Autonomy Without a Global Author: Reflective Reach, Self-Governance Resolution, Diachronic Ownership, Manipulation-Sensitive Authorship, and Autonomous Corrigibility for Embedded Agents in a Consistency-Constrained Information History
Tomoyuki Uchida · Zenodo (CERN European Organization for Nuclear Research) · 2026
Paper 23, Action Without a Global Controller, separated action attribution, control, successful controlled action, agent corrigibility, and the basis for responsibility assessment. This paper pays the resulting autonomy debt without positing a global author with complete introspection, exhaustive self-modeling, unrestricted self-revision, or value outside causal history. The subtitle's five themes are organized by three axes and three standings. Reflective reach and self-governance resolution ground agent-level self-governance capacity (AutCap); diachronic, manipulation-sensitive acquisition grounds state-level ownership (Own); and owned governing support plus mediation on the actual implementation lineage grounds token-level authorship (Author). Ownership admits clean reason-responsive formation, later reflective integration, and autonomous appropriation of a bypassed value. The latter two are genuine departures from the epistemic analogy: values may become one's own after non-reflective acquisition or after repair of a defective history. World-level, attributed, and warranted standings remain distinct. R1-R3 provide elementary scaffolding protecting reach, resolution, closure, and lineage distinctions. R4 locates its philosophical content in matched histories with identical endpoints but different acquisition adequacy. R5 supplies an audit decomposition for autonomous corrigibility: correction recognition and implementation reliabilities lower-bound autonomy-preserving uptake; challenge and refusal reliabilities upper-bound bypass-mediated uptake; and an explicit abstention ceiling prevents ambiguity from masquerading as performance. Under declared coverage, routing, calibration, non-coercion, and module-to-standing assumptions, qualified structure suffices for the method-level target, while the converse fails. Autonomy is neither uncaused self-origination nor resistance to correction; corrigibility is not remote control. ALAC Paper 24. Part of the Consistency-Constrained Information History (Akashic Ledger Architecture Cosmology) series. Continues Paper 23 (Action Without a Global Controller); sets up Paper 25 (Responsibility Without a Global Judge, in preparation).