Work

Machine Governance Beyond Man vs. Machine: Toward a Theory of Situated Machine Knowing, Machine-Operational Policy, Control, and Justified Agency

Year: 2026
Type: article
Abstract: Governance of artificial intelligence is usually narrated as a contest over control between people and machines. This paper develops a governance theory that does not begin from that premise. It asks instead under what conditions a machine is justified in treating a state as known, and under what conditions it is entitled to act on it. We propose four machine-operational constructs. A knowledge property graph (KPG) holds the authority a machine acts under, each claim quoted from the authority's text or attested by a named person. A machine KPG (mKPG) holds what the machine is justified in treating as known about one situation at one time, each assertion bound to the instrument through which it entered. mPolicy is a machine-operational realization of policy, mControl a mechanism that governs machine knowing and acting under it, and mGovernance the governance that results. The normative tier is not only proposed: for the KPG, prior work proves that a committed decision is the authority's decision under premises its witness names, and locates the failed premise when a commitment is wrong. We state eight falsifiable propositions, set them against published evidence that bears on five of them, give a pre-registration-ready design for the proposition on which the epistemic tier rests, and compare the architecture with adjacent frameworks. The paper reports no new experiments. Its contribution is the theory, the formal grounding of its normative tier, an account of what existing evidence already shows, and a program for testing the rest.
Language: English
Open Access status: green
OpenAlex ID: W7236462023
FWCI:
Cites:
Cited by:
Related to: 10