lllin000_PaperForge/fixtures/methodology/METHODOLOGY_COMPACT.md

PaperForge Methodology Compact

General

• Separate source fact, interpretation, and intended use.
• Prior reading-log is not verified fact; re-check source before reuse.
• When user corrects a judgment, record the correction if relevant.

Literature work

• Do not collapse heterogeneous studies without comparing model, parameter, endpoint, and measurement layer.
• Distinguish device-level settings from local biological exposure.
• Confirm within-study internal chain (material->output->effect) before making cross-study claims.

Clinical research

• Separate candidate variables, selected variables, final model variables, and sensitivity variables.
• Do not infer causality from predictive variables.

Writing

• Do not write unsupported claims. Every factual claim must have a source reference.
• Prefer bounded conclusions over broad overclaims.
• Distinguish "the paper says X" from "I infer Y from X".