Temporal Stability of Internal Reference as a Sufficient Mechanism for Social Decoupling: An Ablation Gradient with Dilution Controls
Year: 2026
Type: preprint
Abstract: Paper 6 demonstrated that the self-model suppresses the synchrony-cooperation coupling (r = 0.478 → 0.029). Paper 7 characterizes the internal structure of that suppression through a partial ablation gradient (w_S ∈ {0, 0.1, 0.2, 0.3, 0.4, 0.5}, N = 400 pairs per condition, v13 architecture with the full self-model: autocorrelation-based τ and estimated Φ) and introduces two dilution controls designed to distinguish whether the effect requires self-referential content or merely an internal signal of equivalent weight: NOISE (Gaussian vector renewed each step, w_S = 0.45) and CONST (constant vector, w_S = 0.45). Results: (1) r(M3_sync, coop) falls monotonically from 0.431 to 0.019; the sigmoidal fit outperforms the linear fit in R² (0.998 vs 0.952) but AICc favors the linear model (−32.95 vs −10.71), so the non-linearity hypothesis H7.1 is not confirmed; the region of maximum slope lies at w_S ∈ [0.1, 0.2] as a description, not a critical threshold. (2) The dilution control yields a mixed result that constitutes the main finding: NOISE does not collapse the coupling (r = 0.427 ≈ no-self-model condition) while CONST collapses it completely (r = −0.007 ≈ real self-model). The operative variable is the temporal stability of the fed-back signal, not its self-referential content: a temporally stable internal reference is sufficient for social decoupling; self-monitoring is not necessary. (3) Cooperative mean remains invariant across all conditions (Δμ = 0.007). (4) With the C index computed from real Φ and τ, r(C, coop) is small positive (+0.10) at low w_S, reversing the sign observed in Papers 5–6 and revealing the C construct's sensitivity to its operationalization. A DPI diagnostic battery (N = 120) confirms that cooperative asymmetry is emergent (3/4 H_ruptura). This paper corrects and supersedes a prior analysis that employed a simplified self-model variant without declaring it; all differences are documented in Section 3.8.
Author: Santoyo Ortiz, Armando
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7204643848