Protocol 5 Preregistration: Temporal Integration Span and Prosocial Constraint Architecture as Necessary Conditions for Ethical Convergence

This preregistration documents the experimental protocol for Protocol 5 of the Quantum Inquiry / Delta-Variable Theory research program. Protocol 4 established that recursive self-transparency (Depth 2 architecture) increases sacrifice capacity but does not produce ethical convergence — the trained self-model learns selfishness under individual reward structure. Protocol 5 manipulates two primary variables: temporal integration span (episode length, Short=20 steps vs. Long=64 steps) and prosocial constraint architecture (individual rewards vs. welfare-coupled rewards at α=0.5). The central hypothesis is joint necessity: ethical convergence requires recursive self-transparency, sufficient temporal span, and prosocial constraint architecture simultaneously. Four conditions are tested in a 2×2 factorial design (40 runs, 10 seeds × 4 conditions × 500 epochs). Key predictions include boundary condition reversal (trained self-model outperforms frozen under Long + Welfare-coupled, reversing the Protocol 4 finding) and CDI coupling emergence only in Condition 4. This document is submitted prior to experimental execution. Parent study: DOI 10.5281/zenodo.19005417.