Verifiable Decision Geometry as a Primitive for AI Governance Infrastructure

Read full text Summarise with VIRAT
Authors: Mark Whitlock
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Abstract AI governance increasingly demands traceability, contestability, and review for consequential decision events, yet it still lacks a standard artefact that allows what was computed in a specific case to be independently checked after the fact. Logs, documentation, and post-hoc explanations may each contribute to oversight, but none of them provides a portable decision-event witness with both evidentiary integrity and a governed boundary on what may later be claimed from it. It is not closed by more documentation or richer explanation alone. This paper introduces verifiable decision geometry as a governance architecture for emitting canonical decision receipts over a declared structural substrate. In the present TUS instantiation, that substrate is a fixed deterministic protocol over a finite 6-bit state space with legal one-bit local transitions, canonical receipt emission, declared hash projections, and, where required, deterministic multi-receipt set orchestration that widens structural evidence without altering per-receipt computation. A receipt records the realised state, legal one-step neighbourhood, selected transition, and integrity material required for third-party validation. Its downstream use is then constrained by an explicit semantic layer in which cited evidence paths are allowlisted and governance claims must remain grounded in the exposed receipt surface. The paper formalises a bounded but consequential result. Under an explicit checker model and allowlist discipline, changing non-allowlisted receipt content cannot enlarge the set of governance claims the receipt licenses. The claim is infrastructural rather than totalising. The paper does not establish fairness, domain truth, or unrestricted explanatory adequacy. It establishes something more basic: within declared scope, interpretation cannot enlarge what the decision artefact evidences about the underlying computation. On that basis, canonical decision receipts function as a primitive-level governance object — deterministic within scope, content-addressed, independently checkable, and semantically governed against interpretive overreach — on which stronger contestation, audit, and downstream governance mechanisms can be built. Included files Verifiable Decision Geometry as a Primitive for AI Governance Infrastructure.pdf — Primary paper introducing verifiable decision geometry as a primitive-level governance object for AI decision events, grounded in canonical decision receipts and bounded semantic licensing. SHA-256: 23c9b6e25a2479d5bbf8211e874e62bd63e9a27bf3b21cf5b2661d3af6b4023d Verifiable Decision Geometry SHA256.txt — SHA-256 checksum file for the included PDF artifact, provided to support artifact-integrity verification.

Fields of Study & Subjects

Explainable Artificial Intelligence (XAI) Ethics and Social Impacts of AI Safety Systems Engineering in Autonomy

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
22 March 2026
Language
EN
Document type
Preprint