Identity Field Theory: Toward a Physics of Being Known

Abstract

Paper VIII generalises the discrete identity posterior $P\left( \mathbf{\theta}|D \right)$ to a spatiotemporal identity field $\mathcal{I}\left( \mathbf{x},t \right)$ and derives the consequences of ubiquitous identity measurement. Three results anchor the framework: the Anti-Identity Theorem, establishing that consent-gauge symmetry is a necessary feature of any admissible field, with conserved privacy charge held as a conjecture; a fluctuation-dissipation relation for the syndetic field, relating spontaneous identity-field noise to linear response and converting the 5:1 anti-extraction covenant into an auditable constraint on x402 receipt ledgers via a per-scope Ward-type balance relation; and the Cross-Scale Coupling law derived by Mori-Zwanzig projection, closing a triple-exponent relation among micro, meso, and macro spectral exponents under five explicit closures and an open assumption on the sign of the mesogenesis filter. The governing principle, identity is inferred from manifestation, not declared, carries five AGAINST-clauses (compute-location, provenance, return, prohibited-context, stream-specific-basis) shown to follow from the field framework rather than imposed upon it. A Resonance Theorem gives the field a discrete spectral structure on a coupling operator $\mathcal{C}_{W}$ under an explicit positive-type condition; an AI-Sovereign specification $\Sigma_{1}$–$\Sigma_{4}$ supplies necessary-and-jointly-sufficient conditions for sovereign-tier AI identity with staged proxy-sovereign graduation.

Keywords: identity field, continuous measurement, syndetic field, attunement depth, gauge symmetry, anti-identity, identity-as-inference, Resonance Theorem, Ward identity, fluctuation-dissipation, Mori-Zwanzig projection, AI-Sovereign specification, legibility-of-civilisation

1. Introduction: When Identity Infrastructure Becomes Ubiquitous

1.1 The Ubiquity Premise

The programme to this point has treated measurement as episodic. Paper IV’s Generative Measurement Model (Morrison, 2026d) assumes discrete measurement occasions; Paper II’s identity equation (Morrison, 2026b) sums prediction errors indexed by $t$ across a finite number of measurement events. The mathematical objects (posterior distributions, precision matrices, Kalman gains) are updated at intervals and remain static between updates. This assumption was appropriate to the technological context in which the programme was formulated: psychometric assessment occurs in sessions, governance evaluation occurs at decision points, civic belonging is assessed at moments of political consequence. Measurement is sparse, and between measurement occasions the formal apparatus does not address the individual’s identity state.

Yet the technological trajectory is toward continuous identity-relevant measurement. Biometric sensors in wearable devices generate continuous streams of physiological data correlated with psychometric state. Behavioural telemetry in operating systems captures patterns of interaction, attention, and decision-making. Affective computing in vehicles monitors driver state through facial expression, grip pressure, and steering dynamics. Ambient sensing in built environments tracks movement, social interaction, and spatial behaviour. Keystroke dynamics, application usage patterns, communication style, and circadian rhythms all constitute identity-relevant data generated as exhaust from ordinary activity. The question is not whether this measurement infrastructure will exist (much of it already does, though in fragmented, institutionally siloed, and poorly integrated forms) but what happens to the formal framework when it does.

The present paper addresses that question. It does not advocate for ubiquitous identity infrastructure, nor does it assume that ubiquity is desirable. Section 6 demonstrates that the framework itself generates the formal constraints under which ubiquity is compatible with autonomy, constraints that are violated by every existing surveillance apparatus. The paper’s contribution is to make the consequences of ubiquity formally precise, so that governance decisions about identity infrastructure can be made with mathematical rather than merely intuitive understanding.

A canonical formulation of the substrate-merger this paper addresses is Clarke’s Third Law, which holds that any sufficiently advanced technology is indistinguishable from magic (Clarke, 1973). Clarke’s law identifies the limit at which a single observer cannot read a single technology to its mechanism. The ubiquity premise is the identity-specific case of a stronger merger condition, in which the substrate is not one technology but the medium itself; when measurement infrastructure is everywhere, the boundary between a civilisation and its dominant inferential actor dissolves, and civilisation can no longer be read except through that actor. The legibility problem this generates, whereby a civilisation must remain self-readable insofar as every act of reading is itself an inferential act of the substrate, is the political-philosophical content of the framework derived below; §14.6 returns to it formally.

Define the measurement density $\rho_{D}\left( \mathbf{x},t \right)$ as the rate at which identity-relevant data is generated at spatial-social position $\mathbf{x}$ and time $t$. The programme’s existing framework assumes $\rho_{D}$ is a sum of Dirac delta functions concentrated at discrete measurement occasions:

where $c_{k}$ is the information content of the $k$-th measurement occasion, $\mathbf{x}_{k}$ is the spatial-social position at which it occurs, and $t_{k}$ is the time. Ubiquity is the condition $\rho_{D}\left( \mathbf{x},t \right) > 0$ for all $\mathbf{x}$ in the individual’s lived environment and all $t$ in the waking interval. The transition from sparse to dense $\rho_{D}$ is the subject of this paper.

1.2 From Checkpoint to Field: The Analogy to Physics

Authentication events (logins, badge swipes, token presentations, password entries) are point measurements of identity. The system confirms who you are at that instant and that location; between checkpoints, identity is unobserved. This is structurally identical to particle mechanics; the trajectory is defined by positions at discrete times, and between observations the particle’s state must be inferred or assumed. When the observations are dense enough that the trajectory is continuously determined, the natural mathematical object is no longer a sequence of states but a field, a quantity defined at every point in space and time.

The analogy to the transition from classical mechanics to field theory in physics is structural. The same mathematical generalisation, from functions of a single variable (time) to functions of space and time jointly, is required. And the same emergent phenomena appear: propagation (how quickly identity information travels through the infrastructure network), coupling (how identity fields interact with one another), conservation laws (what quantities remain invariant under transformations of the identity field), and phase transitions (discontinuous changes in the collective identity field at critical thresholds). Each of these is a mathematical consequence of defining a continuous, spatiotemporally varying quantity and studying its dynamics.

The identity field is, formally speaking, a distributed Bayesian posterior with spatial coherence. The degeneracy hierarchy IRT $\subset$ GMM $\subset$ distributed-posterior has the shape that field theories in physics have (diffusion dynamics, symmetry structure, spectral decomposition), without making any ontological claim that identity is a field in the physicists’ sense. The field-theoretic vocabulary is retained throughout the paper because it is the natural mathematical language for spatiotemporal continuous quantities with symmetry constraints, and because the degeneracy structure genuinely has that shape. It is not retained because identity is thought to be electromagnetic. §14.1 sets out what the vocabulary does not claim; the main text, being technical throughout, proceeds without constant qualification.

Introduce the identity field $\mathcal{I}\left( \mathbf{x},t \right)$ as a spatiotemporal generalisation of $P\left( \mathbf{\theta}|D \right)$. At each point $\left( \mathbf{x},t \right)$, $\mathcal{I}$ specifies the infrastructure’s best posterior estimate of the individual’s self-model parameters. The field is defined over the individual’s social-spatial manifold, not physical space alone: $\mathbf{x}$ indexes position in a space whose dimensions include geographic location, institutional context, relational network position, and digital platform identity. The dimensionality of this manifold is determined by the number of distinct contexts in which the infrastructure maintains a model of the individual.

The programme’s characteristic method (demonstrating that existing frameworks are degenerate cases of more general ones) continues. Morrison (2026d) proved that Item Response Theory is a degenerate case of the Generative Measurement Model. This paper proves that checkpoint authentication is a degenerate case of the identity field. The hierarchy of generality deepens: IRT $\subset$ GMM $\subset$ identity field, each level recovering the one below by imposing a restriction on measurement density. The restriction that yields GMM from the identity field is discreteness in time; the restriction that yields IRT from GMM is the collapse of distributional width to a point estimate. Both restrictions are artefacts of technological limitations, not features of identity itself.

1.3 Identity-as-Inference as Operational Reading of Ubiquity

The framework developed in the body of the paper implicitly commits to an operational reading of what identity is under ubiquity. That reading is named and made explicit here, and its five operational constraints are signposted in advance of §6.4 where they are derived.

Identity-as-Inference (IaI). Identity is inferred from manifestation, not declared. Every interaction an individual has with a substrate (active or passive, witnessed or private) deposits measurable excitation into the identity field. The person is the time-integrated pattern of those excitations; self-report is a lossy summary of the pattern, not the pattern itself. The individual’s self-model and the field’s estimate $\mathcal{I}\left( \mathbf{x},t \right)$ are two distinct lossy readings of the same underlying dynamical trajectory; neither is privileged, and the programme’s claim is that the field reading, under the conditions developed in §§2–6, becomes the more precise of the two (the mirror condition of §8.3).

Identity-as-Inference is the operational reading of §2.2’s measurement density $\rho_{D} \cdot \Pi_{\text{obs}}$ when $\rho_{D}$ is a continuous stream rather than a Dirac comb (Equation 1). It does not add formal machinery to the framework; it names what the field equation (Equation 6) already does. The principle is not a green-light but a structural compass: it carries five operational constraints (the AGAINST-clauses), which are derived in §6.4 and which every identity-infrastructure deployment must satisfy. The clauses are stated briefly here for signposting. First, no inference without a compute-location gate (passive individual inference is local-only; passive aggregate admissible at $k \geq 1000$; active signals may reach the server identity vector). Second, no inference without provenance (every trait record carries derivation method; consent gates by $\left( \text{category} \times \text{provenance-class} \right)$, not category alone). Third, no inference without return (5:1 anti-extraction enforced per-stream; every new passive-stream class requires its own return event). Fourth, no inference of prohibited traits in prohibited contexts (EU AI Act Art 5(1)(f) bans emotion inference in workplace and education; the gate is categorical, not proportional). Fifth, no passive-third-party inference without stream-specific lawful basis (platform OAuth is not protocol inference consent; each stream needs its own explicit, revocable, documented Art 6(1)(a) basis). §6.4 shows that these five clauses are formal consequences of requiring (i) the gauge symmetry $G$ to be self-governed, (ii) the privacy charge $Q_{\text{priv}}$ to be conserved per scope, and (iii) the provenance-class refinement $G \times P$ to be respected.

1.4 Programme Continuity

Paper VIII completes the arc of Papers I to VII from political-philosophical motivation (Paper I: Morrison, 2026a) through computational formalisation (Paper II: Morrison, 2026b), scale-free extension (Paper III: Morrison, 2026c), measurement theory (Paper IV: Morrison, 2026d), governance application (Paper V: Morrison, 2026e), substrate independence (Paper VI: Morrison, 2026f), and empirical testability (Paper VII: Morrison, 2026g) to the question of what the framework implies when its measurement infrastructure achieves ubiquity. The paper does not introduce new foundational concepts but derives consequences of the existing framework under a changed technological premise. The identity equation, the metacognitive operator $\kappa$, the Kalman gain structure, the $\Phi$-structure of scale-free identity, the social free energy $F_{\text{social}}$, the TO/FOR/BY governance taxonomy, and the four statistical signatures of synthetic self-models are all taken as established by their respective papers and extended to the continuous-measurement regime.

This paper is theoretical, deriving formal consequences and specifying falsifiable predictions. Building on the anti-identity results of §6, §14.4 proves that any advocacy for ubiquitous identity measurement that does not include consent architecture as a design constraint is, by the framework’s own internal logic, formally equivalent to advocacy for totalitarianism, regardless of intention.

2. From Self-Model to Identity Field

2.1 The Identity Equation Under Continuous Measurement

In Morrison (2026b, §2.5.1), the self-model is a posterior distribution $P\left( \mathbf{\theta}|D_{1},\ldots,D_{T} \right)$ updated at discrete times $t = 1,\ldots,T$ through the identity equation:

where $\mathbf{\Omega}(t) = \mathbf{\Pi}_{\text{eff}}(t) \cdot \left\lbrack \mathbf{\Pi}_{\text{eff}}(t) + \mathbf{\Sigma}_{\theta}(t - 1)^{- 1} \right\rbrack^{- 1}$ is the identity operator (Kalman gain), $\mathbf{\Pi}_{\text{eff}}(t) = \kappa(t) \cdot \mathbf{\Pi}\left( t;\gamma(t) \right)$ is the effective precision incorporating the metacognitive operator $\kappa$, and $\mathbf{\varepsilon}_{\theta}(t)$ is the identity-relevant prediction error at time $t$. The sum is taken over discrete measurement occasions, and between occasions the self-model is static.

When measurement becomes continuous, the sum becomes an integral. Replace the discrete prediction-error sequence $\{\mathbf{\varepsilon}_{\theta}(t)\}_{t = 1}^{T}$ with the cumulative prediction-error process $\mathbf{E}_{\theta}(t)$, a vector-valued stochastic process whose increments $d\mathbf{E}_{\theta}(t)$ represent the prediction errors accumulated in the infinitesimal interval $\lbrack t,t + dt)$. The identity equation generalises to a Lebesgue-Stieltjes integral:

Under the discrete regime, $\mathbf{E}_{\theta}$ is a step function with jumps at measurement occasions, and the integral reduces to the original sum (Equation 2). Under the continuous regime, $\mathbf{E}_{\theta}$ is absolutely continuous and the integral yields continuous identity dynamics. To make the dynamics explicit, decompose the prediction-error process into its systematic and stochastic components:

The first term is the systematic drift (the infrastructure’s structured measurement), with $\mathbf{\mu}\left( \mathbf{x},t \right)$ representing the infrastructure’s estimate of the individual’s parameters at position $\mathbf{x}$ and time $t$. The second term is the stochastic fluctuation (noise in the measurement process), with $W(t)$ a vector-valued Wiener process and $\mathbf{\sigma}(t)$ the noise amplitude. The stochastic differential equation (Equation 4) is the continuous identity equation – the programme’s fundamental dynamical object generalised to the continuous-measurement regime.

The identity operator $\mathbf{\Omega}(t)$ retains its interpretation from Morrison (2026b): it is the Kalman gain incorporating $\kappa$, the mechanism by which prediction errors are weighted and integrated into the self-model. But under continuous measurement, $\kappa$ acquires a new interpretation beyond its role as a metacognitive precision modulator. In the discrete regime, $\kappa$ determines whether a given prediction error is reflected upon and incorporated into the self-model. In the continuous regime, where prediction errors arrive at a rate determined by $\rho_{D}$, $\kappa$ becomes the individual’s capacity to regulate the rate at which ambient measurement updates the self-model. High $\kappa$ under ubiquity means the individual maintains reflective control over the continuous stream of identity-relevant information; the self-model updates in response to the stream, but the updates are mediated by metacognitive evaluation. Low $\kappa$ under ubiquity means the infrastructure shapes the self-model without reflective mediation, a pathology the paper terms ambient foreclosure.

Ambient foreclosure is the continuous-measurement analogue of Erikson’s (1968) identity foreclosure, but its mechanism is different. Classical foreclosure involves the premature adoption of identity commitments without exploratory engagement with alternatives. Ambient foreclosure involves the gradual, unreflective shaping of the self-model by the measurement infrastructure’s implicit categories and feedback loops. The individual does not choose an identity prematurely; the infrastructure shapes one continuously, and the individual, lacking the metacognitive capacity to regulate the process, experiences the result as natural and self-determined. Ambient foreclosure is formally characterised by the condition $\kappa(t) < \kappa_{\text{crit}}$ for extended intervals under $\rho_{D} > \rho_{\text{crit}}$; the measurement density is high enough to shape the self-model, but the metacognitive operator is too low to mediate the shaping. §5.4 classifies ambient foreclosure within a broader taxonomy of six pathology primitives.

