M-channel until acquisition, which then re-ranks the Object-field, while negative-valence installation terminates pursuit while retaining monitoring-Salience.

Component	Primary purpose	Input	Output	Termination	AIM instantiation
Decision Hub (integrator and action-selector)	Action-selection from integrated valuation, the Scalar to the Thinker	Body Manager's output, Observer's output, Thinker's ground-resolution output	Action selection plus tag-stripped integrated Scalar fed back to the Thinker	Continuous selector, no termination condition	Decision Hub (vmPFC and ventral striatum, per Axiom 2)
Body Manager (homeostatic-deficit signaller)	Homeostatic deficit signalling and deficit-repair termination	Internal physiological state	Deficit-scaled valuation signals	Consumption or repair satiates (per Axiom 6)	A, Appetites (per Axiom 1)
Thinker (prior-maintainer and Bayesian ground-resolver)	Prior-maintenance (about Body Manager, itself, Observer, Object field), deliberative inference, and Bayesian ground-resolution on the Scalar	Experience, diagnostic evidence, structural knowledge (for priors); tag-stripped Scalar (for resolution)	Posteriors written to the Belief Register; ground-attribution fed back to Decision Hub	Endogenous termination of engagement, or interruption (per Axiom 6)	I, Intrinsic Motivation (per Axiom 1)
Observer (continuous conditional Object-field Salience assessor)	Continuous conditional Object-field Salience assessment	Belief Register (current state) plus the distinctive input stream (observed pursuits, observed valuations, comparative-position data)	Salience and value signals over Objects	No termination condition (architectural; per Axiom 6)	M input channel (per Axioms 1, 5, 7); Mimetic Desire is the emergent phenomenological output of M-signal integration, tag-stripping, and mis-resolution under an impoverished M-Hypothesis
Belief Register (state component)	Persistent storage of the Thinker's committed posteriors, and salience-conditioning of the Observer	Belief-Register Updates written by the Thinker following Ground-Resolution	Current posteriors over the Hypothesis Space, supplied to the Observer as salience-conditioning configuration	No termination condition; the Belief Register holds state and computes nothing	No AIM-source identity; the store through which I's posteriors reach M's live computation

4.4.11 The Failure Mode, Object Parasitism as Architectural Consequence

Axiom grounding. The failure mode is the joint consequence of Axioms 2, 3, 4, 5, 7, and 8 operating together in the specified component-architecture. Axiom 5 supplies Object Parasitism as M's operating mode. Axioms 2 and 3 supply the tag-stripping that makes the mis-resolution possible. Axiom 4 supplies the confabulated ground-attribution output. Axiom 7 places the M-signal's entry into the loop before conscious scrutiny is possible. Axiom 8 supplies the updating dynamic that makes the failure mode iterated and self-reinforcing rather than one-time.

Function-only specification. The minimal architecture's characteristic failure mode is Object Parasitism, formalised at the fast layer as the Parasitism Theorem (Appendix A, Section A.5.4). Object Parasitism is generated by the specification given above without additional mechanism; it is not an independent claim about the architecture. Object Parasitism follows directly from the Observer's satiation dynamics as specified in 4.4.7. Because the Observer is non-satiating and generates no Objects of its own (Axiom 5), the only Objects available for it to attach Salience to are those the Thinker's categorical priors have already parsed into the field, that is, Objects already carrying A-functional or I-functional roles. A non-terminating, Object-less Salience function operating on a field of A-Objects and I-Objects can do nothing other than displace the valuation of those Objects, upward under a positive reading and downward under a negative one, which is Object Parasitism. The failure mode and the satiation asymmetry of 4.4.12 are therefore the same architectural fact seen from two sides.

The failure mode arises through the following sequence (the Confabulation Rationality Lemma at Appendix A.5.6 establishes that, under an impoverished Hypothesis Space, this sequence is a rational Bayesian output rather than a cognitive error). The Observer emits a Salience signal for an Object, weighted substantially by the Observer's distinctive input stream (observed pursuits and observed valuations of other Individuals, and comparative-position data). The signal enters the integration stream at the Decision Hub. The Decision Hub integrates it with the Body Manager's output and the Thinker's ground-resolution feedback, producing a tag-stripped Common-Currency Scalar. The Thinker receives the tag-stripped Scalar and performs Bayesian ground-resolution on it. Under an impoverished Hypothesis Space that does not adequately include the Observer-driven-by-distinctive-input-stream Hypothesis, the Thinker returns a high-confidence ground-attribution (a Body Manager deficit or a deficit of what a process requires) that is systematically wrong for the Salience at hand. The Decision Hub initiates acquisition or engagement on the wrong ground. Neither deficit-satiation nor engagement-termination engages upon completion of the action, because no deficit of either kind was closed. The acquired Object resets the Model-tracking and comparative-position data that the Observer uses on its next iteration, the Belief Register is updated in ways that lock in the wrong ground-attribution, and the loop restarts.

Four features of the architecture jointly make the failure mode structurally available rather than contingently occasional.

First, the Observer's signal generation is sub-filterable at its operating timescale: it cannot itself refrain from emitting the signal while the distinctive input stream is live, and (in the AIM instantiation) the signal is generated within the 60 to 340 millisecond window before conscious awareness is possible.

Second, the Decision Hub's tag-stripping is irrevocable: the source decomposition is destroyed at the moment of integration and cannot be recovered from the Scalar downstream.

Third, the Thinker's ground-resolution operates on incomplete data. The resolution is Bayesian reasoning on a signal from which source-identity has already been destroyed, and the resolution's correctness depends entirely on whether the Hypothesis Space includes the Hypothesis that would allow the leading source to be inferred from the course of the tag-stripped signal across contrasting conditions. When the Hypothesis Space does not include that Hypothesis, wrong ground-attribution is the predictable output of correct Bayesian reasoning on impoverished Hypothesis data, not a malfunction of the Thinker.

Fourth, no component in the architecture other than the Thinker holds priors. The Body Manager cannot correct the Thinker, because the Body Manager does not bear priors from which corrections could be sent. The Observer cannot correct the Thinker, because the Observer is downstream of the Thinker's priors and its output is not independent evidence against them. The Decision Hub's feedback to the Thinker is source-opaque and cannot distinguish confabulated ground-resolutions from accurate ones. The failure mode is therefore not merely available; it is architecturally protected against within-system correction.

The failure mode is therefore architectural, not contingent. It is available in any system that instantiates the minimal architecture with a Thinker whose Hypothesis Space about the Observer is impoverished. An adequate Hypothesis Space does not abolish the failure mode, because self-coupling still returns the incumbent attribution, but it removes the restriction that compounds it (Appendix A, Section A.5.6). The Thinker's Hypothesis Space about the Observer is therefore the one intervention target from inside the Individual that corrects the Thinker, and the loop's other loci are set out at Section 4.4.13.5 and Appendix A, Sections A.5.1 and A.5.5.

Mapping onto the AIM instantiation. In the AIM Framework, Object Parasitism is the failure mode of the AIM Operational Loop, and it compounds when I's Hypothesis Space about M is impoverished, as it characteristically is in cultures that lack M as a source of wanting in their vocabulary. The impoverishment is what the Camouflage Architecture's Layer 7 (the Vocabulary Gap, or Hilbert Collapse) names at the cultural level, and what the full Camouflage Architecture (developed in Chapter 5, Section 5.8) explains at the civilisational level. The Knowledge Patch, developed in Chapter 18, is the architecturally available intervention because it operates on I's Hypothesis Space about M, which is the only target at which the Thinker's own drift can be corrected (Appendix A, Section A.5.9). Iterated operation of the failure mode under social observation of other Individuals' failure-mode outputs produces Confabulation Cascades at individual and collective scale, which Chapter 11 and Chapter 13 develop at their respective scales. The failure mode's architectural availability is what makes Object Parasitism, Confabulation, scapegoating, and terminal mimetic crisis the structurally predictable outputs of the unpatched architecture rather than contingent cultural pathologies.

4.4.12 Architectural Localisation and Derived Consequences

The minimal architecture is not symmetrical across its components. The preceding sub-sections have specified each component in isolation, the Belief Register that mediates downstream access to the Thinker's posteriors, the loop that couples the components, and the failure mode the coupling produces. What the specification makes available, once all four components and the Belief Register are in view, is an account of the architecture's load-bearing asymmetries and of the consequences those asymmetries have for intervention. This sub-section states the localisations first, then draws the asymmetries out as derivations rather than as additional claims.

Load-bearing localisations. The following localisations must be preserved in any instantiation for the architecture's derivations to follow.

All signalling of homeostatic set-point departure is in the Body Manager. The Body Manager has set-points. The Body Manager has no stored cognitive content. The Body Manager has no upstream channel from any other component, because deficit signalling is generated by underlying state and does not require templates or priors to perform its function.
All cognitive content (priors, templates, Hypothesis Space, categorical vocabulary, causal knowledge) is maintained in the Thinker. This includes priors about the Body Manager, priors about the Thinker itself, priors about the Observer, and priors about the Object field. Downstream operations that run at timescales shorter than the Thinker's deliberative cycle permits access the Thinker's current posteriors through the Belief Register, which is read-only from the downstream side. The Observer's live computation is that operation. Action-selection at habit timescales is not: it runs on the Decision Hub's intrinsic cached-value store, which the Thinker does not write and which repetition conditions, and that independence is what makes a habit insensitive to a posterior the Thinker has since revised. Belief-Register revision is a Thinker function performed via Bayesian updating on the Decision Hub's Scalar. No other component maintains priors of any kind.
All outward template-supply in the architecture is from the Thinker to the Observer, routed through the Belief Register. The Thinker does not supply templates to the Body Manager, because the Body Manager requires none. No component supplies templates to the Thinker, because no other component bears priors.
All Salience computation over the perceived Object field is in the Observer. The Observer's computation is conditional on Thinker-supplied priors as cached, but the Observer's distinctive input stream (observed pursuits, observed valuations, comparative-position data) is the Observer's alone.
All integration and action-selection is in the Decision Hub. The Decision Hub does not interpret; it integrates and selects. The Decision Hub's output to the Thinker is the tag-stripped Scalar, not three separately-tagged source-contributions. At habit timescales, action-selection at the Decision Hub is driven directly by the Decision Hub's intrinsic cached-value (habit) store, bypassing the Thinker's deliberative ground-resolution cycle; the Observer's read of the Belief Register conditions live Salience but is not the route through which habitual selection runs.
All ground-attribution (Bayesian resolution of the Decision Hub's Scalar) is in the Thinker. The Confabulation failure mode is specifically the Thinker's output when its Hypothesis Space about the Observer is impoverished.

