Commitment Conservation — MO§ES

Commitment conservation is the principle that semantic meaning in natural language is preserved under recursive transformation only when governance enforcement is present. It is the measured outcome that the Conservation Law of Commitment predicts and that the MO§ES enforcement architecture produces. This topic hub collects all MO§ES content related to commitment conservation — the theory, the evidence, the enforcement mechanisms, and the practical implications for AI systems.

What Commitment Conservation Is

Every natural language utterance carries commitment — the semantic obligation that binds a speaker to a future action or state. "The system shall reject unauthorized transactions" carries a commitment. "The system should reject unauthorized transactions" carries less. "The system may reject unauthorized transactions" carries almost none. The difference between these sentences is not stylistic. It is the difference between a governance constraint and a suggestion.

When AI systems process language — summarizing, translating, compressing, re-encoding, passing signals between agents — they apply transformations. The Conservation Law predicts that these transformations degrade commitment unless governance enforcement is present. Commitment conservation is the outcome of that enforcement: the commitments survive the transformation intact.

Formally, the Conservation Law states:

C(T(S)) ≈ C(S)  with enforcement
C(T(S)) < C(S)  without enforcement

Where C is the commitment function, T is the transformation operator, and S is the original signal. Commitment conservation is the condition where C(T(S)) ≈ C(S) — the transformed signal retains the commitment of the original.

Why Commitment Conservation Matters

Every AI system that processes natural language is a transformation operator. Summarization tools transform long documents into short ones. Translation systems transform text from one language to another. Multi-agent pipelines transform signals as they pass between agents. LLM chains transform outputs from one step into inputs for the next. Each of these transformations has the potential to degrade commitment.

The experimental evidence is unambiguous. In the published experimental record, unenforced recursive transformation reduced commitment by 15-20% per iteration. After 10 iterations — a typical depth for a multi-agent pipeline — original commitments were degraded by over 80%. With MO§ES enforcement, commitment was conserved at 80-85% of original levels across all 10 iterations.

This is not a marginal improvement. It is the difference between a governance system that works and one that does not. Without commitment conservation, AI systems systematically strip the obligations from the language they process. With it, those obligations survive.

The Evidence

Commitment conservation was tested through seven controlled experiments (EXP-001 through EXP-007) using:

• A 20-signal canonical corpus with known commitment levels
• 10 recursive iterations per signal
• NLI bidirectional entailment for semantic measurement
• Jaccard surface stability for lexical measurement

Results showed that with enforcement, commitment was conserved at 80-85% of original levels across all 10 iterations. Without enforcement, commitment degraded to below 20% by iteration 5 and was effectively zero by iteration 8. The full experimental record is published on Zenodo (DOI 10.5281/zenodo.19105225) under CC-BY-4.0.

A Worked Example

Suppose a governance policy states: "The system shall reject unauthorized transactions within 500ms." This sentence carries a commitment — a binding obligation with a specific threshold. Now run it through 10 rounds of summarization (a common transformation in multi-agent pipelines):

• Iteration 1: "The system should reject unauthorized transactions promptly." — "shall" drifted to "should," and the 500ms threshold is gone.
• Iteration 3: "The system may reject unauthorized transactions." — "should" softened to "may." The obligation is now a suggestion.
• Iteration 5: "The system handles unauthorized transactions." — "may reject" became "handles." The rejection requirement is gone entirely.
• Iteration 8: "The system processes transactions." — "unauthorized" dropped. The original commitment is unrecognizable.

Without enforcement, each transformation strips a layer of commitment. With MO§ES enforcement — pre-execution gating that checks the transformed signal against the original commitment — the 500ms rejection obligation survives all 10 iterations intact. The surface wording may change, but the binding meaning is preserved. That is commitment conservation.

