What Is Work: The Law of Information Conservation and the AI Productivity Paradox in High-Context Knowledge Work
Alexander Vityaz (ORCID 0009-0006-0489-7881) · Corezoid Inc., Dnipro, Ukraine Published: April 2026 · Version: v1 · License: CC BY 4.0
Note. This preprint has no DOI yet; cite via the ResearchGate URL until a Zenodo DOI is minted (see ../../PUBLISHING.md).
The paper asks what work is from an informational point of view, and whether AI actually reduces its volume. It answers with three stated invariants. The Law of Information Conservation holds that artifact quality is a function of the information accounted for, so the information volume I_Q required for a given quality is fixed by the task rather than by the tool — removing a participant redistributes that burden without shrinking it. The Law of the Bottleneck locates the binding constraint in human cognitive throughput: AI generates roughly 200 times faster than a human can read with comprehension, and working memory caps verification at a few chunks per second, so the queue of unread material grows instead of the output. The Law of Factorization applies the author's Vityaz–Ashby theorem, under which a minimal good regulator factorizes into a model of the system and a deterministic noise filter; an error-prone filter makes the regulator either non-minimal or prone to false control actions. From these the paper derives the productivity paradox of high-context knowledge work — the prototype arrives in seconds while total time to a given quality does not fall — along with a context threshold, the collapse from a working pair to a lone initiator, and the resulting accumulation of management debt. It closes by positioning "AImatics" (LLM as interface, executable graph as the deterministic filter) as the path to a restored factorization, and states its own genre limits explicitly: an essay with theorems, whose numerical illustrations demonstrate a method rather than confirm a hypothesis.
| File | Description |
|---|---|
| paper.pdf | Canonical PDF (author's copy of the ResearchGate deposit) |
Vityaz, A. (2026). What Is Work: The Law of Information Conservation and the AI Productivity Paradox in High-Context Knowledge Work. Preprint, ResearchGate. https://www.researchgate.net/publication/403936327
@misc{vityaz2026whatiswork,
author = {Vityaz, Alexander},
title = {What Is Work: The Law of Information Conservation and the AI Productivity Paradox in High-Context Knowledge Work},
year = {2026},
month = apr,
howpublished = {Preprint, ResearchGate},
url = {https://www.researchgate.net/publication/403936327}
}- Builds on On the Necessity of Noise Suppression for Minimal Good Regulators — supplies the Vityaz–Ashby factorization theorem and the determinism requirement on the filter, on which the essay's Law of Factorization and its §§7 and 9 architectural argument rest (reference 20).
- Builds on Management Debt—Part I — supplies the management-debt accounting to which the essay attributes the measurable losses of the degraded human+AI configuration (reference 21).
- Cited by The Compact Company — which uses the information-theoretic argument that verification and correction remain positive work when the first LLM output is fast (reference [59]).
- ResearchGate (no DOI): https://www.researchgate.net/publication/403936327
See CHANGELOG.md.