The satiation-and-termination asymmetry. The three input-bearing components do not merely happen to differ in how their signals terminate. Each component's termination dynamic is entailed by the function the component performs, and could not be otherwise without changing the function. This is the Satiation-Function Entailment, a derived consequence of the specifications in 4.4.2, 4.4.3, and 4.4.7, not an additional claim laid over them.

The Body Manager's function is to signal departure from a homeostatic set-point and to terminate on deficit-repair (4.4.2). A component whose signal is generated by a set-point deficit necessarily ceases to signal when the deficit is corrected and necessarily signals again when the deficit recurs. Episodic satiation followed by cyclic recurrence is therefore not a behavioural regularity observed of A and then attributed to it, it is what a deficit-signalling function does. Remove the set-point and the dynamic disappears, fix the function and the dynamic is fixed with it.

The Thinker's function is to maintain priors and to perform Bayesian ground-resolution, that is, to compute and to drive action (4.4.3). The component performs that function continuously and in private, because computation, recall, problem-solving, and the resolution of attention do not require an audience or an external reward to proceed. It follows that the Thinker's function does not satiate: the deficits the Thinker signals are for the Objects whatever process it is running requires, not for the function itself, so correcting one closes that deficit and leaves the function running, and engagement deepens as competence and inference develop because deepening is the function operating, not a separate reward attached to it. A process the function sets terminates on completion, on the exhaustion of resolution, which is boredom, or on interruption (Section 4.2; Axiom 6).

The I-enabling Object, by contrast, does satiate, because the Object's value is contingent on the particular task the Thinker is performing. Once the task is performed the Object is no longer required, and once functional adequacy for anticipated tasks is reached no further Object is required, the Object remains available for future tasks but generates no further demand. Function-non-satiation and Object-satiation are therefore the same function viewed from two sides, the function is not contingent on completion and so does not satiate, the Object is contingent on the task and so does. Non-satiation of the Thinker's function is architecturally guaranteed by the function specification, not an additional empirical claim on top of it. It differs from the Observer's non-satiation in 4.4.7 in one respect the formalism records: the Thinker's contribution for a given Object closes when the deficit for that Object is corrected, where the Observer's contribution has no closing term (Appendix A, Section A.2, restatement M5b).

The direction of this dynamic can now be stated positively rather than as a mere absence of a termination condition. Because the ground-resolution function is recruited by the unresolved, the process continues for as long as unresolved structure remains and is drawn onward to the next unresolved Object when the present one is exhausted. Novelty and prediction-mismatch are the Thinker's activation condition, not incidental stimuli, and Curiosity is the felt presence of that recruitment. Engagement is the same function continuing to run as resolution deepens, which is why an Object that sustains unresolved structure sustains engagement, and why the I-process is self-sustaining, the deficit it signals being for what the process requires, so that closing it supplies the process rather than ending it.

Boredom is the complementary signature: the felt registration of the exhaustion of the activation condition, the state reached when an Object's novelty and capacity-deepening potential have been consumed and no unresolved residue remains for the ground-resolution function to operate on. Boredom is not a failure of the Thinker but the correct endogenous termination signal of the I-process, and it is the signal that recruits the Thinker toward the next unresolved Object. The impoverishment of a field of anything novel, unusual, or unresolved is the impoverishment of the conditions the ground-resolution function requires, and Boredom is the felt registration of that impoverishment. Curiosity marks the presence of the activation condition; Boredom marks its exhaustion.

The Observer's function is continuous conditional Salience assessment over the perceived Object field, conditioned on the Thinker's cached priors and driven by the distinctive input stream (4.4.7). A conditional assessment over a perceived field cannot complete while perception and the conditioning inputs persist, so the Observer has no termination condition and does not satiate. Acquisition does not end the signal, because Salience attaches to the Object as pursued rather than as possessed: acquisition removes the condition that produced the signal, and the signal collapses at its source without closing any deficit, since none was recorded. The integration that makes this signal felt as a lack is adaptive, since it directs deficit-driven action toward whatever the environment marks as valued; the pathology lies not in the mixing but in the emergent properties the same operation entails once the Observer's domain is unbounded and coupling is high. As 4.4.7 has already established, this non-satiation is architecturally guaranteed by the function specification. The Observer is not deficit-driven, so it cannot satiate in the Body Manager's sense, and it does not depend on conscious engagement, so it cannot terminate in the Thinker's sense. As long as the Individual is conscious and observing, the Observer continues to feed the Decision Hub Salience signals over Objects, conditioned on the Belief Register it is reading at each cycle. On the emission side, M's non-satiation is further entailed by the Salience-Emission Asymmetry: the drive to be Salient has an exogenous, event-shaped terminator and therefore no closable loop, and it recurs positionally per the Salience Ratchet Principle. This complements the incentive-Salience (wanting-liking) account, which explains the urgency of the recurring demand, without displacing it.

The Salience-Emission Asymmetry needs no machinery of its own. The drive to be Salient runs on the channel already specified: the Individual's own executed action re-enters through the same stream that carries everything else the Observer watches, as the responses of others to what was done, so the emission side is that stream closed through the social world rather than a second drive bolted onto the first. The architecture acquires the drive to be Salient in acquiring the capacity to observe, which is why no component had to be added to account for it.

The consequence for the framework is that the three satiation profiles are three readings of three component functions, not three independently posited empirical regularities. This is what protects the satiation dynamics against the charge that they are asserted because the framework requires them. They are required by the framework, but they are not asserted, they are entailed by the component functions specified independently of the uses the framework later makes of them, and they are behaviourally observable for exactly the reason they are introspectively inaccessible. The termination dynamic is a property of the signal's relationship to its generating condition (set-point, engagement, observation), and that relationship is expressed in when the behaviour stops, while Source Opacity (Axiom 3) destroys the source-tag at integration and so denies introspection the same information that behaviour displays. The satiation dynamics are the behavioural surface on which source-discrimination is inferable precisely where introspection cannot recover it, which is why they, and not introspective report, are the load-bearing diagnostic of the entire framework.

Three conditions hold together at the Observer's channel, and each answers a different question.

The first is that the Observer's contribution leaves no phenomenological trace. A deficit signal has a Referent that persists: the hungry Individual goes on feeling hunger, and the Individual who lacks the Object their engagement requires goes on feeling that absence. Those states remain available to notice alongside the wanting, and they are what allows an Appetitive or Intrinsic contribution to be inferred once its tag is gone. The Observer's reading persists in nothing. It is integrated before awareness of the wanting arrives, and Axiom 7 fixes the timing of that in the human substrate, but it is the absence of the residue rather than the speed of the transmission that constrains the Individual: introspection has nothing to work on, because the Mimetic channel is the one input that leaves nothing behind. This ordering of the three channels by what they leave available to notice is the Legibility Gradient.

The second is that the signal so made traceless is a Salience reading and not a deficit. The third is that integration destroys the tag that would have marked it as one.

Each alone would be survivable. A traceless deficit signal would still terminate of its own accord once its Object was obtained. A tagged Salience reading could be held as a reading. A Salience reading that left a trace could be examined against the wanting it accompanied. Holding together, the three entail that a Salience signal is invisible as a Salience signal, unidentifiable by introspection as anything but a lack, and inexhaustible in what it asks. The results derived downstream, non-satiation and the Mimetic Treadmill (the Treadmill), Confabulation and Projection, the Mimetic Premium and its accumulation into Status, and the escalation dynamics that follow from all of them, are consequences of these three conditions obtaining at once.

The channel-count asymmetry. The Integration Equation carries three weighted terms (wA·UA, wI·UI, wM·UM), and its surface symmetry suggests that three sources contribute on equal terms to the integrated value. They do not. The physical channel count is one to one to one (Thinker, Observer, Body Manager), since each component supplies one channel of direct input. The asymmetry is in control, not channel number: the Thinker controls two of the three arguments of the integrated value, its own direct input and the templates under which the Observer's input is classified and sign-coded, the Observer computing on cached I-maintained priors that shape its Salience upstream, while the Body Manager controls one and the Observer controls none. Counted by controlling component, the arithmetic of influence over the integrated value is therefore two to one in the Thinker’s favour, the Body Manager being the one genuinely independent input, and no deadlock between the Body Manager and the Observer against the Thinker is possible because the Observer cannot oppose the Thinker with content the Thinker did not supply. The asymmetry holds at every subcritical operating point, and because loop closure amplifies all three arguments equally, template configuration is most consequential where the loop gain is highest. In the AIM instantiation: I contributes to V through direct UI and through cached I-maintained priors that shape UM upstream; M contributes to V through UM, upstream-conditioned by I's cached priors; A contributes to V through UA alone. The fast-layer formalisation of this asymmetry, and of the four derived consequences below, is at Appendix A.5.9.

This channel-count asymmetry is the architectural arithmetic underlying four derived consequences, none of which is an additional claim over and above the specification above.

The I-Supremacy Result. Because the Thinker's priors shape both its own direct input and the Observer's output, cultural investment in the Thinker's templates loads two of the three channels of direct input to the Decision Hub simultaneously. The Body Manager's channel cannot be similarly amplified, because the Body Manager bears no priors and has no upstream channel for cultural input to modify. The consequence is that cultural configuration of the Thinker's templates can override Body Manager-sourced priorities, whenever the templates code the observation stream at all, at the Decision Hub, even when the Body Manager is reporting a survival-level deficit already inside the body (starvation, injury already occurring).

The A-Disadvantage. The Body Manager has no reciprocal cultural amplifier, because homeostatic set-points are biologically fixed and the Body Manager has no upstream channel to receive cultural shaping. The other two channels of influence over V are both either the Thinker itself or Thinker-shaped. Errors in cultural configuration therefore compound against the Body Manager's channel systematically, not symmetrically. This is the architectural reason that prolonged fasting, voluntary endurance of pain, celibacy against reproductive drive, martyrdom, suicide and other high-A-override cultural practices are structurally achievable at all.

The I-Template Override Result. Dying-for-the-Object outputs are within the reachable output set of any culture whose Thinker-templates, once written to the Belief Register, supply the priors required to make the override region reachable at the Decision Hub. When I's templates have been configured to discount A-signals (the Glory-template, the martyrdom-template, the kamikaze-template, the seppuku-template, the heroic-last-stand-template), the sum wI·UI + wM·UM can exceed wA·UA even for large A-magnitudes, because UI and UM are both I-shaped while UA is not. The Belief Register carries the override at action-execution timescales through the Observer's read of it, the Observer's template-conditioned Salience reaching the Decision Hub within a single operating cycle, which is why the soldier going over the top and the kamikaze pilot in final approach need not reconsider in the moment of action. The override is not a malfunction; it is the minimal architecture's arithmetic operating correctly on the I-templates the culture has supplied and cached.

The I-Corrective Asymmetry and the externality of the Knowledge Patch. Errors in the Thinker's priors compound through both of the Thinker's channels of influence over V, and they cannot be corrected from within the architecture because no component outside the Thinker holds priors from which corrections could be sent, and the feedback the Thinker receives is source-opaque. The Thinker's Hypothesis Space is therefore the only architecturally available intervention target, and the intervention cannot come from within the architecture itself. This is the architectural justification for the externality of the Knowledge Patch developed in Chapter 18: the Patch cannot operate as an internal correction channel because no such channel exists; the Patch must supply the Thinker with priors, procedures, and behavioural diagnostics that the Scalar alone cannot generate. The Belief Register compounds this: even when the Patch has updated the Thinker's deliberative posteriors, the Belief Register must also be revised, through repeated Belief-Register Updates, before the Observer's live computation tracks the updated posteriors. Habit-routed action-selection does not track them at all, the cached-value store not being read against current posteriors, so it is reconditioned only by repeated cycles of fresh integration laying down new values. Patch-supplied insight therefore does not translate into Patch-conforming behaviour without repetition, and the two routes lag for different reasons: the Observer's because the Belief Register has not yet been rewritten, the habit route's because it never consults the Belief Register.

Mapping onto the AIM instantiation. In the AIM Framework, these localisations are instantiated as follows. All signalling of bodily state is in A, which has homeostatic set-points and no stored cognitive content, and which receives no templates from any other component. All cognitive content is maintained in I, including priors about A, priors about I itself, priors about M, and priors about the Object field. All outward template-supply is from I to M, routed through the Belief Register. I does not supply templates to A, because A requires none. All Salience computation over the perceived Object field is in M, conditional on cached I-supplied priors, with M's distinctive input stream (mirror-neuron-mediated Model-observation, observed valuations, comparative-position data) being M's alone and carried in the human substrate by mirror-neuron-mediated pathways operating within the Transmission Window of 60 to 340 milliseconds per Axiom 7. All integration and action-selection is in the Decision Hub, localised to the ventromedial prefrontal cortex and the striatum per the foundational citations to Axiom 2: the ventral striatum carries the value-integration function, and the dorsolateral (sensorimotor) striatum carries the cached-value (habit) store on which habitual selection runs. The dorsolateral localisation is held at best-current-mapping, resting largely on rodent lesion work and on the corticostriatal homologies drawn from it (Yin and Knowlton, 2006; Balleine and O'Doherty, 2010), with the human evidence for a posterior dorsolateral contribution to habit learning more recent and narrower (Tricomi, Balleine and O'Doherty, 2009). What the localisation rests on is the outcome-insensitivity of habitual responding: an instrumental response established under a reinforcer continues after that reinforcer is devalued, where a deliberatively controlled response does not (Adams and Dickinson, 1981; Dolan and Dayan, 2013). That insensitivity is the property the architecture requires of the cached-value store, and it is what makes the cached route a distinct route rather than a fast setting of the deliberative one. How much repetition establishes a habit is not fixed by the architecture and is not settled for the human case, where attempts to induce habit by extended training have repeatedly failed to produce the insensitivity the animal work reports (de Wit et al., 2018). The architecture holds that repetition conditions the store, not that any given quantity of it suffices. The Hypothesis Space over which the Thinker resolves is constituted from the Memory Store, and no component other than the Thinker reads either store. The Decision Hub's output to I is the tag-stripped Scalar V, not the three weighted terms that contributed to it, and at habit timescales the Decision Hub draws on its own intrinsic cached-value store. All ground-attribution is in I, and Confabulation is specifically I's output when its Hypothesis Space about M is impoverished. The Knowledge Patch operates on I's priors about M, which is the only target at which drift in the Thinker's priors can be corrected.

The four derived consequences, in the AIM instantiation, are the I-Supremacy Result (cultural investment in I-templates loads two of the three channels of direct input to the Decision Hub simultaneously, while A has no reciprocal cultural amplifier), the A-Disadvantage (errors in cultural configuration compound against A systematically), the I-Template Override Result (Glory-cultures, kamikaze training programs, martyrdom traditions, honour-regimes, and religious austerity can systematically produce override of A-signalling even when A-magnitude is high, because cultural investment in I-templates loads both the I and M channels simultaneously while A has no reciprocal cultural amplifier, with the Belief Register mechanism carrying the override at action-execution timescales), and the I-Corrective Asymmetry (I is the sole priors-bearing component in the architecture, operating on the Scalar from which source-identity has been stripped, with no internal channel through which A or M could correct I's priors; external supply of I's priors through the Knowledge Patch is therefore the only intervention the architecture admits, a Belief-Register Update is a necessary second step for the Patch to reach M's live computation, and habit-routed selection is reconditioned only by repetition).

These localisations are load-bearing for the treatise's derivations of Object Parasitism, the Status-Reward Asymmetry, scapegoating, Confabulation Cascades, terminal mimetic crisis, and the Knowledge Patch's architecture. Every downstream derivation presupposes them, and every downstream derivation's correctness depends on their being preserved exactly as stated here.

4.4.13 Crisis Geometry — How the Loop Tips Over

The components and the loop have now been specified. Decision Hub, Body Manager, Thinker, Observer, Belief Register, and the operational loop that couples them have each been described in turn, the Failure Mode has been named as Object Parasitism, and the architectural localisations and derived consequences have been worked through in §4.4.12. The loop's dynamics emerge as it runs over time.

A coupled loop does not just have a static description. It has a dynamic. Once the components are talking to each other, every signal one component sends affects what the next component does, which affects what the first component receives on the next cycle. Sometimes those cycles settle down. Sometimes they run away. Appendix A supplies the full mathematical treatment.

4.4.13.1 The Loop Has a Volume Knob

Imagine a microphone close to a loudspeaker. If the volume on the loudspeaker is low, the microphone picks up a small amount of sound, the loudspeaker plays it back at low volume, the microphone picks up an even smaller amount on the next cycle, and the sound dies away. If the volume is high, the microphone picks up the sound, the loudspeaker plays it back louder, the microphone picks up more on the next cycle, the loudspeaker plays it back louder still, and within a second or two the room is filled with the screaming feedback noise that everyone has heard at a school concert. The volume knob does not change what the microphone is doing or what the loudspeaker is doing. It changes how much of each cycle's output becomes the next cycle's input.

The coupled AIM loop has a volume knob of the same kind. It is called the Internal Loop Gain, and the symbol used for it in the formal treatment is G. G is not a single setting. The formal treatment expresses it as the product of four factors: the salience-to-M coupling, $κ_{S}$ (how strongly the ambient Salience field drives the Observer's M signal, upstream of the Decision Hub entirely); the Decision Hub's weight on M, $β$ (how strongly the integrated valuation weights the M-channel once that signal arrives; labelled $w_{M}$ elsewhere in this chapter); the Status feedback gain, $γ$ , from realised valuation to environmental salience; and the slope, $σ^{'} (V^{*})$ , at the current operating point, of the bounded monotone mapping from realised valuation to environmental salience. The slope is steep where a small change in how much an Object is wanted makes a large difference to how visible and status-bearing it becomes, and flat where valuation is too low to register in the field or where the field is already saturated. The fast loop closes from the M signal through the integrated valuation and the Salience field back to the M signal, and does not pass through the Thinker; the Thinker's ground-resolution sits on the slow Belief layer and couples into the fast layer through weights and likelihoods.

Because the four factors combine multiplicatively rather than additively, turning down any one of them turns down G as a whole. The Reafference Pathway specified in §4.4.9.1 is self-observation of executed action through the Observer; it is not the memory self-coupling $κ_{se l f}$ that contributes to $G_{s l o w}$ , and it is not carried as a diagonal term in the appendix's Observability channel (§4.4.9.1). The channels in §4.4.13.5 distinguish the within-Individual fast layer, the slow Belief layer, and the network layer accordingly.

4.4.13.2 Below the Tipping Point and Above It

Stay with the microphone-and-loudspeaker image. There is a particular volume setting, a particular threshold, at which the loop crosses from "fine" to "screaming feedback". Below the threshold, the system is stable: any chance noise dies away. Above the threshold, the system is unstable: any chance noise grows until it dominates everything else. The threshold is not a region. It is a sharp line. Engineers who design audio systems describe sound below the threshold as subcritical and sound above it as supercritical, and the same vocabulary is used here.

The AIM loop has thresholds of the same kind, and there are two questions to ask rather than one. The first is whether the loop is capable of running away at all, which is fixed by how strongly the four factors are set. The second is where the system is sitting, which is fixed by its history. Where the four factors are weak enough, the loop has one resting state at any level of input: perturbations of the fast valuation and salience loop decay, the Observer's Salience signals are absorbed without compounding, and there is no threshold to cross because nothing lies on the other side of it. Where they are strong enough, a second resting state exists, inflated far above the A-I functional baseline, and the loop can be carried into it.

The critical value at which the loop tips is $G_{cr i t}$ , and the formal treatment establishes that it is exactly 1, however fast or slow each stage of the loop responds: those speeds set how quickly a runaway develops, not where the threshold sits. Above it, each iteration of the loop loads the next with more M-charge than it received, and the Individual enters a cycle in which the desire-objects keep growing in apparent significance no matter how much is acquired. (The three-layer additive stability statement of Appendix A.5.7.4, $G_{t o t a l}^{(3)} = G_{e f f} + G_{s l o w} + G_{cr oss} + G_{n e t}^{Λ} + G_{n e t}^{K} < G_{cr i t}^{t o t a l}$ , is a leading-order engineering heuristic for weakly coupled loops: $G_{s l o w}$ carries the leading cross-coupling product of order $O (∥ C_{s f} ∥ \cdot ∥ C_{f s} ∥)$ , and $G_{cr oss}$ stands for the corrections of higher order in the cross-coupling norms. The possession-coupled expansion of Appendix A.9.5 adds the Treadmill and Model-Convergence gains and is written $G_{t o t a l}^{(H)}$ . The exact stability criterion in either case is the largest real eigenvalue of the full block-Jacobian. See Appendix A Sections A.5, A.5.7.4, and A.9.5.) This is not a metaphor. It is the same dynamical pattern as the loudspeaker, applied to a different system.

The runaway does not continue without limit. Real feedback does not either. The howl rises for a moment and then settles into a sustained scream, held at the ceiling of what the equipment can deliver. The AIM loop behaves the same way, because the mapping from realised valuation to environmental salience saturates: large valuations stop producing proportionately larger salience. The level of realised valuation at which that mapping has flattened to nothing is the Salience Ceiling. What the loop reaches after it tips is that second resting state. A system sitting there is not escalating. It has escalated, and it goes on sitting there under no further input, because a resting state is a place a system stays rather than a place it passes through.

The two thresholds are not in the same place, and anyone who has handled a microphone knows this from experience. The feedback starts at one setting of the volume knob. Turning the knob back to that setting does not stop it. To silence the room the setting has to go lower still, past the point where the feedback began, and only then does the howl cut out. The setting at which the loop ignites and the setting at which it releases are two different settings, and the release is the lower of the two. Hysteresis is the property of a loop whose state depends not only on where its inputs sit now but on where they have been, and it holds in the AIM loop for the reason it holds in the room: above the tipping point the loop has two resting states, and which one it occupies is fixed by the path it took. The formal statement is the Ignition-Release Asymmetry at Appendix A.5.3, which fixes the release level below the ignition level and shows the gap widening as the loop's capacity rises.

When the loop does release, the inflated valuation does not vanish. The drop in the integrated Scalar is a departure from what the Belief Register predicts, so it recruits the Thinker, and Ground-Resolution moves the Belief Register off the Hypothesis that had carried the inflated state onto the next most likely one, with recall of the old Hypothesis slowing the move without reversing it. How firmly the new Hypothesis is held depends on the Individual. Under pressure for Cognitive Closure it is adopted with near-certainty, and the templates it writes can direct the Observer's Salience to new Objects. An Individual exercising Epistemic Suspension holds it provisionally, with the competing Hypotheses still open (Appendix A, Section A.5.6).

The loudspeaker image carries one further fact. Feedback systems fail in one of two shapes: they run away, or they swing, overshooting, being pulled back, and overshooting the other way. Swinging has a strict prerequisite. Somewhere in the circuit there must be a component that answers a rise with a push in the opposite direction, the way a thermostat answers heat with cooling. Walk the AIM loop component by component and no such component exists: observation raises the Observer's Salience, Salience raises the Decision Hub's valuation, realised valuation raises Observability and Status, and Status raises Salience again, every connection passing a rise along as a further rise.

The slower Belief machinery does not supply the missing push-back either. The Monotone Belief-Race Lemma, stated formally in Appendix A, Section A.5.6 and developed at Chapter 13, establishes that Hypotheses competing for the Belief Register form a race with a blind referee: the support Evidence lends a Hypothesis does not depend on that Hypothesis's current popularity, so the race sorts steadily toward the best-supported Hypothesis and cannot circle, and recall of past belief, however strong and however delayed, slows the sort without reversing it (Appendix A, Section A.5.6); what entrenches an incumbent is that recall, which returns the prior as Evidence, and not the race itself, and a better-supported late entrant still gains ground from the moment it enters, because the grading stays blind. A circuit built entirely of one-way amplifiers, refereed blind, can sit quietly below its threshold or run away above it. It cannot swing where one Hypothesis dominates the belief state, which is where the Fold is derived; where belief is spread across several Hypotheses, the joint case is a named specialist task (Appendix A, Section A.7, Task 11).

4.4.13.3 Why the Tipping Point Is Hard to See Coming, and What Tipping Costs

The shape of this transition determines how much warning the system gives before it tips and how costly the tip is when it occurs. Some transitions are gradual: as the volume rises, the sound gets noticeably louder for a while, and a careful operator notices and turns it down before feedback begins. Other transitions are abrupt: nothing seems to be happening, the system seems stable, and then with a tiny final increment of the knob it tips suddenly into the supercritical regime, with no rise in level to warn of it. This abrupt kind of transition is called a Saddle-Node Fold, and under the architectural conditions established in Appendix A.5.3 (no restoring force on the M-channel, no error-correction under Source Opacity, monotone concealment until collapse, with the Hopf exclusion derived in closed form there and carried to the slow and network layers at A.5.6 and A.5.7.6), it is the default geometry the AIM loop's three-node skeleton produces. The level gives no warning. What changes is the recovery: near the threshold, each small disturbance takes longer to die away, and the variance of the output rises (Appendix A, Section A.5). That slowing is the only warning the Fold gives, and it can be read only by an operator who is measuring for it.

Two further features of the Saddle-Node Fold determine what recovery looks like after the system tips. The first is that the transition is a collapse rather than a degradation. The stable operating point the system had been running in does not become noisy or unreliable; it disappears. Once the fold has occurred, the operating regime the system used to occupy no longer exists, and the system is forced into a different regime entirely. Recovery is therefore not a matter of removing whatever pushed the system over the edge and waiting for things to settle. The conditions under which the original operating point existed have to be reconstructed before the system can return to it, and that reconstruction takes a long time. During the reconstruction, the system is not operating in a degraded version of its former mode; it is not operating in that mode at all. The colloquial verb that fits is "breaks": the system breaks at the fold, and rebuilding it back to the optimal operating range is a slow and effortful process distinct from the failure itself.

The second is that this collapse is genuinely different from the alternative dynamical regime, the architecturally inaccessible Hopf Sub-Case (Appendix A.5.3 establishes that Hopf requires introducing a new state variable with intrinsic negative feedback, or reversing the sign of one of the loop's four couplings, and is unreachable by any positive-coupling parameter choice within the three-node skeleton). Audio equipment separates the two conditions with which collapse is confused, and it separates them cleanly. When a signal is pushed harder than the equipment can cleanly reproduce, the peaks of the waveform flatten against the equipment's upper limit and the sound becomes harsh. What is holding the level here is the ceiling, not anything regulating the signal. This is clipping, and it is the everyday form of Saturation Hold: the system is not escalating, because it has nowhere left to rise, and it is not being governed, because nothing is pushing back.

A limiter does something else. It reads the level and reduces gain in proportion before the ceiling is reached, so the signal never flattens, the output stays clean, and headroom is preserved. This is the everyday form of a Governed Band, and it is what a restoring node supplies when it introduces a state variable of its own, acts on a coupling in the loop gain, and acts against rather than with, at moderate gain and lag. The Hopf Sub-Case is the limiter. It is not the clipping.

The difference between the two decides what happens next. A limiter runs indefinitely: it holds the level below the ceiling, and when the input falls the limiter stops acting and the system returns by the path it came. Clipping does neither. Equipment held against its ceiling eventually fails, a component overheats, a driver tears, a transformer burns out, and that failure is the breakage the Saddle-Node Fold names. Nor does clipping release the way a limiter does. The signal has to fall further than the level at which the flattening began before clean reproduction resumes, which is hysteresis in audible form. A system in Saturation Hold is therefore not in a milder condition that a further push would worsen. It is in the supercritical condition already, held there by its own ceiling, with the fold beneath it and the way back longer than the way in. Sustained societal conflict is what this looks like at scale, and it is the period during which collapse is approaching rather than the period during which collapse has been averted.

The breakage the Saddle-Node Fold names, at the civilisational scale at which the AIM Framework's later chapters operate, is not abstract. The human system has nuclear weapons. The collapse-mode available to a sufficiently escalated supercritical regime at the civilisational scale is nuclear war, and nuclear war is a Saddle-Node Fold from which the operating point of organised human civilisation does not recover within any timescale that matters to the Individuals alive at the moment of collapse. The Saddle-Node Fold is the default geometry for the AIM instantiation, and it is the geometry the rest of the AIM Framework's crisis-derivations presuppose; the Hopf Sub-Case is the lower-stakes alternative that is architecturally inaccessible within the three-node skeleton (Appendix A.5.3), reachable only by augmenting the architecture with a fourth state variable carrying intrinsic negative feedback, or by reversing the sign of a coupling in the loop. Where that augmentation is installed and correctly tuned it holds the system in a Governed Band, and a Governed Band is a destination. Saturation Hold is not, and the two are routinely confused because both look bounded from outside. When the AIM Framework speaks of subcriticality and of Subcriticality Maintenance, it is speaking of keeping the system below the threshold at which sustained conflict ignites the Fold that ends civilisation in nuclear war.

4.4.13.4 Why the M-Channel Is Architecturally Parasitic

The Failure Mode subsection at §4.4.11 named the failure mode as Object Parasitism. Five properties of the M-channel as it operates in the loop together are what the formal treatment calls the Parasitism Theorem.

First, the M-channel does not turn itself off. The Body Manager turns itself off when the deficit it signals is repaired (the hunger goes away when the Individual eats). The Thinker's contribution closes when the process it is running completes, the state it was for being reached, and a process also ends on the exhaustion of resolution, which is boredom, or on interruption (Section 4.2). The M-channel has nothing equivalent. There is no condition the M-channel can reach that makes it stop firing.

Second, the M-channel does not have its own dedicated apparatus. The Body Manager has the body. The Thinker has its priors. The M-channel works by borrowing the operating channels of the other components: it loads the Observer's input stream, it rides the Decision Hub's integration, and it rides the mapping from realised valuation to Observability and Status. It has no equipment of its own; it uses everyone else's.

Third, the M-channel does not stabilise the loop it is part of. It loads the loop without damping it. A signal that loads a loop without damping it is a signal that increases gain without contributing anything that pulls the loop back toward equilibrium.

Fourth, the M-channel has no say over its own activation threshold. The threshold $G_{cr i t}$ is fixed at 1 by the topology of the loop, and the gain that is measured against it is made entirely of host parameters (Appendix A, Section A.5.4, Property 4). M cannot decide to activate itself or to refrain from activating itself. Both the threshold and the gain are set elsewhere.

Fifth, once the loop is above threshold, the M-channel does not need its original trigger any more. The cycle sustains itself once ignited. The Model that was being observed at the time can disappear, and the cycle continues; removing Models lowers the operating point but does not remove the cycle, provided the point at which the inflated state would release lies below the forcing the remaining Appetitive and Intrinsic inputs supply; where it lies above, removing Models releases the system to its functional baseline (Appendix A, Section A.5.4, Property 5). This is why removing the originating trigger does not stop the failure mode: the cycle has become parasitic on the host architecture, not on the trigger.

These five properties together are what makes Object Parasitism architectural rather than contingent. They are not extra claims; they are five different ways of reading the same loop equations.