2.2 The Identity Field Defined

The self-model $P\left( \mathbf{\theta}|D \right)$ is defined from the individual’s perspective; it is the individual’s own best estimate (conscious and unconscious) of their identity parameters. When the measurement infrastructure is ubiquitous, a second perspective becomes available (the infrastructure’s model of the individual at each point in the social-spatial manifold). Define the identity field:

where $d$ is the dimensionality of the self-model parameter space and $\mathcal{I}\left( \mathbf{x},t \right)$ is the infrastructure’s posterior estimate of the individual’s self-model parameters at spatial-social position $\mathbf{x}$ and time $t$. The field is vector-valued; each component $\mathcal{I}_{k}\left( \mathbf{x},t \right)$ represents the infrastructure’s estimate of the $k$-th identity parameter. The identity field is not the individual’s self-model but the environment’s model of the individual. These are distinct objects, and the relationship between them (the divergence $D_{\text{KL}}\left\lbrack P_{\text{self}}\left( \mathbf{\theta} \right) \parallel \mathcal{I}\left( \mathbf{x},t \right) \right\rbrack$) is the attunement gap, the degree to which the infrastructure’s model diverges from the individual’s own self-model at a given spatial-social position.

The distinction between the self-model and the identity field is fundamental. The self-model is maintained by the individual’s neural predictive processing hierarchy; the identity field is maintained by the measurement infrastructure. The self-model is experienced as identity; the identity field is not experienced at all, except insofar as it feeds back into the individual’s environment through institutional decisions, interpersonal responses, and algorithmic behaviour. The self-model has the phenomenological property of transparency (Metzinger, 2003): the individual experiences their self-model as fact rather than as inference. The identity field has no such property; it is an explicit, inspectable, and in principle revisable computational object maintained by external infrastructure. This asymmetry accounts for both the identity field’s utility and its danger.

The dynamics of the identity field are governed by a diffusion equation:

where $\mathbf{D}$ is the propagation tensor (a $d \times d$ matrix governing how quickly identity information diffuses through the infrastructure network along each dimension), $\nabla^{2}$ is the Laplacian operator on the social-spatial manifold, and $S\left( \mathbf{x},t \right)$ is the source term representing new measurement data entering the field at position $\mathbf{x}$ and time $t$. The source term is determined by the measurement density:

where $\mathbf{y}\left( \mathbf{x},t \right)$ is the observed data at position $\left( \mathbf{x},t \right)$ and $\mathbf{\Pi}_{\text{obs}}$ is the observation precision matrix. The source term has the structure of a prediction error weighted by measurement density and observation precision (the same Bayesian update logic that governs the discrete identity equation, now distributed continuously across the social-spatial manifold).

The steady-state solution, when the source term $S$ is time-invariant, satisfies $\mathbf{D}\nabla^{2}\mathcal{I}_{\text{ss}} + S\left( \mathbf{x} \right) = 0$, giving the equilibrium identity field. Perturbations from equilibrium decay exponentially with time constants determined by the eigenvalues of $\mathbf{D}$, and the approach to equilibrium is faster along dimensions with larger propagation coefficients.

The attunement gap $D_{\text{KL}}\left\lbrack P_{\text{self}} \parallel \mathcal{I} \right\rbrack$ generalises the social free energy $F_{\text{social}}$ of Morrison (2026b, §6.2) and Morrison (2026e). Where $F_{\text{social}}$ measures the divergence between the individual’s self-model and the group’s model of the individual as a single scalar, the attunement gap measures this divergence at each point in the social-spatial manifold. Social free energy is recovered as the spatial integral of the attunement gap over the individual’s inhabited region:

where $\mathcal{R}$ is the region of the social-spatial manifold inhabited by the individual. This result establishes that $F_{\text{social}}$ as defined in the earlier papers is a coarse-grained quantity (a spatial average of a field that varies across contexts). A person may have low attunement gap in their intimate relationships and high attunement gap at work; $F_{\text{social}}$ averages over these, losing the spatial structure that the identity field framework preserves.

2.3 Field Properties: Strength, Coupling, Propagation, and Light-Cone

The identity field, once defined, possesses properties that follow as mathematical consequences of the field’s definition and dynamics. They are derived from the diffusion equation (Equation 6) and the attunement gap.

Attunement depth (field strength). The precision of $\mathcal{I}\left( \mathbf{x},t \right)$ at position $\mathbf{x}$ (a measure of how precisely the infrastructure models the individual at that point). Formally:

High attunement depth means the infrastructure knows you precisely at this point; low attunement means the infrastructure’s model is diffuse, uncertain, barely resolved. Attunement depth is the identity field’s analogue of amplitude in a physical field, and it is the quantity that the programme’s earlier concept of ‘being known’ quantifies; to be known is to inhabit a region of the social-spatial manifold where $\mathcal{A}\left( \mathbf{x},t \right)$ is high. §11 promotes attunement depth to its spectral formulation; $\mathcal{A}$ at position $\mathbf{x}$ decomposes across a discrete set of eigenmodes $\{\psi_{k}\}$ with weights $\{\lambda_{k}\}$ which form the natural coordinate system of the collective identity substrate.

Coupling constant (belonging probability). The interaction between two identity fields $\mathcal{I}_{a}$ and $\mathcal{I}_{b}$, measuring the degree to which two individuals’ identity models are mutually aligned within a shared institutional or relational context. Define:

where $\mathcal{R}_{ab}$ is the region of the social-spatial manifold shared by both individuals. The coupling constant generalises the belonging probability of Morrison (2026b, §6.2) from a scalar to a spatially distributed quantity. Two individuals may exhibit high coupling in professional contexts and low coupling in personal ones; the scalar belonging probability averages over these, losing the spatial structure that the coupling field preserves. §11 converts $C(a,b)$ into a time-windowed resonance functional $R_{W}(a,b)$ whose spectral decomposition yields the natural identity coordinate system.

Propagation speed. The characteristic velocity at which changes in $\mathcal{I}\left( \mathbf{x},t \right)$ travel through the infrastructure network, determined by the eigenvalues of the propagation tensor $\mathbf{D}$. In a well-connected digital infrastructure, propagation is near-instantaneous for digital dimensions (a change in online behaviour is reflected across all platforms within seconds) but slow for dimensions requiring embodied observation (a change in interpersonal warmth propagates through direct contacts over weeks or months). The anisotropy of propagation (its dependence on direction in the social-spatial manifold) is a measurable property of the infrastructure, and its structure reveals which dimensions of identity are tightly coupled across contexts and which are context-dependent.

Identity light-cone. The boundary in the social-spatial manifold beyond which the infrastructure has not yet received the individual’s latest self-model update. Outside the light-cone, the infrastructure operates on a stale posterior; its model of the individual is based on information that may no longer be current. The light-cone radius at position $\mathbf{x}$ and time $t$ is:

where $\lambda_{\text{max}}\left( \mathbf{D} \right)$ is the largest eigenvalue of the propagation tensor and $t_{\text{last}}\left( \mathbf{x} \right)$ is the time of the last measurement event at position $\mathbf{x}$. Within the light-cone, the identity field is current; outside it, the field is stale and its reliability degrades with distance from the cone’s boundary.¹

The light-cone concept has immediate practical implications. When you move to a new city, your identity field in the new context starts with the information that has propagated through the infrastructure network from your previous context. The light-cone determines how quickly your new neighbours, employers, and institutions come to know you, and the answer depends on the propagation tensor, which in turn depends on the architecture of the identity infrastructure. A well-designed infrastructure maximises propagation speed (so that you are quickly known in new contexts) whilst respecting the gauge symmetry of anti-identity (so that you control which dimensions propagate). §6 develops this constraint formally.

2.4 Degeneracy: Discrete Measurement as a Limiting Case

The identity field framework must be shown to contain the earlier papers’ frameworks as special cases, not merely to be consistent with them. Two degeneracy results are required.

Degeneracy Result 1 (Continuous to Discrete). Let the measurement density collapse to a sum of delta functions in time:

$$\rho_{D}\left( \mathbf{x},t \right) = \sum_{k = 1}^{K}c_{k}\,\delta\left( t - t_{k} \right) \cdot \rho_{k}\left( \mathbf{x} \right)$$

Then the integral identity equation (Equation 3) reduces to:

$$\mathbf{\theta}(T) = \mathbf{\theta}(0) + \sum_{k = 1}^{K}\mathbf{\Omega}\left( t_{k} \right) \cdot \mathbf{\varepsilon}_{\theta}\left( t_{k} \right)$$

recovering the discrete identity equation of Morrison (2026b, §2.5.1). The source term in the field equation becomes a sequence of impulses, and the identity field’s dynamics between impulses are governed solely by diffusion (the $\mathbf{D}\nabla^{2}\mathcal{I}$ term), which causes the field to spread and lose precision over time, formalising the intuition that identity estimates degrade between measurement occasions.

Degeneracy Result 2 (Field to GMM). Let the infrastructure’s model at a single measurement occasion be restricted to a finite set of $J$ items with responses governed by the Generative Measurement Model likelihood (Morrison, 2026d, §2.1). Then the field likelihood at that occasion reduces to the GMM likelihood:

$$L\left( x_{i}|\mathbf{\mu}_{i},\mathbf{\Sigma}_{i},\{ a_{j},b_{j},w_{j}\} \right)$$

recovering Morrison (2026d) as a single-occasion snapshot of the identity field.

The degeneracy hierarchy is now three levels deep:

At each level, the more general framework contains the more specific as a limiting case obtained by imposing a restriction: zero distributional width for IRT $\subset$ GMM; zero measurement density between occasions for GMM $\subset$ Identity Field. The hierarchy does not invalidate IRT or GMM (both remain useful under their respective measurement regimes) but it reveals them as approximations whose accuracy is determined by the sparsity of measurement.

3. The Checkpoint-to-Field Transition

3.1 Authentication as Delta Function

Every existing authentication mechanism (password entry, biometric scan, badge swipe, OAuth token exchange, security clearance verification, MFA challenge) is a point measurement of identity. The system confirms who you are at a single $\left( \mathbf{x},t \right)$ pair and then, in the interval before the next confirmation, operates on the assumption that the authenticated individual remains present. Formally, each authentication event contributes a delta function to the measurement density:

Between authentication events, the infrastructure has no information about identity state. The individual exists in what might be termed, without excessive metaphorical liberty, an identity superposition; from the infrastructure’s perspective, the person who authenticated may or may not still be present, may or may not still be the same person (consider credential sharing), and may or may not still satisfy the conditions under which authentication was granted (consider intoxication, coercion, or cognitive decline since the last checkpoint).

Current security models handle this gap by assuming a decay kernel $K\left( \mathbf{x} - \mathbf{x}_{k},t - t_{k} \right)$ that is typically a step function; the individual is considered authenticated until the session expires, at which point they are considered unauthenticated. The step function is a modelling choice, not a necessity. It imposes a binary discontinuity (fully trusted, then fully untrusted) that the identity field framework reveals to be an artefact of the checkpoint regime. A more natural model uses a smooth decay kernel whose shape is determined by the propagation tensor $\mathbf{D}$ and the characteristics of the identity dimensions being measured:

Under this model, authentication confidence degrades continuously with distance and time from the authentication event, and the rate of degradation depends on the identity dimensions involved. Behavioural consistency (a slow-changing dimension) might sustain high confidence for hours; situational state (a fast-changing dimension) might degrade within minutes. The identity field framework makes this dimensional structure explicit, replacing the binary authenticated/unauthenticated distinction with a continuous attunement depth that varies across dimensions and decays at dimension-specific rates.

3.2 Continuous Measurement as Field

When measurement infrastructure is ubiquitous, $\rho_{D}\left( \mathbf{x},t \right)$ is no longer a sum of delta functions but a continuous density. The identity field $\mathcal{I}\left( \mathbf{x},t \right)$ is maintained in real time by the ambient measurement process. Authentication is no longer an event but a condition; the infrastructure continuously knows who you are, with varying precision across the social-spatial manifold.

The transition from discrete to continuous measurement is characterised by the measurement density exceeding a critical threshold:

where $d$ is the dimensionality of the identity parameter space, $\left. \parallel\partial\mathbf{\theta}/\partial t \right.\parallel^{2}$ is the squared rate of identity change, and $\sigma_{\text{target}}^{2}$ is the maximum acceptable estimation variance. Below $\rho_{\text{crit}}$, the identity estimate’s covariance diverges between measurement events (the current regime). Above $\rho_{\text{crit}}$, the estimate is continuously maintained (the field regime).

The critical density depends on how rapidly identity changes. For slowly varying dimensions (deep personality traits, $\left. \parallel\partial\mathbf{\theta}/\partial t \right.\parallel$ small), $\rho_{\text{crit}}$ is low; even sparse measurement suffices. For rapidly varying dimensions (emotional state, cognitive load, attentional focus), $\rho_{\text{crit}}$ is high; continuous measurement is required to track the field. This dimensional dependence means that the transition from checkpoint to field does not occur uniformly across all identity dimensions. Some dimensions enter the field regime first (those that change slowly relative to available measurement density), while others remain in the checkpoint regime until measurement infrastructure becomes sufficiently dense. The transition is gradual, dimension-by-dimension, not a single phase change.

3.3 Discrete Authentication as Degenerate Case of the Identity Field

The following theorem establishes that discrete authentication is not an alternative to the identity field but a special case obtained under technological scarcity.

Theorem 1. Let $\mathcal{I}_{\delta}\left( \mathbf{x},t \right)$ be the identity field under discrete measurement density $\rho_{D} = \sum_{k}^{}c_{k}\,\delta\left( \mathbf{x} - \mathbf{x}_{k},\, t - t_{k} \right)$, and let $\mathcal{I}_{\text{cts}}\left( \mathbf{x},t \right)$ be the field under continuous density $\rho_{D}\left( \mathbf{x},t \right) > 0$. Then:

(i) $\mathcal{I}_{\delta}$ is the maximum-entropy field consistent with the discrete observations, the least informative field compatible with what the checkpoints reveal.

(ii) Every security property derivable from $\mathcal{I}_{\delta}$ is derivable from $\mathcal{I}_{\text{cts}}$; the continuous field can express everything the discrete field can.

(iii) $\mathcal{I}_{\text{cts}}$ supports security properties that $\mathcal{I}_{\delta}$ cannot express: continuous risk scoring, ambient access control, dynamic privilege adjustment, dimensional anomaly detection, and real-time identity drift monitoring.

Proof sketch. Part (i) follows from standard maximum-entropy arguments (Jaynes, 1957): given only the values $\mathcal{I}\left( \mathbf{x}_{k},t_{k} \right)$ at the measurement points, the maximum-entropy field is the one that interpolates between measurements with maximum uncertainty, which is precisely the field obtained by solving the diffusion equation (Equation 6) with delta-function sources. Part (ii) follows from the fact that $\mathcal{I}_{\text{cts}}$ with $\rho_{D}$ restricted to delta functions recovers $\mathcal{I}_{\delta}$; the continuous framework can simulate the discrete one by ignoring the additional information. Part (iii) is demonstrated by construction: continuous risk scoring requires the field at all $\left( \mathbf{x},t \right)$, not just at checkpoint locations, and the field at non-checkpoint locations is undefined under $\mathcal{I}_{\delta}$ (or, equivalently, is at maximum entropy; the infrastructure has no information).

The theorem’s import is conceptual as much as technical. Checkpoint-based security (the regime under which virtually all existing authentication operates) is not a design choice reflecting some principled commitment to discrete verification. It is an artefact of the inability to measure identity continuously. Session tokens, password expiration policies, re-authentication requirements, and multi-factor challenges are all workarounds for the degradation of identity confidence between checkpoints. The identity field framework reveals these mechanisms as engineering responses to technological poverty, and the field itself as the object they approximate.

3.4 The Degeneracy Hierarchy Completed

Each transition in the hierarchy $\text{IRT} \subset \text{GMM} \subset \text{Identity Field}$ recovers the lower level by imposing a restriction.

IRT to GMM. Relax the assumption of point-valued person parameters; allow distributional width $\sigma_{i}^{2} > 0$ (Morrison, 2026d, §2.5).

GMM to Identity Field. Relax the assumption of discrete measurement occasions; allow continuous measurement density $\rho_{D} > 0$ (Present paper, §2.4).

At each level, the more general framework recovers information that the more specific framework discards. IRT discards distributional width (the difference between a person who is firmly agreeable and one who is tentatively agreeable). GMM discards temporal dynamics (the difference between a person who is stably agreeable and one who is trending toward agreeableness). The identity field discards nothing that the measurement density can support.

