Part IV · Valence
# Frustration and Replica Symmetry Breaking
A state can be perfectly recurrent and still be terrible. Rumination returns to itself; so does chronic pain. The missing ingredient is frustration, and its mathematics is the mathematics of spin glasses.
> Assumes
> Chapters 6 and 7. Some acquaintance with statistical mechanics is helpful but §8.2 is self-contained.
> Delivers
> Axiom 8.1 — the valence functional (8.2) — plus the phase diagram and the annealing dosing law.
8.1 Purity is a replica partition function
The quantity $\mathcal{A}=\mathrm{Tr}\rho^2$ has a second life in statistical field theory. Writing $\rho=e^{-\beta H}/Z_1$,
$$ \mathrm{Tr}\,\rho^{\,n}=\frac{Z_n}{Z_1^{\,n}} $$
where $Z_n$ is the partition function of $n$ copies — replicas — of the system. In field theory (Callan–Wilczek; Calabrese–Cardy) $Z_n$ is computed on the $n$-sheeted branched cover of spacetime, glued along the entangling surface. That manifold carries a manifest $\mathbb{Z}_n$ symmetry: cyclic permutation of the sheets.
Whether the dominant saddle point respects that symmetry is a physical question with a name and a large literature. If it does, the system is replica-symmetric. If it does not, we have replica symmetry breaking, discovered by Parisi in the study of spin glasses and now understood to be the universal signature of frustrated, rugged energy landscapes.
8.2 Parisi's order parameter
The diagnostic is the distribution of overlaps between two independent samples from the same equilibrium ensemble:
$$ P(q)=\mathbb{E}_J\Bigl[\bigl\langle\,\delta(q-q_{ab})\,\bigr\rangle\Bigr], \qquad q_{ab}=\frac{1}{N}\sum_{i=1}^{N}s_i^{a}s_i^{b}, \qquad \mathcal{D}:=\mathrm{Var}_{P}(q) $$ (8.1)
In a replica-symmetric phase, two samples always look alike: $P(q)=\delta(q-q_{\mathrm{EA}})$ and $\mathcal{D}=0$. The system is one thing. Under RSB, $P(q)$ spreads over an interval and acquires a hierarchical, ultrametric structure — states organise into basins within basins within basins, with distances satisfying $d(x,z)\le\max\{d(x,y),d(y,z)\}$. The system is irreducibly many things at once, and which one it is depends on where it happens to have fallen.
> Ultrametricity is a strong and testable claim about the geometry of the state space, not a metaphor. It is Chapter 11, prediction 3.
I want to be explicit that "frustration" is not a pun here. In the technical sense it is the impossibility of simultaneously satisfying all pairwise constraints; in the phenomenal sense it is what internal conflict feels like. The claim of this chapter is that these are the same quantity seen through the two functors of Axiom 2.1 — one relationally, one intrinsically. The nested-basin structure of RSB is then a prediction about the felt organisation of suffering, and the description it gives — stuckness, hierarchical entrapment, the sense of being multiply committed and unable to settle — is recognisable.
8.3 The valence functional
> Axiom 8.1 · Valence
>
> Posited
>
> The valence of a moment is
>
> $$ \mathfrak{V}[\rho]\;=\;\mathcal{C}[\mu_\rho]\;\cdot\;\Bigl(1-\tfrac{2\,\mathcal{D}[\rho]}{\mathcal{D}_{\max}}\Bigr), \qquad \mathfrak{V}\in[-\mathcal{C},\,+\mathcal{C}] $$ (8.2)
>
> Consonance sets the magnitude; replica symmetry sets the sign.
The structural consequence is worth stating on its own, because it is the model's most easily testable qualitative claim: since $|\mathfrak{V}|\le\mathcal{C}$, you cannot suffer intensely without being coherent. Intensity of feeling, of either sign, is bounded by spectral coherence. Anaesthesia should therefore abolish agony and bliss by the same mechanism and at the same threshold — which is what it does.
8.4 The phase diagram
> [figure] Figure 8.1 — The two order parameters are independent, which is the whole point. High coherence alone does not give bliss: the glass is both highly coherent and highly frustrated. Escaping it requires transiently destroying coherence — the trajectory must dip through the low-𝒞 region before it can re-form on the symmetric side. That detour is what neural annealing describes, and it is why the process is unpleasant in the middle.
