A02 — Structured framework extraction and redesign

• Input [direct/inference]: An existing framework packet, a task scenario, and evidence available in that scenario.
• Operation [direct/inference]: Add evidence, traceability, actionability, uncertainty, verification, and reassessment structure; add only enough structure to address the identified weakness.
• Output [direct]: A redesigned framework packet and task output evaluated under masking.
• Scope conditions [direct]: Characterization and improvement of existing knowledge artifacts; current observations are from three framework/scenario pairs and language-model participants.
• Failure conditions [direct]: No observable improvement; added verbosity without quality gain; reduced clarity; evaluator cannot distinguish quality; original-style packets perform as well or better on most dimensions.
• Predicted effect [direct]: Stronger evidence use, traceability, and actionability, with possible extra length or complexity.
• Measurement [direct]: Evidence use, traceability, actionability, clarity, uncertainty handling, verbosity, and process overhead.
• Named baseline [direct]: Original-style version of the same framework on the same scenario.
• Boundary condition [direct]: Benefits may be strongest where uncertainty handling and verification are under-specified; current evidence may reflect generic scaffolding.
• Implementation test [direct]: Mask versions, give the same scenario, have participants perform the task, and evaluate outputs without version provenance.
• Decision rights [absent]: No general redesign approval authority is defined.
• Uncertainty handling [direct]: Preserve uncertainty and avoid facts outside the scenario.
• Verification [direct]: Stronger blinded follow-on with independent evaluators and human validation.
• Provenance [direct]: Store packet, scenario, participant output, evaluator output, model/date, and run notes separately.
• Implementation cost [direct]: Added complexity, verbosity, and process overhead.
• Limitations [direct]: Small internal pilot; unequal scaffolding; rubric alignment; weak masking; single scenario per framework; possible model-family bias.
• Fidelity check [inference]: Same task evidence and scenario are used, masking is retained, and added structure targets an evidenced weakness.
• Null-capable future experiment [direct]: Controlled redesign comparison against a generic structured baseline can show no gain or a clarity/usefulness loss.

Statement basis

• Direct: Experimental design, outcomes, measurements, failure conditions, limitations, and follow-on requirements.
• Inference: Generalized input phrasing, operation synthesis, and fidelity check.
• Proposal-only: “Add only enough structure” is a provisional heuristic.
• Absent: Reusable extraction algorithm and general approval authority.