Atomic Decision Boundaries: A Structural Requirement for Guaranteeing Execution-Time Admissibility in Autonomous Systems
Fuente:
Zenodo
Enregistré dans:
| Auteur principal: | |
|---|---|
| Format: | Recurso digital |
| Langue: | anglais |
| Publié: |
Zenodo
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866902017404829696 |
|---|---|
| author | Fernández, Marcelo |
| author_facet | Fernández, Marcelo |
| contents | <div>This paper introduces the atomic decision boundary as a structural requirement</div> <div>for admission control systems that must enforce admissibility at execution time.</div> <div> </div> <div>Existing governance mechanisms — including RBAC, policy engines such as Open</div> <div>Policy Agent (OPA), and pre-execution policy evaluation — share a common</div> <div>architecture: they evaluate policies separately from the state transition they</div> <div>authorize. This separation creates a structural gap that no policy enrichment,</div> <div>external state, or re-evaluation can close.</div> <div> </div> <div>We formalize agent execution as a labeled transition system (LTS) and define</div> <div>two classes of admission systems: atomic systems, where evaluation and state</div> <div>transition occur as a single indivisible LTS step, and split evaluation</div> <div>systems, where they are separate transitions that may be interleaved by</div> <div>environmental actions. Under realistic concurrent environment assumptions,</div> <div>we prove that no construction can make a split system equivalent to an atomic</div> <div>system with respect to admissibility across all execution traces. This is a</div> <div>structural impossibility — independent of policy expressiveness, state</div> <div>availability, or implementation sophistication.</div> <div> </div> <div>We extend the analysis to the Escalate outcome, absent from classical</div> <div>time-of-check/time-of-use (TOCTOU) analyses, and prove that escalation</div> <div>transfers rather than removes the atomicity obligation: any system that</div> <div>resolves an escalated request must itself be atomic at the point of resolution.</div> <div> </div> <div>Key results:</div> <div>- Theorem (Structural Impossibility): No split evaluation system can guarantee</div> <div>execution-time admissibility under all traces in a concurrent environment.</div> <div>- Corollary (Enrichment Futility): Adding policies, external state, or</div> <div>re-evaluation steps to a split architecture cannot close the structural gap.</div> <div>- Theorem (Escalation Atomicity): The resolution of an Escalate outcome is</div> <div>itself subject to the atomic boundary requirement.</div> <div>- Mapping: RBAC and OPA are classified as split systems; the Agent Control</div> <div>Protocol (ACP, Paper 1) is shown to instantiate the atomic boundary via its</div> <div>evaluate-then-mutate pipeline with serializable ledger and atomic commit.</div> <div> </div> <div>This paper is the formal foundation of a four-paper series on agent governance.</div> <div>Paper 1 (ACP) instantiates the atomic boundary as a concrete protocol.</div> <div>Paper 2 (IML) detects behavioral drift above the enforcement layer.</div> <div>Paper 3 (Fair Atomic Governance) addresses fair allocation of access to the</div> <div>boundary. Paper 4 proves the four-layer architecture is irreducible under</div> <div>finite observability.</div> <div> </div> <div>Series repository: https://github.com/chelof100/decision-boundary-model</div> <div>ACP (Paper 1, published): https://arxiv.org/abs/2603.18829</div> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_19642166 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Atomic Decision Boundaries: A Structural Requirement for Guaranteeing Execution-Time Admissibility in Autonomous Systems Fernández, Marcelo autonomous agents, admission control, labeled transition systems, atomicity, TOCTOU, runtime enforcement, agent governance, split evaluation <div>This paper introduces the atomic decision boundary as a structural requirement</div> <div>for admission control systems that must enforce admissibility at execution time.</div> <div> </div> <div>Existing governance mechanisms — including RBAC, policy engines such as Open</div> <div>Policy Agent (OPA), and pre-execution policy evaluation — share a common</div> <div>architecture: they evaluate policies separately from the state transition they</div> <div>authorize. This separation creates a structural gap that no policy enrichment,</div> <div>external state, or re-evaluation can close.</div> <div> </div> <div>We formalize agent execution as a labeled transition system (LTS) and define</div> <div>two classes of admission systems: atomic systems, where evaluation and state</div> <div>transition occur as a single indivisible LTS step, and split evaluation</div> <div>systems, where they are separate transitions that may be interleaved by</div> <div>environmental actions. Under realistic concurrent environment assumptions,</div> <div>we prove that no construction can make a split system equivalent to an atomic</div> <div>system with respect to admissibility across all execution traces. This is a</div> <div>structural impossibility — independent of policy expressiveness, state</div> <div>availability, or implementation sophistication.</div> <div> </div> <div>We extend the analysis to the Escalate outcome, absent from classical</div> <div>time-of-check/time-of-use (TOCTOU) analyses, and prove that escalation</div> <div>transfers rather than removes the atomicity obligation: any system that</div> <div>resolves an escalated request must itself be atomic at the point of resolution.</div> <div> </div> <div>Key results:</div> <div>- Theorem (Structural Impossibility): No split evaluation system can guarantee</div> <div>execution-time admissibility under all traces in a concurrent environment.</div> <div>- Corollary (Enrichment Futility): Adding policies, external state, or</div> <div>re-evaluation steps to a split architecture cannot close the structural gap.</div> <div>- Theorem (Escalation Atomicity): The resolution of an Escalate outcome is</div> <div>itself subject to the atomic boundary requirement.</div> <div>- Mapping: RBAC and OPA are classified as split systems; the Agent Control</div> <div>Protocol (ACP, Paper 1) is shown to instantiate the atomic boundary via its</div> <div>evaluate-then-mutate pipeline with serializable ledger and atomic commit.</div> <div> </div> <div>This paper is the formal foundation of a four-paper series on agent governance.</div> <div>Paper 1 (ACP) instantiates the atomic boundary as a concrete protocol.</div> <div>Paper 2 (IML) detects behavioral drift above the enforcement layer.</div> <div>Paper 3 (Fair Atomic Governance) addresses fair allocation of access to the</div> <div>boundary. Paper 4 proves the four-layer architecture is irreducible under</div> <div>finite observability.</div> <div> </div> <div>Series repository: https://github.com/chelof100/decision-boundary-model</div> <div>ACP (Paper 1, published): https://arxiv.org/abs/2603.18829</div> |
| title | Atomic Decision Boundaries: A Structural Requirement for Guaranteeing Execution-Time Admissibility in Autonomous Systems |
| topic | autonomous agents, admission control, labeled transition systems, atomicity, TOCTOU, runtime enforcement, agent governance, split evaluation |
| url | https://doi.org/10.5281/zenodo.19642166 |