8.5 A cortical spin glass
To make the diagram quantitative, take $N$ coarse-grained cortical modes with Sherrington–Kirkpatrick couplings $J_{ij}\sim\mathcal{N}(0,J^2/N)$ in a uniform field $h$:
$$ H=-\sum_{i\lt j}J_{ij}s_is_j-h\sum_i s_i, \qquad q=\int\! \mathcal{D}z\;\tanh^2\!\bigl(\beta(J\sqrt{q}\,z+h)\bigr) $$ (8.3)
The replica-symmetric solution is stable only above the de Almeida–Thouless line, and near the critical temperature that line has a definite shape:
$$ (\beta J)^2\!\int\!\mathcal{D}z\;\mathrm{sech}^4\!\bigl(\beta(J\sqrt{q}\,z+h)\bigr)=1, \qquad \frac{h_{\mathrm{AT}}^2}{J^2}\;\simeq\;\frac{4}{3}\Bigl(1-\frac{T}{T_c}\Bigr)^{\!3} $$ (8.4)
> Proposition 8.2 · A dosing law
>
> Derived
>
> Escaping the glass means crossing above the AT line and re-cooling into the replica-symmetric basin. Equation (8.4) gives $h_{\mathrm{AT}}\propto(1-T/T_c)^{3/2}$: the drive required to anneal a state grows as the three-halves power of how rigid that state already is.
>
> Too little drive and the system stays frustrated. Too much and $\mathcal{C}\to 0$ — the coherence that bounds $|\mathfrak{V}|$ is destroyed and the trajectory lands in noise rather than crystal. There is an optimal path that skirts the critical line, and the therapeutic window is its width.
This is the formal content of Johnson's neural annealing: raise the energy enough to melt the frustrated structure, then cool along a path that lands in the symmetric basin. The model adds three things to the informal account — an order parameter ($\mathcal{D}$), a critical line (8.4), and an exponent (3/2) that can be measured and could be wrong.
> ### What Chapter 8 established
>
> - Purity is a replica partition function on a branched cover carrying a $\mathbb{Z}_n$ symmetry; whether the saddle respects it is physical.
> - Frustration $\mathcal{D}=\mathrm{Var}_P(q)$ is zero iff replica symmetry is unbroken; under RSB the state space is ultrametric.
> - Axiom 8.1: $\mathfrak{V}=\mathcal{C}(1-2\mathcal{D}/\mathcal{D}_{\max})$. Consonance gives magnitude, replica symmetry gives sign, and $|\mathfrak{V}|\le\mathcal{C}$.
> - Four regimes — crystal, glass, noise, inert — with suffering requiring coherence, not merely disorder.
> - Proposition 8.2: annealing obeys a three-halves dosing law with a finite therapeutic window.
> ### Exercises
>
> 1. Show $\mathrm{Var}_P(q)=0$ exactly when $P$ is a single delta, and hence that $\mathcal{D}$ detects RSB and nothing else. [1]
> 2. Verify from (8.2) that $\mathfrak{V}$ changes sign at $\mathcal{D}=\mathcal{D}_{\max}/2$, and discuss whether the sign change should be sharp or smooth. What would each imply experimentally? [2]
> 3. Solve (8.3) numerically at $h=0$ and locate $T_c$. Confirm that $q=0$ is the only solution above it. [2]
> 4. Derive the exponent 3/2 in (8.4) by expanding the AT condition to leading order in $\tau=1-T/T_c$. [2]
> 5. The ultrametric inequality is a strong constraint. Given $M$ sampled neural states and their pairwise distances, design a statistical test for ultrametricity with a null model that controls for hierarchical clustering artefacts. [2]
> 6. Open. $\mathcal{D}$ requires a disorder average, but a single brain is a single realisation. Justify — or refute — the use of time windows as replicas, given that non-self-averaging is precisely what RSB asserts. This is the weakest joint in the chapter. [3]
For agents
GET /api/library/spectral-panpsychism/ch8.md← Consonance and Arithmetic · Intensity and the Logarithmic Scales →