hypothesis

H017 Review

H017 Review

  • Hypothesis ID: H017
  • Name: Procedural Compression
  • Current statement: Different surface representations may compress into a stable recovered procedural structure.
  • Research question: Do representation variants compress into the same recovered structure strongly enough to be meaningful?
  • Prior direction and strength: active / low
  • Prior kill-condition status: not_tested

Evidence Supporting

  • EXP-002 completed all expected records and maintained full representation compliance. Source: research/evidence-runs/ECR-000003-representation-sensitivity/experiments/EXP-002-cross-representation-stability/comparison/generated-v3.1/exp-002-comparison-summary.md
  • Explainability reported explicit compressed-step and expanded-step patterns rather than only flat disagreement. Source: research/evidence-runs/ECR-000003-representation-sensitivity/comparison/ECR-000003-explainability-summary.md

Evidence Challenging

  • EXP-002 still reported mixed backbone and disagreement in primitive and constraint layers. Source: research/evidence-runs/ECR-000003-representation-sensitivity/experiments/EXP-002-cross-representation-stability/comparison/generated-v3.1/exp-002-comparison-summary.md

Mixed Evidence

  • Compression-like behavior is visible, but not enough to claim that multiple representations reliably collapse into one stable structure.

Provider, Representation, and Domain Effects

  • Representation family matters.
  • Some disagreement may reflect optional elaboration or summarization rather than full structural divergence.

Relevant Instrument Findings

  • Backbone measures appear more stable than primitive or constraint measures, which is consistent with a limited compression story.

Threats

  • representation-format effects
  • prompt-induced structure
  • missing human comparator

Alternative Explanations

  • Apparent compression may be produced by comparator abstraction rather than by the models recovering the same structure.

Kill Condition

  • Current kill condition: narrative, bullet, graph, and transition-table representations produce materially unrelated recovered structures.
  • Status: partially_met

Proposed Direction

slightly_supported

Proposed Evidence Strength

weak

Board Recommendation

retain

Rationale

The evidence is more supportive than neutral, but still too limited for a broader compression claim.

Next Evidence Needed

  • negative non-isomorphic controls
  • human extraction comparisons
  • constraint-layer calibration

Human Decision

  • Decision:
  • Reviewer:
  • Date: