A derivation written as a sequence does not by itself establish that every downstream assertion class remains authorized at the end of it.
Research Poster · LiO v0.1
LiO v0.1: Deterministic Permission Composition Across Indispensable Inferential Bridges
Micheal Charles Preble · Independent Researcher · SSRN 7424499
Problem · Background · Research Question
Builds on truth-maintenance systems, abstract interpretation, three-valued logic, and access-control policy combination (XACML) without claiming to invent any of them.
Given upstream-adjudicated bridge permissions, what authorization survives composition, deterministically and non-compensatorily?
Framework: three separated problems
What assertion classes and semantic distinctions belong in the universe. Upstream of LiO; not addressed by this paper.
What each indispensable bridge permits, excludes, or leaves unresolved. Upstream of LiO; not addressed by this paper.
Given those supplied inputs, what authorization deterministically survives every indispensable bridge. This is what LiO v0.1 addresses.
Central diagram: the formal composition rule
Auth(a | D) = ∧b∈ID vb(a) — one resolved exclusion is sufficient for non-authorization; all-permit yields authorization; any remaining unresolved case yields UNRESOLVED.
Propositions / Observable Implications
- Non-expansion: composed permission can never exceed the intersection of every indispensable bridge's permission.
- Unresolvedness preserved: an unresolved bridge is never silently converted to approval or rejection.
- Weakest link: support elsewhere in the derivation cannot compensate for one bridge's explicit restriction.
Evidence Required
E12 Review Cycle 1: reviewer model Mistral Medium 3.5 blind-reimplemented LiO from the frozen spec, reproduced all 7 public cases exactly, then matched all 12 creator-frozen held-out cases (H001–H012) exactly on the two implemented core fields. A raw-output hash discrepancy was disclosed rather than hidden; it did not affect the semantic match.
Limitations
| Category | Claim | Status |
|---|---|---|
| C-CORE-001 | Cross-model reproducibility on the Cycle 1 corpus | Established by E12 Cycle 1: 7/7 public, 12/12 held-out exact |
| C-CORE-002 | Resolved and partially unresolved composition matched creator-frozen expectations | Established, no semantic disagreement observed |
| C-CORE-003 | Explicit unresolvedness preserved in tested cases | Established, within the tested corpus only |
| C-UPSTREAM-001–004 | Assertion-space adequacy; reproducible non-circular adjudication; local separability; dependency completeness | Required but not validated by this paper |
Falsifiers / Open Questions
RQ-001: which semantic distinctions must be represented to prevent assertion-class laundering? RQ-002: can evidence and semantic transformation be mapped reproducibly to bridge-level permissions? RQ-003: when does bridge-wise powerset permission stop being sufficient?
Citation / QR
Suggested citation
Preble, Micheal Charles. “LiO v0.1: Deterministic Permission Composition Across Indispensable Inferential Bridges” Available at SSRN 7424499, 2026. CC BY 4.0.
Preble, Micheal Charles. “LiO v0.1: Deterministic Permission Composition Across Indispensable Inferential Bridges” Available at SSRN 7424499, 2026. CC BY 4.0.