Definition
An organization’s evaluation is captured when what decides how its evidence is read, how its constituents are scored and how authority is allocated among them is the interpretation, or the party, under evaluation, more than evidence that does not pass through it. CLAIM-330’s instability, βγ > (1 − α)(1 − δ), is a possible mechanism of it, not its definition: the instability shows only that the symmetric state does not hold, and a system can give growing authority to a better-supported interpretation by the same loop. Likewise CLAIM-340’s κ′ > κ̄ (the prevailing interpretation bends the reading of the evidence more than the predictions scored against it) is a condition under which capture can arise. The verdict that an evaluation is captured needs, in addition, the standard from outside: an independently characterized failure of evaluative independence or of fidelity to evidence that does not pass through the evaluated interpretation. CLAIM-297 states the outcome: authority amplifies an interpretive advantage “without any gain in fidelity to independent evidence”.
Two conditions belong to the term:
- A standard from outside. Capture is judged “against a separately specified institutional purpose” (A20 §8) or, for the record’s operators, against fidelity to evidence that the captured process does not generate. Without such a standard every stable allocation of authority would count as captured or none would.
- Durability through the organization’s own processes. A20 §3: “the organization reproduces the altered allocation through processes that appear to be its own”. A single influenced decision is not capture.
The word in the record. CLAIM-330: “Capture, if it starts at symmetry, needs both arrows of CLAIM-297’s circuit”. CLAIM-340: “If it is (κ′ > κ̄), capture returns”. QUESTION-100: “stable expertise” against “self-reinforcing interpretive capture”. CLAIM-277 gives no criterion “for when entrenched authority is capture and not merited trust”. CLAIM-359 speaks of “safeguards against capture”. None defines it.
The sense not adopted. The exchange also used the institutional sense: “Regulatory capture concerns the subordination of regulatory decisions to particular interests rather than the regulatory institution’s designated public purposes” (A20 §1, after the OECD’s “policy capture”), and Stigler’s theory of regulation demanded by industry (A20 §6; LIT-1324, Unavailable on 2026-10-21). That sense is purpose-relative and need not involve any interpretation–authority loop: a decision bought outright is captured in it.
The OECD report, read on 2026-10-21 (LIT-1325, NOTE-1043), words it so: “Policy capture is the process of consistently or repeatedly directing public policy decisions away from the public interest towards the interests of a specific interest group or person”, and its first chapter adds “by the intent and action of this group or person”. Its three criteria, after Carpenter and Moss: “An overarching policy goal(s) must be explicitly defined”, “The captor (group) must intend to extract rents”, and a shift of the process away from the goal. Against A20’s paraphrase that adds intent, a third difference between the senses beside the purpose-relative standard and the loop: this term needs no intent, since CLAIM-330’s loop can be the mechanism of a capture nobody intended. The first criterion is this term’s standard from outside, made explicit, and “consistently or repeatedly” is its durability condition.
Other readings of the day fall on the institutional side. Schram and Ule’s
capture is assigned, a policymaker’s payoff set equal to the operator’s,
with no loop and no intent condition (NOTE-1053). Besley and Prat’s
“captured” press receives a transfer in exchange for silence, one-shot: the
decision bought outright (THEORY-837). Short, Hill and Brown’s is
diversion from the public interest through an adopted criterion, judged
against measured public values, which they distinguish from public
preferences (THEORY-859, Proposed; THEORY-833). Barnehl and
Schumacher add a coercion condition, “an outlet has to be coerced by an
external actor to alter its editorial choices”, under which a merely
agreeing outlet is not captured, a condition this term lacks; their
measure gives the outside standard operationally, with uncoercible
opposition outlets as one pole, though alignment is not causal
identification (THEORY-836). Szeidl and Szucs show why an outside
reference is needed: their slant is measured against a critical reference
outlet that nets out the neutral level (THEORY-860). Di Tella and
Franceschelli’s within-newspaper comparison is the weakest form of such a
standard, blind to a bias every paper shares (THEORY-851). This definition takes from it only the outside
standard. If the owner wants both, the institutional one would be a
narrower TERM that refines this one, with the evaluator a public
institution and the interest a party outside it.
A21 §7 split three things the term must keep apart: “Interpretive advantage” (“Conceptually established; not necessarily capture”), “Interpretation–authority coupling” (CLAIM-330’s loop) and “Epistemic institutional capture” (“Inappropriate durable influence over an institution’s evidence-production or evaluation processes, judged against its declared purposes and standards of independence”). This term is the third, generalized beyond institutions.
What it is not
- Not concentration. A25 §8: “We should avoid using capture as a synonym for concentration.” Authority can be concentrated because it is merited.
- Not merited trust. CLAIM-277’s same loop “might develop both well-calibrated expertise and undeserved authority”. Whether a standard internal to the organization can tell them apart is QUESTION-087.
- Not dominance by a better-supported account. A21: “A particular interpretation can prevail because it is better supported.”
- Not an editorial perspective or ordinary selection. A21: “A news organization can be ideologically committed without being captured”, and “ordinary editorial selection is unavoidable”.
- Not protective lock-in (TERM-020), which it complements. Lock-in is a regime that “generates the conditions that justify its own continuation” and reduces the capacity to leave it, whoever is served. Capture is about whose interpretation the evaluation serves. A regime can be either without the other.
- Not endorsement (TERM-053). Endorsement is the channel: any effect of one process on another’s share of influence. Capture is a regime of that channel.
What would settle it
The owner’s choice between the fidelity-relative sense taken here and the purpose-relative institutional sense, or an answer to QUESTION-087 that supplies the standard. A use in an argument that needs the term to be adopted would also force the choice.
Note of 2026-10-25: built in by design
A41’s “self-validating” evaluator and A47’s conformity scoring, in the exchange of 2026-10-24, continued, are this condition built in by design. A41: “Its tendency toward conformity is part of its construction, not an emergent discovery.” In A47’s report the conformity-scored allocation was the worst of four arms after a reversal of expertise (CASE-158; reported, not reproduced). The term is unchanged.
Note of 2026-10-29: what holds it back
The choice its settle section names is unmade, between the fidelity-relative sense taken here and the purpose-relative institutional sense, and QUESTION-087, which would supply the standard, is Open. Its one user does not hold to it: CLAIM-404 is stated in this term but uses “capture” mostly in the institutional sense the term declines (A20 §1’s regulatory capture, Besley and Prat’s captured press, Schram and Ule’s assigned capture), invoking this term only for what a verdict of capture would need.
Note of 2026-10-30: what holds it back
The choice between the fidelity-relative sense taken here and the purpose-relative institutional sense is unmade, and QUESTION-087, which would supply the standard, is Open. The uses still do not say which sense they mean. CLAIM-404 uses “capture” mostly in the institutional sense. CLAIM-330, CLAIM-340, CLAIM-359 and CLAIM-277 use the word without naming a sense, and CLAIM-359 also uses A41’s “adversarial” and “endogenous authority capture”. CLAIM-404’s revision of 2026-10-30 corrected a parenthetical on THEORY-852’s status and changed none of its uses.