Publications

Compliance dashboards stay green while mandate alignment silently erodes. DASR is a 30-day structured audit protocol that measures the divergence between an AI system's authorized mandate and its current operational state — before the gap becomes a governance incident. Delivers Drift Magnitude, Drift Velocity, and Cumulative Exposure as audit-grade metrics.

Existing frameworks verify that decisions were made correctly. DCF addresses a different question: is this decision still legitimately authorized to execute right now? Defines three closure conditions — Authority Closure, Assumption Closure, Evidence Closure — that must hold continuously, not just at approval time. Includes Authority Rebinding and Failure Transition Control.

AI governance assessments lack a unified framework for measuring both capability and compliance as system conditions change. GCCL provides a structured evaluation framework with drift detection, autonomy levels, and witnessable certification. Submitted to the EU AI Office Expert Forum as a candidate governance standard (Contribution ID: 510c3274).

Most reliability assessments require access to model internals. Output-Only Diagnostics provides a black-box framework for detecting multi-turn instability and structural unreliability in language models from observable outputs alone — applicable in production environments where internal state is not accessible.

AI systems can remain statistically stable while their internal meaning structures shift in ways that affect decision identity. This paper establishes SSE as the scientific field that makes internal semantic states, drift trajectories, and meaning evolution formally observable. The canonical field definition that underlies all Decision Integrity analysis.

A system can pass all certification checks and maintain behavioral correctness while its decision identity has already drifted. This paper provides the formal separation showing why certification alone is insufficient for governance — and why decision identity requires its own invariants and witnessability conditions.

Adaptive systems undergo internal meaning-state changes that are invisible to standard monitoring. This paper provides the dynamical framework for analyzing how meaning states evolve, when they shift regimes, and how long internal structure remains interpretable — the operational foundation for drift detection in Decision Integrity applications.

Foundational claim and timestamped category anchor for Semantic Stability Engineering. Establishes SSE as independent from statistical concept drift, alignment research, and interpretability tooling — and defines the conditions under which meaning-state analysis becomes necessary for governance.

Research artifact and implementation package for reflexive audit architecture and semantic drift stabilization. Demonstrates that Decision Integrity principles are operationally implementable — not only theoretically specified.