4.4.13.5 The Five Channels Through Which the Volume Can Be Turned Down

G is not the only quantity that determines whether the loop runs away. G governs the loop within a single Individual at the fast valuation timescale, and it is the product of four factors: the strength with which Observer-borne Salience loads the Decision Hub's M-input, the Decision Hub's weight on the M-channel, the Status feedback gain from realised valuation to environmental salience, and the slope of the valuation-to-salience mapping at the current operating point. The Reafference Pathway is self-observation through the Observer and is not identified with the slow-layer memory self-coupling $κ_{se l f}$ (§4.4.9.1, §4.4.13.1). That is the within-Individual fast layer.

There are two further layers at which the same loop logic operates, the within-Individual slow layer at which Bayesian Belief revision runs on a longer timescale, and the network layer at which Beliefs and observed actions propagate between Individuals, and each of those layers has its own loop-gain quantity governed by its own architecture. To leading order, the stability condition combines the three layers additively; the exact condition is the largest real eigenvalue of the full block-Jacobian. Subcriticality requires the total to sit below the threshold, not each individual layer in isolation, and an intervention that reduces the within-Individual fast loop gain G is one of the five channels through which the Knowledge Patch turns the total down. The architecture's full set of loci (Appendix A, Sections A.5.1 and A.5.5) also includes reducing the observed Model stream itself, the Status feedback gain, and the salience-to-M coupling, and the Institutional Buffer that sits outside the loop.

The first channel works inside the Individual at the fast valuation timescale, and it is the one the I-Override gate developed in Chapter 18 installs. The Patch introduces a gate between the Decision Hub's integrated valuation and motor output at which the Individual's diagnostic competence can intercept an action that the integrated valuation would otherwise produce. The diagnostic competence does not work by introspectively identifying the felt valuation as M-loaded, which Source Opacity makes impossible; it works on situational pattern-recognition, the Individual's learned capacity to recognise the kinds of situations in which M-loading is statistically likely (the presence of Models, the Salience of comparative-position information, the absence of a homeostatic deficit that would warrant the A-reading, the absence of an Object-requirement of the Thinker's that would warrant the I-reading), and to apply heightened scrutiny in those situations even though the source of the felt valuation cannot be directly perceived. The heightened scrutiny is the recruitment of the Thinker: a strong valuation in a situation recognised as M-likely presents the prediction-mismatch that recruits the Thinker before motor output, and only a recruited Thinker can perform the I-Override (Section 4.3.3). The pattern-recognition is itself the cached output of the slow Bayesian work the next two channels install; channel one is the fast-layer deployment of the diagnostic competence that channels two and three build at the slow layer, with the Belief Register of §4.4.8 supplying the bridge between them.

When the gate intercepts, the effective weight the Decision Hub places on the M-input drops, and because the within-Individual fast loop gain G is proportional to that weight, G drops with it. The formal treatment calls this β-reduction, because β is the formal name for the Decision Hub's weight on M (labelled $w_{M}$ elsewhere in this chapter and Chapter 9), and the Patch reduces it in proportion to the Patch-Interception Probability, the probability that the diagnostic competence intercepts a candidate action given the integrated valuation and the environmental salience (Appendix A.6.4): the effective weight is the share of β that survives interception, and the reduction is operative whenever the gate is active. The reduction is not a structural rewiring of the Decision Hub; it is the architectural consequence of an Individual exercising the situational pattern-recognition the Patch installs at the moment between valuation and action.

The second channel works inside the Individual on the slow Bayesian layer rather than the fast valuation layer, and it is the architecturally prior channel: nothing channels one or three accomplish is possible without it. The Thinker's Bayesian ground-resolution runs over a Hypothesis Space, the set of candidate explanations available to the Thinker for the felt valuation it has received from the Decision Hub. If the Thinker does not have the Hypothesis available that the felt valuation may have been produced by Mimetic Desire rather than by a homeostatic deficit or a deficit of what a process the Thinker is running requires, the Bayesian update concentrates posterior weight on the Hypotheses it does have, and the resulting Evidence the Thinker emits into the Belief Register is more strongly M-driven than the Thinker can detect. AIM Vocabulary Acquisition, the explicit learning of the architectural vocabulary stated in this treatise, enlarges the Hypothesis Space by adding the M-Hypothesis to the set the Thinker can entertain. The formal treatment expresses this as a reduction in the Restriction-Induced Concentration Factor that previously inflated the Bayesian layer's contribution to total loop gain.

Three consequences follow. First, enlarging the Hypothesis Space attenuates that Bayesian-layer contribution directly, which lowers the total. Second, with the M-Hypothesis available as a live direction, the Two-Question Diagnostic becomes deployable: it can route probability mass through the expanded subspace toward H_M where the underlying motivation is M-driven, which it could not do in the restricted subspace. Third, the Thinker's posteriors about which Objects in the Individual's experience are M-charged become accurate where they were previously concentrated on the wrong Hypotheses, and the Belief Register of §4.4.8 propagates those updated posteriors to the Observer's live Salience computation at preconscious timescales. The Observer, operating on accurate priors, attaches Salience appropriately rather than inflating Salience over Objects the cached priors have not yet diagnosed. This is the architectural mechanism by which slow-layer Bayesian work calibrates fast-layer Salience computation, and it is the route by which channel two's work reaches channel one.

The third channel also works inside the Individual at the slow Bayesian layer, alongside the second, but it operates on a different feature of Bayesian updating rather than on the Hypothesis Space. Once the M-Hypothesis is in the Hypothesis Space, the Thinker can hold open the question of which Hypothesis best explains the felt valuation, rather than collapsing immediately onto whichever Hypothesis the prior favours. Holding the question open is Epistemic Suspension, and Epistemic Suspension has an architecturally measurable consequence: the residual coherence between competing Hypotheses (the Interference Term in the formal treatment) is sustained for longer rather than being prematurely resolved. A sustained Interference Term reduces the rate at which the Thinker emits high-confidence Evidence propositions into the Belief Register, which is the slow-layer analogue of the I-Override gate's effect at the fast layer. The formal name is a reduction in the Evidence-Emission Coefficient.

The Two-Question Diagnostic of Chapter 18 deploys across channels two and three together: it routes probability mass through the subspace channel two has expanded, under the maintained interference that channel three sustains, which together prevent the narrative-closure drive from collapsing the Bayesian update onto the wrong Hypothesis before the diagnostic can complete. As with channel two, the Patch-corrected posteriors that result from this work are cached and propagated to the Observer at preconscious timescales, so channel three's slow-layer contribution also reaches the fast loop through the Belief Register. Over time, as the Patch becomes habitual, the work that was initially deliberative becomes preconscious, the diagnostic operates faster than the action it is diagnosing, and the gate's interception decisions ride on a Belief Register whose contents the slow-layer work has continuously updated. This is the architectural picture of the Patch as a discipline: not a single intervention but a continuous coupling between slow-layer Bayesian work and fast-layer salience-and-action computation, mediated by the Belief Register.

The fourth channel moves out of the Individual and onto the network. Individuals do not update their Beliefs only from their own internal valuations; they update from the Evidence propositions other Individuals emit into shared discourse, and the weight an Individual places on a given other's Evidence depends on the relationship structure between them. The aggregate of those between-Individual weights is the off-diagonal Evidence-coupling matrix, and its spectral radius, the long-run factor by which Evidence circulating through the whole network grows or shrinks with each round of transmission, governs the network's contribution to total loop gain. When Individuals have acquired Patch-competent Evidence intake, the weight they assign to others' M-driven Evidence is attenuated, because they can recognise the M-driven character of that Evidence and discount accordingly. The aggregate effect across the population is a reduction in the spectral radius of the off-diagonal coupling matrix. This is the slow-layer analogue of the second and third channels lifted to the population scale, and the formal treatment names it as a reduction in the network Evidence-channel loop gain. Individual Patch adoption across channels one to four is necessary but not sufficient for network-scale subcriticality (Appendix A, Section A.5.7.7).

The fifth channel also works at the network scale, but on the Observability matrix rather than the Evidence-coupling matrix. Executed actions and observed Models propagate between Individuals through the Observability Pathway, and the spectral radius of the Observability matrix governs the rate at which a realised valuation in one Individual becomes an observed Model signal for another. Observability Governance, the institutional and platform-design interventions that attenuate this propagation (Blind Pricing, audience-removal architectures, visibility-suppression of Model classes, ), reduces the spectral radius of the Observability matrix and lowers the corresponding contribution to total network loop gain. The formal name is a reduction in the network Observability-channel loop gain. The fifth operates at a scale no Individual's intervention reaches, and it is the complement that closes the gap between individual Patch efficacy and network-scale stability (Appendix A, Section A.5.7.7). Anonymising consumption outcomes and decoupling valuation from its Observability signature act instead on the Status feedback gain inside each Individual's loop (Appendix A, Section A.5.1).

The five channels are not redundant. Each operates on a different parameter, each has different practical preconditions for installation, each is robust to different categories of failure in the others, and each addresses a scale or layer the others do not. Channels one through three are the Patch's three-component specification at the within-Individual scale, the fast-layer I-Override gate, the slow-layer enlargement of the Hypothesis Space, and the slow-layer Epistemic Suspension protocol. Channel four is exercised by Individuals at the network scale, through Patch-competent Evidence intake, and channel five is the institutional and collective-action lever required for population-scale Subcriticality. Subcriticality Maintenance is a multi-channel discipline operating across the within-Individual fast layer, the within-Individual slow layer, and the network layer; it is not a single-instrument intervention, and no single channel is sufficient on its own.

4.4.13.6 Subcriticality Maintenance, the Realistic Goal

Turning down G is not enough on its own, for two reasons. A loop resting anywhere is already resting at a gain below its threshold, so a low present gain is what resting looks like rather than what safety requires. The demand is either that the four factors are brought low enough that no inflated resting state exists at any level of input, which removes the fold rather than deferring it, or, where that is out of reach, that the system is held on the low state with margin enough to absorb foreseeable perturbation. The system encounters perturbations (a Reference Set shifts, a new Model appears, a piece of cultural content elevates the M-charge of some Object-class), and a configuration that rests on the low state may be carried across by one of them. The Total Stability Condition states the threshold the margin is measured from.

The second reason is that human Individuals do not exist alone. They exist in networks, and the same dynamic operates at the network scale: when inter-Individual coupling is high enough, Beliefs cascade through the population in self-reinforcing rather than self-correcting ways. The threshold value of inter-Individual coupling above which this happens is called the Critical Coupling Strength $κ_{c}$ . Formal treatment of the Total Stability Condition is at Appendix A.5.7.4, and of the Critical Coupling Strength at Appendix A.5.7.5; Chapter 13 develops the population-scale $κ$ and $κ_{c}$ framework on those formal foundations.