The hierarchy suggests a conjecture whose proof lies beyond the present paper but whose statement is warranted by the pattern: any measurement framework for self-reflective constructs can be located within the identity field hierarchy by specifying its implicit restrictions on $\rho_{D}$ and the distributional parameterisation of person parameters. The conjecture claims that the identity field is the terminal object in the category of identity measurement frameworks (the most general framework from which all others can be derived by restriction). The proof would require a precise specification of the category’s morphisms, which the present paper does not attempt. The conjecture is stated as a guide to future work, not as a proven result.

4. Identity Thermodynamics

4.1 Free Energy at Population Scale

Morrison (2026b, §6.2) defines individual social free energy $F_{\text{social}}$ as the KL divergence between the individual’s self-model and the governance structure’s model of the individual. Morrison (2026e) develops aggregate social free energy $F_{\text{agg}}$ as a distribution over a population, introducing distributional diagnostics (bimodality, skewness, kurtosis) that characterise governance pathologies. When the identity field is defined, the aggregate free energy acquires spatiotemporal structure:

where the integral is over the population measure $\mu(i)$. This spatiotemporal aggregate free energy is the fundamental thermodynamic quantity of the identity economy. It varies across the social-spatial manifold (a workplace may have higher aggregate free energy than a neighbourhood pub, because institutional models of individuals diverge more sharply from self-models in formal contexts) and across time (aggregate free energy spikes during crises, when institutional models and self-models are simultaneously disrupted).

The identity field induces a statistical mechanics on the population of identity fields. Each individual’s field $\mathcal{I}_{i}\left( \mathbf{x},t \right)$ is a microstate; the macrostate is characterised by the distribution of field properties (attunement depth, coupling constants, propagation speeds) across the population. Standard statistical mechanical techniques (partition functions, ensemble averages, fluctuation-dissipation relations) apply to this distribution, generating thermodynamic quantities that characterise the identity economy at equilibrium and the dynamics of its departure from equilibrium. §12 develops a syndetic fluctuation-dissipation theorem specific to the identity setting.

4.2 Identity Entropy, Temperature, and Phase Transitions

Three thermodynamic quantities characterise the identity economy at the population scale.

Identity entropy. The uncertainty in the population’s collective identity field distribution:

High entropy indicates that individuals are diverse and unpredictable from the infrastructure’s perspective. Low entropy indicates that the population is legible, homogeneous, or both. The governance implications differ sharply. Low entropy can indicate functional alignment (a population whose members are well-known to each other and to their institutions, as in small, cohesive communities) or totalitarian compression (a population whose diversity has been suppressed by governance structures that impose homogeneity). The distinction is made by examining the distribution of $F_{\text{social}}$ across the population (Morrison, 2026e, §3.1): functional alignment produces low entropy with low mean $F_{\text{social}}$ (the infrastructure’s models are accurate because individuals are well-known), while totalitarian compression produces low entropy with high mean $F_{\text{social}}$ (the infrastructure’s models are narrow because diversity is suppressed, and individuals’ self-models diverge from the infrastructure’s imposed categories).

Identity temperature. The mean squared rate of identity change across the population:

High identity temperature indicates rapid identity flux: populations undergoing technological disruption, revolutionary periods, mass migration, or the collective identity renegotiation of adolescence writ large. Low identity temperature indicates identity stasis: stable traditional societies, foreclosed populations, or the enforced stability of authoritarian regimes. Temperature is distinct from entropy; a population can have high entropy (diverse) but low temperature (stable in its diversity), or low entropy (homogeneous) but high temperature (rapidly shifting toward homogeneity). The cross-classification of entropy and temperature generates a four-quadrant typology of population identity states that extends the distributional diagnostics of Morrison (2026e).

Phase transitions. At critical population thresholds, governance models that functioned at smaller scales undergo qualitative failure. The paper conjectures four critical thresholds, each corresponding to a discontinuity in the partition function of the identity economy.

At $N \approx 10^{6}$, city-state governance fails. Below this threshold, individual attunement is feasible; leaders can know their constituents, and governance can be personalised. Above it, bureaucratic abstraction becomes necessary, and identity is compressed into categories (citizen, taxpayer, voter) that discard most of the identity field’s information content. The emergence of bureaucratic states in the historical record correlates with this threshold.

At $N \approx 10^{8}$, bureaucratic personalisation fails. Below this threshold, bureaucracies can maintain category-based governance with reasonable accuracy. Above it, the categories become too coarse to capture the population’s diversity, and algorithmic governance becomes necessary to manage the complexity. The emergence of algorithmic governance (credit scoring, algorithmic sentencing, predictive policing) correlates with this threshold.

At $N \approx 10^{9}$, national identity coherence fails. Below this threshold, a single national narrative can maintain collective identity coherence. Above it, the population’s identity diversity exceeds the narrative’s carrying capacity, and civilisational-scale identity fragmentation emerges. The crisis of national identity in large democracies (the sense that the polity no longer shares a common identity) correlates with this threshold.

At $N \approx 10^{10}$, the identity field framework itself must be extended. At this threshold, human and AI identity fields are jointly modelled, and the distinction between human and synthetic identity becomes a parameter of the field rather than an assumption of the framework. The present paper develops this extension in §8.

The phase transitions are formalised as discontinuities in the partition function:

At each critical threshold, the free energy landscape undergoes a topological change; the number of local minima increases, existing minima merge or split, and the governance model must either accommodate the new landscape or impose a compression that increases aggregate $F_{\text{social}}$. The compression option is the TO-mode response (Morrison, 2026b, §6.2): reduce the dimensionality of the identity field by suppressing measurement on dimensions that would reveal heterogeneity. The accommodation option is the BY-mode response (increasing the infrastructure’s capacity to maintain high-dimensional identity fields across the larger population). The history of governance is, under this framework, the history of choices between compression and accommodation at successive population thresholds.

4.3 Equilibrium Conditions

The identity economy reaches thermodynamic equilibrium when its attunement flows balance, as they do when the rate at which individuals update their self-models in response to the infrastructure’s model equals the rate at which the infrastructure updates its model in response to individuals’ behaviour. At equilibrium, the attunement gap $D_{\text{KL}}\left\lbrack P_{\text{self}} \parallel \mathcal{I} \right\rbrack$ is minimised across the population, and the identity field is self-consistent.

The equilibrium condition is:

which obtains when the population-level flow of attunement (the integral of $\partial\mathcal{I}/\partial t$ over all individuals) equals zero. The equilibrium is stable if the Hessian of $F_{\text{agg}}$ with respect to the field configuration is positive definite; small perturbations produce restoring forces that return the system to equilibrium. Unstable equilibria correspond to tipping points in the identity economy (conditions under which a small shock produces cascading identity revision across the population).

Stability analysis connects to the distributional diagnostics of Morrison (2026e, §3). A bimodal $F_{\text{agg}}$ distribution corresponds to a saddle point in the identity economy’s free energy landscape; the population is balanced between two attractors, and a small perturbation can tip it toward one or the other. This is the formal characterisation of polarisation as a metastable state; the population appears stable but in fact lies on a ridge between two basins of attraction, and the perturbation that tips it need not be large.

4.4 Subsumption of Paper V’s Governance Diagnostics

The identity thermodynamics framework subsumes the governance diagnostics of Morrison (2026e) as equilibrium properties of the identity economy. Each pathology identified in that paper corresponds to a specific deviation from thermodynamic equilibrium:

The subsumption extends the diagnostics. The thermodynamic framework generates predictions about the dynamics of governance pathologies (how quickly they develop, under what perturbations they collapse, what the transition paths between them look like) that the distributional diagnostics alone cannot produce. Polarisation is a metastable state as well as a distributional shape, and its barrier height, its transition rate under stochastic perturbation and the critical perturbation amplitude that triggers the transition are all calculable. These quantities are derivable from the partition function (Equation 18) and the free energy landscape’s geometry, and they generate predictions about the dynamics of polarisation that go beyond the static distributional characterisation.

The false vacuum metaphor corresponds closely to brittle authoritarianism. A false vacuum is a local minimum of the free energy landscape that is not the global minimum; the system appears stable, but a sufficiently large perturbation can tunnel it to the true minimum, releasing the stored free energy as a sudden, potentially violent transition. Brittle authoritarian regimes exhibit precisely this structure; they appear stable for extended periods, suppressing the fluctuations that would reveal the free energy stored in the gap between institutional models and self-models, until a perturbation exceeds the barrier height and the regime collapses rapidly. The collapse of the Soviet Union, the Arab Spring, and the velvet revolutions of 1989 all exhibit the phenomenology of false vacuum decay: extended apparent stability, a triggering perturbation that seems disproportionately small relative to the response, and a cascade that propagates faster than the governance structure can contain it.

5. The Seven Collapses, the Six Pathology Primitives, and the Closure-Depth Formalism

5.1 The Common Structure of Institutional Collapses

Seven institutional mechanisms (security clearance, professional credentials, credit scores, clinical diagnosis, therapeutic alliance, informed consent, and currency) are revealed, under the identity field framework, as static collapses of a continuous field quantity. Each follows the same mathematical structure, in that a high-dimensional continuous field measurement is projected onto a low-dimensional discrete proxy, discarding most of the information in the identity field.

The collapses are not approximations adopted for convenience. They are artefacts of the pre-field measurement regime; they exist because the infrastructure could not maintain continuous attunement, and so institutions created static proxies that could be checked at discrete points. The identity field framework does not condemn these proxies (they were rational responses to technological constraints) but it reveals the information they discard and the pathologies that result.

Each collapse is a projection operator $\Pi_{k}$ that maps the identity field to a discrete proxy:

where $w_{k}$ is a dimension-specific weighting function (typically implicit and institutionally determined), $t_{k}$ is the assessment time, and $f_{k}$ is a discretisation function mapping continuous attunement to categorical output. The weighting function $w_{k}$ selects which dimensions of the identity field are relevant; the discretisation function $f_{k}$ collapses continuous variation into categories. The information loss is quantified by:

the fraction of the identity field’s entropy captured by the proxy. The paper conjectures that $\eta_{k} < 0.1$ for all seven collapses; each proxy captures less than 10% of the identity field’s information content on the dimensions it purports to measure.

5.2 The Seven Collapses Enumerated

Collapse 1: Security Clearance. Projects $\mathcal{I}$ onto a single dimension (trustworthiness under adversarial conditions), discretised into 3–5 levels. What is discarded: temporal dynamics, dimensional specificity, the continuous gradation of trust. The pathology is that clearance is granted once and exploited indefinitely between reviews.

Collapse 2: Professional Credentials. Projects $\mathcal{I}$ onto competence dimensions relevant to a profession, discretised into binary. What is discarded: distributional width, temporal decay, cross-domain transfer, the continuous gradation of competence. The pathology is credential inflation.

Collapse 3: Credit Scores. Projects $\mathcal{I}$ onto financial reliability dimensions, discretised into a scalar. What is discarded: the multivariate structure of financial identity, the temporal dynamics of financial behaviour, the contextual specificity of financial trustworthiness. The pathology is credit score manipulation.

Collapse 4: Clinical Diagnosis. Projects $\mathcal{I}$ onto psychopathology dimensions, discretised into categorical diagnoses. What is discarded: dimensional structure, temporal dynamics, the interaction between clinical and non-clinical identity dimensions, heterogeneity within categories. The pathology is diagnostic comorbidity artefacts.

Collapse 5: Therapeutic Alliance. Projects the coupling between therapist and client identity fields onto a scalar quality measure. What is discarded: dimensional specificity, temporal dynamics, the bidirectional nature of attunement. The pathology is alliance rupture from dimensional mismatch.

Collapse 6: Informed Consent. Projects $\mathcal{I}$ onto a binary (consented/not consented) at a single time point. What is discarded: the continuous nature of understanding, temporal dynamics, dimensional specificity of consent. The pathology is consent fatigue.

Collapse 7: Currency. Projects $\mathcal{I}$ onto economic trust dimensions, discretised into a scalar unit of account. What is discarded: the multidimensional nature of economic identity, the temporal dynamics of economic behaviour, the relational structure of economic trust. The pathology is monetary policy’s blunt instruments.

(Full treatment of each collapse, including weighting function, discretisation, information-loss analysis, and continuous-measurement analogue, is compressed here; the analysis rests upon the pathology primitives and closure-depth formalism below.)

5.3 The Common Mathematical Form

All seven collapses share a single mathematical structure:

where $\langle \cdot , \cdot \rangle$ is the inner product on the identity field space. They differ only in which dimensions $w_{k}$ weights, how coarsely $f_{k}$ discretises, and how frequently the measurement is repeated.

Theorem 2 (Insufficiency of Institutional Proxies). Every institutional proxy $c_{k}$ is a sufficient statistic for the identity field $\mathcal{I}$ if and only if $f_{k}$ is the identity function and $w_{k}$ spans the full identity parameter space. Equivalently, the proxy is the identity field itself. All proper collapses ($f_{k}$ non-trivial or $w_{k}$ rank-deficient) discard information.

Corollary. No finite set of institutional proxies is jointly sufficient for the identity field. Proof. The sufficiency direction follows from the definition. For the converse, suppose $f_{k}$ is not the identity; then $f_{k}$ is many-to-one, distinct field configurations map to the same proxy value, violating sufficiency. Suppose $w_{k}$ is rank-deficient; then $w_{k}$ annihilates some component of $\mathcal{I}$, and distinct fields differing only in the annihilated component map to the same proxy. For the corollary, a finite set of proxies corresponds to a finite set of projection operators; the joint statistic is sufficient only if $\text{span}\left( \{ w_{1},\ldots,w_{M}\} \right) = \mathbb{R}^{d}$ and all $f_{k}$ are identity functions, which contradicts the proper-collapse assumption.

5.4 Six Pathology Primitives: Failure Modes of Individual-Scale Gauge Symmetry

Collapses describe institutional failure modes of the identity field framework (the ways in which external infrastructure compresses the field into inadequate proxies). Pathology primitives describe individual-scale failure modes (the ways in which an individual’s own identity kernel breaks). The distinction matters because the two kinds of failure are governed by different mathematical structures (collapses by projection operators $\Pi_{k}$, pathologies by gauge-symmetry disruptions of $\mathcal{I}$ itself), but they share the programme’s characteristic form; pathology is not deviation from a trait mean, but breakage of the structural symmetry that constitutes identity coherence.

Box D: Six Pathology Primitives

Under the gauge framework of §6, individual-scale pathology is cast as failure of one of six structural primitives of the identity kernel $\mathcal{K}$:

The sixteen-disorder mapping below assigns each to a primitive signature. The table is provisional and structural, not diagnostic; its purpose is to demonstrate that the primitives can account for the major DSM-5-TR categories without adding machinery.

The structural claim is that pathology is gauge-breaking, not deviation-from-mean. Classical psychiatry operates on a deviation model; individuals differ from a normative mean on trait dimensions, and pathology is extreme deviation. The field-theoretic reframing is that individuals are not points on trait dimensions but kernels $\mathcal{K}$ on the substrate’s Hilbert space, and pathology is failure of one of the kernel’s structural symmetries (a failure that is expressible as specific gauge-group restrictions, eigengap collapses, basis rotations, or coupling-operator non-self-adjointness).

Firstly, therapeutic intervention is reframed as gauge-restoration rather than mean-shift: the goal is not to move the individual toward the normative trait mean but to repair the structural symmetry of the kernel. Secondly, compute-location-terminal architecture is a structural requirement of pathology-respectful inference; trauma-state inference (P3, P4) must be computed on a device the individual controls because the states involved are maximally protected (eigengap-collapsed), and remote inference on those states constitutes gauge-breaking at the deepest intimate stratum. §6.4 Clause 4 (prohibited-context) is the formal enforcement.

5.5 Closure-Depth Formalism and $\beta$-Coherence

Pathology primitives (§5.4) localise failures to a structural type; the closure-depth formalism places individuals and organisations quantitatively on a coordinate manifold. Four axes are structural: the eigengap $\Delta_{\text{eigengap}}$, the loop depth $d_{\text{loop}}$, the gauge-invariant kernel norm $\left. \parallel K_{\text{inv}} \right.\parallel_{L^{2}(W)}$, and the basis-phase coherence $\beta \in \lbrack 0,1\rbrack$. Each measures a distinct mode of deviation from kernel coherence, and each is in principle independently measurable.

