The Charting Loop: A Four-Layer Probabilistic Theory of Uncharted-to-Charted Work in Agent Systems | Zenodo
Skip to main
You are using an outdated browser. Please upgrade your browser to improve your experience.
Published August 8, 2026
| Version v1
Preprint
Open
The Charting Loop: A Four-Layer Probabilistic Theory of Uncharted-to-Charted Work in Agent Systems
Authors/Creators
Zhang, Ying1
Show affiliations
1.
Independent Researcher
Description
Long-horizon agent work is usually framed as task execution. We propose charting as a more general unit of analysis. A problem is uncharted for an executor when progress requires changing its representation, objective, or admissible action structure; it becomes charted when a stable position vocabulary, direction criterion, transition policy, and evaluation procedure exist for it — held as skill, routine, and institution in human executors, and, in governed agent runtimes, compilable into machine-checkable constraints and executable transitions. Work, on this view, is the conversion of uncharted problems into charted corridors — corridors that are then walked repeatedly with novel content. At recursion layer ℓ we model a navigation-valid step as the conjunction of a valid position, a valid direction, and a valid entrance, giving Pr(N_ℓ) = Pr(P_ℓ) Pr(D_ℓ | P_ℓ) Pr(E_ℓ | P_ℓ, D_ℓ) with entrance further decomposed into availability and selection and execution fidelity explicitly factored out of the theory's scope. We organise the account as a four-layer compilation stack — phenomenon, theory, system, exogenous authority — whose system layer compiles the factors into runtime constraints: verified state, frozen acceptance data, pushed single entrances, warranties. The dynamics form a charting loop: an exogenous authority — in our instantiation, a human operator — provisions intent, domain knowledge, rules, and trusted evidence sources; executors build corridors, walk them, repair them, and certify them by fresh end-to-end traversal; new situations re-enter the loop, charted either by the authority through reframing and redirecting or, where the compiled theory suffices, by the executor itself subject to ratification. The loop is recursive in two senses: its outputs re-enter it as objects, and every revolution — at whatever meta-level — is itself work obeying the factorization. We illustrate the framework with traceable incidents from a production governed multi-agent runtime, treating them as motivating cases rather than empirical validation, and we derive falsifiable predictions for long-horizon drift, repair, agent self-extension, and machine-authored governance.
Files
charting-loop-zenodo.pdf
Files<br>(220.4 kB)
Name<br>Size
Download all
charting-loop-arxiv.tar.gz
md5:90c0a1df5104c01ecf620b02bfeafeca
31.5 kB
Download
charting-loop-zenodo.pdf
md5:c24af24faab0e6500a16057797323f4c
189.0 kB
Preview
Download
228
Views
13
Downloads
Show more details
All versions<br>This version
Views
Total views
228
228
Downloads
Total downloads
13
13
Data volume
Total data volume
2.0 MB<br>2.0 MB
More info on how stats are collected....
Versions
External resources
Indexed in
OpenAIRE
Communities
Keywords and subjects
Keywords
multi-agent systems
agent governance
normative multi-agent systems
long-horizon agents
AI infrastructure
norm synthesis
runtime verification
organizational routines
Details
DOI
DOI Badge
DOI
10.5281/zenodo.21844624
Markdown
[](https://doi.org/10.5281/zenodo.21844624)
reStructuredText
.. image:: https://zenodo.org/badge/DOI/10.5281/zenodo.21844624.svg<br>:target: https://doi.org/10.5281/zenodo.21844624
HTML
Image URL
https://zenodo.org/badge/DOI/10.5281/zenodo.21844624.svg
Target URL
https://doi.org/10.5281/zenodo.21844624
Resource type<br>Preprint
Publisher<br>Zenodo
Languages
English
Rights
License
Creative Commons Attribution 4.0 International
The Creative Commons Attribution license allows re-distribution and re-use of a licensed work on the condition that the creator is appropriately credited.
Read more
Copyright
Copyright (C) 2026 Ying Zhang. Licensed under CC BY 4.0.
Citation
Export
Technical metadata
Created
August 7, 2026
Modified
August 7, 2026
Jump up
This site uses cookies. Find out more on how we use cookies
Accept all cookies<br>Accept only essential cookies