TitlePhase III Protocol: Confounder-Ruling-Out Analysis for the Phase II Dependence Finding

This document is a pre-registered research protocol specifying, in advance of execution, the statistical tests, data sources, and falsification conditions of Phase III of the Systems of Designed Intent empirical validation program. No analytical results are reported herein. The purpose of pre-registration is to bind the analysis to specifications made before the relevant data has been examined under the relevant tests, so that subsequent execution produces a result that cannot be reverse-engineered to confirm the framework's prior conclusions. Phase II (2026) applied maximum mean discrepancy and the Hilbert-Schmidt Independence Criterion to 1,035 unique domain pairs drawn from 46 federal agencies across 1970–2024, at n = 55 annual observations per domain. The result was universal pairwise dependence at HSIC 100 percent significance, with bandwidth robustness across 11 sigma values and stress-test confirmation across the maximum-structural-independence cross-branch pairs. The Phase II result empirically falsified the independence assumption that had been grounded in the official U.S. Government Organizational Chart. A methodological concern arises on the opposite margin. Statistical dependence is a necessary but not sufficient condition for the architectural capture hypothesis. Dependence is also consistent with shared exogenous drivers — macroeconomic cycles, political cycles, war and recession shocks, shared regulatory framework changes, and shared measurement regimes — that would produce co-movement across federal agencies without requiring a designed control architecture. The Phase II analysis does not formally exclude these confounders. Phase III closes this gap. The protocol specifies three parallel analytical paths, each drawing exclusively from government-sourced or government-collected data, and each producing an independently falsifiable test of the capture hypothesis against the shared-confounder hypothesis. Path One applies Partial HSIC conditional independence testing across three sequential conditioning rounds (macroeconomic core, political and fiscal, full confounder vector). Path Two uses major U.S. institutional reorganizations as natural experiments via Bai-Perron structural breaks and Callaway-Sant'Anna staggered difference-in-differences. Path Three constructs private-sector and foreign-government placebo comparators using exclusively government-collected data. Falsification thresholds are stated in advance: strong falsification of the capture hypothesis at less than 25 percent Partial HSIC survival under the full conditioning set; partial falsification at 25 to 50 percent; confirmation at 75 percent or higher. Reform-event predictions, placebo-domain expectations, and federal-versus-private dependence comparisons are similarly pre-specified. The framework commits, in advance, to incorporating Phase III execution results into its revision record. Document classification: Pre-Registration Protocol, Execution Pending. HPRS Version 1.0 governance. Phase III of the Systems of Designed Intent Research Series — companion to Doc 15 (Phase II Results) and the natural successor in the framework's empirical-validation program. Independent Phase III execution by qualified researchers is welcomed. — — — Note on document position: This is the pre-registration of Phase III. The Phase III Results document will be deposited as a separate Zenodo record upon execution, following the same HPRS compliance structure. Independent execution of any or all three paths is welcomed. Code and data extracts will be published alongside the Results document for replication. Related to Phase II Results (Doc 15 — HSIC empirical falsification of the independence assumption); Open Challenge to Peer Review — Phase I (Doc 1 — compounded probability framework); Game-Theoretic Confirmation (Doc 14 — mechanism-design analysis). All Phase II and Phase III documents operate under the Hybrid Prosecutorial Research Standard (HPRS) Version 1.0.