The case
Two hypotheses with comparable evidence in a fixed corpus; two operators, one better at finding evidence bearing on each; queries conditioned on the currently dominant hypothesis; endorsement updated from apparent retrieval success, with an adjustable memory timescale.
Regime A scores operators against the current dominant interpretation. Regime B scores them by independently assessed relevance and role (CLAIM-281). Controls (A23 XI.B): initial framing asymmetry, coupling strength, exploration rate, endorsement timescale, independence of evaluation, role specificity.
Measures (A22 §7): meaning concentration, influence concentration, evidential fidelity, hysteresis. Analysis targets: fixed points, stability of the symmetric state, multiple attractors, transitions between adaptive and captured regimes.
What each outcome would show
A23 XI lists five candidate predictions, among them amplification of small framing asymmetries relative to a fixed-influence model, and influence concentrating while independently scored fidelity stays flat. Its failure criteria: “If the second influence field yields no distinctive predictions beyond a standard adaptive-retrieval process, the additional organizational structure is not explanatory.”
Main threat
CLAIM-288: the operators’ objectives and the score of successful inquiry are the modeller’s, so the model cannot show self-constitution, only dynamics. A23 XIV.F names the objection. Ablations against fixed weights and against a matched single policy are what separate the coupling from ordinary adaptive retrieval.
Note of 2026-10-16: the linearization, and the analyses proposed for it
A14 of the exchange of 2026-10-16 on sports fandom (workbench entry of 2026-10-16) read this case correctly and linearized it at the symmetric state. The checked result is CASE-099 and CLAIM-330: with each field stable alone, only the two-way loop can destabilize symmetry, at βγ > (1 − α)(1 − δ), and the growing direction mixes meaning and authority. Regime B, which sets γ near zero, removes that instability.
A14 §9 proposed six analyses beside the four measures above: the eigenmodes of the joint dynamics against ablations of each arrow; effective information of macrovariables spanning both fields, against either field alone and a matched full-state model (THEORY-222); predictive closure (THEORY-257); persistence under periodic replacement of operators; role-indexed against scalar endorsement (CLAIM-281); and ablation against simpler models (CLAIM-278). A15 named the simpler models: reflected appraisal (DeGroot–Friedkin), adaptive networks and persuasion bias, as the baseline before any claim of novelty (QUESTION-117). They are recorded here as proposals; the case is unchanged.
Note of 2026-10-23: what Friedkin’s model already shows
A5 of the exchange of 2026-10-23, “Framework Overview Request” cited Friedkin’s reflected-appraisal theory (LIT-1404, registered unread) and its DeGroot–Friedkin analysis (LIT-1146, unread): members appraise one another’s control over prior outcomes, and that appraisal reshapes the influence network, which by the abstract “generates transitions of the influence network toward concentrated power”. A5: “effective authority need not be imposed once and held fixed”. What stays for this model is whether interpretive authority and practical influence form a coupled system, which reflected appraisal does not by itself test; A5 asked that the simulation be compared with DeGroot–Friedkin, adaptive-network and recommender models, and not only reproduce CLAIM-330’s instability. CASE-144 folds this model into a larger design.
Note of 2026-10-29: two reflected-appraisal baselines, not one
Friedkin’s reflected-appraisal theory, LIT-1404, which the note of 2026-10-23 calls registered unread, is now read (NOTE-1374), and the DeGroot–Friedkin analysis, LIT-1146, was read on 2026-10-24 (NOTE-1135). So “unread” no longer holds for either. The note treats them as one baseline. They are two, with different predictions.
- Friedkin’s rule, with the Friedkin–Johnsen single-issue step that anchors members on their initial positions, concentrates power almost always. Re-run as printed on 300 random complete networks each of sizes 5, 10 and 20, every one ended with one self-weight above 0.99.
- The DeGroot–Friedkin update, with DeGroot consensus on each issue, kept power spread on random complete networks of sizes 5 and 10: median largest self-weight 0.29 and 0.13, never above 0.40.
The comparison is filed as THEORY-1161, Proposed. A comparison of this model with “DeGroot–Friedkin” should say which single-issue step it is compared with. Friedkin’s paper is also a worked case of CLAIM-297’s shape, authority on the next round set by success on the last, in which that update concentrates authority. The design is unchanged.