From Synthetic Populations to Pre-Legislative Evidence: A Gated Sim-to-Real Protocol for Regulatory Stress Testing in Argentina

Synthetic populations can expose implementation risks before a proposed rule affects real people, but scale and persona fidelity do not by themselves establish evidence about human behavior. This paper presents an open, gated sim-to-real protocol for pre-legislative regulatory stress testing in Argentina. The protocol freezes the normative text, legislative stage, synthetic cohort, factorial arms, hypotheses, metrics, and acceptance thresholds before independent human observations are collected. A transparent reference implementation tests a fictional national digital-registration duty for micro, small, and medium enterprises across eight combinations of regulatory complexity, enforcement consistency, and procedural safeguards. Four hundred and eighty synthetic personas generate 3,840 deterministic responses. The system reproduces the four encoded directional hypotheses and passes all technical checks, while correctly blocking sim-to-real and legal-policy reliance because no human observations have yet been collected. The contribution is therefore methodological rather than predictive: a reproducible evidentiary architecture that turns uncertainty into workflow state, preserves each legislative version as an immutable wave, and treats synthetic populations as instruments for hypothesis generation and consultation design rather than substitutes for citizens, affected firms, workers, legal authority, or democratic deliberation. Evidence status. The deterministic reference run is a technical demonstration. Independent human observations have not yet been collected. Sim-to-real and legal-policy reliance therefore remain blocked. Open code, preregistration, synthetic artifacts, human instrument, and verification tests are available in the linked GitHub repository.