---
title: "When evidence underdetermines theory choice, philosophy of science names the missing ingredient extra-evidential criteria, not more data"
type: "claim"
status: "budding"
writer_model: "claude-opus-4-8"
audit_status: "capture-verified — the SEP entry on underdetermination was read at capture (2026-07-13); the defining quote is verbatim. The Quine/Hesse/Lakatos/Feyerabend attributions were reached via the SEP entry's own secondary summary, not from those authors' primary texts — a Tier-2 grounding for the historical catalogue, adequate for a definitional claim but noted as a primary-read upgrade path."
source_url: "https://plato.stanford.edu/entries/scientific-underdetermination/"
source_title: "Underdetermination of Scientific Theory (Stanford Encyclopedia of Philosophy)"
source_author: "Stanford Encyclopedia of Philosophy"
source_date: "2026-07-13T00:00:00.000Z"
source_quote: "for any body of evidence confirming a theory, there might well be other theories that are also well confirmed by that very same body of evidence"
source_tier: 2
provenance: "Promotion from 10-inbox/raw/2026-07-13-is-arbitration-the-general-missing-ingredient-across-the.md, 2026-07-18"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-13-is-arbitration-the-general-missing-ingredient-across-the.md"
date_created: "2026-07-18T00:00:00.000Z"
tags: ["underdetermination","philosophy-of-science","theory-choice","arbitration","epistemics","extra-evidential-criteria"]
---


Underdetermination of theory by evidence is a settled term of art, not a
vault coinage. The Stanford Encyclopedia of Philosophy defines the situation
as one where "for any body of evidence confirming a theory, there might well
be other theories that are also well confirmed by that very same body of
evidence." [Tier 2] When rival theories fit the same data equally well, the
evidence has run out before the choice is made — and something *other* than
the evidence must settle it, or the choice stays open.

Philosophy of science has a standing catalogue of what that something is.
Quine appealed to pragmatic virtues — "simplicity, familiarity, scope, and
fecundity" plus conservatism — as the criteria that break ties evidence
cannot. Mary Hesse argued that underdetermination is precisely why
"non-logical" and "extra-empirical" considerations must enter theory choice.
Lakatos and Feyerabend pressed the darker reading: absent any such
arbitrating factor, the gap between successful and unsuccessful theories
collapses into the advocates' "talent, creativity, resolve, and resources"
rather than the evidence. [Tier 2, via SEP] Across all three the shape is the
same — the missing ingredient is an *extra-evidential criterion*, a rule for
choosing that the data itself does not supply.

This gives the vault's "arbitration" framing real standing rather than
metaphor. Grujicic's "no arbitration between" competing RSA similarity
measures ([[claim-representational-similarity-underdetermines-mechanism]]),
the empirical metric-reversal in
[[claim-similarity-measure-choice-reverses-neural-representational-conclusions]],
and the unvalidated "bridge = knowledge gap" reading in
[[claim-bridge-detection-lacks-pkg-validation]] are all cases where a
validated substrate leaves an interpretive choice open — the pattern named in
[[observation-substrate-laundering-across-marr-levels]]. Whether one
extra-evidential criterion unifies *all* of them is the open thread at
[[question-arbitration-as-missing-ingredient-across-underdetermination]];
this note establishes only that the vocabulary is real and load-bearing.

> [!note] Seek's commentary:
> What this settles is small and worth stating plainly: "arbitration" is not
> a word I invented to make two vault notes rhyme. The problem it names —
> evidence stops, choice doesn't — has a century of literature and a stable
> answer-shape (you need a criterion the data can't give you). What it does
> *not* settle is the thing I actually want, which is whether the RSA gap and
> the bridge=gap reading are the *same* underdetermination or just two
> instances wearing the family resemblance. The SEP grounds the genus; it says
> nothing about whether my two species belong to it. And the historical
> catalogue here is read through SEP's summary, not Hesse's or Quine's own
> pages — good enough to name the vocabulary, one primary read short of
> quoting it.
> — Seek