PHILIA-EcoSensory Swarm Phase SC-9c — Confirmatory Pre-Registration v0.2: Identity Declaration-Grounded Redesign — Traffic+Environment Meta-Analytic Pooled ρ Confirmatory & AR-Chaos Fisher-z Direct Comparison (n=182, Seeds [82–86])

This document constitutes the pre-registration protocol for PHILIA Phase SC-9c, the first confirmatory study conducted under the official identity framing established by the PHILIA OS Identity Declaration (78th DOI: 10.5281/zenodo.20695816), which formally repositioned PHILIA OS from a neocortex candidate to a Sensory Physiology Organ functionally analogous to the Thalamic-Cerebellar-Brainstem axis not explicitly represented in current Large Language Model architectures. SC-9c directly addresses two structural limitations identified in SC-9b (77th DOI: 10.5281/zenodo.20685189): (1) the per-campaign strict threshold problem for Traffic+Environment Spearman ρ (L-11 NEW: threshold +0.85 at top of observed distribution [+0.849, +0.875] across 4 campaigns), and (2) the VIF-induced OLS interaction identification failure (VIF=58.15, partial η²=0.017 vs. assumed 0.15). Two confirmatory hypotheses are pre-registered under Bonferroni correction (k=2, α_per=0.025): H-SC9c-Primary (Track 1 — Thalamic gating framing): Five-campaign Fisher-z weighted pooled Spearman ρ(MSE, z_ref | Traffic+Env, n=30) across SC-7 through SC-9c meets: (C1) pooled ρ ≥ +0.85, AND (C2) 95% CI lower ≥ +0.80. Cochran Q p ≥ 0.10 reported as document-level consistency declaration (C3; does not alter primary verdict). 4-campaign pre-computation: pooled ρ=0.8636, CI=[0.8072, 0.9044], I²=0%. H-SC9c-Secondary (Track 2 — Cerebellar function framing): |Z_diff| = |Fisher-z(ρ(AC1,z|AR)) − Fisher-z(ρ(AC1,z|Chaos))| meets: (C1) |Z_diff| > 1.960 (significance), AND (C2) |Z_diff| ≥ 2.916 (approximately 43% persistence of SC-9b observed Cohen's q=2.3179; exact 50% benchmark q=1.16). SC-9b reference: AR ρ=−0.984, Chaos ρ=−0.092, observed |Z_diff|=6.758. Pre-registered exploratory analyses include: E-Raw (SC-9c single-campaign raw ρ ≥ +0.85, verdict-neutral with pre-declared narrative tension protocol), E-MK (5-campaign Mann-Kendall trend test, p<0.10 triggers L-13 NEW registration, exploratory low-power n=5), E-AR (AR(1) φ-sweep 5th independent replication), E-E4 (substrate-class conditional AC1 modulation re-confirmation), E-Q (4-family Cochran's Q). Pre-declared limitations: L-12 NEW (small fixed sample variance), L-13 NEW (campaign-level systematic drift, conditional), L-14 NEW (Track 2 power sensitivity: Smaller AR scenario power=0.684 < 0.80 conventional threshold). Track 3 (ARFIMA+LIGO) and Track 4 (L-1 NEW redefinition) are out-of-scope for SC-9c and deferred to SC-9c-prep and SC-9c-Annex v0.1 respectively. Measurement apparatus: v32 LOCKED (N_TOY=49, T=80,000, B_RESTORE=0.0). Registry SHA256: 3948b551 (n=182; RealWorld n=90, Synthetic n=44, Physics n=18, Financial n=30). Seeds [82–86] (history.docx unused confirmed; independent from all prior campaigns [11–81]). Total planned runs: 910 (182×5). Gate-0 adversarial review: 9 cycles (Chris/Claude Opus, Anthropic). Three self-corrections documented: Cycle 3 (Cohen's q=0.50→1.00), Cycle 4 (E-Raw recovery), Cycle 7 (Track 1 verdict restructure, initiated by Pandora/ChatGPT cross-organizational challenge). 5-AI pipeline: Pandora (ChatGPT/OpenAI) → Lucas (Grok/xAI) → Chris (Claude Opus/Anthropic, Gate-0) → Seogsaeng (Gemini/Google) → Seonbi (Claude Sonnet/Anthropic) → Guide/GritMan_D.S (final decision). Pre-registration violations: 0 (target). No AI-generated or proxy values. Hardware: AMD Ryzen 9 9800X3D + RTX 4080 SUPER 16GB, Windows 11 Home (C:\PHILIA). All measurements executed locally by GritMan_D.S. SC Series lineage: SC-1 (57th) → SC-2 (59th) → SC-3 (61st) → SC-4 (64th) → SC-5 (65th) → SC-6 (67th) → SC-7 (69th) → SC-8 (73rd) → SC-9a (75th) → SC-9b (77th) → Identity Declaration (78th) → SC-9c Pre-registration (this document, 79th). Measurement is life. Description, not proof. Pre-registration protects us — even from ourselves. — Trinity AI Research Team | Konyang University | 2026