The objective is not the elimination of the M-channel. The architecture does not permit M-elimination. The Observer's distinctive input stream, the Decision Hub's tag-stripping integration, and the Thinker's Bayesian ground-resolution under an impoverished Hypothesis Space are load-bearing features of human cognition. They cannot be removed without disabling the architecture's ability to navigate a social world at all. The objective is to keep G below $G_{cr i t}$ and κ below $κ_{c}$ indefinitely, through ongoing application of the five channels, while accepting that M-driven Salience will continue to enter the loop at every cycle.

This is what Subcriticality Maintenance means, and the stakes named at the close of §4.4.13.3 are why it matters. The Saddle-Node Fold available to the human system at the civilisational scale is nuclear war. Sustained societal conflict is not an oscillation that precedes the Fold, because the three-node loop cannot oscillate. It is Saturation Hold with the Fold beneath it, held only by a ceiling, and every perturbation it absorbs is taken from the margin that separates it from the Fold (Section 4.4.13.3; Appendix A, Section A.5.3). Subcriticality Maintenance is the architecturally available discipline by which the human system is prevented from reaching the threshold at which the conflict that ends civilisation ignites. The aim is not the elimination of conflict, the elimination of M, or the transformation of human nature. The aim is vast improvement over the architecture's default supercritical trajectory and indefinite deferral of the terminal mimetic crisis the supercritical regime would otherwise produce.

Hysteresis fixes the price of the second of the two demands, and the price is not symmetric with the price of prevention. Because the release threshold sits below the ignition threshold, the reduction required to bring an ignited system back exceeds the reduction that would have prevented the ignition, and the further the system has travelled, the wider that difference becomes. Subcriticality Maintenance is hard for this reason rather than because the individual interventions are difficult, though some of them are. A system that has crossed does not return along the path it took.

Humanity is currently in a supercritical state. The four factors are set such that an inflated resting state exists, and the system is sitting in it rather than in the low one. What it exhibits instead of immediate collapse it owes to the institutional structures a durable polity accumulates, which supply some part of the restoring structure the loop itself lacks, at pressures chosen for other reasons by people who are not reading the quantity that governs the terminus. The system is clipping, which is to say it is in Saturation Hold: bounded by its own ceiling rather than by anything governing it, with the fold beneath.

Partial uptake of the Knowledge Patch is sufficient to reduce G, but it is not sufficient on its own. Ongoing maintenance is required to keep the reduction in place against parameter drift, against rising stakes, and against the technological reality that the collapse-mode the human system can now reach has consequences the species cannot recover from. The detailed operational specification of the five channels and of Subcriticality Maintenance as a civilisational discipline is the work of Chapter 18. The architectural foundation on which that work depends is the crisis geometry stated here.

4.4.14 Forward Pointer — Formal Treatment in Appendix A

The Minimal Architecture Specification has been set out at the level at which the rest of this treatise depends on it: the reading-surface level at which the four components, their signed valuations, the Belief Register mechanism, the operational loop, and the architecture's characteristic failure mode are specified in functional terms. Readers with a mathematical, engineering, economic, or formal-modelling background who wish to see the operation of the architecture at each level expressed in formal terms, including the integration mapping at the Decision Hub, the many-to-one summation that yields Source Opacity, the Body Manager's contribution treated as an exogenous (or quasi-exogenous) drive relative to the fast M–V–S loop — the Body Manager tracks physical and material Referents, so for loop-gain analysis its signal enters as external forcing rather than as a node of the internal cycle; the full Body Manager ODEs lie outside the minimal skeleton — the Bayesian update mechanics of the Thinker, the conditional Salience function of the Observer, and the coupling of these operations under the operational loop, are directed to Appendix A, which provides the formal treatment.

Appendix A is supplementary to the main argument; the chapters that follow do not require it, and a reader without the relevant background loses nothing by reading past it. For readers who have the background, Appendix A makes precise what this Section 4.4 has specified functionally, and provides the formal apparatus on which several of the later derivations, in particular the Source Opacity Theorem in Chapter 9 and the κ and τ analyses developed in subsequent chapters, depend.

4.4.15 How Far the Neural Commitments Extend

The architecture just specified makes contact with the brain at two points, with a third held at best-current-mapping. The reach of the neural commitment is narrower than the architecture's explanatory ambition might suggest, and the narrowness is deliberate.

The first point of contact is the integration locus. The claim that the three sources converge on a single common-currency valuation, localised in the human brain to the ventromedial prefrontal cortex and the ventral striatum, is an established finding, replicated across single-neuron recording, functional imaging, and meta-analysis. The second point of contact is the existence of three dissociable upstream pathways feeding that locus, which lesion, pharmacological, and imaging studies have repeatedly shown can be disrupted independently of one another. These two claims, convergence on the Decision Hub and dissociability of the inputs, are the whole of what the architecture asserts as established neuroscience.

Everything else in the specification is a description of what the system does, not a claim about how the brain is wired to do it. When the foregoing sections speak of the Body Manager, the Thinker, and the Observer, they name functional roles that are separately identifiable in behaviour and in association dynamics to the precision the framework's derivations require. They do not name anatomically bounded modules that could be excised one from another. The underlying neural systems overlap, interact continuously, and are more entangled than the four-box loop diagram of this chapter depicts. A reader who took the diagram for a wiring schematic would be reading more into it than the architecture claims, and more than the neuroscience would support.

The upstream localisations that can be offered (hypothalamic and interoceptive circuits for bodily deficit, the dopaminergic midbrain projection for intrinsic engagement, and mirror-neuron and social-reward systems for the observation of others) are the best current account of where these functions are realised, but the architecture does not rest on them. It survives their revision, because its load-bearing claims are about the functions and their integration, not about their addresses in the cortex. The relaxability of these substrate assumptions, including the mirror-neuron substrate and the preconscious timing window, is made precise in Appendix A.6, which shows that the qualitative behaviour of the architecture is preserved when they are relaxed.

Source Opacity occupies a different position again. It is not an empirical neuroscientific claim at all, and so it is not exposed to neuroscientific refutation. It follows by necessity from the fact that integration reduces three source signals to one Scalar: once three numbers have been summed into one, the summands cannot be recovered from the total. This is the Data Processing Inequality applied to the integration step, a mathematical entailment rather than a contingent feature of brain design. The only way to deny Source Opacity is to deny that integration is many-to-one, and that is the very thing the established common-currency finding affirms.

That immunity is not a self-granted exemption. Source Opacity is not exposed to the record directly, but the premise it follows from is: many-to-one integration at the Decision Hub is an empirical claim about the human system, and it stands or falls on the evidence. Withdraw the premise and Source Opacity falls with it. The result is protected against refutation by a finding about self-knowledge, not against refutation as such.

The same holds of the architecture as a whole. The functional descriptions of this chapter are a legibility layer over a falsifiable base, not a device for insulating the framework from it. Because the framework's substantive results are derived from the empirical record rather than merely made consistent with it, they are exposed to falsification by it: a finding that overturns a load-bearing premise refutes the derivations resting on that premise, and the affected structure must then be revised or abandoned. This is a different case from a change of citation: a finding that is overturned takes its derivations with it, whereas a better-sourced account of an output that still stands changes only what is cited for it. And where the record shows more structure than one of these functional descriptions represents, that structure is real and the description is the deliberate simplification, not a competing claim.

The upshot is that this architecture accepts the empirical record of the neuroscientific subdisciplines it draws on, while declining to adopt or to contradict any one discipline's preferred theoretical architecture. The findings are taken as given; the architecture built around them is treated as a separable matter, open to refinement as the science develops. Where a better account of an established output supersedes a citation relied on here, the citation is what changes, not the architecture. No claim made in this chapter runs against the empirical record as it stands.

A note for readers with a formal background. The compatibility position stated here has a precise mathematical counterpart. Readers from systems engineering, control theory, applied mathematics, and related fields who wish to see exactly how the architecture specified in this chapter behaves when its substrate assumptions are made explicit and then relaxed are directed to Appendix A, where the integration mapping at the Decision Hub, the many-to-one summation that yields Source Opacity, and the coupling dynamics of the operational loop are given formal treatment. In particular, Appendix A.6 sets out which assumptions are load-bearing and which are relaxable, and demonstrates that the loop-gain stability condition does not depend on the timing window or the anatomical specifics by which the functions are realised (Appendix A, Section A.6.2); the bifurcation type depends on the form of the valuation-to-salience mapping, which Appendix A, Section A.5.3 constrains architecturally. The formal apparatus is what allows the claim "the architecture survives substrate revision" to be stated as a derived result rather than an assurance. Appendix A is supplementary: the argument of the chapters that follow does not require it, and a reader without the relevant background loses nothing by reading past it.

4.5 The Two Outputs of Integration — Scalar and Affect

The account of the Motivational Architecture given so far describes integration as emitting a single Common-Currency Scalar that drives action-selection. This is accurate as far as action is concerned, but it is incomplete. A single integration event emits two distinct products, not one. The first is the Common-Currency Scalar, the felt urge to act, the magnitude whose threshold-crossing determines action. The second is Phenomenal Affect, the felt state that accompanies it. These are joint products of one integration, computed from overlapping inputs, but they are not the same quantity, and they can dissociate. For action-selection, which is the AIM Framework's primary concern, the Scalar alone is sufficient; where the subject is felt experience, both outputs are required. The AIM Framework uses "affect" rather than "emotion" throughout this discussion because affect is the broader category, encompassing sensations, moods, and felt bodily states as well as discrete emotions; the architecture emits felt states of all these kinds, and the wider term is required to cover them.

Two consequences of the architecture bear on affect, and both are entailed rather than merely observed. The first is that the Source Opacity Theorem applies to Phenomenal Affect exactly as it applies to the Scalar. The tags identifying the Appetitive, Intrinsic, and Mimetic contributions are destroyed at integration, and both outputs are emitted after that destruction. The Individual therefore experiences a felt state whose provenance is architecturally unrecoverable by introspection. The state is real, fully formed, and mute as to its own cause. The two outputs differ, however, in duration. The Scalar resolves rapidly at action-selection and leaves no standing object behind, whereas Phenomenal Affect persists beyond the ignition event that produced it. Because the affect endures, the Thinker has a standing object to examine, revisit, and correlate against context and memory, which is why a richer, though still indirect and inferential, understanding can be built of an affect's source than of a motivation's source, even though both were equally tag-stripped at integration.

