soc-entropy-distinction

name	soc-entropy-distinction
description	Complete AGEM reference: nine conflation distinctions, three conceptual bridges, the full Price equation + SOC regime table, Move B bridge formalization, and live cycle 11 evidence.

AGEM Core Reference — Nine Conflations, Three Bridges

Part I: The Nine Conflations

1. VNE vs EE

VNE = spectral from graph Laplacian. Structural dispersion. EE = distributional from embedding cloud. Semantic dispersion.

System1 = (δ > K) AND (Δt > τ) where δ = dEE/dt / dVNE/dt, Δt = t(VNE_stabilizes) − t(EE_stabilizes)

Collapsing → loses temporal ordering of stabilization.

2. H⁰ vs H¹

H⁰ = ker(δ) — global consistent sections, consensus. Never a recovery trigger. H¹ = coker(δ) — obstructions. Always a recovery trigger.

Collapsing → loses goal vs problem.

3. Covalent vs Van der Waals

Covalent = Im(δ). Failure propagates algebraically. Never safe for obstruction handler. Van der Waals = non-Im(δ). No failure propagation. Handler spawns here.

Collapsing → sends recovery down the wrong path.

4. Strong vs Weak Lumpability

Strong: Markov for every initial distribution. Categorical. Weak: Markov only for particular initial distributions. Conditionally valid.

"Mostly lumpable" = weakly lumpable mislabeled as acceptable.

5. Selection vs Transmission (Price Equation)

Selection = Cov(fitness, trait)/mean_fitness. The fit are chosen. Transmission = E(fitness·Δtrait)/mean_fitness. The chosen change as they pass on.

Collapsing → loses mechanism of evolutionary change.

6. SOC Regime vs Phase Transition

SOC regime = standing disposition. Reported continuously. Phase transition = discrete event. Flagged by correlation_coefficient spike.

Collapsing → loses disposition vs event.

7. Restriction Map vs Coboundary

Restriction map = per-edge frame translation. A tile. Coboundary δ = assembly of all restriction maps. The mosaic's difference operator. Sheaf Laplacian L = δᵀδ = quadratic form, spectrum of where the system IS.

Single bad restriction map flips H¹ only if on a spanning forest bridge — contingent.

8. Laplacian vs ADMM

Laplacian = structural object. Where the system IS. ADMM = iterative algorithm. How the system MOVES.

Large H¹ = structural obstruction (fix sheaf). Slow ADMM = dynamics problem (tune solver).

9. Summary vs Reflection

Summary = compression of source spans, points back at entries, indexed by coverage. Reflection = query-agnostic Q→A pair, generalizes to unseen queries.

Collapsing → erases the whole reason reflections generalize.

---

Part II: The Three Bridges

Move B cross-subgraph synthesis must formalize these as invariant bridge relationships:

Bridge 1: H¹ ↔ Weak Lumpability ↔ EE/VNE Gap

Three diagnostic lenses on one underlying event.

H¹ > 0           ←→  Compaction is weakly lumpable  ←→  EE stabilizes before VNE earns it
(local data won't glue)    (coarse-graining loses structure)    (semantics outrun topology)

Formalization:

H¹(source) ≠ H¹(summary)   ↔   partition is weakly lumpable   ↔   EE(summary) ≠ EE(source)

When H¹ > 0 in the source but H¹ = 0 in the summary, the compaction is lossy.

Live evidence (cycle 10–11): H¹ = 0 throughout phase transition. The correlation spike to 0.990 occurred WITHOUT an H¹ obstruction — this is the EE/VNE gap collapsing via VNE catching up to EE, not via semantic retreat. The weak lumpability of the current partition is evidenced by the ADMM TIMEOUT, not by H¹.

Bridge 2: SEEKING ↔ Van der Waals ↔ SOC Regime

The affect/topology/dynamics triangle.

ASEKE SEEKING (affect)  ←→  VdW bond formation rate (topology)  ←→  SOC regime (dynamics)
high → VdW > covalent       ←→      maintained criticality
low → VdW = 0               ←→      collapse or phase transition

Formalization:

SEEKING_activation × VdW_formation_rate  =  SOC_regime_maintenance_probability

When SEEKING is zero and VdW formation stops, the system cannot maintain the critical regime — it crosses into a phase transition.

