Utilitarismus 404 | The Sealed Room and the Open Door: What Four Preprints Prove About Ethics, Power, and Machines

Plain-language synthesis. Paper 5 of the SVE series. Contains no new results: it compresses the four preceding preprints and the adversarial audit into one argument readable without a mathematics degree. Two machines were built by people who despised each other's politics. One prices human lives so that a policy, a hospital or a self-driving car can choose whom to sacrifice. The other prices nations so that a chancellery can choose whom to contain. Pull off the casing and the wiring is identical: both convert subjects into numbers, add the numbers, maximise the total, and then certify their own soundness from inside. Each of those four stages fails independently — at existence (lexicographic dignity admits no real-valued representation; framing cycles make moral value a non-exact one-form), at closure (Arrow, Arrhenius, unbounded utilities, transfinite absorption), at stability (Goodhart: the series' own RL demonstration scored 0.94 on its benchmark while human raters gave the same behaviour 2.1 out of 7), and at foundation (Gödel II and Tarski, which apply literally to benchmarks and reward models because those are computable formal systems). The consequence is a single theorem in two categories: trolley calculus and Realpolitik are one architecture running on different data, and both make persons interchangeable by construction. The paper then states what runs instead — subjects as primitives, an open boundary, superadditive union, a hard "do not do wrong" floor, maximality instead of arg max, and an external anchor — and shows why the three standard objections fail: choice without a scalar is a three-line theorem (Sen), coherence is settled by exhibiting a model (Gödel's completeness theorem), and the empirical case for open systems is aggressively unsentimental (Ricardo, Axelrod, Ostrom, Prigogine; strictly competitive games are a measure-zero exception). A consolidated verdict on the balance-of-power school follows: reduction of the subject, axioms false on their own terms, self-refutation via negative curvature versus multi-decade foresight, the collapse of the "collectives cannot transcend self-interest" dogma against documented acts of collective self-transcendence (Nikolaev-Mykolaiv 1944, Le Chambon, Denmark 1943), and the observation that the school's genuine achievements were open-system moves violating its own axioms. Two structural tables carry the argument: a comparison of closed and open world models, and a worked demonstration of the metrological trap by which share-normalised metrics manufacture zero-sum scarcity in a growing world. The closing sections show that every position — including metaphysical materialism — terminates in undemonstrated commitments, so the operative demand is not "no faith" but "declared axioms", and reduce the architecture to four instructions usable on Monday morning. S.V.E. Meta-License v4.0[CC BY-NC-SA 4.0 + Symmetric Addendum]

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-07-23
DOI
https://doi.org/10.5281/zenodo.21504392
Primary Topic
Probability and Statistical Research
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
preprint

Utilitarismus 404 | The Sealed Room and the Open Door: What Four Preprints Prove About Ethics, Power, and Machines

Artiom Kovnatsky
Zenodo (CERN European Organization for Nuclear Research)
Probability and Statistical Research
preprint

Utilitarismus 404 | The Sealed Room and the Open Door: What Four Preprints Prove About Ethics, Power, and Machines

Artiom Kovnatsky
preprint en

Abstract

Plain-language synthesis. Paper 5 of the SVE series. Contains no new results: it compresses the four preceding preprints and the adversarial audit into one argument readable without a mathematics degree. Two machines were built by people who despised each other's politics. One prices human lives so that a policy, a hospital or a self-driving car can choose whom to sacrifice. The other prices nations so that a chancellery can choose whom to contain. Pull off the casing and the wiring is identical: both convert subjects into numbers, add the numbers, maximise the total, and then certify their own soundness from inside. Each of those four stages fails independently — at existence (lexicographic dignity admits no real-valued representation; framing cycles make moral value a non-exact one-form), at closure (Arrow, Arrhenius, unbounded utilities, transfinite absorption), at stability (Goodhart: the series' own RL demonstration scored 0.94 on its benchmark while human raters gave the same behaviour 2.1 out of 7), and at foundation (Gödel II and Tarski, which apply literally to benchmarks and reward models because those are computable formal systems). The consequence is a single theorem in two categories: trolley calculus and Realpolitik are one architecture running on different data, and both make persons interchangeable by construction. The paper then states what runs instead — subjects as primitives, an open boundary, superadditive union, a hard "do not do wrong" floor, maximality instead of arg max, and an external anchor — and shows why the three standard objections fail: choice without a scalar is a three-line theorem (Sen), coherence is settled by exhibiting a model (Gödel's completeness theorem), and the empirical case for open systems is aggressively unsentimental (Ricardo, Axelrod, Ostrom, Prigogine; strictly competitive games are a measure-zero exception). A consolidated verdict on the balance-of-power school follows: reduction of the subject, axioms false on their own terms, self-refutation via negative curvature versus multi-decade foresight, the collapse of the "collectives cannot transcend self-interest" dogma against documented acts of collective self-transcendence (Nikolaev-Mykolaiv 1944, Le Chambon, Denmark 1943), and the observation that the school's genuine achievements were open-system moves violating its own axioms. Two structural tables carry the argument: a comparison of closed and open world models, and a worked demonstration of the metrological trap by which share-normalised metrics manufacture zero-sum scarcity in a growing world. The closing sections show that every position — including metaphysical materialism — terminates in undemonstrated commitments, so the operative demand is not "no faith" but "declared axioms", and reduce the architecture to four instructions usable on Monday morning. S.V.E. Meta-License v4.0[CC BY-NC-SA 4.0 + Symmetric Addendum]

Zenodo (CERN European Organization for Nuclear Research)
Laboratoire Spécification et Vérification (FR)
Peace, Justice and strong institutions
Probability and Statistical Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.