The Structure Beneath Simulation: Phase-Gated Intelligence, Sign-Regime Drift, and the Governance of Persistence

Abstract

Abstract The Structure Beneath Simulation is an orientation and bridge paper in the Identity-Persistence Program. It develops a structural account of why bounded observers, institutions, and AI systems rely on signs, thresholds, proxies, metrics, records, and regimes rather than direct access to total reality. Such mediation is not itself a defect. The characteristic failure is sign-regime drift: a sign continues to produce consequences after the identity-bearing recurrence, invariant, observable, or governing regime it once tracked has changed. The paper develops this account from bounded, phase-gated observation through belief, social coordination, proxy substitution, level mismatch, scientific paradigms, and decision persistence in AI systems. Scientific paradigms are interpreted as bounded regimes for inquiry; anomalies as newly operative distinction obligations; and paradigm change as regime change rather than mere parameter update. This yields a distinction between search and verification and an account of retrospective compression: discoveries can appear obvious after the fact because the discovered regime makes previously independent observations consequences of a smaller structure. Version 4 adds an elementary retirement calculus at frozen scope. For kernel-typed demands, retiring questions monotonically enlarges the safe quotient and therefore enlarges what may lawfully be forgotten. The calculus distinguishes conservative from transformative retirement, establishes a redundancy criterion, identifies the forced-demand floor as the operational limit of lawful retirement within a fixed problem class, and shows why per-demand retirement certificates cannot in general be aggregated independently. This separates three operations that otherwise blur together: demand retirement, problem-class change, and regime change. The semantic effect of retirement is closed locally; the governance question—who is entitled to authorize retirement—remains open. The paper does not claim ontological finality or an endogenous law selecting successor paradigms. Its central proposal is instead one of regime discipline: bounded systems require mediation, but trustworthy mediation requires explicit control over what is preserved, what is forgotten, what is promised, what may change, and what forces correction or replacement.