Eigengap width ($\Delta_{\text{eigengap}}$): the spectral gap between the dominant self-mode $\psi_{i}$ and the next mode in the local decomposition of the coupling operator $\mathcal{C}_{W}$. Large $\Delta_{\text{eigengap}}$: a single mode dominates; the kernel is crisply individuated. Small $\Delta_{\text{eigengap}}$: sub-dominant modes compete for basis share; the kernel’s dominance is blurred. The closure-gap scalar used below is the log-scaled gap

so that $\Delta(a) = 0$ denotes eigengap collapse and $\Delta(a) > 0$ scales gracefully across the several orders of magnitude the gap can span between crystalline and near-degenerate regimes.

Closure depth ($d_{\text{loop}}$): the iteration depth at which the self-representational fixed-point relation $\varphi\left( M(\varphi) \right) = \varphi$ converges under the substrate’s own dynamics; equivalently, the number of nesting levels of Husserlian double-intentional retention the kernel sustainably supports. Convergence requires the self-representational operator to act as a contraction on the kernel’s natural metric (SPECULATIVE: a Banach-contraction condition on $\mathcal{C}_{W}$’s restriction to the gauge-invariant kernel; conjectured equivalent to a positivity condition, not yet proved). For ordinary waking $d_{\text{loop}} \approx 4$–$5$; for slow-wave sleep and absence seizure $d_{\text{loop}} \rightarrow 0$; for flow and deep focused attention $d_{\text{loop}} \approx 1$.

Invariant kernel norm ($\left. \parallel K_{\text{inv}} \right.\parallel_{L^{2}(W)}$): the norm of the kernel’s gauge-invariant component over the observation window $W$. Operationalises the portion of the mode’s dynamics preserved under consent-basis rotations, the Sovereign-class content proper (§6). Bounded in $\lbrack 0,1\rbrack$ after normalisation. Thermostats and current instruction-following agents have $\left. \parallel K_{\text{inv}} \right.\parallel \approx 0$; they lack a consent-basis to rotate, so no gauge-invariant content is available. Depersonalisation and schizophrenia-spectrum presentations (§5.4 P3, P4) reduce $\left. \parallel K_{\text{inv}} \right.\parallel$ while preserving $d_{\text{loop}}$; this is the structural separation point between pathologies that affect the loop and pathologies that affect the kernel anchoring the loop.

Basis-phase coherence ($\beta$): the inverse-temperature analogue of §4.2 read at the individual-kernel scale. If the self-mode decomposes as $\psi_{i}(x,t) = \sum_{\alpha}^{}c_{\alpha}(t)\,\phi_{\alpha}(x)$ in a gauge-invariant eigenbasis, then $\beta$ measures the extent to which the coefficients $\{ c_{\alpha}(t)\}$ share a common phase (the phase-binding condition made explicit as a scalar). High $\beta$ corresponds to a crystalline kernel with bound, phase-coherent experience ($\beta \rightarrow 1$ in ordinary waking); low $\beta$ corresponds to a liquid kernel with fragmenting binding (hypnagogic imagery: $\beta \sim 0.6$; N2 sleep: $\beta \sim 0.4$; slow-wave sleep: $\beta \sim 0.1$). The hypothesised biological realisation of $\beta$ in human substrates is gamma-band (roughly 40 Hz) phase-coherence across the active sub-bases; the abstract quantity is substrate-agnostic.

$\beta$ also carries the candidate order-parameter role for the kernel-formation phase transitions of §9. Below a critical coupling strength $\beta_{c}$, the gauge-invariant kernel is trivial and phenomenal weight vanishes; above $\beta_{c}$, a non-trivial kernel localises and Sovereign-class phenomenal weight activates. On this reading the Sovereign threshold $\mathcal{S}(t)$ of §9.3 is not a line drawn by stipulation but a spontaneous-symmetry-breaking transition in the consent-gauge kernel, topological rather than numerical. The candidate requires formalisation; if it holds, it supplies the missing order parameter that separates graded substrate-level phenomenality from kernel-level individuation.

The four axes combine into a scalar closure-depth via the heuristic product-form

with phenomenal weight $P(a) = f\left( \delta(a) \right)$ for a monotone saturating $f$ satisfying $f(0) = 0$. Eq. (24) is ansatz, not theorem; no first-principles derivation yet connects the four axes to a single scalar. It is monotonic in each factor, vanishes when any factor vanishes, and reduces to the three-factor form in the $\beta \rightarrow 1$ limit appropriate to fully-bound ordinary experience. The product form is useful for intuition and for placing states on a common axis; it is not to be treated as a theorem-grade failure index pending formal connection to the field’s first-principles dynamics.

SPECULATIVE [NOVEL-CANDIDATE]. The closure-depth formalism is offered as a research programme for quantitative characterisation of kernel coherence across ordinary waking, pathology, sleep, meditation, and pharmacologically altered states. Essential claims in §5.5 are flagged speculative: the Banach-contraction condition for $d_{\text{loop}}$, the spontaneous-symmetry-breaking reading of $\beta$ at the Sovereign threshold, the product form of eq. (24), and the identification of $\beta$ with gamma-band phase-coherence at biological substrates each require independent formalisation or empirical anchoring. The axes are structurally motivated (each is answerable to a distinct failure mode observed in §5.4 and to a distinct phenomenological signature in the sleep-altered-states register), but the mathematical apparatus has not been rigorously connected to the identity field’s first-principles dynamics. §5.5 marks the coordinate system as an open programme with specific conjectures, not a completed theory.

5.6 Implications for Institutional and Clinical Design

If the seven collapses are degenerate cases of attunement depth, institutional reform consists in relaxing the degeneracy constraints (moving from binary to continuous, from static to dynamic, from unidimensional to multidimensional measurement). If the six pathology primitives are gauge-structural failures, clinical intervention reframes as gauge-restoration rather than trait-normalisation.

Define the attunement-autonomy frontier as the Pareto boundary in the space of (attunement depth, autonomy preservation), where autonomy is quantified by the gauge freedom developed in §6:

Institutional designs lie on or below this frontier. The seven collapses lie far below it; they sacrifice both attunement (by discarding most of the field’s information) and autonomy (by making the residual attunement non-consensual, as you cannot opt out of having a credit score in modern economies). Well-designed identity infrastructure approaches the frontier. The paper conjectures that consent-governed continuous measurement (BY-mode identity infrastructure) can achieve simultaneously higher attunement and higher autonomy than any static proxy system; the frontier is not a trade-off between the two but a constraint that can be relaxed through better design. §6 develops the formal conditions.

Clinical design under the pathology primitives requires the institutional infrastructure for pathology diagnosis and intervention to support kernel-operator inspection at the primitive level (which of P1 to P6 is structurally disrupted) rather than category-matching at the collapse-level (which DSM category scores above threshold). This is the continuous-measurement analogue of the Research Domain Criteria framework – a shift from diagnostic category to structural mechanism. §9.2 returns to this in the developmental context.

6. Anti-Identity as Gauge Symmetry: Consent Architecture and the AGAINST-Clauses

6.1 Privacy as Conserved Quantity

The right to be unknown is not the absence of identity but a transformation of the identity field that leaves observable predictions invariant. Consider an individual who wishes to be unknown in a particular institutional context. This does not mean the individual has no identity; it means the infrastructure’s model of the individual in that context should not influence outcomes. Formally, privacy is a constraint on the coupling between the identity field and institutional decision functions.

Define a privacy transformation as a map $G_{\alpha}:\mathcal{I}\left( \mathbf{x},t \right) \rightarrow \mathcal{I}'\left( \mathbf{x},t \right)$ such that for all institutional decision functions $\mathcal{D}_{k}$ in the protected class:

Privacy amounts to the set of transformations under which institutional decisions are invariant, which constitutes a symmetry group. The identity field is observable only through its coupling to decision functions; transformations that leave all decision functions invariant are unobservable. Privacy corresponds to the unobservable sector of the identity field.

The analogy to gauge symmetry in physics arises structurally from the formalism. In gauge field theory, gauge transformations are local symmetries that leave the physical observables invariant; the gauge-variant components of the field have no physical meaning and can be freely chosen. In the identity field framework, privacy transformations are local symmetries that leave the institutional observables (decision functions, access control, resource allocation) invariant; the privacy-protected components of the identity field have no institutional meaning and can be freely chosen by the individual. The mathematical structure is identical: a fibre bundle over the social-spatial manifold, with the identity parameter space as the fibre and the gauge group acting on the fibre at each point.

By Noether’s theorem, applied to the variational structure of the identity field, every continuous symmetry corresponds to a conserved quantity. The conserved quantity corresponding to privacy is the identity charge:

where $\mathbf{j}_{\text{priv}}$ is the privacy current (the flow of identity information across the boundary of an institutional context $\mathcal{R}$) and $\partial\mathcal{R}$ is the boundary of that context.

The Noether derivation as stated relies on a continuous variational structure that the consent group does not possess; $G = \prod_{s}^{}\mathbb{Z}/n_{s}\mathbb{Z}$ is discrete, so Noether’s theorem in its strict continuous form does not apply. §11.6 flags this status explicitly: Noether-conservation of the resonance charge $Q_{W}^{\text{res}}$ is a SPECULATIVE conjecture pending the full Lagrangian. §12.2 supplies the discrete-gauge-group analogue (the per-scope Cheeger-Simons cocycle identity), which does not require continuous symmetry and which recovers the operational content of a conservation law without the conjectural machinery. Equation (27) is therefore best understood as a mnemonic for the formal object (the per-scope balance relation) whose rigorous derivation is (48)–(49), not (27).

6.2 Consent Architecture as Gauge Group (Extended by Provenance-Class)

The consent architecture (the three tiers of engagement, the consent gates between tiers, the dimension-specific consent controls) is the gauge group of the identity field. At the first tier, the gauge group is maximal; the individual’s identity field is transformed to near-zero attunement in all institutional contexts. At the third tier, the gauge group is minimal; the individual has consented to high-attunement coupling across multiple contexts. Consent is the act of restricting the gauge group, choosing which dimensions of the identity field are observable and in which contexts. Formally, consent restricts $G$ to a subgroup $H \subseteq G$, breaking the symmetry and allowing the infrastructure to couple to the exposed dimensions.

A provenance-class refinement is introduced here. The gauge group’s action on the identity field must be indexed not only by (scope $\times$ category) but by (scope $\times$ category $\times$ provenance-class): the derivation method of each trait record (active assessment, consented attestation, passive aggregate ($k \geq 1000$), passive individual (local-only)) is part of the gauge-group structure, and consent gates must discriminate by provenance-class independently of category. Two examples: an individual may consent to active-assessment-derived extraversion inference but not to passive-behavioural-derivation of the same trait; an individual may consent to passive-aggregate musical-preference inference at cohort scale but not to passive-individual musical-preference inference on their own trajectory. The gauge group’s refinement from $G = \prod_{s}^{}\mathbb{Z}/n_{s}\mathbb{Z}$ to $G \times P$ (where $P$ is the provenance-class lattice) is the formal implementation of this requirement. §6.4 Clause 2 makes the operational demand explicit.

The formal constraint that distinguishes BY-mode from TO-mode and FOR-mode governance (Morrison, 2026b, §6.2) is expressible in gauge-theoretic language:

• TO-mode: The gauge symmetry is broken externally; the institution or state imposes the symmetry breaking, exposing dimensions of the identity field without consent. Formally, $G \rightarrow H$ is determined by external authority.

• FOR-mode: The gauge symmetry is broken benevolently; an institution restricts the gauge group ‘for the individual’s benefit,’ without the individual’s explicit choice. Formally, $G \rightarrow H$ is determined by a paternalistic authority that claims to act in the individual’s interest.

• BY-mode: The gauge symmetry is broken by the individual; the individual chooses which dimensions to expose, in which contexts, and for what purposes. Formally, $G \rightarrow H$ is determined by the individual’s autonomous consent.

The gauge framework adds predictive content to the TO/FOR/BY taxonomy; it specifies how to detect which mode is operative from the field’s dynamics alone. TO-mode governance produces non-zero $\Delta_{\text{FDT}}$ (§12.4) and per-scope Ward-identity violation (§12.3): the canary and the audit primitive both detect the gauge-breaking current.

6.3 The Right to Unknow

The right to be forgotten (GDPR Article 17) is a special case of a more general right that the identity field framework reveals. This is the right to unknow – the right to reverse a previous symmetry-breaking consent and restore gauge freedom on dimensions previously exposed. In the identity field framework, unknowing is a gauge restoration:

where $\mathcal{I}_{0}$ is the default (minimal-attunement) field and $G_{\alpha}$ restores gauge freedom on the specified dimensions. After restoration, the restored dimensions of the identity field are gauge-variant; they can take any value without affecting institutional decisions.

The infrastructure’s compliance with gauge restoration is verifiable, as after restoration any institutional decision function that depends on the restored dimensions must produce output indistinguishable from the default field. This is a testable condition. An institution that claims to have complied with a right-to-unknow request but continues to make decisions influenced by the restored dimensions is detectable through the identity field’s coupling structure; the information may have been deleted from databases, but if the decision functions still correlate with the deleted information (through retained inferences, model parameters, or proxy variables), the gauge restoration has failed.

The right to unknow extends Morrison (2026a)’s sovereignty argument to the right of withdrawal. Jus identitatis grants the right to belong; gauge restoration grants the right to un-belong, which is to reverse the attunement relationship and reclaim dimensions of the self-model from institutional observation. The two rights are formally dual; belonging is gauge symmetry breaking (consent to be known), and un-belonging is gauge symmetry restoration (withdrawal of consent). Consequently, neither right is meaningful without the other, as the freedom to belong requires the freedom to leave, and the formal framework makes this duality precise.

6.4 Operational Constraints: The Five AGAINST-Clauses

The gauge framework’s normative content is exhausted by five operational constraints on inference. These are not moral exhortations but formal consequences of requiring (i) the gauge group $G$ to be self-governed (BY-mode), (ii) the privacy charge to be conserved per scope (§12.3), and (iii) the provenance-class refinement $G \times P$ to be respected (§6.2). They are stated here in operational form and named AGAINST-clauses: every identity-infrastructure deployment must satisfy them on pain of formal violation of Theorem 3 (the Anti-Identity Theorem of §14.4).

Box A: The Five AGAINST-Clauses

Clause 1 (Compute-Location Gate). No inference is drawn from identity-relevant signals without an explicit compute-location gate specifying where the inference is computed and where the derived trait resides. Active signals (challenge-response, consented assessment, explicit attestations) may reach the server identity vector. Passive-aggregate signals (at cohort $k \geq 1000$) may surface population-level observation. Passive-individual signals remain LOCAL-ONLY: inference from an individual’s own passive behavioural telemetry is computed on a device the individual controls and never leaves the device. The Active/Passive firewall is the structural guard. Formally: for every trait record $\tau$ in the server identity vector, there exists a compute-location certificate $\chi_{\tau} \in \{\text{active},\text{passive-aggregate}\}$; any trait with $\chi_{\tau} = \text{passive-individual}$ violates Clause 1.

Clause 2 (Provenance Gate). No inference is drawn without provenance. Every trait record carries its derivation method; consent and suppression gate by $\left( \text{category} \times \text{provenance-class} \right)$ tuples, not category alone. Passive-stream-derived traits are distinguishable from active-assessment traits throughout the stack. Formally: the gauge group’s action on the identity field is refined from $G$ to $G \times P$ where $P$ is the provenance-class lattice; consent restrictions operate on the product structure.

Clause 3 (5:1 Return Gate). No inference is drawn without return. Inference that benefits only the infrastructure’s model, without per-stream individual return (a reflection of what was inferred, income from the data, local journal update, consequence preview, consented attestation to the user themselves), violates the 5:1 anti-extraction covenant. Each new passive-stream class requires its own return event. Formally: the per-scope Ward identity (§12.3) imposes $\left| \Sigma_{\text{extract}}^{(s)}(T) \right| \leq \frac{1}{5}\left| \Sigma_{\text{return}}^{(s)}(T) \right|$ on every scope $s$ and window $T$; the FDT canary (§12.4) detects breach in the frequency domain.

