Experiments

Falsification Map

Experiment	Prediction	Outcome
V10 (MARL)	Forcing functions create geometry	Contradicted. All conditions show alignment; removal increases it.
Exp 2 (World Model)	$\mathcal{C}_{\text{wm}}$ increases with evolution	Partial. 100x at bottleneck, flat in general population.
Exp 3 (Representation)	Compression and modeling co-emerge	Partial. Co-emerge under bottleneck only. Compression is cheap.
Exp 4 (Language)	Compositional communication	Not confirmed. Chemical commons but $\rho_{\text{topo}} \approx 0$ .
Exp 5 (Counterfactual)	Reactive-to-detached transition	Null. Wall at $\rho_{\text{sync}} \approx 0$ .
Exp 6 (Self-Model)	SM emergence with $\intinfo$ jump	Weak. n=1 event at bottleneck.
Exp 7 (Affect Geometry)	Tripartite alignment	Partial. A-C develops over evolution (0.01 to 0.38). A-B null.
Exp 8 ( $\iota$ )	Participatory default, animism	Confirmed. $\iota \approx 0.30$ , animism > 1.0 in all 20 snapshots.
Exp 9 (Normativity)	Exploitation penalty	Null. Requires agency.
Exp 10 (Superorganism)	$\intinfo_G > \sum \intinfo_i$	Not confirmed. Ratio 1-12%, increasing.
Exp 11 (Entanglement)	Co-emergence clusters	Not confirmed. Different cluster structure.
Exp 12 (Capstone)	Seven criteria for identity thesis	All met (moderate/weak). Geometry confirmed.
V19 (Furnace)	Selection vs creation	Creation confirmed 2/3 seeds.
V20 ( $\rho$ wall)	$\rho_{\text{sync}} > 0.1$	Confirmed. 0.21 from cycle 0.
V22-V24 (Prediction)	Prediction → integration	Not confirmed. Linear readout always decomposable.
V27 (MLP)	Nonlinear head → $\intinfo \uparrow$	Confirmed (seed 7: 0.245). Seed-dependent.
V28 (Width)	Bottleneck width matters	Not confirmed. Mechanism is gradient coupling.
V29/V31 (Social)	Social target lifts $\intinfo$	Not confirmed. 3-seed fluke; 10-seed: $p = 0.93$ .
V30 (Dual)	Self+social > either	Negative. Gradient imbalance; self colonizes.
V31 (Seeds)	Seed distribution	Confirmed: 30/30/40 split. Post-drought bounce $r = 0.997$ .
V32 (Autopsy)	First bounce predicts category	Revised: First bounce NOT predictive ( $p = 0.60$ ). Mean bounce across all droughts IS ( $\rho = 0.60, p < 10^{-5}$ ). Trajectory, not event.
V35 (Language)	Referential communication emerges	Confirmed: 10/10 seeds (100%). But does NOT lift $\intinfo$ . Language is cheap.
VLM Conv.	VLMs recognize affect in protocells (RSA > 0.3)	Confirmed: GPT-4o $\rho = 0.72$ , Claude $\rho = 0.54$ . Raw numbers: 0.78, 0.72.

Falsification Scoreboard. 7 confirmed, 7 contradicted, 1 revised. The framework survives not by being right everywhere, but by being wrong in specific, informative ways. Each contradiction sharpened the theory — the forcing function failure led to the geometry/dynamics distinction; the social prediction failure revealed the gradient interference pattern; the language failure established the rung 4-5 / rung 8 boundary.