Quick Review
1. Core State Variables
Every node \(i\) at time \(t\) is governed by four interacting state variables:
| Variable | Notation | Definition | Biological Analog |
|---|---|---|---|
| Continuous Stress | \(S_i(t)\) | Instantaneous operational load on node \(i\). | Membrane Potential |
| Cascade Activation | \(F_i(t)\) | Rate of externalizing overflow stress. | Action Potential |
| Dynamic Capacity | \(C_i(t)\) | Immediate structural resilience. | Synaptic Depression |
| Baseline Target | \(C_i^*(t)\) | Long-term homeostatic capacity goal. | Homeostatic Plasticity |
2. The Four Continuous Dynamical Laws
These coupled, non-linear differential equations govern the microscopic behavior of the socio-technical system.
Law I: Stress Evolution
Governs the accumulation, dissipation, and externalization of operational stress:
\[\frac{dS_i}{dt} = L_i(t) - \gamma S_i(t) + \sum_{k} \sum_{j} W_{ji}^{(k)} F_j^{(k)}(t) - \sum_{k} F_i^{(k)}(t)\]Law II: Differentiable Cascade Activation
Replaces the rigid step-function of classical SOC with a bounded, differentiable sigmoid threshold:
\[F_i^{(k)}(t) = \kappa^{(k)} (S_i(t) - C_i(t)) \sigma(a(S_i(t) - C_i(t)))\](Where the logistic sigmoid function is \(\sigma(x) = \frac{1}{1+e^{-x}}\))
Law III: Dynamic Capacity Degradation
Calculates the immediate mathematical degradation of capacity (the burnout penalty) during an active stress cascade:
\[\frac{dC_i}{dt} = \mu(C_i^*(t) - C_i(t)) - \eta \sum_{k} F_i^{(k)}(t)\]Law IV: Homeostatic Adaptation
The algorithmic governor that drives the system toward self-organized criticality by tuning the baseline capacity against global activity \(A(t) = \frac{1}{N}\sum_i F_i(t)\):
\[\frac{dC_i^*}{dt} = \zeta(A_0 - A(t))\]3. Network Topology and Conservation
Multiplex Directed Graph
The organization is modeled across \(k\) interaction layers (e.g., formal hierarchy vs. informal email networks):
\[G = (V, E^{(k)})\]Strict Normalization Condition
Ensures stress is not artificially lost during a cascade transfer:
\[\sum_i W_{ji}^{(k)} = 1\]Global Leaky Conservation
Summing Law I over all \(N\) nodes cancels out internal propagation, proving the system is bounded exclusively by exogenous input and natural leak:
\[\sum_{i=1}^{N} \frac{dS_i}{dt} = \sum_{i=1}^{N} L_i(t) - \gamma \sum_{i=1}^{N} S_i(t)\]4. Macroscopic Thermodynamics and Stability
Systemic Entropy (\(H\))
Quantifies the macroscopic disorder and spatial diffusion of stress. A violent spike (\(\frac{dH}{dt} \gg 0\)) serves as an early warning for network percolation:
\[H(t) = - \sum_{i=1}^{N} p_i(t) \ln(p_i(t))\](Where stress probability mass is \(p_i(t) = \frac{S_i(t)}{\sum S_j(t)}\))
Normative Free Energy (\(\mathcal{F}_{NSD}\))
A pseudo-Lyapunov function measuring distance from critical energy depletion. Temperature \(\tau\) is strictly proportional to the variance of the exogenous load (\(\tau \propto \text{Var}(L)\)):
\[\mathcal{F}_{NSD}(t) = \sum_{i=1}^{N} C_i(t) - \tau H(t)\]Stability Boundary Equation
The time derivative of Free Energy defines absolute Lyapunov bounds:
\[\frac{d\mathcal{F}_{NSD}}{dt} = \sum_{i=1}^{N} \left[ \mu(C_i^* - C_i(t)) - \eta \sum_{k} F_i^{(k)}(t) \right] - \tau \frac{dH}{dt}\]Saddle-Node Burnout Bifurcation
Setting capacity degradation to equilibrium (\(\frac{dC_i}{dt} = 0\)) reveals the deterministic collapse boundary:
\[\mu(C_i^* - C_i) = \eta \kappa \left( S_i - C_i \right) \sigma(a(S_i - C_i))\]5. Macroscopic Approximations & Network Metrics
Mean-Field Approximation
Collapses the \(N\)-dimensional network to isolate global behavior:
\[\frac{d\langle S \rangle}{dt} = \langle L \rangle - \gamma \langle S \rangle\] \[\frac{d\langle C \rangle}{dt} = \mu(\langle C^* \rangle - \langle C \rangle) - \eta \langle F(\langle S \rangle, \langle C \rangle) \rangle\] \[\frac{d\langle C^* \rangle}{dt} = \zeta(A_0 - \langle F \rangle)\]Key Theoretical Control Parameters
- Systemic Control Parameter: \(\rho = \frac{\langle L \rangle}{\langle C \rangle}\)
- Active Branching Ratio: \(b = \langle k \rangle P_f\)
- Scale-Free SOC Avalanche Distribution: \(P(s) \sim s^{-\alpha}\)
- Random Graph Failure (Exponential Cutoff): \(P(s) \sim s^{-\alpha} e^{-s/s_c}\)