Limitations and Open Questions

The Conservation Law of Commitment is a proposed law, not an independently established result. It has not been peer-reviewed or replicated by independent researchers. The current evidence base has several limitations:

• Single-author preprint: The law and its experimental record are authored by Deric J. McHenry and published on Zenodo. No independent replication has been published.
• Small corpus: The experimental record uses a 20-signal canonical corpus. Whether the results generalize to larger, more diverse corpora is an open question.
• Measurement proxies: Commitment is measured using NLI bidirectional entailment and Jaccard surface stability. These are proxies for semantic meaning, not direct measurements. The choice of proxy affects the measured conservation rate.
• Enforcement scope: The experiments test MO§ES enforcement specifically. Whether other enforcement mechanisms produce similar conservation rates is not yet demonstrated.
• Self-citation: The one OpenAlex-recorded citation to the public harness (W7139122947) is from the author's own SigRank dataset paper. No independent citing work has been identified.

The law is falsifiable — it makes specific predictions about commitment degradation under transformation — and the experimental record is public and reproducible. But falsifiability is a property of the framework, not proof of its correctness. Independent replication and peer review are needed before the law can be considered established.

Relationship to MO§ES and the Products

The Conservation Law of Commitment provides the theoretical foundation for MO§ES™ — the governance enforcement framework. MO§ES is the mechanism that makes commitment conservation operational: it enforces the law's predictions at execution time. The products built on this stack — SigRank (operator evaluation), Signomy (governed marketplace), and the Upsilon enterprise measurement engine — implement MO§ES governance in their respective domains.

The connection is supported: MO§ES was designed to enforce the Conservation Law, and the experimental record tests MO§ES enforcement specifically. But the theory itself is the author's proposed framework, not an independently validated result. The products work as described regardless of whether the Conservation Law is eventually accepted as a general principle — they enforce commitment preservation because that is what they are built to do, not because the law has been proven.

The Original Paper

The Conservation Law of Commitment is formally stated in:

McHenry, D. J. (2026). A Conservation Law for Commitment in Language Under Transformative Compression and Recursive Application (V.05) [Preprint]. Zenodo. https://doi.org/10.5281/zenodo.20029607

ORCID: 0009-0002-9904-5390 · License: CC-BY-4.0 · Patent pending 63/877,177. The public recursive transformation harness is indexed in OpenAlex (W7139122947). The prototype test harness is open-source on GitHub. An independently written summary is available on Grokipedia.

Related Concepts

• Conservation Law of Commitment: The falsifiable law predicting commitment degradation without enforcement
• Commitment Conservation: The detailed concept page for commitment conservation
• Governance Enforcement: The mechanism that makes commitment conservation possible
• Lineage Claw: The cryptographic binding that preserves commitment provenance
• Recursive Compression: The transformation operator that threatens commitment

Related Guides

• How to Enforce Commitment Conservation: A practical guide to implementing conservation in AI pipelines
• Measuring Commitment: How to quantify commitment levels using NLI and Jaccard methods

Related Comparisons

• Enforced vs. Unenforced Transformation: Quantitative comparison of commitment outcomes
• MO§ES vs. No Enforcement: The measured difference across 10 iterations

How MO§ES Produces Commitment Conservation

MO§ES produces commitment conservation through four enforcement mechanisms that operate at execution time — before a transformation is applied, not after:

1. Pre-execution gating: Every transformation is checked against governance constraints before it is applied. Transformations that would degrade commitment below an acceptable threshold are rejected.
2. Lineage binding: Every transformed signal is cryptographically tied to its origin through the Lineage Claw mechanism, creating an immutable chain of custody.
3. Resonance thresholding: Transformations must maintain resonance with their origin signal above a defined threshold. Below that threshold, the transformation is rejected.
4. Audit trails: SHA-256 hashes of every transformation create a verifiable record that can be audited at any time.

With these mechanisms in place, commitment is conserved. Without them, the Conservation Law predicts — and the experiments confirm — systematic degradation. The MO§ES architecture makes the law operational.

FAQ