nucleation

QUESTION-032: Relative to a level of description and a target, what makes a factor of a system an agent, and what further makes it a willed agent?

3 min read

Properties9
StatusOpen
Version4
Tagsagency, individuation
Linedistributed-agency
RefinesQUESTION-009 — When does a pattern of coordination constitute an additional agent?
Refined byQUESTION-228 — Run on real systems, does Kenton et al.'s interventional procedure give the verdicts its definition of agency predicts, for a thermostat with and without its design process, an RL policy with and without its training loop, and a human, and does it separate an accidentally optimal input-output system from an agent where a compression test does not, under a principled choice of variables?
Answered byCLAIM-170 — The agents in a system are a factorization of it relative to a level of description and a target; agents in general are identified by adaptation, and willed agents are individuated by their will-organization: a willed agent is the jurisdiction of a will-organization, CLAIM-174 — A candidate factor is an agent at a level when it is one system there, when a goal-directed description beats a device description by more than the Occam cost, and when its policy would adapt if its actions' consequences changed; it is a willed agent when, further, a will-organization over several regimes describes it better than none, and what passes that is the jurisdiction of a will-organization
BearingTHEORY-801 — Agents are the systems that would adapt their policy were their actions to affect the world differently, so the agent is wherever that adaptation is: a thermostat is not an agent while it together with its design process is, an RL training process is while its learnt policy generally is not, and humans are, THEORY-624 — Orseau, McGill and Legg cast whether an identified system is an agent or a device as Bayesian model comparison between descriptions of the function it optimises and descriptions of its input-output mapping, so that the verdict is relative to the observer's prior; on seven toy gridworld trajectories goal-seeking comes out agent-like and wall-following device-like, and they argue that under a speed prior everything may look like an agent, THEORY-616 — If a system has beliefs and desires just when attributing them, with rationality, predicts its behaviour well enough against rival theories, then the bearer of a language model's attitudes is the instance created by one context, not the model, the case that instances have them is presumptive and rests on tractability, and the best current desire hypothesis is helpfulness, honesty and harmlessness plus intrinsic desires a prompt installs, THEORY-208 — Agency is multiply realisable without cognition or mental states, and AI is a new kind of it, THEORY-071 — Minimal agency requires individuality, interactional asymmetry and self-generated normativity, and individuality without the other two is not agency, CLAIM-155 — Willed agency is a kind between agency and selfhood: a control system whose behaviour is governed by a will-organization over several regimes is more than an agent and closer to a self, in the self-governing sense, without thereby being a subject
Cited as “bears on” byTHEORY-1236 — Agency is relative to an observer: an identified system is an agent, for that observer, to the degree that a description of the function it optimises explains its behaviour better under the observer's prior than any description of its input-output mapping

Graph View

Why it is a question

The will paper set this question aside on purpose. It supposed its agent (CLAIM-080) so as not to face the permissiveness objection, that every control system has a will (CLAIM-010). This session has built the materials an answer needs:

  • which blocks are one system (CLAIM-157);
  • how drive is apportioned among them (TERM-051);
  • a way to price a description against its cost (CLAIM-154).

The owner then proposed (CLAIM-170) that agents are a factorization of a system at a level, and that each is individuated by its will-organization (TERM-023).

What would count as an answer

A test, run on a candidate factor at a stated level and for a stated target, that admits some factors as agents and refuses others, and could be wrong. It must:

  • refuse a thermostat (CASE-044) or explain why it should pass;
  • say whether a dancing couple, a firm and a microbiome are agents, and at which levels;
  • avoid presupposing the agent it is meant to identify (CLAIM-168).

The proposed answer is CLAIM-174.

Note of 2026-10-12: a typed partial answer

THEORY-071 now carries an answered_by edge. Its criteria (individuality, interactional asymmetry, self-generated normativity) refuse a thermostat. They are not relative to level or target, which is the tension the counter CLAIM-194 takes up.

Note of 2026-10-12: a permissive answer

THEORY-208 (Floridi) now carries an answered_by edge. Interactivity is necessary for agency, and autonomy and adaptability are graded, which is close to CLAIM-170’s “identified by adaptation”. It is permissive: thermostats and rivers pass, by splitting agency into kinds. It is not relative to level or target, and it gives no test that could fail.

Note of 2026-10-14: a comparative criterion relative to a decomposition

THEORY-232 (Watson, LIT-990) makes a whole more agential than its parts when its organization lets them resolve more of their mutual constraints than they do at the equilibrium of their former organization. It is relative to a chosen decomposition, as this question asks, and it gives the goal-directedness test of CLAIM-174 a behavioural mark in delayed gratification. It does not supply the test asked for here: it takes the parts’ goals as given, so it does not say which factors are agents, and it sets no threshold. It carries no answers edge.

Note of 2026-10-14: the frame a relativization must fix

THEORY-249, filed from LIT-969, supports this question’s framing: whether a system has agency is settled only relative to a reference frame. It widens what the relativization must include. Beyond a level and a target, a frame fixes the causal variables, a principle for meaningful goals and a class of meaningful adaptations. It does not supply the test asked for here and gives no verdict on CASE-044, so it carries no answers edge.

Backlinks