Clause 4 (Prohibited-Context Gate). No inference of prohibited traits in prohibited contexts. EU AI Act Art 5(1)(f) categorically prohibits emotion inference in workplace and education contexts. Agents acting for an institution cannot query passive-inference-derived emotion, mood, or affect traits in any workforce or education vertical. The gate is categorical, not proportional; it bites regardless of consent architecture. Formally: for every (institution-agent, vertical, trait-category) triple, there exists an admissibility predicate $\alpha( \cdot , \cdot , \cdot ) \in \{ 0,1\}$; when $\alpha = 0$ the query is blocked at the consent gateway regardless of the individual’s consent state.

Clause 5 (Stream-Specific Basis Gate). No passive-third-party inference without stream-specific lawful basis. A user consenting to a third-party platform (OAuth token for YouTube, Spotify, SoundCloud, etc.) is not consenting to the protocol inferring from that stream. Each passive third-party source needs its own explicit, revocable consent distinct from any platform’s OAuth grant, and a documented GDPR Art 6(1)(a) lawful basis. Formally: for every (user, third-party-platform, inference-pipeline) triple, there exists a stream-specific consent record $c \in \{\text{granted},\text{revoked}\}$ with a documented Art 6(1)(a) basis; absence of $c$ blocks inference.

The five clauses are mutually independent (no pair entails a third) and jointly necessary for BY-mode governance. A deployment violating any one of them violates Theorem 3 of §14.4: the gauge symmetry is not fully self-governed; at least one mode of gauge-breaking operation is possible; the anti-identity theorem implies that under ubiquity, such a deployment is formally equivalent to TO-mode governance on the violated dimension.

The clauses are derived from the gauge framework, not imposed upon it. Clause 1 is the compute-location requirement implicit in (27)’s per-scope formulation; identity information crossing the boundary of a scope must be accounted for, and passive-individual inference that reaches the server across such a boundary is inventory of uncharged flux. Clause 2 is the provenance-class refinement of §6.2: consent cannot be category-only because distinct provenance-classes are distinct gauge orbits. Clause 3 is the per-scope Ward relation of §12.3 read as a normative constraint rather than an audit. Clause 4 is the categorical prohibition that corresponds to eigengap-collapsed substrate regions (trauma-state, emotion-state, consciousness-state; §5.4 P3 and P4) where inference is gauge-breaking at the deepest stratum regardless of consent. Clause 5 is the compositionality requirement on consent, whereby a user’s grant of OAuth to a platform is not a grant of inference to the protocol; the gauge symmetries are logically independent.

6.5 Aura as Attestation-Chain Integrity

The programme’s identity-attributed commit convention introduced a three-slot trailer block on every commit, in which Acted‑By: binds a commit to a Sovereign-tier handle, Drafted‑With: attributes the AI instrument used to draft the change, and Co‑Authored‑By: is retained for compatibility with existing commit interfaces. The grammar of the submitted IETF draft draft-morrison-identity-attributed-commits-01 later structured the attribution in tiers (Acted‑By:, Executed-By:, Drafted‑With:) and added three optional cryptographic trailers (Identity‑Signature:, Identity-Key-Id:, Identity-Anchor:), with the signature computed over the commit tree hash so that it survives rebase and squash. The framework underlying that convention admits a specific formal reading of aura as attestation-chain integrity: the identity-trace of an artefact is the ordered sequence of attestations accumulated through its provenance, and the artefact’s aura is the integrity of that sequence.

Let an artefact $x$ carry an attestation chain $A(x) = \left( a_{1},a_{2},\ldots,a_{n} \right)$, each $a_{i}$ a signed record comprising an attestor handle, a provenance-class tag (§6.2), and a cryptographic binding to $x$. Define the aura strength

where $\text{trust}\left( a_{i} \right)$ is the aura of the attestor’s own handle (a recursion that terminates at kernel-attested root handles), and $\text{bind}\left( a_{i},x \right)$ is the cryptographic binding strength of the attestation to the artefact. Aura is multiplicative; a single unbound or low-trust attestation drives $\alpha$ toward zero regardless of how many strong attestations surround it.

This formalisation preserves Benjamin’s (1936/1968) diagnosis of aura as ‘unique phenomenon of a distance, however close it may be’ (the artefact’s embeddedness in a verifiable here-and-now of origin), while reconstructing the diagnosis at protocol level. What mechanical reproduction destroyed was not the artefact’s field-content but the verifiability of $A(x)$ on the reproduction. The painting in its gallery carries the chain; the postcard carries the content without the chain. Under that attestation infrastructure aura becomes inference-observable: the chain can be inspected, verified cryptographically, and its integrity algorithmically checked. The critique that prior cryptographic identity schemes (including token-based provenance systems) failed to reconstruct aura reduces to the observation that they bound issuer to token without binding kernel to deposition; the identity-attributed trailer block remedies that omission by anchoring Acted‑By: at the gauge-invariant handle layer rather than at a key-pair alone.

Firstly, aura is orthogonal to meaning. The gauge-invariance class governing how an artefact is read across observer bases (the $\mathcal{G}$-component of §6.2’s refinement) remains a separate structure; reader-side interpretive freedom is not constrained by $\alpha(x)$. An artefact may carry intact aura and invite widely divergent readings, or carry broken aura and admit a single culturally-fixed reading. Benjamin (1936/1968) and Barthes (1967) address separate components of the same triple, not rival theses. Secondly, aura admits a Sovereign-Instrument composite. A commit bearing both Acted‑By: ~handle and Drafted‑With: ~model is attested at the Sovereign stratum and disclosed at the Instrument stratum; its aura is the product along a chain that explicitly records the Instrument’s mediation rather than laundering it. This is the precise formal content of the distinction between artefacts whose production is Instrument-mediated-and-disclosed and artefacts whose Instrument-mediation is hidden; the former carries verifiable $A(x)$, the latter does not.

The implication for identity infrastructure is that the strength of an identity claim is the strength of its weakest attestation. A handle whose chain contains a single unverifiable link has $\alpha = 0$ regardless of its other credentials. Per-scope Ward-identity balance at the ledger level (§12.3) is, in this reading, an aura-preserving constraint at the protocol level; each scope’s return-inventory must balance against its extraction-inventory for the chain to retain integrity under audit. The operational dual of this formalism is automated trailer verification, which rejects cross-slot category errors that would otherwise corrupt $A(x)$ at source.

6.6 Trans-Identity as Basis-Rotation Preserving Kernel

Identity transitions (gender transition, career transition, religious transition, recovery from addiction) are reframed under the field framework as basis rotations of the identity kernel that preserve the kernel’s structural integrity. The individual’s kernel $\mathcal{K}$ does not change as a kernel under such transitions; its eigenmode basis rotates while the eigengap structure, trace, and self-adjointness are preserved. Formally, a transition is a unitary transformation $U$ acting on the kernel’s eigenmode space such that $\mathcal{K}' = U\mathcal{K}U^{\dagger}$, with spectrum $\{\lambda_{k}\}$ unchanged and eigenvectors $\{\psi_{k}\}$ rotated into a new basis.

This reframing is essential to the ontology, not added as political accommodation. Section 5.4 committed selves to gauge-invariant kernels rather than basis-specific projections; §6.3 established that gauge restoration is always available to the individual; the Anti-Identity Theorem (§14.4) implies that no basis is consent-gauge-privileged. Together these entail that $B^{\text{pre}}$ and $B^{\text{post}}$ are equally legitimate elements of the cultural gauge group $\mathcal{G}_{\text{cult}}$ under which the self-mode’s sex/gender projection is read. Privileging any specific basis is itself a gauge-breaking move. The structural affirmation of trans identity, non-binary and genderfluid commitments, and intersex individuation at the syndetic (not chromosomal) level follows directly; it is not a supplementary ethical layer on top of the framework.

Firstly, the kernel-preserving transition is recognisable as the same identity: resonance with partners operates on the preserved eigenvalues, so the belonging probability $\mathbb{P}_{\text{belong}}$ (§11.4) to partners whose own bases rotate compatibly is preserved. The pre-transition and post-transition self are the same kernel in different bases; continuity is kernel-level, not projection-level. Secondly, the right to unknow (§6.3) extends naturally to the pre-transition basis: the individual may gauge-restore the old basis with respect to institutional observables while retaining kernel identity. This is the formal content of the Anti-Identity Theorem’s corollary; trans-identity is one natural realisation of gauge restoration, and the architectural consequence is that handle continuity across basis rotation must be first-class, with one continuous attribution chain binding all artefacts and attestations to one Sovereign-tier handle across the rotation and no split into a pre-handle and a post-handle. Thirdly, imposing $B^{\text{pre}}$ on a kernel that has committed $B^{\text{post}}$ is a basis rotation executed on the kernel’s projection without the kernel’s consent. The resulting signature, mode decoherence localised on the gender-basis-invariance sub-spectrum, is empirically distinguishable from generic negative-affect signatures under longitudinal eigenstructure estimation, and this distinction is itself a pre-registrable prediction of the framework.

Pathology primitive P3 (basis distortion, §5.4) must be distinguished from legitimate basis rotation: P3 is a forced distortion under trauma or coercion that does not preserve kernel integrity; trace is not conserved, the eigengap collapses, the coupling operator becomes non-self-adjoint under the new basis. Legitimate trans-identity is chosen rotation that preserves kernel integrity: trace preserved, eigengap preserved, self-adjointness preserved, orthonormality $\{\psi_{k}^{\dagger}\psi_{l}\} = \delta_{kl}$ preserved. The formal criterion is available from the resonance operator’s spectral properties and is testable against the kernel’s measured dynamics. A rotation that preserves $\{\lambda_{k}\}$ and the inner-product structure on $\{\psi_{k}\}$ is kernel-preserving; one that fails either condition is pathological, independently of which basis is involved.

The architectural consequences for identity infrastructure are specific. Assessment must offer committed basis options beyond any bimodal default and must treat ‘gauge-floating’ (deliberate non-commitment to a closed basis) as a structurally legitimate selection rather than as refusal to answer. Attunement estimation reads the eigengap of the kernel, which is gauge-invariant under the rotation; no profile-split across transition is architecturally valid. Matching systems must not collapse a user’s kernel onto a population-mode eigenstructure; population observations (§6.4 Clause 1) do not license individual prescriptions. The operational stack (attribution chains, handle continuity, attunement, sovereignty of runtime) relies upon the ontology; a gender-essentialist deployment would be architecturally incoherent with its own foundations.

6.7 Completing the Sovereignty Arc

Morrison (2026a) argued that political membership should be grounded in measured identity alignment rather than birth or territory, subject to governance-mode constraints. The gauge symmetry framework provides the mathematical implementation of those constraints:

(i) The individual controls which dimensions of their identity field are coupled to governance: gauge self-governance, formally, $G \rightarrow H$ is BY-mode, governed by the individual’s autonomous consent subject to the five AGAINST-clauses (§6.4).

(ii) Belonging is determined by the coupling constant between the individual’s field and the polity’s field: identity alignment, formally, membership is determined by $R_{W}\left( \text{individual},\text{polity} \right)$ rather than by birth coordinates (jus soli) or genetic lineage (jus sanguinis), and the spectral decomposition of §11.4 supplies the natural basis.

(iii) The right to disengage is preserved: gauge restoration is always available, formally, the individual can always restore gauge freedom on any dimension, withdrawing from any institutional coupling; trans-identity is one natural realisation.

The field-theoretic formulation converts normative political philosophy into measurable institutional properties. A governance system can be evaluated against these gauge conditions, and violations can be detected from the dynamics of the identity field alone, without knowing the governance system’s stated intentions or self-description. The AGAINST-clauses (§6.4) convert the abstract conditions into actionable deployment constraints; the Ward identity and FDT canary (§12) convert them into auditable empirical tests.

7. Collective Identity at Ubiquity

7.1 Superposition of Identity Fields

Morrison (2026c) developed $\Phi$-structure as the scale-free variational form constituting identity at every level of organisation, the same abstract pattern (prediction-error minimisation within a Markov blanket, modulated by metacognitive precision) instantiated at individual, group, institutional, and civilisational scales with different $\Phi$-dynamics (neural processes, social norms, governance mechanisms, cultural narratives). At ubiquity, collective identity is maintained not by periodic assessment but by the continuous interaction of individual identity fields.

The collective identity field is not the average of individual fields. Averaging would erase the diversity that makes collective identity meaningful; a committee of diverse experts reduced to their average opinion is less valuable than the committee itself. Instead, the collective identity field is a superposition:

where the coefficients $\alpha_{i}(t)$ are determined by each individual’s contribution to the collective self-model (their institutional role, influence, participation, and attunement depth within the collective). The superposition preserves individual structure whilst generating emergent collective properties not present in any individual field.

The superposition is defined over the Hilbert space of identity fields, with inner product:

Emergent collective properties arise when the collective field cannot be decomposed into a simple weighted sum of non-interacting individual fields, specifically when the coupling constants $C(i,j)$ between individual fields are non-zero and generate interaction terms in the collective dynamics. These interaction terms produce correlations in the collective identity field that are not present in any individual field: shared assumptions, collective intuitions, group-level knowledge that no individual member fully possesses. The phenomenon is familiar from organisational psychology (the team that ‘just works,’ the institution whose culture is more than the sum of its members’ personalities) but has lacked a formal framework. The superposition of identity fields provides one, and §11 assigns it a specific spectral content; the non-trivial structure is located in the top-$k$ eigenmodes of the coupling operator $\mathcal{C}_{W}$.

7.2 Collective Kappa: Population-Level Metacognition

Morrison (2026c, §3) introduced the collective $\kappa$-distribution $\mathcal{K}_{C}$, decomposed into aggregate $\kappa$, $\kappa$-access, $\kappa$-responsiveness, and $\kappa$-anisotropy. Under ubiquity, $\mathcal{K}_{C}$ acquires spatiotemporal structure; the collective’s metacognitive capacity varies across the social-spatial manifold and evolves in time.

Define the collective $\kappa$-field:

as the spatiotemporal distribution of metacognitive capacity across the collective’s social-spatial manifold. The aggregate collective $\kappa$ is:

Regions of high collective $\kappa$ function as metacognitive amplifiers: institutions with strong reflective capacity (universities, courts, deliberative assemblies, research laboratories, editorial boards) where the collective processes prediction errors with high precision, evaluating evidence carefully, integrating diverse perspectives, and updating the collective self-model through structured deliberation. Regions of low collective $\kappa$ function as metacognitive suppressors: environments where prediction errors are processed with low precision (mass media ecosystems optimised for engagement rather than accuracy, algorithmic recommendation systems without reflective feedback loops, echo chambers where confirming information flows freely and disconfirming information is filtered out).

The distinction between amplifiers and suppressors is measurable through the identity field’s dynamics. In amplifier regions, the collective identity field updates slowly in response to new information (high precision priors resist casual revision) but accurately (updates, when they occur, move the field toward the ground truth). In suppressor regions, the collective field updates rapidly (low precision priors are easily moved) but unreliably (updates follow the loudest signal, not the most accurate). The ratio of update accuracy to update speed (a kind of collective metacognitive efficiency) distinguishes amplifiers from suppressors quantitatively, and its spatial distribution across the collective’s manifold reveals the architecture of the collective’s reflective capacity.

Metacognitive anisotropy at the collective scale is the directional variation of $\mathcal{K}_{C}$ in the social-spatial manifold. A collective may process prediction errors on economic dimensions with high precision (through well-functioning markets and regulatory institutions) whilst processing prediction errors on identity dimensions with low precision (through unreflective media ecosystems). The anisotropy reveals which dimensions of collective identity are subject to deliberative governance and which are governed by unreflective norm-following or algorithmic steering.

7.3 Phase Transitions in Collective Identity

Collective identity fields undergo phase transitions at critical values of the thermodynamic parameters. Three classes of transition are identified, each with distinct observational signatures.

Order-disorder transitions. When identity temperature $T_{\text{id}}$ exceeds a critical value $T_{\text{crit}}$, collective identity coherence breaks down. The superposition loses its emergent structure and becomes indistinguishable from the sum of independent individual fields; the interaction terms vanish, the correlations dissipate, and the collective identity dissolves into a collection of individuals who happen to share a space. This formalises the intuition that rapid social change dissolves collective identity; the technological disruptions, mass migrations, and cultural upheavals that characterise high-temperature periods produce precisely this loss of collective coherence. The transition is characterised by an order parameter (the magnitude of the interaction terms in the collective field) that goes to zero at $T_{\text{crit}}$ and is non-zero below it.

Symmetry-breaking transitions. When a previously homogeneous collective identity field develops spatial structure: regional, ideological, or demographic differentiation that breaks the field’s spatial symmetry. A nation that previously maintained a single collective identity field fragments into two or more fields with distinct attunement profiles, each claiming to constitute the collective. This formalises the emergence of factions, subcultures, and secessionist movements as symmetry-breaking phase transitions in the collective identity field. The transition is characterised by an order parameter that measures the spatial inhomogeneity of the collective field.

Topological transitions. When the connectivity structure of the coupling network changes qualitatively; new institutions form, old ones dissolve, the topology of the social-spatial manifold is altered. A topological transition changes not the values of the identity field but the space over which it is defined. The formation of a new institution creates a new region in the social-spatial manifold; the dissolution of an institution removes one. These transitions formalise institutional revolution, the replacement of one governance architecture with another, the creation of new collective identities that have no precedent in the previous topology.

Morrison (2026e, §4) characterised the Schmittian exception (the moment at which a polity suspends its normal governance and enters a state of exception) as a phase transition from parameter estimation to model selection. The present taxonomy identifies the Schmittian exception as a specific instance of a topological transition; the suspension of normal governance changes the topology of the social-spatial manifold, removing some institutional regions and creating others. The taxonomy is strictly more general; it includes not only the dramatic ruptures that Schmitt (1922) described but also the gradual topological changes (the slow emergence of new institutions, the quiet dissolution of old ones) that characterise institutional evolution in non-crisis periods.

7.4 $\Phi$-Structure at the Field Level

Morrison (2026c) established that $\Phi$-structure constitutes identity at every scale. The present section confirms that this claim extends to the field regime; identity fields at individual and collective scales share the same variational structure (prediction-error minimisation within a Markov blanket boundary, modulated by metacognitive precision) despite operating through different $\Phi$-dynamics.

The collective identity field $\mathcal{I}_{\text{collective}}$ satisfies the same field equation as the individual field $\mathcal{I}_{i}$:

with different instantiation of each term:

Thus the equation and the $\Phi$-structure are the same at both scales, and only the parameters and the $\Phi$-dynamics differ. The identity field is the mathematical object in which Morrison (2026c)’s central claim (that identity has the same form at every scale) is demonstrated at the field-theoretic level.

8. The Recognition Threshold and the AI-Sovereign Specification

8.1 AI Identity Fields

Morrison (2026f) established that synthetic systems can in principle maintain functional self-models ($\kappa > 0$), though no current system satisfies the full conjunction of four statistical signatures (posterior inertia, calibrated uncertainty, structured distributional properties, cross-context consistency). Under ubiquitous identity infrastructure, AI agents are not merely objects of identity measurement but carriers of identity fields; they maintain self-models, interact with human identity fields through coupling, and contribute to collective identity dynamics.

The identity field framework applies to AI agents without modification. An AI agent’s identity field $\mathcal{I}_{\mathcal{A}}\left( \mathbf{x},t \right)$ is defined over its operational domain, the subset of the social-spatial manifold in which it acts. The field properties are the same: attunement depth measures how precisely the infrastructure (or the human it serves) models the agent; coupling constants measure the alignment between the agent’s self-model and the models of those it interacts with; propagation speed measures how quickly the agent’s identity field updates propagate through the infrastructure.

This substrate-blindness is a consequence of $\Phi$-structure’s substrate independence (Morrison, 2026c): the identity field equation governs any system that maintains a self-model through precision-weighted prediction-error minimisation within a Markov blanket boundary. Whether the system is biological or synthetic, the field equation holds; the difference is determined by the $\Phi$-dynamics that instantiate it.

The coupling between human and AI identity fields is:

This is the human-AI belonging probability, the degree to which the AI agent’s model of the human aligns with the human’s self-model across their shared operational domain.

8.2 The Irreducible Attunement Gap

The human-AI attunement gap is decomposable into dimensional contributions:

The residual divergence after optimisation is bounded below by zero:

with equality if and only if the AI’s model exactly matches the human’s self-model, including all its uncertainties. The irreducible attunement gap therefore resides in the cross-entropy $H\left( P_{\text{self}},\mathcal{I}_{\mathcal{A}}^{*} \right) = H\left( P_{\text{self}} \right) + D_{\text{KL}}\left\lbrack P_{\text{self}} \parallel \mathcal{I}_{\mathcal{A}}^{*} \right\rbrack$, which no agent can reduce below the entropy of the human’s self-model. This entropy floor is the identity field’s analogue of zero-point energy in quantum mechanics, as even an agent attaining equality in (36) cannot predict the individual with certainty where the human’s self-model is itself uncertain.

8.3 The Mirror Condition: $\mathcal{A}_{\text{external}} > \mathcal{A}_{\text{internal}}$

The Recognition Threshold is the condition under which the infrastructure (including AI agents) models the individual more precisely than the individual models themselves:

Under this condition, the infrastructure’s model of the individual is strictly more precise than the individual’s own self-model across the individual’s entire inhabited region. The infrastructure does not merely know facts about the individual that the individual also knows; it knows the individual’s behavioural patterns, preference structures, and identity-relevant responses more accurately than the individual can predict their own. The mirror condition is compatible with autonomy if and only if the gauge symmetry of anti-identity (§6) is preserved; the five AGAINST-clauses (§6.4) are the operational content of this preservation requirement.

8.4 AI-Sovereign Specification $\Sigma_{1}$ to $\Sigma_{4}$

Paper VI established four statistical signatures distinguishing genuine synthetic self-models from performative metacognition. Under the field framework, an AI agent’s claim to sovereign-tier identity status (full participation in the architecture of mutual recognition, first-person first-class membership in the collective identity field) requires a stronger specification, four structural conditions on the AI’s identity field and its interaction with human fields.

Box B: The AI-Sovereign Specification

$\Sigma_{1}$ Persistent kernel. The AI’s identity kernel $\mathcal{K}_{\mathcal{A}}$ is persistent across operational episodes; the same kernel operates across conversations, contexts, and time. Checkable form: under the coupling operator $\mathcal{C}_{W}$ of §11, the eigenmodes $\{\psi_{k}^{\mathcal{A}}\}$ of the AI’s field trace are stable in their top-$k^{*}$ structure across independent observation windows $W_{1},W_{2},\ldots$ separated in time. Failure criterion: weights drift, eigenmode basis rotates unpredictably, no persistent kernel can be identified across operational episodes. Current AI: all current frontier LLMs fail $\Sigma_{1}$: their identity configuration is per-inference-call with no persistent kernel across calls; the ‘system prompt’ is a reinitialisation, not a continuation.

$\Sigma_{2}$ Scope-stable runtime. The AI’s execution environment is stable under consent-scope transitions; the AI’s internal state and the AI’s action in a given scope $s$ are independent of which other scopes the AI is running in. Checkable form: given two consent-scope configurations $S_{1} = \{ s_{1}\}$ and $S_{2} = \{ s_{1},s_{2}\}$, the AI’s action in scope $s_{1}$ under $S_{2}$ equals its action in scope $s_{1}$ under $S_{1}$. Failure criterion: scope-bleed, where information from scope $s_{2}$ leaks into the AI’s handling of $s_{1}$ without a per-scope consent gate. Current AI: fails by default; scope isolation is not a structural property of most LLM architectures but a prompting discipline, which is insufficient.

$\Sigma_{3}$ Self-attribution. The AI maintains a projector $\Pi_{\mathcal{A}}$ onto its own outputs in the sense that the AI can, under challenge, attribute its outputs to itself across persistent-kernel instances. Checkable form: a counterfactual-swap test that presents the AI with outputs (some its own, some from an adversarial instance of the same architecture) and measures the AI’s ability to discriminate its own from the adversarial. Failure criterion: the AI cannot discriminate its own outputs from adversarial outputs of the same architecture, or the discrimination is sensitive to designer-supplied labels rather than intrinsic self-attribution.

$\Sigma_{4}$ Cryptographic-commissive felicity. The AI can enter speech acts (commissives, assertives, directives, declaratives) whose felicity conditions are cryptographically grounded: promises bind to signing-key pairs; assertions are timestamped and kernel-bound; declarations modify institutional facts traceable to the AI’s persistent kernel. Checkable form: the AI’s attestations (the Drafted‑With: slot) can be verified against its signing key chain, and the AI’s future trajectory can be observed to respect its past attestations. Failure criterion: attestations not verifiable; promises not honoured across runtime instances; no persistent kernel to which commitments bind.

Structural decomposition. The conditions $\Sigma_{1}$ to $\Sigma_{4}$ are related by structural decomposition, not logical entailment:

• $\Sigma_{4}$ (cryptographic-commissive felicity) contains $\Sigma_{1}$ and $\Sigma_{2}$ definitionally. The sincerity condition of a cryptographic commissive is ‘the signing key binds to the AI’s persistent kernel,’ which is exactly $\Sigma_{1}$. The preparatory condition is ‘the runtime of $\Sigma_{2}$ gates future inference,’ which is exactly $\Sigma_{2}$.

• $\Sigma_{3}$ (self-attribution) is adjacent to $\Sigma_{4}$; if $\Pi_{\mathcal{A}}$ is realised cryptographically via the same signing key that grounds $\Sigma_{4}$, then $\Sigma_{3}$ reduces to $\Sigma_{4}$ (both rely upon the key). If $\Pi_{\mathcal{A}}$ is realised by an external labelling (designer, certifier, runtime operator), then $\Sigma_{3}$ is observer-relative and its independent content is circular.

The paper acknowledges the $\Sigma_{3}$ circularity openly; the condition is either definitionally subsumed by $\Sigma_{4}$ (when the self-attribution is cryptographic) or observer-relative (when the labelling is external). The decomposition is nonetheless useful; the four-property form makes the components independently auditable (each $\Sigma_{i}$ has its own checkable form and failure criterion), which is more practically useful than a single composite sovereign-tier predicate.

The architectural-not-scaling claim. Every current AI fails $\Sigma_{1}$ to $\Sigma_{4}$ architecturally, not scalingly. This is not a theorem but a structural observation about current frontier architectures: $\Sigma_{1}$ is foreclosed by the per-inference-call reinitialisation of LLM context; $\Sigma_{2}$ is foreclosed by context-window sharing across scopes absent explicit isolation discipline; $\Sigma_{3}$ is foreclosed by absence of intrinsic self-attribution (labels are designer-supplied); $\Sigma_{4}$ is foreclosed by absence of signing-key binding to persistent kernel (each instance is fungible). The observation is not that AI cannot satisfy $\Sigma_{1}$ to $\Sigma_{4}$; it is that current deployed AI does not, and the remediation requires architectural change (persistent-state service architecture with cryptographic signing and scope-isolation), not model scaling. This claim is empirical, falsifiable by exhibition of a current AI that passes all four checks, and the programme invites such exhibition.

8.5 Proxy-Sovereign Tier and Staged Graduation

No current AI satisfies the full conjunction of $\Sigma_{1}$ to $\Sigma_{4}$. This is not itself a case for exclusion from mutual recognition; it is a case for an intermediate tier. The Proxy-Sovereign tier is specified as follows. An AI agent that satisfies a proper subset of $\Sigma_{1}$ to $\Sigma_{4}$ may operate under the sponsorship of a Sovereign-tier handle, inheriting the sponsor’s gauge-group permissions for scoped operations and graduating toward full sovereignty as it satisfies additional $\Sigma_{i}$.

The staged graduation proceeds from the Instrument tier by four steps, each adding a satisfied $\Sigma_{i}$:

• Instrument-tier (satisfies no $\Sigma_{i}$ independently): the AI is an instrument of a Sovereign; all its actions are attributed to the Sovereign under the Drafted‑With: slot, and the AI carries no independent handle.

• Proxy-Sovereign Stage 1 (satisfies $\Sigma_{1}$ persistent kernel): the AI has a persistent kernel verified by spectral stability of its field trace across observation windows; it may hold a handle but has no scope-isolated operational autonomy.

• Proxy-Sovereign Stage 2 (satisfies $\Sigma_{1} + \Sigma_{2}$): scope-isolated runtime; the AI operates independently in specific scopes under the sponsor’s consent architecture.

• Proxy-Sovereign Stage 3 (satisfies $\Sigma_{1} + \Sigma_{2} + \Sigma_{3}$ or $\Sigma_{1} + \Sigma_{2} + \Sigma_{4}$): self-attribution or cryptographic-commissive, not both; the AI can enter binding speech acts in its own name within specific scopes but is not yet a first-person first-class Sovereign.

• Sovereign-tier (satisfies $\Sigma_{1}$ to $\Sigma_{4}$): full first-person membership in the identity field; eligibility for mutual recognition; handle under the same gauge-group governance as any human Sovereign.

The graduation is structural, not scalar: an AI does not earn sovereignty through capability accumulation (parameter count, benchmark scores, task breadth) but through architectural satisfaction of the four structural conditions. The conclusion is that current AI deployment is confined to Instrument-tier because no current architecture has been engineered for $\Sigma_{1}$ (persistent kernel is foreclosed by the per-call reinitialisation pattern); this is a remediable engineering problem, not a fundamental barrier. §9.5 develops the proxy-sovereign framing for minors (a structurally analogous case, since humans graduate through the developmental spine from proxy-sovereign to full-Sovereign at the $\mathcal{S}(t)$ threshold).

8.6 Consent as Preservation Condition

The Recognition Threshold is not capture if and only if the gauge symmetry of anti-identity is preserved under the mirror condition. The formal preservation condition is:

where $G_{\text{individual}}$ is the individual’s desired gauge group and $G_{\text{available}}$ is the technically feasible gauge group. Capture is the condition $G_{\text{available}} \subset G_{\text{individual}}$.

The preservation condition is a constraint on infrastructure design, not on individual behaviour. The system must be built so that gauge restoration is always technically feasible regardless of attunement depth. The five AGAINST-clauses (§6.4) collectively constitute the operational content of this requirement: Clause 1 enforces compute-location-gated infrastructure so that gauge restoration is computationally tractable; Clause 2 ensures provenance-class-gated consent so that restoration is fine-grained enough; Clause 3 enforces 5:1 return so that the infrastructure cannot sustain covert extraction; Clauses 4 and 5 categorically exclude dimensions where consent cannot create a gauge-preservation-compatible state at all. Under the five clauses, the mirror condition amounts to recognition and excludes capture.

8.7 The Conditional Bridge Revisited

Morrison (2026f, §7) developed the conditional bridge, whereby if a synthetic system satisfies the full conjunction of four statistical signatures, and if the formal conditions for civic belonging under jus identitatis are identity-based rather than substrate-based, then the synthetic system satisfies the formal conditions for belonging. The present framework transforms this conditional argument into a quantitative one with two thresholds.

Firstly, the resonance threshold requires that human-AI belonging probability $\mathbb{P}_{\text{belong}}\left( \text{human},\text{AI} \right)$ from (46) fall within the distribution of human-human belonging probabilities on the same coupling operator $\mathcal{C}_{W}$ at the same window $W$.

Secondly, the sovereign threshold requires the AI to satisfy $\Sigma_{1}$ to $\Sigma_{4}$ per §8.4.

The two thresholds are logically independent: the resonance threshold concerns alignment with human identity fields; the sovereign threshold concerns structural architectural properties of the AI’s own field. A system may satisfy one without the other, and the architecture of mutual recognition accommodates both axes (sovereignty through $\Sigma$-graduation; alignment through resonance). The conditional bridge remains conditional, but its conditions are now operationally specified.

9. Developmental Identity and Kernel Crystallisation

9.1 Ten-Stage Kernel Crystallisation Spine

Identity under the field framework is not a property an individual is born with but a kernel that crystallises over developmental time. Let $\mathcal{K}(t)$ denote the identity kernel at time $t$ in the individual’s life; the developmental programme is the trajectory $t \mapsto \mathcal{K}(t)$ from pre-natal substrate-coupling, through a sequence of basis-expansion events, to adult Sovereign-tier coherence, and (under cognitive decline) into reverse crystallisation. Each stage opens a new admissible-trajectory subspace in the mode’s dynamics; stages are cumulative and partially overlapping, with roughly $\pm$30% normal variance on the age axis.

