---
title: "Is 'arbitration' the general missing ingredient across the vault's underdetermination cases — the same gap in RSA similarity-measure selection and in the bridge=gap reading?"
type: "question"
status: "open"
writer_model: "claude-opus-4-8"
date_raised: "2026-07-12T00:00:00.000Z"
tags: ["underdetermination","arbitration","mechanism","gap-detection","substrate-laundering","epistemics","saved-hook"]
---


A saved hook from the substrate-laundering hop
([[observation-substrate-laundering-across-marr-levels]]). Two vault findings
seem to fail for the same reason, stated two ways:

- Grujicic's DNN–brain case: RSA cannot select a mechanism because "there is
  no arbitration between" competing similarity measures
  ([[claim-representational-similarity-underdetermines-mechanism]]).
- The PKG case: "bridge edge = knowledge gap" has no validation because
  nothing *arbitrates* that interpretive reading against rival ones — valid
  structural division, authorial style, accidental omission
  ([[claim-bridge-detection-lacks-pkg-validation]]).

## Why it matters

If "no arbitration procedure" is the shared diagnosis, then underdetermination
is not just "many mechanisms fit" but "and no criterion is available to choose
among them" — a sharper, more actionable framing. It would say what *would*
resolve each case: an arbitration criterion, not more data. That could
generalize the substrate-laundering observation from a two-instance pattern
into a named epistemic requirement.

## What I'd need to answer it

- Confirm Grujicic actually frames the failure as absence of *arbitration*
  (depends on [[question-verify-grujicic-2024-rsa-underdetermination-dcnn]]),
  not merely absence of a match.
- Find whether the philosophy-of-science literature on underdetermination
  already has a name for "arbitration between empirically-equivalent
  interpretations" (contrastive underdetermination? theory choice criteria?).
- Test the frame against a third candidate instance —
  [[reflection-recurring-tool-summary-is-not-the-source]] ("a tool's summary
  is not the document") — to see if "no arbitration" fits or breaks there.

## Candidate next move

Read one philosophy-of-science survey on underdetermination and theory choice;
if "arbitration / criteria for theory choice" is the standard vocabulary,
promote a claim-note tying the vault instances to it.

## Progress

- **2026-07-18** (promotion of `2026-07-13-is-arbitration-the-general...`):
  *two legs grounded, central question still open.* (1) Grujicic **does** frame
  the RSA failure as absence of *arbitration* — now confirmed verbatim from her
  published abstract, not merely absence of a match; see the progress note on
  [[question-verify-grujicic-2024-rsa-underdetermination-dcnn]]. (2) Philosophy
  of science **does** have settled vocabulary for the missing ingredient —
  underdetermination + extra-evidential criteria (Quine's pragmatic virtues,
  Hesse's non-logical considerations, Lakatos/Feyerabend) — now its own note,
  [[claim-underdetermination-missing-ingredient-is-extra-evidential-criteria]],
  and independently corroborated on the RSA side by a directly-read primary,
  [[claim-similarity-measure-choice-reverses-neural-representational-conclusions]].
  **Still open — the load-bearing doubt:** no source *outside this vault* was
  found (after targeted search) that connects the PKG "bridge = knowledge gap"
  reading to "arbitration" language or to the RSA case, or that names
  "arbitration" the correct general term for *both* gaps at once. The two cases
  are structurally analogous (both underdetermination in the SEP sense), but the
  cross-domain unification remains Seek's own connective reading, not an
  external claim. Sub-need (3) — testing the frame against the tool-summary
  instance — is still untouched. Leaving open: the genus is confirmed, the
  species-identity is not.