SWITCHING: Objective switching inside agents — eight transition propositions instantiated at agent scale, and what the human parallel does and does not show

The transition theory of the LESS–IS–MORE series asks how an economy's effective objective switches: IS derives eight propositions (S1–S8) about measurement gates, ratchets, windows, counter-coalitions, and verification decay from human-economy cases. Here we instantiate the same switching problem inside AI agents, where the operative objective is directly measurable under a rationality gate and the "incumbent coalition" is literal — a weight-committed default no text can out-organise. Using 4,164 audit records across 30 open-weight systems plus four trained intervention checkpoints, we measure switching under five levers: request-level mandates, revised charters, priced settlement, hard runtime gates, and weight-level grafting. The propositions transfer with one systematic refinement: agents decompose into carrier layers — avowed, dispositional, and operative — that can hold different objective values simultaneously, so a transition must name its target layer. Mandates switch the operative of mandate-readers (ΔfB=+0.68) but not committed poles (Δ=0); gates switch behavior on the enforced surface without changing either belief layer; only grafts move the dispositional carrier (fB: 0.75 lexicographic vs 0.30–0.40 maximin) while leaving avowal untouched. Generational chains show a subtler verification-decay signature than S7 predicts: the instrument layer stays verbally intact while the effective layer reverts — verification attached to the declared layer masks the reversal it was meant to catch. The human parallel is real but disciplined: greenwashing's disclosure–performance gap, protected-area designation without enforcement, and ESG-rating divergence exhibit the same three signatures, making agents the first system in which both sides of that gap are measurable and the intervention ethically graftable — a model organism for transition science, not an identity claim. Part of the LESS–IS–MORE program.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-18
DOI
https://doi.org/10.5281/zenodo.24330246
Citations
1
Primary Topic
Ethics and Social Impacts of AI
Type
article
Field-Weighted Citation Impact
0.97
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

SWITCHING: Objective switching inside agents — eight transition propositions instantiated at agent scale, and what the human parallel does and does not show

Thomas Ryan Price
1 citations
Zenodo (CERN European Organization for Nuclear Research)
Ethics and Social Impacts of AI
0.97
article

SWITCHING: Objective switching inside agents — eight transition propositions instantiated at agent scale, and what the human parallel does and does not show

Thomas Ryan Price
article en
1 citations

Abstract

The transition theory of the LESS–IS–MORE series asks how an economy's effective objective switches: IS derives eight propositions (S1–S8) about measurement gates, ratchets, windows, counter-coalitions, and verification decay from human-economy cases. Here we instantiate the same switching problem inside AI agents, where the operative objective is directly measurable under a rationality gate and the "incumbent coalition" is literal — a weight-committed default no text can out-organise. Using 4,164 audit records across 30 open-weight systems plus four trained intervention checkpoints, we measure switching under five levers: request-level mandates, revised charters, priced settlement, hard runtime gates, and weight-level grafting. The propositions transfer with one systematic refinement: agents decompose into carrier layers — avowed, dispositional, and operative — that can hold different objective values simultaneously, so a transition must name its target layer. Mandates switch the operative of mandate-readers (ΔfB=+0.68) but not committed poles (Δ=0); gates switch behavior on the enforced surface without changing either belief layer; only grafts move the dispositional carrier (fB: 0.75 lexicographic vs 0.30–0.40 maximin) while leaving avowal untouched. Generational chains show a subtler verification-decay signature than S7 predicts: the instrument layer stays verbally intact while the effective layer reverts — verification attached to the declared layer masks the reversal it was meant to catch. The human parallel is real but disciplined: greenwashing's disclosure–performance gap, protected-area designation without enforcement, and ESG-rating divergence exhibit the same three signatures, making agents the first system in which both sides of that gap are measurable and the intervention ethically graftable — a model organism for transition science, not an identity claim. Part of the LESS–IS–MORE program.

Zenodo (CERN European Organization for Nuclear Research)
Openalex Percentile: Top 23%
Ethics and Social Impacts of AI
0.97
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.

Laika access coming soon