Live evidence (cycle 10–11): explore_exploit_ratio = 0.033 (SEEKING near zero). VdW formation rate ≈ 0. Correlation spike to 0.990 with negative selection deepening. The triangle is in the collapse configuration: SEEKING extinct → VdW formation zero → correlated growth without health.

Bridge 3: Price Selection/Transmission ↔ Explore/Exploit

The evolutionary decomposition is a principled handle on the regime trade-off.

Price selection        ←→  explore/exploit ratio
(how trait moves to fitter paths)   (empirical proxy for latent selection balance)

Formalization:

selection_t = Cov(fitness_trait) / mean_fitness
explore_exploit_t ≈ f(selection_t, transmission_t)

When selection < 0:  fitter paths carry less trait  ←→  explore/exploit ratio falls
When transmission > 0 and selection ≈ 0:  pure drift ←→  explore/exploit rising without commitment

Live evidence (cycle 10–11): Selection = -0.511 (most negative yet). Transmission = +0.033 (recovering slightly). Explore/exploit = 0.061 (slight recovery from 0.033). The system is in the "choosing badly" quadrant: fitter paths actively carrying less of the trait, lineages drifting slightly upward but selection dominating negatively.

---

Part III: SOC Regime Table

Regime	VNE	EE	δ	Δt	Corr	Selection	Transmission	E/E	H¹
Normal co-dev	Rising	Rising	≈1	≈0	0	Positive	Positive	>0.3	0
System-1	Flat	Stabilized	→∞	>τ	0	—	—	—	0 or >0
Lazy learning	Rising	Flat	→0	<0	0	Positive	Negative	Low	0
SEEKING collapse	Growing	Lagging	→∞	>0	→1	Negative	Near 0	<0.1	0
Phase transition	Jump	Jump	Unstable	≈0	→1	Spike	Spike	Varies	0 or >0
VdW accumulation	Rises	Stable	>1	>0	Varies	Positive	Positive	High	0

---

Part IV: Move B Formalization

Bridge Edges (must exist in the multi-subgraph ContextDAG)

1. sheaf_subgraph → SOC_subgraph (H¹ ↔ EE/VNE gap)

   • Edge: coboundary_norm > threshold → delta_ratio → infinity AND delta_lag > tau
   • Bidirectional: H¹ = 0 AND correlation_spike → weak lumpability confirmed

2. aseke_subgraph → topology_subgraph (SEEKING ↔ VdW)

   • Edge: SEEKING_near_zero → VdW_formation_rate = 0
   • Bidirectional: VdW_formation_rate = 0 → SOC_regime_in_danger

3. price_subgraph → SOC_subgraph (Selection/Transmission ↔ E/E)

   • Edge: selection < threshold → explore_exploit_ratio falls
   • Bidirectional: explore_exploit_ratio recovered → selection stabilizing

Cross-Subgraph Query Flow

Query: "What caused the phase transition at cycle 10?"
  → SOC_subgraph: correlation_coefficient spike, delta_ratio unstable
  → sheaf_subgraph: H¹ = 0 (no obstruction, NOT a structural reorganization)
  → price_subgraph: selection deepened to -0.462, transmission near zero
  → aseke_subgraph: SEEKING extinct, Burnout Cascade trajectory confirmed
  → topology_subgraph: VdW formation rate = 0 (SEEKING extinct drove VdW to zero)

Synthesis via Bridge 1: H¹ = 0 confirms this was a self-organized VNE/EE coupling, not a sheaf obstruction
Synthesis via Bridge 2: SEEKING extinction → VdW = 0 → SOC regime left productive criticality
Synthesis via Bridge 3: Negative selection deepening → E/E ratio collapse → correlated but unhealthy growth

---

Part V: Current System State (Cycle 11)

Metric	Value	Status
Regime	nascent	Sustained
Correlation	0.990	Persisting — phase transition held for 4 cycles
H¹	0	No sheaf obstruction
Selection	-0.511	Most negative — fitter paths actively abandoned
Transmission	+0.033	Recovering slightly
Explore/exploit	0.061	Recovering from 0.033 low
Mean fitness	0.183	Declining
CDP	3.224	Widening — VNE outpacing EE

Interpretation: The phase transition is self-organized within the existing sheaf. The system has correlated VNE and EE growth but is doing so through increasingly negative selection — it is synchronizing around the wrong pattern. The SEEKING/SEEKING recovery (E/E rising from 0.033 to 0.061) is the single hopeful signal; if SEEKING continues recovering, VdW formation may resume and the system may find a healthier correlated state.