Furkejuvvon:
Bibliográfalaš dieđut
Váldodahkki: Santoyo Ortiz, Armando
Materiálatiipa: Recurso digital
Giella:
Almmustuhtton: Zenodo 2026
Fáttát:
Liŋkkat:https://doi.org/10.5281/zenodo.20265133
Fáddágilkorat: Lasit fáddágilkoriid
Eai fáddágilkorat, Lasit vuosttaš fáddágilkora!
Sisdoallologahallan:
  • <p>We present PROSENTIR Phase 4, introducing the self-model as the<br>central architectural innovation and the C index as the first<br>operational measure of proto-consciousness in the PROSENTIR<br>framework (Phase 1: 10.5281/zenodo.19912924; Phase 2:<br>10.5281/zenodo.20019260; Phase 3: 10.5281/zenodo.20090917).</p> <p>The self-model S = [M3_norm, tau_norm, Phi_norm] implements<br>reflexive information integration: the agent monitors its own<br>global state, temporal persistence, and inter-module integration<br>in real time, feeding this self-perception back into its Level-3<br>global integrator with dominant weight w_S = 0.45. The C index<br>(version 1.0) is defined as:</p> <p>  C = 0.45·S + 0.25·Phi + 0.15·tau + 0.15·A</p> <p>with weights fixed a priori and explicitly justified as<br>provisional. A pre-registered comparative experiment (N=60 pairs<br>CON self-model vs N=60 SIN) finds the primary hypothesis not<br>confirmed (CON=0.352 vs SIN=0.359, p=0.747).</p> <p>The principal finding — the Amplification Hypothesis — is that<br>the self-model does not raise mean cooperation but substantially<br>amplifies the system's capacity for extreme states: 5 CON pairs<br>exceed 0.60 cooperation vs. 0 SIN pairs (post-hoc descriptive);<br>the CON maximum (0.831) far exceeds the SIN maximum (0.482).</p> <p>Retrospective analysis of the 5 successful pairs reveals that 4<br>of 5 exhibit early M3 synchrony (self-model stabilizes<br>pre-existing alignment); one pair shows genuinely emergent<br>synchrony (seed 1778480933, r_late=0.581 > r_early=0.518,<br>|M3_A-M3_B| decreasing 0.1015→0.0756).</p> <p>Individual simulation analysis reveals three dynamical regimes:<br>when dynamic synchrony is present, C correlates strongly with<br>cooperation (r up to 0.759); when absent, C provides no<br>predictive power. The C index is robust (mean≈0.55, SD≈0.04)<br>but insensitive to the self-model as a specific intervention,<br>motivating empirical weight optimization in Phase 5.</p> <p>Phase 5 is specified as: (1) Learning-Action coupling L via<br>explicit predictor submodule P generating M3_pred(t+1)=W_P·S(t);<br>(2) empirical weight optimization via gradient descent on<br>r(C, mutual_cooperation); (3) formal mapping of PROSENTIR<br>signatures to the 3D-MIND hybrid biological-electronic system<br>(Princeton, 2026, Nature Nanotechnology).</p> <p>Reproducible code: prosentir_v13.py (included).</p>