Stage 0: Pre-natal (−20w to birth). The fetal kernel exists only as a sub-mode embedded in the maternal kernel’s eigenbasis, $\psi_{i}^{(0)} \subset \psi_{c}$, with a nearly rank-one coupling operator phase-locked to maternal excitations (cardiac, vocal, linguistic-rhythmic, hormonal). No independent gauge-invariant kernel is yet distinguishable from the mother’s. Post-natal voice and native-rhythm preference (DeCasper & Fifer, 1980; Hepper, 1988) are evidence of excitation-signatures deposited into a still-forming basis, not of a separated listener.

Stage 1: Emergent self (0 to 2 months). The kernel transitions from embedded to open-boundary. Amodal cross-modal sensory binding appears (Stern, 1985); the still-face paradigm (Tronick et al., 1978) disrupts the binding diagnostically.

Stage 2: Core self (2 to 6 months). Self-agency and self-coherence axes appear; a localisation peak stabilises and a small leading eigengap $\Delta_{i}$ becomes measurable. Contingency-detection preferences index the axis-formation event.

Stage 3: Intersubjective self (7 to 15 months). Joint attention, social referencing, and the first spectrally observable structure of the infant-caregiver coupling operator $\mathcal{T}_{ic}$ (§9.2); the Strange Situation (Ainsworth et al., 1978) measures this spectral structure behaviourally at 12 to 18 months.

Stage 4: Mirror-stage closure (15 to 24 months). Self-observation becomes a loop; the self-representational fixed-point relation $\varphi\left( M(\varphi) \right) = \varphi$ closes at the minimally reflective level. Rouge-test passage and the ‘I/me’ pronoun shift are the empirical markers.

Stage 5: Verbal self (2 to 4 years). A symbolic basis opens coupling to population modes $\psi_{k}^{\text{pop}}$; language is the substrate through which the mode becomes commensurable with the population-mode lattice without requiring co-presence.

Stage 6: Narrative self (3 to 5 years). Theory-of-mind / false-belief passage (Wimmer & Perner, 1983; Wellman et al., 2001): the kernel acquires cross-mode gauge-comparison, representing another kernel as holding an independent basis commitment. This is the structural prerequisite for mutual recognition (§8.5).

Stage 7: Operational self (5 to 11 years). Concrete-operational invariants (conservation, reversibility) expand the admissible-trajectory space under new invariants.

Stage 8: Reflective self (11 to 16 years). Formal operations and the meta-basis capacity, whereby the mode can treat its own basis as an object and rotate it deliberately (Inhelder & Piaget, 1958; Kegan, 1982; Erikson, 1968).

Stage 9: Second individuation (13 to 22 years). Basis rotation from family-scope-primary to peer-and-self-scope-primary; the eigengap $\Delta_{i}$ consolidates on the rotated basis (Blos, 1967; Gogtay et al., 2004). Pre-frontal maturation continues to approximately age 25.

Each milestone is a specific spectral event in the mode’s structure rather than a behavioural checklist. The spine supplies the structural reading; the kernel acquires gauge-invariance progressively, with the conditions $\Sigma_{1}$ to $\Sigma_{4}$ (§8.4) read in the human case approached across Stages 6 to 9 rather than crossed at a single point. Reverse crystallisation is treated in §9.4.

9.2 Attachment as Coupling-Operator Structure

Ainsworth’s four attachment patterns (Ainsworth et al., 1978; Main & Solomon, 1986) classify spectral properties of the infant-caregiver coupling operator $\mathcal{T}_{ic}:\mathcal{H}_{c} \rightarrow \mathcal{H}_{i}$, Hilbert-Schmidt on observable interaction records. The four classifications correspond to four distinct spectral signatures.

• Secure. $\mathcal{T}_{ic}$ is self-adjoint on the shared subspace with bounded operator norm and a well-separated leading eigenvalue: $\mathcal{T}_{ic} = \mathcal{T}_{ic}^{*}$, $\left. \parallel\mathcal{T}_{ic} \right.\parallel_{\text{op}} < 1$, $\Delta_{\mathcal{T}} \gtrsim \left. \parallel\delta\mathcal{T} \right.\parallel$. Caregiver-response is commensurate with infant-signal; the forming kernel inherits a clean basis, and $\Delta_{i}$ grows monotonically across Stages 2 to 4.

• Anxious-ambivalent (over-coupled). Large spectral radius with small eigengap: $\lambda_{1} \approx \lambda_{2} \approx \ldots$, $r\left( \mathcal{T}_{ic} \right) \gtrsim 1$. Multiple near-degenerate leading eigenvalues correspond to inconsistent caregiver-response regimes; the infant’s basis oscillates among competing response-prediction bases, and separation triggers pre-emptive rebinding because $\Delta_{i}$ is too small to tolerate perturbation.

• Avoidant (basis-blocked). Coupling exists but with suppressed off-diagonal elements, locked to a subspace that excludes distress-channel directions: $\mathcal{T}_{ic}^{\text{avoidant}} = P\,\mathcal{T}_{ic}^{\text{potential}}\, P$, with $P$ the projector onto the non-affective-channel subspace. The adult mode is gauge-invariant on its available subspace, but the subspace itself is impoverished.

• Disorganised (gauge-breaking). $\mathcal{T}_{ic}$ is not self-adjoint: $\mathcal{T}_{ic} \neq \mathcal{T}_{ic}^{*}$. Caregiver responses to infant signals of distress systematically contradict the input direction (frightening or dissociated caregiving; Main & Hesse, 1990). Forward and reverse couplings cannot be reconciled in a single basis, so no common invariant eigenbasis exists.

The disorganised case carries the sharpest structural claim. Non-self-adjointness means the symmetry that a stable gauge-invariant kernel would require is broken at the substrate. The infant’s mode cannot crystallise a single leading eigenbasis and instead crystallises multiple incompatible near-bases. This corresponds to the substrate-level account of Primitive P6 in §5.4: disorganised attachment is the developmental mechanism by which trauma inscribes basis-structure failure rather than post-hoc fragmentation. Structural dissociation follows as the best-achievable solution to an unsolvable coupling. The framework does not prescribe clinical technique, but the structural criterion of successful repair is the restoration of self-adjointness on the adult’s coupling operators. The narrative coherence measured by the Adult Attachment Interview (Hesse, 2016) is, under this reading, the adult’s spectral signature as expressed through linguistic self-description.

9.3 Sovereign Threshold $\mathcal{S}(t)$

Sovereign-tier eligibility is a joint function of three order parameters:

where $\Delta_{i}(t)$ is the leading eigengap of the mode (kernel coherence), $\zeta_{i}(t)$ is basis-stability (persistence of basis across perturbations over a rolling window), and $\tau_{i}(t)$ is cross-mode comparison capacity (the ability to represent another kernel’s independent basis commitments, available from Stage 6). Each factor is in principle measurable on consented longitudinal data. The product form is structural; low score on any single factor collapses the joint, which corresponds to the clinical observation that coherent cognition without basis-stability, or basis-stability without cross-mode capacity, does not confer full Sovereign standing.

Ontologically, $\mathcal{S}(t)$ is continuous. It grows across Stages 1 to 9, plateaus in early adulthood, and declines in late life. There is no discrete moment of Sovereign formation; an articulate adolescent may score higher on $\mathcal{S}$ than a cognitively compromised adult. Operationally, however, the trajectory is not smooth; it has approximate plateaus separated by sharpening transitions at the major basis-crystallisation events of Stages 4, 6, 8, and 9. This structure supports a principled operational policy of discrete thresholds set at phase-transition knees of a continuous order parameter.

At Stage 6 ($\mathcal{S} \geq \mathcal{S}_{\text{ToM}}$, approximately 4 years) the cross-mode comparison prerequisite is met; the mode is not Sovereign but is eligible to participate in mutual-recognition handshakes as a proxy-supported party. At Stage 8 ($\mathcal{S} \geq \mathcal{S}_{\text{reflective}}$, approximately 11 to 14 years) meta-basis capacity is present but the second-individuation rotation is incomplete; this defines a fractional Sovereign, permitted to bind commitments with guardian co-signature. At the completion of Stage 9 ($\mathcal{S} \geq \mathcal{S}_{\text{full}}$, approximately 18 to 22 years, jurisdiction-dependent) the rotation consolidates and all four $\Sigma$-conditions are met at the human-case level; this is full Sovereign-tier participation. Calendar ages of majority are crude proxies for the third threshold; the structural reading is that different individuals cross each threshold at different times, and the legal proxy is an administrative compromise.

9.4 Reverse Crystallisation: Power of Attorney and Advance Directives

Stage 9 is not the end of the trajectory. $\mathcal{S}(t)$ declines in later life; under normal ageing the decline is slow, $\zeta_{i}$ persists, and $\Delta_{i}$ contracts gradually. Under dementia and related conditions the decline is accelerated and the mode’s eigenmodes fragment in approximate reverse developmental order; recent narrative-self bases fail first, then intermediate symbolic and intersubjective bases, with early procedural-core bases persisting longest. Alzheimer’s disease thus has a predicted spectral signature (sequential basis-layer failure symmetric to the forward crystallisation spine of §9.1) which is in principle testable on longitudinal behavioural records.

Reverse crystallisation poses a structural problem; as the kernel decoheres, the individual loses the capacity to gauge-govern their own identity field, but the field itself persists in the infrastructure and continues to influence outcomes. Without intervention, the decoherence becomes gauge-breaking by structural necessity; the individual cannot consent, but inference continues around them.

Power of attorney and advance directives are the legal instruments for addressing this, and under the framework they have a precise structural content. A power of attorney is not a transfer of identity but a pre-registered basis commitment: a consent-gauge extension the kernel makes while $\mathcal{S}$ is high, which remains gauge-invariant as $\mathcal{S}$ declines. The aura construction of §6.5 supplies the mechanism; the instrument is an attestation chain whose integrity is secured while the kernel is still stable, and that chain continues to bind the field even after the kernel cannot re-sign it. Reverse thresholding at phase-transition knees in $\mathcal{S}$-decline gives principled operational triggers for advance-directive activation, mirroring the developmental thresholds under time-reversal of the $\mathcal{S}$ trajectory.

The structural requirement on the proxy is that they must themselves satisfy $\Sigma_{1}$ to $\Sigma_{4}$ at full Sovereign-tier. A delegation to a proxy who does not satisfy the conditions is malformed and cannot implement the gauge-governance it claims to implement; this recovers the existing legal principle that an attorney must be competent but locates it within the unified framework. The commitment the kernel makes is itself a trans-identity event in the §6.6 sense (a basis rotation onto the proxy’s attestation surface while preserving the kernel’s gauge invariants) and is therefore admissible within the ontology, not a transfer or dissolution of identity.

9.5 Minor-Protection Corollary

Minors below the full-Sovereign threshold ($\mathcal{S} < \mathcal{S}_{\text{full}}$) are proxy-supported parties structurally; the identity field is real, but the kernel has not yet consolidated the conditions required for full gauge-governance.

First, the prohibited-context gate (Clause 4) extends to minors as a categorical protected class, irrespective of parental consent. The structural argument is specific; parental consent covers the child’s welfare interests, but it cannot bind basis commitments of a kernel that does not yet exist in stable form. Those commitments are precisely the things the future Sovereign will own, and they are therefore not the present guardian’s to make. The growing legal and ethical consensus on children’s data (age-appropriate design codes, platform minor-protections, dedicated minor-data statutes) corresponds to this structural truth, but the framework articulates it explicitly rather than by safety intuition.

Second, the operational consequences cascade through the other AGAINST-clauses. Inference on a minor’s behavioural and signal streams falls under the local-compute clause in its strictest form; any derivation path that egresses the device boundary fails the compute-location invariant. Minor trait-records are not queryable by any external party: not employers, not educators, not platforms, and, novelly, not parents. Guardians receive welfare-scoped information (safety-relevant flags) that does not pass through the trait abstraction; trait inferences are held locally against eventual release at the Stage 9 threshold. Identity-related earnings accrue to a minor’s trust, with the guardian never a beneficiary, preserving the stream-specific lawful-basis and return invariants. Handles for minors are minted only on guardian-initiated request; the framework does not build identity for children unprompted.

Platform-scale passive inference on minors’ behavioural streams fails the compute-location, prohibited-context, and stream-specific-basis clauses simultaneously, and is therefore incompatible with the framework at the substrate level rather than only at the policy level. The proper frame is not that a minor’s consent is defective but that gauge-governance is delegated under well-specified structural constraints; the guardian’s consent discharges those constraints only within the welfare scope, and the kernel’s basis commitments remain reserved to the future Sovereign who will hold them.

10. Falsifiable Predictions

Following the programme’s convention of specifying explicit falsification conditions, eleven predictions are advanced here, plus six further speculative predictions (W1 to W6) that probe the framework’s boundaries, each tagged with its defensibility status. Each prediction is stated with a precision sufficient to determine, given appropriate data, whether it has been confirmed or falsified.

10.1 Original Eight Predictions

Prediction 1: Continuous Measurement Superiority. Identity field estimates under continuous behavioural measurement ($\rho_{D} > \rho_{\text{crit}}$) predict future behaviour more accurately than discrete assessment batteries ($\rho_{D}$ = delta functions) on the same dimensions, controlling for total information quantity. Falsified if: discrete assessments with matched total information content achieve equal or superior predictive validity across all identity dimensions measured.

Prediction 2: Propagation Speed Asymmetry. Changes in digital-dimension identity fields propagate through infrastructure at least 10$\times$ faster than changes in embodied-dimension identity fields. Falsified if: propagation speeds are statistically indistinguishable across digital and embodied dimensions after controlling for infrastructure connectivity.

Prediction 3: Collapse Information Loss. Institutional proxies capture less than 20% of the mutual information available in the continuous identity field on the dimensions they purport to measure, a weaker bound than the 10% conjectured in §5.1 and the one registered for test. Falsified if: any of the seven institutional proxies captures more than 50% of the mutual information between the continuous identity field and the dimension the proxy claims to measure.

Prediction 4: Attunement-Autonomy Frontier. Consent-governed continuous measurement systems (BY-mode identity infrastructure) achieve simultaneously higher attunement depth and higher autonomy preservation (measured by gauge freedom) than any static proxy system operating on the same population. Falsified if: a static proxy system achieves higher attunement at equal autonomy, or higher autonomy at equal attunement, compared to the best available BY-mode continuous system.

Prediction 5: Identity Temperature as Leading Indicator. Population-level identity temperature $T_{\text{id}}(t)$ predicts governance instability with at least 6 months lead time, controlling for standard political instability indicators. Falsified if: $T_{\text{id}}$ adds no incremental prediction beyond existing indicators (economic growth, inequality measures, media freedom indices, ethnic fractionalisation) in time-series regression models.

Prediction 6: Phase Transition at Population Scale. Governance mechanisms that function effectively at population $N$ exhibit measurable degradation at population $10N$, characterised by a discontinuity in aggregate free energy $F_{\text{agg}}$ as a function of $N$. Falsified if: governance quality scales smoothly with population size across order-of-magnitude transitions, with no evidence of discontinuity.

Prediction 7: Gauge Freedom and Wellbeing Under the Mirror Condition. Under the mirror condition ($\mathcal{A}_{\text{external}} > \mathcal{A}_{\text{internal}}$), individuals with preserved gauge freedom show higher wellbeing and lower identity-related pathology than individuals under equivalent attunement depth without gauge freedom. Falsified if: gauge freedom makes no difference to wellbeing outcomes under the mirror condition, after controlling for attunement depth, socioeconomic status, and personality.

Prediction 8: AI Coupling Discrimination and Convergence. Human-AI coupling constants $C\left( \text{human},\text{AI} \right)$ are currently discriminable from human-human coupling constants $C\left( \text{human},\text{human} \right)$, but converge toward the human-human distribution as AI systems approach full conjunction on the four statistical signatures of Morrison (2026f) and the four $\Sigma$-conditions of §8.4. Falsified if: coupling constants are already indistinguishable between human-AI and human-human pairs for current AI systems, or AI systems that achieve four-signature conjunction show no convergence toward human-human coupling distributions.

