A claim's warrant does not determine its fit in a working theory

Type: kb/types/note.md · Tags: learning-theory, evaluation

A claim used in a working theory faces at least two different evaluations. One asks whether the claim is true, formally valid, source-supported, or otherwise warranted over its stated scope. The other asks whether the claim belongs in this working theory in a form that improves explanation, search, prediction, modification, or revision. Evidence for one question does not settle the other.

A warranted claim may fit poorly. It can be irrelevant to the theory's purpose, duplicate a claim already present, state the right fact at the wrong abstraction level, use a scope that does not match the decisions the theory must guide, or add detail that increases maintenance without changing any relevant inference. Independent warrant does not imply that retaining the claim improves the working theory.

Apparent fit does not imply warrant either. A false, overgeneral, or weakly supported claim may integrate smoothly because the current implementation, vocabulary, or neighboring claims already assume it. It may even improve local search or repair while encoding a misconception that later evidence exposes. System coherence around a claim can therefore be evidence about integration without being independent evidence that the claim is warranted.

The distinction also changes how contradiction should be interpreted. When a warranted new claim conflicts with the working theory, rejecting the claim is not the only possible repair. The contradiction may identify a defect in another claim, a scope boundary, a bad decomposition, or a larger theory that needs revision. Theory fit is therefore relational: it concerns how warranted claims compose into an operative explanatory structure, not whether each claim passes its independent warrant test in isolation.

The evaluations can inform but not substitute for each other

Fit evidence can expose warrant problems. A claim that repeatedly fails to predict, guide recovery, transfer, or survive later demands may deserve renewed factual or formal scrutiny. Warrant evidence can likewise force a fit revision: a newly established result may require restructuring the working theory even when that restructuring initially makes the theory less convenient.

The prohibition is substitution. "This claim helped the system" cannot silently become evidence that the claim is true, and "this claim is true" cannot silently become evidence that it deserves a retained role in this theory. The two evaluations may exchange evidence while retaining different questions and failure conditions.

Scope

  • Warrant is used broadly here because claims in the knowledge base may be empirical, formal, source-derived, or procedural. Literal factual truth is not the only relevant independent standard.
  • Fit is intentionally not given a complete scalar definition. Relevance, non-redundancy, scope, abstraction level, explanatory integration, and operational usefulness may contribute differently in different working theories.
  • Objectives, commitments, and grants of authority can affect whether a claim belongs in an operative theory without themselves being settled by empirical truth tests.
  • The claim does not imply that warrant and fit are statistically independent. It says neither evaluation determines the other.

Relevant Notes: