The Prioritization Regress
Why the Pull To Act Cannot Be Read, and Why It May Be Felt
The prioritization argument stated on its own. The main theoretical paper sets out the whole framework; this paper isolates the argument the framework rests on, states it as compactly as it can be stated, and tests it against the strongest computational alternatives, active inference above all. It is written for philosophers and cognitive scientists, and it can be read without the rest of the corpus.
Pace, J. C. (2026). The Prioritization Regress: Why the Pull To Act Cannot Be Read, and Why It May Be Felt. PhilPapers. https://philpapers.org/rec/PACTPR-2
Abstract
Accounts of consciousness often conflate value, how good or bad something is taken to be by the organism, and urgency, the pull to act. Prioritization is a matter of urgency. Any account on which prioritization produces a score that something separate reads faces a regress: the reader's rule is either fixed, so the scores' significance for action was settled elsewhere, or adaptive, so it needs a further evaluator. Systems with no separate reader avoid it, but where tension resolves directly into action they cannot deliberate. Standard planners, active inference included, deliberate but fall back in. What both deliberates and ends the regress is a held pull: urgency that persists undischarged, whose change under each rehearsed strategy is that strategy's evaluation. I propose, as an identity thesis, not a deduction, that the system-wide state of the field holding such pulls is identical with consciousness, and state how each claim could fail.
Two distinctions
The argument rests on two distinctions. The first is between value, how good or bad something is taken to be, and urgency, the pull to act. They often rise together, but they come apart: someone you love matters enormously and may exert no pull at all when nothing is wrong, while an itch pulls hard on something that hardly matters. The question an organism answers is not "what is most valuable?" but "what is urgent, right now?"
The second is between reflexive and deliberative settlement. In reflex, tension resolves directly into action. In deliberation, an urgent demand is held unresolved while ways of meeting it are weighed against it -- carrying a scorching dish to the table instead of dropping it, or deciding what to do about hunger in the middle of a meeting.
Two claims, with different standing
Claim A, a structural result. Any account on which prioritization produces a score that something separate must read faces a regress: the reader's rule is either fixed, so a demand's significance for action was settled before the conflict it is meant to settle, or adaptive, so it needs a further evaluator, and the question recurs. Systems with no separate reader escape, but only by settling reflexively. Planners deliberate, but their scores are read. What both deliberates and ends the regress is a held pull: urgency that persists undischarged, whose change as each strategy is rehearsed is that strategy's evaluation. The paper gives a test for when a system has a separate reader, and applies it to reinforcement learning, multi-objective agents and active inference, and to its own account.
Claim B, a proposal. The paper proposes, as an identity thesis and not a deduction, that the system-wide state of the field in which such pulls are held is identical with consciousness. A reader who accepts Claim A and rejects Claim B has understood the paper and disagreed with it in the right place.
How it could fail
Each claim is given a condition under which it fails. The regress fails by counterexample: a working system that deliberates with a separate reader and still lets the present conflict settle what its values mean for action, or a reader's rule that is neither fixed nor adaptive. The account fails for human deliberation if a neural gain, dopaminergic signaling being the obvious candidate, proves separable from valuation, changing which option governs while leaving every option's evaluation unchanged. The identity fails if, in a system that deliberates without a separate reader, a demand governs deliberation and nothing about it or its resolution is felt, tested by felt relief when the demand is resolved, since a demand can be felt without being reported. The paper also states a signed dual-task prediction that has not been run.