The second consequence follows from the nature of the Thinker. The Thinker is, by definition, the slow, serial, deliberative faculty. It cannot, by its own functional character, act on either output of integration before that output exists, because deliberation is posterior to the fast automatic process that produces the outputs. This lateness is analytic. It is entailed by what the Thinker is, not contingent on any empirical measurement, and the two-stage structure of emotional processing found in the affective sciences corroborates it without being its ground.

The Thinker's relation to an affective episode is therefore fixed: it receives Phenomenal Affect as an accomplished fact and can operate only downstream of the experience, in two ways. It can process the affect after the fact, stripping the fast confabulated meaning and installing an accurate account (the Confabulation Race), and it can revise the prior carried by the loop so that the next Perturbation of the same type is integrated against a more accurate Hypothesis. This prospective revision can both attenuate the magnitude of subsequent affect, by lowering the Salience the Observer assigns to the triggering event, and shorten the Decoherence Time of its abatement. What the Thinker cannot do, in principle and not merely in practice, is prevent, eliminate, or act upon the initial affect once emitted, because that affect is upstream of everything the Thinker is constituted to do.

The two outputs also differ in what becomes of them. The Scalar converts to action (4.4.1), and the action is where it becomes observable. Phenomenal Affect undergoes no such conversion. Phenomenal Affect is emitted and it persists, and it persists as an internal object, available to the Thinker to examine and to no one else except through what the Individual subsequently does or says. This fixes where the AIM Framework's diagnostics attach. The Behavioural Signatures read the transduced output, which is why they can be applied by a third party and why they survive an Individual's inaccurate account of themselves; anything read off affect passes through the Thinker's account of it, which is the channel Confabulation occupies. Both outputs were tag-stripped at the same integration and neither can be run backwards to its sources. They differ not in what they retain but in whether they leave the body.

What leaves the body has to be followed, and the junctions at which it is converted from one form into the next are called couplings. The term carries this sense throughout, and the formal treatment names the same junctions. Four of them carry the account from the integration event specified in this chapter to the scale at which civilisations hold or fail.

Three source signals are converted into one Common-Currency Scalar at the Decision Hub. This is the integration event, and it is where the source tags are lost.
The Scalar is then converted in two directions at once. Outward it becomes action, which is to say the direction in which the body's available energy is spent. Inward it becomes an item of Evidence on which the Thinker's Bayesian updating operates, which is the junction given formal treatment at Appendix A.5.6.
The action, once outside the body, is observed by others. Observation of another's goal-directed pursuit is an input to the observer's own architecture and not merely a report about the world: it enters as an observed Model signal, through the Observability channel, and what the actor says of it enters the observer's Belief Register as Evidence, weighted by Coupling Strength. The formal statement of both channels is at Appendix A, Section A.5.7.3.
Evidence flowing between Individuals accumulates into institutional structure, and that structure then governs the supply of Evidence reaching every Individual inside it.

The chain reaches civilisational scale without changing its subject matter. Each Individual's output is another Individual's input, so the argument about the quality of what enters a body is not a claim about one person in isolation; it is a claim about a circuit that closes. A population is a set of bodies converting each other's actions into their own Evidence, and the collapse in which those dynamics terminate is not a new topic taken up in the later chapters. It is the same signal, several couplings downstream, still obeying the conversions specified at the first one. The formal counterpart is derived rather than asserted: the Saddle-Node Fold established for the single Individual is preserved when the account moves to a coupled population, and preserved by derivation rather than by analogy, at a decohered operating point under one-signed coupling (Appendix A.5.7.6). Nothing has to be added at civilisational scale that was not already present at the neural one, and that is why one derivation chain is enough to run the whole distance.

4.6 From Foundation to Argument

The foundational chapter closes here. Sections 4.1 through 4.4 have done four things.

Section 4.1 established why AIM is presented axiomatically rather than narratively: because the conclusions are confronting and must be examinable, because the claims must be falsifiable and must be precise, because the foundation must be parsimonious and must be minimal, because the derivation chain is the theory and must be visible, and because the domain includes the reader's own cognition and cannot rely on introspective assent.

Section 4.2 reconstructed the derivation by which the nine axioms were arrived at, beginning from the concept of Mimetic Desire as a triangular structure, identifying the mirror neuron system as a candidate substrate, locating the integration locus at which the mirror neuron system converges with the homeostatic-deficit signalling system and the dopaminergic intrinsic-engagement system, deriving the satiation and termination dynamics that distinguish the three convergent systems, deriving Source Opacity as a structural consequence of integration, deriving the phenomenological diagnostics from how each system's signal presents in conscious experience, and deriving the remaining axiomatic structure from the architectural and dynamic foundations.

Section 4.3 stated the nine axioms in their canonical form, with foundational citations identifying the empirical basis on which each axiom rests.

Section 4.4 specified the Minimal Architecture Specification (MAS): the Decision Hub at which integration occurs, the Body Manager whose homeostatic-deficit signalling is the substrate of Appetites, the Thinker whose conscious-cognitive engagement and Bayesian inference is the substrate of Intrinsic Motivation, and the Observer whose continuous conditional Object-field Salience assessment is the substrate of Mimetic Desire, together with the Belief Register mechanism that mediates downstream access to the Thinker's posteriors at preconscious timescales, the operational loop that couples the components, and the architectural localisations and derived consequences that follow.

Taken together, these four sections fix what the human motivational system is, at the level of the Dashboard Gauge Principle articulated in Section 4.4. They fix the system as a four-component architecture in which three motivationally distinct sources converge on a single integration locus, in which the integration is a many-to-one map that strips source-tags from the integrated signal, in which the three sources operate on different timescales and terminate by different mechanisms, in which Mimetic Desire is acquired preconsciously and operates by inflating the perceived value of Objects already valued by the other two sources, in which the conscious-cognitive system updates Bayesianly on an evidence base that is itself partly produced by the architecture, and in which the architecture's characteristic failure mode is the propagation of confabulated narratives about the causes of choice, both within Individuals and across the social networks in which Individuals are embedded. These claims are not yet the disaggregation of human motivation as an argument; they are the architectural premises from which that argument is mounted.

The disaggregation argument is the work of Chapter 5. It is a separate undertaking because it is doing separate work. The foundational material specifies what the system is. The disaggregation argument specifies why the system must be carved into Appetites, Intrinsic Motivation, and Mimetic Desire as three distinct sources, rather than treated as undifferentiated wanting, treated as a two-source system, or treated as a four-or-more-source system.

Two points bear on the relationship between Chapter 4 and the rest of this treatise.

The first point is that the axioms in Section 4.3 and the Minimal Architecture Specification in Section 4.4 are load-bearing. Every chapter from Chapter 5 onward derives its claims from these foundations, not from independent observations. The Source Opacity Theorem in Chapter 9, the analysis of Intrinsic Motivation in its dedicated chapter, the analysis of Mimetic Desire and the Camouflage Architecture in greater depth, the treatment of Bayesian Belief dynamics under the coupling-strength and decoherence-time parameters, the analysis of institutional capture and civilisational risk, and the Knowledge Patch that supplies the compensatory mechanism for routing around Source Opacity, are all derivations from the axioms and the architecture stated in this chapter. The reader who finds a later claim surprising or contestable is invited to trace the claim back to the axioms and the architectural specification on which it depends, and to identify whether the disagreement is with the derivation or with the foundation. If with the derivation, the derivation can be re-examined; if with the foundation, the foundation can be contested directly. This is what the axiomatic presentation makes possible, and it is what later chapters depend on. The challenge is welcomed.

The second point is about what this foundational chapter does not yet do. The architectural specification establishes that the human motivational system has the structure described, but it does not establish what the consequences of that structure are at the scale of the Individual's life, the institution's operation, the network's dynamics, or the civilisation's trajectory. Those consequences are the substantive concern of the rest of this treatise, and they are the work of the chapters that follow.

The three motivational systems specified here, Appetites (A), Intrinsic Motivation (I), and Mimetic Desire (M), together with the integration architecture that combines them, the satiation and termination dynamics that distinguish them, and the Source Opacity, Confabulation, Object Parasitism, and Bayesian-Belief-dynamics consequences that follow from their interaction, provide a principled basis for the disaggregation of human motivation. The disaggregation is not an interpretive choice imposed on undifferentiated motivational data; it is the architectural fact that the motivational system actually has three distinct sources operating on different principles, and that the difference between the sources is consequential for what the Individual does, what the Individual reports about what they have done, and what consequences follow at every scale at which Individuals act. Chapter 5 mounts the argument that establishes this disaggregation against the alternatives. The architecture stated here is what makes the argument possible.

What Would Falsify This?

The claims of this chapter are the framework's axioms and the minimal architecture, so their falsification conditions are the framework's exposed surface rather than a single chapter's. That surface is consolidated at Appendix B, Section B.2, as six load-bearing commitments, each with its disconfirming observation: three-source exhaustiveness (a genuine fourth source passing the decisive test of Chapter 5); the source-distinctness and preconscious transmission of M; Source Opacity and its introspective inaccessibility; Differential Satiation; the criticality dynamics of Appendix A (including the fold-not-oscillation signature); and the confabulation predictions with their agent-based default. The two operational falsifiers that carry the near-term burden, Interaction Necessity and Temporal Mediation, are stated at Chapter 8 (What Would Falsify This?), and the confabulation cluster's criteria at Chapter 11. In addition, this chapter's own instantiation claim is separately falsifiable as stated in Section 4.4: the human system might fail to instantiate the minimal architecture (for example, if Salience computation were found not to be conditional on priors, or prior-maintenance were found distributed across the A and M components) while the architecture remained a valid specification for some other system.

References

Adams, C. D., & Dickinson, A. (1981). Instrumental responding following reinforcer devaluation. Quarterly Journal of Experimental Psychology Section B, 33(2b), 109–121.

Allport, G. W. (1954). The Nature of Prejudice. Addison-Wesley.

Balleine, B. W., & O'Doherty, J. P. (2010). Human and rodent homologies in action control: Corticostriatal determinants of goal-directed and habitual action. Neuropsychopharmacology, 35(1), 48–69.

Barrett, J. L. (2004). Why Would Anyone Believe in God? AltaMira Press.

Bartra, O., McGuire, J. T., & Kable, J. W. (2013). The valuation system: A coordinate-based meta-analysis of BOLD fMRI experiments examining neural correlates of subjective value. NeuroImage, 76, 412–427.

Berridge, K. C. (2004). Motivation concepts in behavioral neuroscience. Physiology & Behavior, 81(2), 179–209.

