The claim
The owner, verbatim, on 2026-10-11: “or maybe a control system identifiable by a will is a special kind that we should differentiate from ‘agency’? This might be the distinction I was trying to touch on with my ‘self’-ness question. we can have agency without a ‘self’, but a control system which can be properly be characterized as possessing a ‘will’ might be worth differentiating, at the very least insofar as it might be closer to a ‘self’ than a thermostat.” On the name: “willed agent works”.
The tiers nest:
| tier | test | example that reaches it and no further |
|---|---|---|
| control system | feedback toward a setpoint | a bare thermostat |
| agent (TERM-044) | system, description, adaptation | a thermostat with its designers; an RL training process |
| willed agent (TERM-048) | plus a will-organization over several regimes | perhaps a firm such as Northstar (CASE-023) |
| self (TERM-045) | further unities, sense stated | a person, in the subject sense |
Why this narrows the owner’s earlier thesis
CLAIM-170 said agents are identified by their will-organization (TERM-023). The readings show that agency is wider. Kenton et al.’s and Orseau et al.’s tests admit systems with no governance over several regimes. So:
- agents in general are identified by adaptation and description;
- willed agents are identified by the jurisdiction of their will-organization.
The factorization half of the owner’s thesis, relative to a level and a target, stands for both tiers. CLAIM-170 version 2 narrows its individuation half to willed agents.
Why “closer to a self”, and in which sense
A will-organization gives a system standards that decide among its own regimes, and with them the persistence of that governance through changes in what drives (CLAIM-316). That is near ADR-024’s self-governance unity (TERM-045 sense 5) and supports persistence (sense 3). It gives no point of view (sense 1). That is why the firm is the hard case: it may be willed and self-governing without being a subject. And it is why the thermostat is far from a self on every sense.
The permissiveness objection, answered on its merits
The will paper’s CLAIM-010 said that on its account every control system has a will. The paper set this aside by supposing its agent (CLAIM-080). On the tiers it is simply false: control systems include agents, agents include willed agents, and the inclusions are strict. CLAIM-080 remains a sound scoping decision for the paper. It is no longer the only answer.
What it does not say
- That will is required for selfhood in the subject sense. A subject without governance over several regimes is not ruled out.
- That any firm is a willed agent. That is the test, applied.
Note of 2026-10-29: what holds it back
CLAIM-485, Active since the twelfth status review, holds that allocating authority among one’s own processes by standards one cannot revise makes a system no more self-governing, in TERM-045’s sense 5, than an external scheduler would. This claim’s case for “closer to a self, in the self-governing sense” is that a will-organization supplies “standards that decide among its own regimes”, which is that allocation. Until the claim is narrowed to willed agents whose will-organization can revise those standards (TERM-023’s transformation dimension), or argues against CLAIM-485, the comparison in sense 5 is unsupported. TERM-048’s “What it is not”, which gives a willed agent “something like the self-governance unity”, carries the same tension.