What question did this study set out to answer?

The aim is to establish a valid execution framework for autonomous systems operating under partial observability, highlighting limitations of current methods.

April 22, 2026Open Access

Reconstructive Authority Model: Runtime Execution Validity Under Partial Observability

Key Points

The aim is to establish a valid execution framework for autonomous systems operating under partial observability, highlighting limitations of current methods.
Introduced the Reconstructive Authority Model (RAM) to separate integrity from coverage.
Proved necessity via two theorems and three corollaries.
Evaluated the performance of RAM against attestation-based systems using a large experimental setup.
RAM achieves integrity error rate (IER) = 0 at all coverage levels.
Attestation-based systems demonstrate IER = 0.423 at low coverage and IER = 0.233 at full coverage.
RAM effectively eliminates failures that existing attestation mechanisms cannot address.

Abstract

Autonomous systems increasingly operate under partial observability where the true execution-relevant state is never fully accessible. Existing governance mechanisms — trusted execution environments, oracle-signed state proofs, cryptographic attestation — enforce the integrity of computation and state projections. This paper shows that this is structurally insufficient: an authenticated projection of state is necessary, never sufficient. We introduce the Reconstructive Authority Model (RAM), which separates integrity from coverage. The RAM reconstruction gate reasons over an explicit coverage envelope — comprising the proven state, declared assumptions, and the acknowledged unobservable residual — and permits execution only when coverage is adequate for the specific action class. When coverage is partial, the gate narrows privileges dynamically; when coverage is insufficient, it fails closed. Attestation proves trust in measurement, not completeness of execution-relevant reality. We formalize RAM, prove its necessity via two theorems and three corollaries, and present a hybrid RAM + Attestation architecture with privilege-narrowing. Theorem 1 (Attestation Insufficiency) establishes that no attestation-based system can guarantee execution validity with respect to the real state unless authority is reconstructible at runtime. Theorem 2 (Necessity of RAM) establishes that reconstruction from the real state is a necessary condition for execution validity guarantees. Appendix A provides the complete constructive proof of Theorem 1 via Lemma A.1 (gap existence) and Lemma A.2 (execution-critical gap). Experimental evaluation (N = 100,000 steps, seed = 42) confirms: RAM achieves IER = 0 at all coverage levels, while attestation-based systems exhibit IER = 0.423 at low coverage and IER = 0.233 even at full coverage — a semantic failure arising from treating undefined state as benign. The class of failures eliminated by RAM cannot be addressed by stronger attestation or more comprehensive oracles alone. This paper is Paper 5 (operational closure) of the Agent Governance Series: - Paper 0 — Atomic Decision Boundaries. DOI: 10.5281/zenodo.19642166 - Paper 1 — Agent Control Protocol (ACP) v1.30. DOI: 10.5281/zenodo.19642405 - Paper 2 — Invariant Measurement Layer (IML). DOI: 10.5281/zenodo.19643761 - Paper 3 — Fair Atomic Governance. DOI: 10.5281/zenodo.19643928 - Paper 4 — Irreducible Multi-Scale Governance. DOI: 10.5281/zenodo.19643950 RAM operationalizes the observability impossibility of Paper 2 (IML) and completes the execution semantics of the four-layer architecture of Paper 4. Source code, LaTeX source, and experimental figures are available at: https://github.com/chelof100/reconstructive-authority-model

Reconstructive Authority Model: Runtime Execution Validity Under Partial Observability

Key Points

Abstract

Cite This Study