Block, R. A., & Grondin, S. (2014). Timing and time perception: A selective review and commentary on recent reviews. Frontiers in Psychology, 5, 648.

Cabanac, M. (1971). Physiological role of pleasure. Science, 173(4002), 1103–1107.

Caggiano, V., Fogassi, L., Rizzolatti, G., Casile, A., Giese, M. A., & Thier, P. (2012). Mirror neurons encode the subjective value of an observed action. Proceedings of the National Academy of Sciences, 109(29), 11848–11853.

Campbell-Meiklejohn, D. K., Bach, D. R., Roepstorff, A., Dolan, R. J., & Frith, C. D. (2010). How the opinion of others affects our valuation of objects. Current Biology, 20(13), 1165–1170.

Cannon, W. B. (1932). The Wisdom of the Body. W. W. Norton.

Cannon, W. B., & Washburn, A. L. (1912). An explanation of hunger. American Journal of Physiology, 29(5), 441–454.

Chartrand, T. L., & Bargh, J. A. (1999). The chameleon effect: The perception–behavior link and social interaction. Journal of Personality and Social Psychology, 76(6), 893–910.

Chib, V. S., Rangel, A., Shimojo, S., & O'Doherty, J. P. (2009). Evidence for a common representation of decision values for dissimilar goods in human ventromedial prefrontal cortex. Journal of Neuroscience, 29(39), 12315–12320.

Csikszentmihalyi, M. (1990). Flow: The Psychology of Optimal Experience. Harper & Row.

Davidson, D. (1963). Actions, reasons, and causes. The Journal of Philosophy, 60(23), 685–700.

Deci, E. L., & Ryan, R. M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. Plenum Press.

de Wit, S., Kindt, M., Knot, S. L., Verhoeven, A. A. C., Robbins, T. W., Gasull-Camós, J., Evans, M., Mirza, H., & Gillan, C. M. (2018). Shifting the balance between goals and habits: Five failures in experimental habit induction. Journal of Experimental Psychology: General, 147(7), 1043–1065.

Dolan, R. J., & Dayan, P. (2013). Goals and habits in the brain. Neuron, 80(2), 312–325.

Easterlin, R. A. (1974). Does economic growth improve the human lot? Some empirical evidence. In P. A. David & M. W. Reder (Eds.), Nations and Households in Economic Growth (pp. 89–125). Academic Press.

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.

Gazzaniga, M. S. (2000). Cerebral specialization and interhemispheric communication: Does the corpus callosum enable the human condition? Brain, 123(7), 1293–1326.

Girard, R. (1961). Mensonge romantique et vérité romanesque. Grasset. [English: Deceit, Desire, and the Novel (Y. Freccero, Trans., 1965). Johns Hopkins University Press.]

Girard, R. (1972). La violence et le sacré. Grasset. [English: Violence and the Sacred (P. Gregory, Trans., 1977). Johns Hopkins University Press.]

Girard, R. (1978). Des choses cachées depuis la fondation du monde. Grasset. [English: Things Hidden Since the Foundation of the World (S. Bann & M. Metteer, Trans., 1987). Stanford University Press.]

Hagger, M. S., Chatzisarantis, N. L. D., Alberts, H., Anggono, C. O., Batailler, C., Birt, A. R., Brand, R., Brandt, M. J., Brewer, G., Bruyneel, S., Calvillo, D. P., Campbell, W. K., Cannon, P. R., Carlucci, M., Carruth, N. P., Cheung, T., Crowell, A., De Ridder, D. T. D., Dewitte, S., … Zwienenberg, M. (2016). A multilab preregistered replication of the ego-depletion effect. Perspectives on Psychological Science, 11(4), 546–573.

Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review, 108(4), 814–834.

Heider, F., & Simmel, M. (1944). An experimental study of apparent behavior. American Journal of Psychology, 57(2), 243–259.

Helmholtz, H. von. (1867). Handbuch der physiologischen Optik. Voss.

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.

Lebreton, M., Kawa, S., Forgeot d'Arc, B., Daunizeau, J., & Pessiglione, M. (2012). Your goal is mine: Unraveling mimetic desires in the human brain. Journal of Neuroscience, 32(21), 7146–7157.

Levy, D. J., & Glimcher, P. W. (2012). The root of all value: A neural common currency for choice. Current Opinion in Neurobiology, 22(6), 1027–1038.

Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. (1983). Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). Brain, 106(3), 623–642.

Lim, J., & Dinges, D. F. (2010). A meta-analysis of the impact of short-term sleep deprivation on cognitive variables. Psychological Bulletin, 136(3), 375–389. https://doi.org/10.1037/a0018883

McEwen, B. S. (1998). Protective and damaging effects of stress mediators. New England Journal of Medicine, 338(3), 171–179.

McNamee, D., Rangel, A., & O'Doherty, J. P. (2013). Category-dependent and category-independent goal-value codes in human ventromedial prefrontal cortex. Nature Neuroscience, 16(4), 479–485.

Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84(3), 231–259.

Parvizi-Wayne, D., Sandved-Smith, L., Pitliya, R. J., Limanowski, J., Tufft, M. R. A., & Friston, K. J. (2024). Forgetting ourselves in flow: An active inference account of flow states and how we experience ourselves within them. Frontiers in Psychology, 15, 1354719.

Pekrun, R. (2006). The control-value theory of achievement emotions: Assumptions, corollaries, and implications for educational research and practice. Educational Psychology Review, 18(4), 315–341.

Rizzolatti, G., & Craighero, L. (2004). The mirror-neuron system. Annual Review of Neuroscience, 27, 169–192.

Rosset, E. (2008). It's no accident: Our bias for intentional explanations. Cognition, 108(3), 771–780.

Ryan, R. M., & Deci, E. L. (2000). Intrinsic and extrinsic motivations: Classic definitions and new directions. Contemporary Educational Psychology, 25(1), 54–67.

Salkovskis, P. M. (1991). The importance of behaviour in the maintenance of anxiety and panic: A cognitive account. Behavioural Psychotherapy, 19(1), 6–19. https://doi.org/10.1017/S0141347300011472

Sterling, P., & Eyer, J. (1988). Allostasis: A new paradigm to explain arousal pathology. In S. Fisher & J. Reason (Eds.), Handbook of Life Stress, Cognition and Health (pp. 629–649). Wiley.

Tenenbaum, J. B., Kemp, C., Griffiths, T. L., & Goodman, N. D. (2011). How to grow a mind: Statistics, structure, and abstraction. Science, 331(6022), 1279–1285.

Tricomi, E., Balleine, B. W., & O'Doherty, J. P. (2009). A specific role for posterior dorsolateral striatum in human habit learning. European Journal of Neuroscience, 29(11), 2225–2232.

Veblen, T. (1899). The Theory of the Leisure Class. Macmillan.

Wegner, D. M. (2002). The Illusion of Conscious Will. MIT Press.

Wright, S. C., Aron, A., McLaughlin-Volpe, T., & Ropp, S. A. (1997). The extended contact effect: Knowledge of cross-group friendships and prejudice. Journal of Personality and Social Psychology, 73(1), 73–90. https://doi.org/10.1037/0022-3514.73.1.73

Yin, H. H., & Knowlton, B. J. (2006). The role of the basal ganglia in habit formation. Nature Reviews Neuroscience, 7(6), 464–476.

Zacks, J. M., & Swallow, K. M. (2007). Event segmentation. Current Directions in Psychological Science, 16(2), 80–84.

Zink, C. F., Tong, Y., Chen, Q., Bassett, D. S., Stein, J. L., & Meyer-Lindenberg, A. (2008). Know your place: Neural processing of social hierarchy in humans. Neuron, 58(2), 273–283.

The Axiomatic and Architectural Foundation of AIM

Position Statement

Dependency Statement

Derivation Preview

Chapter Preview

4.1 Why AIM Requires Axiomatic Foundations

4.2 How the Nine Axioms Were Arrived At

4.3 The Nine Core Axioms

4.3.1 Axiom 1: Three-Source Taxonomy

4.3.2 Axiom 2: Common-Currency Integration

4.3.3 Axiom 3: Source Opacity

4.3.4 Axiom 4: Confabulation

4.3.5 Axiom 5: M as Amplifier (Object Parasitism)

4.3.6 Axiom 6: Differential Satiation Dynamics

4.3.7 Axiom 7: Preconscious Transmission

4.3.8 Axiom 8: Bayesian Belief Dynamics

4.3.9 Axiom 9: Agent-Based Causal Default (Cognitive Closure through Agent Attribution)

4.3.10 Three Layers Beneath the Agent-Based Causal Default: Update-Rate, Formation, and Scaling

4.3.11 Content and the Structure-Content Distinction

4.3.12 Summary: The Axiomatic Foundation

4.4 The Minimal Architecture and Its AIM Instantiation

4.4.0 The Signed Valuation Principle

4.4.1 The Decision Hub (Integrator and Action-Selector)

4.4.2 The Body Manager (Homeostatic-Deficit Signaller)

4.4.3 The Thinker (Prior-Maintainer and Bayesian Ground-Resolver)

4.4.4 The Observer-Quiescence Corollary and the Architectural Basis of Thinker Privacy

4.4.5 Flow as the Low-Crowding Limit of Observer-Driven Crowding-Out

4.4.6 How the Order of Experience Shapes the Thinker

4.4.7 The Observer (Continuous Conditional Object-Field Salience Assessor)

4.4.7.1 The Minimum Installation Condition

4.4.8 The Belief Register

4.4.9 The Operational Loop

4.4.9.1 The Reafference Pathway, A Second Fast Loop Through the Observer

4.4.10 Summary Table

4.4.11 The Failure Mode, Object Parasitism as Architectural Consequence

4.4.12 Architectural Localisation and Derived Consequences

4.4.13 Crisis Geometry — How the Loop Tips Over

4.4.13.1 The Loop Has a Volume Knob

4.4.13.2 Below the Tipping Point and Above It

4.4.13.3 Why the Tipping Point Is Hard to See Coming, and What Tipping Costs

4.4.13.4 Why the M-Channel Is Architecturally Parasitic

4.4.13.5 The Five Channels Through Which the Volume Can Be Turned Down

4.4.13.6 Subcriticality Maintenance, the Realistic Goal

4.4.14 Forward Pointer — Formal Treatment in Appendix A

4.4.15 How Far the Neural Commitments Extend

4.5 The Two Outputs of Integration — Scalar and Affect

4.6 From Foundation to Argument

What Would Falsify This?

References