bias-assessor

Bias Assessor

Triggers & routing

• Trigger: bias, risk-of-bias, RoB, evidence quality, bias assessment, evidence quality.
• Use when: evidence-review generated papers/extraction_table.csv, bias/quality fields need to be completed before synthesis.
• Skip if: not an evidence/systematic review, or there is not yet papers/extraction_table.csv。
• Network: none.
• Guardrail: use a simple verifiable scale (low/unclear/high) + brief notes; keep fields consistent.

Enriches the extraction table with lightweight risk-of-bias fields for evidence-review.

Input

• papers/extraction_table.csv

Output

• updated papers/extraction_table.csv

Contract

The table should expose stable RoB columns:

• rob_selection
• rob_measurement
• rob_confounding
• rob_reporting
• rob_overall
• rob_notes

Allowed values are fixed:

• low
• unclear
• high

Script boundary

scripts/run.py should:

• add missing RoB columns
• assign a conservative overall score
• keep notes brief

Acceptance

• RoB columns exist
• all included rows have values
• values stay within the fixed scale

Non-goals

• deep methodological critique
• rewriting extraction contents