nucleation

CLAIM-344: Resolve is a maintained organization of precedence, commitment and revision that keeps a temporally extended policy effective across changes in locally dominant evaluations while keeping the capacity to revise it; mere slowness or persistence is not resolve

6 min read

Properties13
StatusActive
Version3
Rolethesis
Warrantinterpretive
Tagsself-governance, behavioral-integration, agency
Linewill-organization, distributed-agency
Rests onCLAIM-277 — Durable authority among an organization's processes need not be a primitive: trust can be the temporal regularity of endorsement that policies earn by being right often, and the same loop entrenches whatever gained influence early, CLAIM-331 — A level of organization has a temporal facet as well as a compositional one: higher-level variables are characteristically slower and longer-lived than the activity that produces them, though slowness is a mark of a level and not its definition, CLAIM-099 — Maintaining a regime of effective action and maintaining the capacity to govern among regimes are not the same achievement
GroundsTHEORY-688 — In an actor-critic learner that can choose temporally extended options as well as primitive actions, options that fit the task speed learning, while options with ill-placed subgoals slow it, and options that fit an earlier layout keep the agent from finding a newly opened shortcut, so hierarchy helps only with the right subroutines
ComplementsCLAIM-443 — A learned prior over which processes govern, P(g | z, H) ∝ exp(βQ(g, z) + b_H(g, z)), its bias b_H trained by reward-modulated Hebbian updates to the relations by which one process recruits or endorses another, is a candidate mechanism for durable authority and for amortized governance; but persistence is not amortization, which is shown only when earlier work replaces later computation, communication or recruitment, and a prior over authority differs from a prior over actions only when what is appropriate changes
Corroborated byTHEORY-067 — Self-control over time needs no faculty of will: interests that dominate at different delays bargain, and resolve is the recursive staking of expected future choices on the present one
Illustrated byCASE-128 — The voluntary delegation: a person gives a friend the password to a distracting application during a calm period and is refused it later under temptation, while another person lives under the same restriction imposed without consent
Cited as “complements” byCLAIM-523 — For a normatively relevant discrepancy to change what an organization does, three distinct achievements are needed and any can fail while the others hold: legibility, by which the discrepancy becomes recognizable; volitional efficacy, by which a corresponding demand can move action; and correctability, by which a fitting revision can be carried out within deadline and resources; none establishes that the demand or the revision is justified, and whether the norm-satisfying histories the agent can still realize are empty depends partly on what its earlier actions destroyed
Cited as “illustrates” byCASE-188 — A proposed study: two agents keep the same long-term commitment and decide alike under ordinary conditions, one protecting it from irrelevant fluctuation while letting specified defeaters revise it, the other growing ever more efficient at enforcing it while evidence against it loses its way in; varying the environment, the legitimacy of the original commitment and what the evaluative mechanisms can see, to tell corrigible resolve from entrenchment without building the answer into the diagnostic or the score

Graph View

The claim

A36: “Resolve can emerge when current decisions become linked to expectations concerning future decisions.” The owner, at U37: “I think emergence of “resolve” could be an interesting frame within which to examine our will-organization framework.”

A37 §3: “Resolve is a maintained organization of precedence, commitment, and revision that preserves the practical efficacy of a temporally extended policy across foreseeable changes in locally dominant evaluations.” It set resolve beside five other ways a long-term policy B can win over a momentarily favoured A: preference reversal, bundling, precommitment, endorsement and suspension. Resolve is the one in which “A persistent organization sustains B’s practical authority across the relevant changes in local evaluation”. Then: “mere slowness isn’t resolve. A slow parameter could also be irrelevant, externally imposed, or maladaptively fixed. The crucial properties are causal governance, retention across relevant disturbances, and appropriate capacity for revision.”

The claim has three parts, each resting on a claim the record holds.

  1. Durable authority, earned. The commitment’s precedence is held by influence relations that persist. CLAIM-277 says durable authority can be “the temporal regularity of endorsement”, and that “the same loop entrenches whatever gained influence early”, which is the rigidity risk inside the mechanism.
  2. Slowness is not enough. CLAIM-331: higher-level variables are slower, but “slowness is a mark of a level and not its definition”. A37 on the two-timescale update rule (Heusel et al. 2017, held in the anthology as ANTH-LIT-611): “the slower process in TTUR is the generator, not a higher-order will-controller. Slowness alone supplies no authority or motivational hierarchy.”
  3. Revision kept. CLAIM-099: maintaining a regime of effective action and maintaining the capacity to govern among regimes are not the same achievement. Resolve needs both. A37: “To prevent rigidity from masquerading as resolve, the commitment conditions must also include an appropriate revision policy.”

