nucleation

ARG-026: Judged pragmatic observables that predict human judgements of fidelity or shared kind better than text similarity support the central theses of the pragmatic-transport line

3 min read

Properties8
StatusRejected
Formevidence
Tagsphilosophy-of-science, psychometrics, translation, philosophy-of-language
Linepragmatic-transport
GroundsCASE-184 — The central defeat test as stated: human judgements of fidelity or shared kind, predicted by the theory's judged measures against a text-similarity baseline
ConcludesCLAIM-115 — What survives translation is the set of communicative distinctions and relations a receiving system can still reconstruct and act upon, not the wording or the proposition, CLAIM-139 — A communicative kind's identity is grounded in its relational organization rather than in any expression or encoding; statistical methods recover aspects of that organization from restricted observations, translation transports it, and fidelity is which independently identifiable, decision-relevant relational distinctions persist, CLAIM-005 — Pragmatic fidelity is the match between distributions of context-indexed communicative judgements across a correspondence of source and target contexts, not the similarity of one canonical meaning vector; the match constrains a translation only once one outcome correspondence is fixed across many items, and summed over source contexts alone it charges for loss and not for addition, CLAIM-078 — The theory succeeds if its observational and decision measures predict changes of communicative identity that text similarity misses and follow its compositional predictions across reconstructions; it needs neither quantum contextuality nor exact Noether laws
Cited as “undercuts” byCLAIM-232 — Human ratings of stance or footing and human ratings of fidelity of the same renderings share rater construal, scale use and a general impression, which a machine similarity score shares with neither, CLAIM-233 — The central theses' defeat conditions pit them only against embedding, propositional or surface similarity, and two rivals the record holds, functional equivalence as Nida states it and a proposition with its illocutionary force, also predict that judged pragmatic observables beat text similarity
Rendered intodocs/arguments-by-conclusion.md

Graph View

The inference

CASE-184 is the test the four defeat conditions state: human judgements of fidelity or of shared kind, predicted by the theory’s measures and by a text-similarity score. CLAIM-115, CLAIM-139, CLAIM-005 and CLAIM-078 are each defeated if text similarity predicts the criterion as well as the theory’s measures do. If the theory’s measures win, the theses survive, and the survival is taken as support.

Critical questions

  • Would the evidence be expected if the hypothesis were false? Yes. A pass is what every rival account predicts (the next question), and the predictor shares a method with the criterion (the third).
  • Does it tell the hypothesis apart from its rivals? No. CLAIM-233: the defeat conditions “pit them only against embedding, propositional or surface similarity, which no rival account holds, so surviving them would not tell the thesis from the functional and relevance-theoretic accounts the record says it must be told apart from”. CLAIM-023 says the contribution is “not that translation is pragmatic or that speech acts change contexts, but a specified, testable account of which communicative distinctions a translation preserves”. A functional-equivalence account, Gutt’s interpretive resemblance and plain speech-act pragmatics all predict that ratings of stance and footing beat embedding similarity on renderings built to dissociate the two. So does the rival CLAIM-115 itself names, a global object enriched with force. The line’s distinctive machinery (CLAIM-105, CLAIM-050, CLAIM-117, CLAIM-056) has defeat conditions of its own, and the thesis-level ones depend on none of them.
  • Does the measure measure what it is said to? No. CLAIM-232: “Predictor and criterion in the central defeat tests are both human pragmatic judgements of the same renderings while the baseline is a machine score, so the theory’s measures can win by shared method variance alone.” Two human judgements of one rendering share the rater’s construal, scale use, a general impression and the context of elicitation, and an embedding score shares none of that. “The theory’s measures can beat the baseline without carrying any information the baseline lacks about what the renderings preserve.” The manuscript’s split annotators (§10, as QUESTION-005 reports) remove the shared rater, not the shared construal.

Rejected as an inference to the four theses. It does not say the theses are false, and beating text similarity is the right first test: failing it would defeat them. CLAIM-232 and CLAIM-233 undercut this argument with what they say about the subject matter: what the rival accounts predict, and what two human ratings share.

The weakness was seen once before, for an earlier thesis: CLAIM-083 says the manuscript’s Case I criterion “is weighed against embedding similarity, not against static pragmatic similarity”, and CASE-005’s three-way selection design had the stronger comparison.

The repair

The record conceded both objections on 2026-10-10. CASE-055 restores the three-way selection with human receivers, adds a proposition-plus-force rival, a behavioural criterion and predictors rated by other raters. The revised theses CLAIM-220, CLAIM-202 and CLAIM-204 are stated against it. QUESTION-021’s latent act variable, “a latent act variable that adds nothing once the relational observables are measured”, is the test the record names that would discriminate. CLAIM-115’s and CLAIM-139’s own defeat conditions are unchanged.

Filed 2026-10-28 by the ADR-061 migration, from CLAIM-232 and CLAIM-233.

Note of 2026-10-30: one rival moved

CLAIM-233 was narrowed at the nineteenth status review. Gutt’s interpretive resemblance, named above among the rivals that predict a pass, is now CLAIM-501, Proposed, since Gutt is unread (LIT-784). The critical question still fails: functional equivalence as Nida states it (CASE-056) and the proposition-plus-force rival both predict a pass, and CLAIM-232 is untouched. The argument stays Rejected.

Backlinks