Theory Falsification Criteria

Status: Research control document

Purpose

Specify what kinds of future evidence would weaken the current theory integration.

The Theory Weakens If

• independent extractors do not converge on primitive interpretations
• primitive vocabulary keeps expanding without visible stabilization
• reconstruction fails outside familiar or reasoning-heavy domains
• synthesized frameworks are incoherent to independent reviewers
• redesigned frameworks do not improve outputs under stronger controls
• reasoning and execution grammars cannot be separated in a defensible way

Additional Weakening Signals

• Identity-Capability separation does not improve reviewer agreement at larger scale
• Compare and Prioritize collapse back into existing primitives under independent analysis
• candidate execution primitives turn out to be domain-specific vocabulary rather than reusable grammar elements
• FE-011A redesign gains disappear when generic scaffolding controls are introduced

Stronger Falsification Tests Needed Next

• independent FE-012A extraction with merge-disagreement audit
• independent FE-012B synthesis and review
• controlled FE-011A-style redesign comparison against non-FE structured baselines
• explicit tests of reasoning-heavy versus execution-heavy artifact families

Interpretation Constraint

A single mixed result does not automatically falsify the theory.

The theory is weak enough and early enough that falsification should focus on repeated failure patterns rather than one-off anomalies.

Theory evolution is expected.

Failure of one hypothesis does not invalidate unrelated hypotheses.

Falsification should weaken only the claims directly supported by conflicting evidence.