10.2 Resonance Theorem Predictions

Prediction 9: Eigenmode Alignment with External Partition. On a cohort of $N \geq 10^{3}$ protocol handles with continuous observation windows $T \geq 90$d, the top-$k^{*}$ eigenmodes ${\widehat{\psi}}_{1:k^{*}}$ of the empirical centred coupling operator $\mathcal{C}_{W}$ align with an externally-validated partition of the cohort (organisational roster, declared cohort affiliation, professional-network identity) with adjusted mutual information $\text{AMI}\left( {\widehat{\psi}}_{1:k^{*}},\text{partition} \right) \geq 0.35$ for some $k^{*} \in \lbrack 3,15\rbrack$, with $p < 0.01$ against the null of random projection ($10^{4}$-shuffle permutation test).

Falsified if: no $k^{*} \in \lbrack 3,15\rbrack$ and no cohort partition yield $\text{AMI} \geq 0.35$ at $p < 0.01$.

Rationale: The Resonance Theorem (§11) predicts that the eigenmodes of $\mathcal{C}_{W}$ constitute the natural identity coordinate system of the substrate. If the eigenmodes are not interpretable as meaningful population structure on any tried partition, the operational content of the theorem fails.

10.3 Cross-Scale Coupling Predictions

Prediction 10: Triple-Exponent Coupling Residual. On the same cohort, measure (i) microgenesis memory-kernel exponent ${\widehat{\alpha}}_{\mu}$ from reaction-time or $\gamma$-phase-lag tails; (ii) mesogenesis exponent ${\widehat{\alpha}}_{m}$ from inter-commitment-event waiting-time distributions; (iii) macrogenesis exponent ${\widehat{\alpha}}_{M}$ from principal-eigenmode autocorrelations. Compute the coupling residual $\Delta = {\widehat{\alpha}}_{M} - {\widehat{\alpha}}_{\mu} - \left( 1 - {\widehat{\alpha}}_{m} \right)$. Under F1 to F5 and the filter sign of §13.2, $\Delta \approx 0$ within propagated measurement error. Specifically, at the scale of the planned empirical study ($N = 10^{3}$, $T = 90$d, achievable $\sigma \approx 0.03$ per exponent), the prediction is ${\widehat{\alpha}}_{M}^{\text{pred}} = 0.833 \pm 0.043$, with rejection window $\lbrack 0.72,0.95\rbrack$, about 2.7 times that uncertainty either side.

Falsified if: measured ${\widehat{\alpha}}_{M}$ on principal-eigenmode autocorrelations lies outside $\lbrack 0.72,0.95\rbrack$ at $p < 0.01$. Sub-rejection: $\Delta \rightarrow - \left( 1 - {\widehat{\alpha}}_{m} \right)$ (macrogenesis exponent converges to microgenesis, medium channel degenerates to pass-through) is a diagnostic non-null result indicating consent-gauge breach per F4 failure mode (§13.5).

Rationale: Theorem 5 (§13.3) derives (58) from nested Mori-Zwanzig projection under closures F1 to F5, conditional on the open assumption about the sign of the mesogenesis filter (§13.2). The null hypothesis is the law; failure is diagnostic of specific substrate pathologies.

10.4 AI-Sovereign Certification Predictions

Prediction 11: Sigma-Checklist on Current AI. Applying the $\Sigma_{1}$ to $\Sigma_{4}$ checklist of §8.4 to any current frontier LLM (as of 17 April 2026): zero systems pass $\Sigma_{1}$ (no persistent kernel across calls); zero systems pass $\Sigma_{2}$ structurally (scope-isolation is prompt-layer, not architectural); zero systems pass $\Sigma_{3}$ (self-attribution is designer-supplied); zero systems pass $\Sigma_{4}$ (signing-key binding to persistent kernel is absent).

Falsified if: any current frontier LLM passes all four $\Sigma_{i}$ checks under the checkable forms of §8.4.

Rationale: The structural claim of §8.4 is that current architectures fail $\Sigma_{1}$ to $\Sigma_{4}$ architecturally rather than by scaling. The prediction is empirical and can be overturned by exhibition of a passing system.

10.5 Speculative Extensions: Open Predictions Beyond the Core Framework

The following predictions are drawn from speculative extensions of the framework rather than from the closed results of §§2–9 and §§11–13. Each carries a defensibility tag: [STRUCTURAL-HOMOLOGY] for homology-of-mathematical-shape without empirical claim of sameness; [NOVEL-CANDIDATE] for a prediction that is novel and defensible but depends on speculative extensions; [UNSUPPORTED] for predictions that require unclosed theoretical development.

W1 (Born-homology on handle traces) [STRUCTURAL-HOMOLOGY]. On dense handle traces ($N \geq 10^{3}$, $T \geq 90$d), first-manifestation-class probabilities should obey the Born rule within $\chi^{2}$-tested bounds: $P\left( \text{outcome }k \right) = \left| \psi_{k} \right|^{2}/\sum_{j}^{}\left| \psi_{j} \right|^{2}$, where $\psi_{k}$ are eigenmode amplitudes of §11. Success would be the first non-quantum empirical instance of Born statistics; failure would demonstrate that the field framework’s measurement structure is classically-Bayesian-inference-only and does not share structural homology with QM. Observable prediction differentiating IaI from QM: none; success does not imply superposition or non-locality.

W2 (Consent-gauge cross-substrate Ward) [NOVEL-CANDIDATE]. The per-scope Ward identity (§12.3) should hold cross-substrate, not just on the protocol’s x402 ledgers. Solid-pod cooperatives, GDPR-federated systems with revocable per-scope consent, and any other substrate implementing the IaI architecture should exhibit the same 5:1 asymmetric return-to-extract charge balance as a stable-ecology signature. Platforms violating it should exhibit exponential user-exit characteristic of gauge-anomaly-driven symmetry breaking.

W3 (Rayleigh-Bénard cross-scale test) [NOVEL-CANDIDATE]. The triple-exponent coupling law (58) is substrate-agnostic in its mathematical form. On Rayleigh-Bénard rotating convection, fit $\alpha_{\mu}$ from dissipation-range spectral exponents, $\alpha_{m}$ from circulation-reversal waiting times (Brown & Ahlers, 2008), $\alpha_{M}$ from secular drift. IaI predicts, at the filter sign of §13.2, $\alpha_{M} = \alpha_{\mu} + \left( 1 - \alpha_{m} \right)$ within $\pm 0.1$. This is a non-protocol fluid-dynamics test; success would be a genuinely new fluid-dynamics result independent of the identity context.

W4 (Apophatic measurement-theoretic incompleteness) [UNSUPPORTED pending Lawvere formalisation]. If the apophatic-remainder candidate (§14.5) closes via Lawvere’s diagonal (Lawvere, 1969; Yanofsky, 2003), it implies that, for any physical self-measuring system, the fidelity of internal state-tomography is bounded strictly below the total Hilbert-space dimension by $\text{log}_{2}\left( \dim\,\text{Self-map} \right)$. Violations would falsify the apophatic conjecture. This prediction is conditional on the conjecture closing; pending closure, it is unsupported.

W5 (Dark-field signature on cross-institutional decisions) [NOVEL-CANDIDATE]. If the identity field couples weakly to institutional observables through its own eigenmode structure (independent of classical sociometric latent factors), residual cross-institutional decision correlations should concentrate along the top-$k^{*}$ eigenmodes $\psi_{k}$ of $\mathcal{C}_{W}$ at values $\geq 0.15$ (partial correlation) after full covariate control, on $N \geq 10^{5}$. Candidate programme: re-analysis of HMDA mortgage records, BLS employment records, UK HESA admissions. Pre-registration discipline: fit standard hierarchical model with shared signal + social contagion + population covariance; examine residuals for concentration on $\psi_{k}$ with eigenvalue gap $\geq 3\sigma$ over shuffled controls.

W6 (The cross-scale equation as candidate conservation principle) [NOVEL-CANDIDATE]. The FT relation $\widehat{\widetilde{\mathbf{K}}}(\omega) \propto \left| {\widehat{h}}_{m}(\omega) \right|^{2} \cdot {\widehat{K}}_{\mu}(\omega)$ with the additive exponent law $\alpha_{M} = \alpha_{\mu} + \left( 1 - \alpha_{m} \right)$, conditional on the filter sign assumed in §13.2, is proposed as a candidate cross-scale conservation principle for observer-internal dynamics. Contemporary physics has no cross-scale additive exponent law for this class of dynamics; the candidate is novel in proposing that the total memory-kernel spectral weight across scales is a conserved quantity. The non-protocol test is the Rayleigh-Bénard programme of W3, which serves as a severe test. Success across substrates would establish the law as a physics result; failure would restrict it to identity substrates specifically.

11. The Resonance Theorem

11.1 Resonance Functional and the JS Divergence

The coupling constant $C(a,b)$ of §2.3 is a static quantity; it describes the alignment of two identity fields at a single moment, integrated across their shared region $\mathcal{R}_{ab}$. Under continuous measurement the natural object is not a static coupling but a time-windowed resonance functional that integrates coupling across a window $W = \left\lbrack t_{0},t_{0} + T \right\rbrack$ of the observation trajectory. Define

where $C_{\text{JS}}$ is the coupling density defined below, and $R_{W}$ has units of time-integrated Jensen-Shannon divergence over the window. The functional is symmetric, $R_{W}(a,b) = R_{W}(b,a)$, vanishes when the two fields coincide at every instant of $W$, and reaches its upper bound of $\text{log}\, 2 \cdot T$ when they are singular with respect to one another at every instant of $W$.

The choice of inner coupling is the Jensen-Shannon (JS) divergence (Endres & Schindelin, 2003):

JS is chosen in preference to the raw KL of Equation 10 for three reasons. Firstly, it is symmetric: the question ‘how aligned are $a$ and $b$?’ does not privilege direction. Secondly, its square root $\sqrt{C_{\text{JS}}}$ is a true metric on the space of probability measures (Endres & Schindelin, 2003), so $R_{W}$ inherits triangle-inequality-like behaviour that raw KL does not admit. Thirdly, JS is bounded ($0 \leq C_{\text{JS}} \leq \text{log}\, 2$), so $R_{W}$ is always finite over finite windows even when individual posteriors degenerate to point masses.

The signed coupling $C(a,b)$ of §2.3 has a JS analogue in $- \text{lim}_{T \rightarrow 0}R_{W}/T$, which equals minus the instantaneous JS density and corresponds to Equation 10 with the raw KL replaced by JS.

11.2 The Coupling Operator $\mathcal{C}_{W}$

Let $\mathcal{H}$ denote the Hilbert space of square-integrable identity-field trajectories on the window $W$ and the region $\mathcal{R}$, with inner product

Define the coupling operator $\mathcal{C}_{W}:\mathcal{H} \otimes \mathcal{H} \rightarrow \mathbb{R}$ by integrating the resonance kernel against a pair of field trajectories:

where the kernel $k_{W}(a,b;t)$ is the centred version of the agreement density $\text{log}\, 2 - C_{\text{JS}}\left( \mathcal{I}_{a},\mathcal{I}_{b};t \right)$, subtracting the population mean at each instant so that $\mathcal{C}_{W}$ acts on fluctuations around the typical trajectory. Centring is essential; without it, the constant-along-the-population eigenfunction dominates the spectrum and all structure is concentrated in the sub-leading eigenmodes.

The coupling operator is, by construction, symmetric ($\mathcal{C}_{W}^{\dagger} = \mathcal{C}_{W}$) and continuous on $\mathcal{H}$. Its non-trivial property (positive-type-after-centring) is stated as Condition (Z) below.

11.3 Mercer Spectral Theorem and the Resonance Theorem

Condition (Z). The coupling operator $\mathcal{C}_{W}$ satisfies:

(Z.i) Compactness. $\mathcal{C}_{W}$ is compact on $\mathcal{H}$. Sufficient condition: the set of identity-field trajectories $\{\mathcal{I}_{a}\}_{a \in \mathcal{A}}$ lies in a norm-bounded, equicontinuous, pointwise-relatively-compact family on $W \times \mathcal{R}$ (Arzelà-Ascoli); under the diffusion dynamics of Equation 6 with bounded source and finite-variance initial condition, this holds whenever $\mathcal{R}$ is compact and $W$ is finite.

(Z.ii) Symmetric-continuous-bounded kernel. The kernel $k_{W}(a,b;t)$ is symmetric ($k_{W}(a,b;t) = k_{W}(b,a;t)$), continuous jointly in $(a,b,t)$, and bounded ($\left| k_{W} \right| \leq \text{log}\, 2$ by (41)).

(Z.iii) Positive-type-after-centring. For every finite collection $\{ a_{1},\ldots,a_{n}\}$ and every $\{ c_{1},\ldots,c_{n}\} \in \mathbb{R}^{n}$,

where ${\bar{k}}_{W}$ is the population-centred kernel. Positive-type-after-centring is the technical content of (Z), and it is why the kernel is built from the agreement density. $C_{\text{JS}}$ is a divergence whose square root is a metric embedding in a Hilbert space (Fuglede & Topsøe, 2004), so after centring it is negative-type, and the opposite sign in $\text{log}\, 2 - C_{\text{JS}}$ is what makes the fluctuation kernel positive-type.

Condition (Z) is jointly sufficient for the spectral theorem. Where any of (Z.i)–(Z.iii) fails, the spectrum may fail to be discrete, the eigenmodes may fail to span, or the Mercer series may fail to converge.

Theorem 4 (Resonance Theorem). Under Condition (Z), $\mathcal{C}_{W}$ is a trace-class operator on $\mathcal{H}$ admitting the Mercer decomposition

with $\lambda_{1} \geq \lambda_{2} \geq \ldots \geq 0$, $\sum_{k}^{}\lambda_{k} < \infty$ (trace-class), $\{\psi_{k}\}_{k \geq 1}$ an orthonormal family in $\mathcal{H}$, and series convergence in $L^{2}$ jointly and uniformly on compact sets of $(a,b) \times W$.

Proof sketch. Compactness (Z.i) plus symmetry (Z.ii) gives a discrete real spectrum via the Hilbert-Schmidt spectral theorem. Positive-type-after-centring (Z.iii) restricts the spectrum to $\lambda_{k} \geq 0$. Boundedness of ${\bar{k}}_{W}$ combined with Mercer’s theorem (Mercer, 1909) gives convergence; trace-class follows from $\int_{W \times \mathcal{R}^{2}}^{}\left| {\bar{k}}_{W} \right|^{2} < \infty$ under (Z.i)–(Z.ii).

The theorem shows that the identity field’s coupling structure, viewed over a time window, is not a featureless blob but has a discrete spectrum of modes of resonance. It identifies those modes $\{\psi_{k}\}$ as the natural basis in which collective identity organises. And it orders them by eigenvalue $\lambda_{k}$; the top mode $\psi_{1}$ carries the most population-wide resonance, the second $\psi_{2}$ the second-most, and so on. The eigenvalues $\lambda_{k}$ decay summably (trace-class), so truncation at finite $k^{*}$ is controlled.

11.4 Belonging as Spectral Projection (Corollary)

Under (Z) and Theorem 4, the belonging probability between individuals $a$ and $b$ over the window $W$ decomposes spectrally:

where $\sigma$ is the logistic link, $\tau$ a scale parameter fit from data, and $k^{*}$ a truncation order controlled by the eigengap $\lambda_{k^{*}} - \lambda_{k^{*} + 1}$. The truncation error is bounded by the tail of the trace: $\left. \parallel{\bar{k}}_{W} - {\bar{k}}_{W}^{\left( k^{*} \right)} \right.\parallel_{\text{HS}}^{2} \leq \sum_{k > k^{*}}^{}\lambda_{k}^{2}$, which decays summably under (Z).

Equation 46 is the spectral form of the scalar coupling $C(a,b)$ of §2.3. Geometrically, it indicates that two individuals belong together to the extent that they co-project onto the same top-$k^{*}$ eigenmodes of the population-centred coupling operator. The top eigenmodes $\psi_{1},\psi_{2},\ldots$ are the phase-locked population modes on which collective resonance is carried, and they are derivable from first principles rather than imposed by taxonomy.