The grades

The exchange separated the grades several times, in nearly the same terms.

  • A36 on learned features: “Passive retention”, “Selective retention”, “Regulated retention”, tested by “active causal maintenance”.
  • A37: “Persistence: A representation survives. Resolve-like governance: Its future efficacy is actively protected through a maintained influence structure. Adaptive resolve: The protection itself is revisable in response to appropriate evidence and changing constraints.” A40 §5 kept these as retention, resolve-like governance and adaptive resolve.
  • A43: “Passive persistence: A representation or policy remains unchanged. Regulated persistence: A maintained organization actively protects its efficacy. Resolve: A temporally extended commitment retains practical precedence through relevant disturbances. Adaptive resolve: The organization preserves that commitment while retaining an effective, context-sensitive procedure for justified revision.”

What the claim calls resolve is A37’s “adaptive resolve” and A43’s top grade. The lower grades are persistence.

Two ways to fail

A41, from the insider-threat material (CLAIM-359): “resolve should not be so easy to manipulate that a constituent can redirect it toward an unauthorized objective, nor so rigid that legitimate corrections cannot reach it.” Its three cases: “Fragility: Internal disturbances readily destroy or redirect commitments. Rigidity: Commitments remain stable even when revision is warranted. Adaptive robustness: Commitments resist inappropriate interference while remaining responsive to legitimate evidence and correction.”

A42 added what keeps resolve on the adaptive side (CLAIM-349): “Adaptive resolve requires more than persistence of commitment or resistance to internal disturbance. A governing organization must distinguish access to evaluative signals from their authority over action, preserve the capacity to tolerate unresolved conflict without automatically suppressing it, and maintain mechanisms for revising commitments and governance when warranted.” A44 put the same point as a route: in “Reflectively available resolve”, “Competing evaluations remain accessible, but their control over action is conditionally governed”.

THEORY-688 is the record’s documented case of the rigid side in a learner. In Botvinick, Niv and Barto’s actor-critic, options fitted to an earlier layout kept the agent from a newly opened shortcut: primitive-only agents found it on 75% of runs, doorway-option agents never used it. Learned structure that sped learning protected itself into rigidity.

Resolve without exertion

A45: “Resolve need not consist in an executive process continually exerting effort against resistance. It can consist in a durable organization that has made repeated executive exertion less necessary.” A47 §8: “The achievement of governance may be to make its repeated exercise unnecessary.” Resolve is then “the persistence of an effective commitment through the organization of its constituent processes, rather than the continual expenditure of a superior will overpowering them”. The revision condition stays: “the maintained capacity to determine when persistence, adaptation, and relinquishment are appropriate”.

Against THEORY-067 and CLAIM-159

The record already holds a theory of resolve. Ainslie (THEORY-067): resolve is “the recursive staking of expected future choices on the present one”; treating a choice as a precedent bundles the series, and a lapse costs the rule its credibility. That is A37’s “bundling” row, one mechanism among six. This claim is not a rival mechanism. It says what organization the mechanism must sit in to count as resolve: one that keeps the bundled policy effective and can also revise it. Ainslie’s own reading predicts the failure the revision condition guards against. As NOTE-443 reads it (C6), reliance on personal rules predicts legalism, lapse districts, motivated misperception and over-control, a claim argued and fitted to clinical observation, not tested.

CLAIM-159 puts the record’s rival accounts of will in drive-profile cells: Ainslie’s interests aggregate or emergent, the expected value of control top-down, will-organization the maintained organization that governs how the drive profile changes. Resolve in this claim’s sense belongs to the third cell. A45’s and A47’s point is that it need not be top-down: a mature organization keeps a commitment with less specification from above, not more.

What it does not say

  • It does not give the standard by which revision is warranted. A37 said the revision policy “introduces an independently justified standard for when preservation versus revision is warranted”. Whether a standard internal to the organization can tell warranted persistence from entrenchment is QUESTION-087’s open problem.
  • It does not say that networks in training have resolve. A37: a consolidation mechanism supplied by the training algorithm “may demonstrate resolve-like organization of the training system, not volition owned by the neural network alone”. A40’s Study D, with its control of perturbing or ablating “the slow influence structure while keeping task information and representational capacity comparable”, is the test, and it has not been run.
  • It does not say resolve is good. A commitment kept adaptively can be a commitment to something that ought not to persist (CLAIM-075).
  • It does not say every persistent commitment is rigid. Persistence is sometimes the flexible response (CLAIM-349); rigidity is persistence when revision is warranted.

Backlinks