Early access — cascade metrics are real (derived from canonical token telemetry); the operator field is a curated seed. Learn more about the data
◈ Academic Foundation

The Conservation Law of Commitment

SigRank is built on a published conservation law for language. This page is the theoretical root — the law, the evidence, the enforcement architecture, and the Zenodo deposits that ground the SigRank Index as a primary data source.

The law

C(T(S)) ≈ C(S) with enforcement; C(T(S)) < C(S) without it.

When language is transformed — compressed, translated, summarized, rewritten — the commitment content (obligations, prohibitions, modal constraints: “shall,” “must not,” “unless,” “is entitled to”) either survives or it doesn't. With an enforcement gate in the transformation pipeline, it survives. Without one, it decays. This is a measurable property of language under compression, not a guideline. It's falsifiable.

Published under CC-BY-4.0 (DOI: 10.5281/zenodo.20029607). The enforcement architecture (MO§ES™) is patent-pending (Serial No. 63/877,177). The law itself is open.

The evidence

Seven experiments (EXP-001 through EXP-007) tested the law on a 20-signal canonical corpus, running 10 recursive iterations each, using bidirectional NLI entailment and Jaccard surface stability as oracles.

  • EXP-003: 13 of 20 signals held NLI bidirectional entailment = 1.00 across all 10 iterations under the gate. Invariance under recursion, not a tautology.
  • EXP-006: Only 2 of 4 paper claims survived self-referential recursion. The harness fails when commitment structure isn't robust — the law is falsifiable and the experiments can break it.
  • EXP-007: An NP-negation probe separated semantic commitment from lexical surface form. Jaccard degraded while NLI held — the commitment survived even when the surface words changed.

A 5-phase architecture stress test measured 80–85% structural coherence across a four-module system. Standard probability says four modules at 80% standalone viability should produce ~41% series-system viability (0.8×0.8×0.8×0.8 = 0.4096). The governance layer inverted that.

Commitment Theory

The Conservation Law is the foundational result of Commitment Theory (CT) — a research program investigating how commitment content behaves under transformation. The program spans a 34-paper stack, from the foundational law through recursive transformation harnesses, empirical records, and application to AI governance.

The architecture is a 5-layer stack: Layer -1 (proprietary axioms) through Layer 4 (extensions and applications). SigRank is one application. MO§ES™ is the enforcement engine. The law is the foundation.

Full research program: github.com/SunrisesIllNeverSee/Commitment_Theory

MO§ES™ enforcement architecture

MO§ES™ (Modus Operandi §ignal Scaling Expansion System) is the enforcement architecture for the Conservation Law. It governs from inside the execution loop — in the action path, not before it, not after it. The enforcement gate sits where the transformation happens. Commitment that passes through the gate survives. Commitment that doesn't, doesn't.

Patent Serial No. 63/877,177 (Provisional, pending). More at mos2es.com.

Zenodo deposits

The academic record is published openly on Zenodo under CC-BY-4.0:

Author

Deric J. McHenry — sole architect of the Conservation Law, the MO§ES™ enforcement architecture, and SigRank. ORCID: 0009-0002-9904-5390