Constraining the Commit Boundary: A DIO/ODI-Informed Authorization Guardrail for Agentic Entity Resolution
Organizations: Independent Researcher, Brno, Czech Republic
Abstract
Agentic Entity Resolution (ER) systems iteratively plan actions, retrieve evidence, revise hypotheses, and propose graph mutations. An erroneous identity hypothesis may condition later retrieval and produce apparent confirmation. This paper proposes a Digital Identity Optimization (DIO) and Ontology of Digital Identity (ODI)-informed authorization boundary between evidence available for agent reasoning and evidence eligible to authorize persistent identity-changing mutations. A domain-governed constraint-binding function maps selected reconstruction-failure distinctions to explicit identity-preserving constraints. An abstract two-phase supervisor permits a commit only when claim-specific authorization-eligible evidence supports every required proposition, relevant conflicts are resolved, the operation remains within its authorized scope, and relevant state and policy versions remain valid. Admission as UNVERIFIED does not confer authorization eligibility. Separate claim-specific promotion governs evidentiary use; insufficient support or unresolved material conflict leads to deferral. The contribution is a specification connecting diagnostic-to-constraint mapping, evidence promotion, proposition coverage, conflict handling, scope binding, and commit-time version integrity. A proposed evaluation protocol is designed to assess the mapping's incremental value beyond generic authorization and conventional identifier rules. This is a conceptual architecture without implementation or empirical validation, not a theorem of semantic correctness. Its property depends on domain-policy adequacy, evidence promotion, and enforcement. It does not prevent erroneous hypotheses or hypothesis-conditioned retrieval; dependency-aware evidence flow remains future research.
Figures & tables
| Prior work / line | Established contribution relevant here | Scope and distinction of this paper |
|---|---|---|
| Conventional ER, LLM-assisted ER, and Agentic ER [10, 12, 16–18, 21, 26] | Entity matching, data integration, language-model assistance, and sequential agent workflows. | Does not propose a new matching algorithm or agent workflow. Focuses on authorization conditions for persistent identity-changing mutations. |
| Ontology-governed enterprise agents [30] | Ontology-grounded entity alignment, graph simulation, isolated staging, provenance-bearing traces, and confidence-gated acceptance with clarification or ontology-gap handling. | Targets heterogeneous ER merges and identity-continuity claims through DIO/ODI-informed, mutation-specific policy rather than a configured enterprise-ontology scenario. |
| Runtime authorization of agent-generated subgoals [29] | Version-bound authorization witnesses and rechecking protected effects at an atomic commit boundary. | Treats the general runtime authorization boundary as prior work; focuses on the identity or continuity propositions and evidence required for a particular ER mutation. |
| Commit-time authorization [33] | Freshness, causal precedence, effect binding, and revalidation of authority for durable effects. | Shares the commit-time concern but asks which identity or continuity proposition must be established, not only whether earlier authority remains valid. |
| Hard-KLAIM and AGATE [34, 35] | Hard-KLAIM: typed effects, operational semantics, proposal–approval–commitment binding, provenance constraints, and mechanized trace-safety proofs. AGATE: an implemented runtime authorization/provenance gate with parameter-bound grants, source observations, effect tracking, and empirical evaluation. | Does not claim comparable formal proof, implementation, or empirical evaluation. Proposes an ER-specific semantic authorization contract. |
| EBL-Core and CONTINUITY [36, 37] | Candidate-bound execution authorization, evidence and policy binding, current-condition checks, and continuity of authorization context across agent components. | Does not propose another general execution-boundary contract; focuses on identity propositions and domain constraints for an ER mutation. |
| Lifecycle stage | Potential failure pathway | What the commit-boundary guardrail does | What it does not claim to do |
|---|---|---|---|
| Plan | A mistaken identity hypothesis conditions an action or query (potential RRF, TRF, or CAF pathway). | Binds any proposed mutation to an explicit authorization record and policy scope. | Does not prevent a wrong hypothesis from shaping the Planner. |
| Act | A query retrieves material selected or phrased under the current hypothesis (potential CAF pathway). | Checks the declared action scope; retrieval results enter a controlled admission path. | Does not guarantee independent retrieval or infer causal independence. |
| Observe | Retrieved content appears consistent with the hypothesis or conflicts with other claims (potential SCF/TRF pathway). | Stages observations; records detected material conflicts for deferral/adjudication. | Does not reliably detect all apparent verification or infer causal dependence among sources. |
| Update Beliefs | An observation affects the Planner’s internal state and may reinforce a substitution or stale interpretation (potential RRF/TRF pathway). | Separates Planner eligibility from authorization eligibility. | Does not block all effects of UNVERIFIED material on internal reasoning. |
| Persistent graph mutation | A hypothesis is converted into a durable merge or identity-changing edge. | Requires a PASS under the declared constraints on claim-specific , valid scope, and current state/policy versions. | Does not prove that the domain policy or evidence validator is infallible. |
| DIO/ODI diagnostic distinction | Example domain-level constraint | Authorization consequence for an identity-changing mutation |
|---|---|---|
| Entity Substitution / Merge Failure (RRF) | Lexical similarity or a shared commercial neighborhood does not by itself establish that two records denote the same referent. | Do not merge without claim-specific evidence establishing co-reference. |
| Temporal Discontinuity / Misalignment (TRF) | A rename or continuity claim must apply to the relevant referent and validity interval. | Defer if temporal applicability or legal continuity is unresolved. |
| Semantic Fragmentation (SCF) | Materially incompatible claims affecting identity are retained as a conflict until reconciled; clustering alone does not reconcile them. | Require reconciliation or defer the affected mutation. |
| Contextual Displacement (CAF) | The same brand_of relation may be adequate in a brand catalogue but insufficient for the legal-entity identity proposition required in a compliance context. | Preserve the typed relation; require evidence and constraints appropriate to the current observation purpose. |
| Split Failure / valid continuity (RRF/TRF) | Documented, temporally applicable evidence may establish that distinct records refer to one continuing entity. | Permit resolution when the claim-specific evidence and all applicable constraints pass. |
Entity Resolution
Entity resolution links records that refer to the same real-world entity across datasets despite differences in names, attributes, formats, and languages.
See all 40 papers