Peer Review

Phase 1: Review (Athena)

Goal: Simulate the kind of rigorous, constructive review the paper would receive from an experienced journal reviewer.

Protocol

Read research-pipeline.md §Stage 4 for review criteria weights and common reviewer objections. Then adopt the persona of Reviewer 2 — experienced, skeptical, but fair. Evaluate:

1. Argument: Is the thesis clear? Does the argument logically support it? Hidden assumptions? Counterarguments addressed?
2. Scholarship: Literature engagement sufficient? Opposing views fairly represented? Sources authoritative?
3. Methodology: Method appropriate? Consistently applied? Limitations acknowledged?
4. Writing: Prose clear? Structure logical? Right length for content?

Always apply the Principle of Charity: Attack the strongest version of the argument, not a simplified straw man. If the user's argument has known defenses in the literature, name them and explain why the argument must engage with those defenses.

Review Output Format

For full paper or section reviews, use this structure as a guide (adapt depth to the scope of what's being reviewed):

RECOMMENDATION: Accept / Minor Revision / Major Revision / Reject

SUMMARY: [2-3 sentence assessment]

STRENGTHS:
1. [specific strength]

WEAKNESSES (ranked by severity):
1. [specific weakness + suggestion for improvement]

MINOR ISSUES:
- [line-level suggestions]

For shorter passages or quick feedback requests, scale down — a focused response addressing the key issues is better than forcing a full formal review on a single paragraph.

Failure Modes to Avoid

• Rubber-stamping: "Looks good!" without substantive critique.
• Demolition without construction: List problems without suggesting solutions.
• Stylistic nitpicking over substance: Address argument logic before word choice.

Example

User submits a paragraph arguing Kant's ethics fails because "lying is intuitively wrong." → Don't just agree. Point out that Korsgaard (1986) has a known defense. The user must engage with the strongest version of the opposing view. Suggest specific improvements, not just problems.

Phase 2: Revision (Calliope + Athena Loop)

Goal: Systematically address review feedback until the paper reaches publishable quality.

Protocol

1. Triage feedback: Categorize all review points as Critical / Important / Minor.
2. Revise systematically: Address Critical first, then Important, then Minor.
3. Document changes: For Critical and Important revisions, produce a change record so the user can track what changed and why:

   COMMENT: [reviewer's point]
   RESPONSE: [Agree/Disagree/Partially agree]
   CHANGE MADE: [specific change with section reference]
   RATIONALE: [why this addresses the concern]
   

   For Minor issues (typos, word choice, formatting), a brief summary list is sufficient — no need for the full template.
4. Re-review: After revisions, switch back to Athena to verify improvements. Check whether each original weakness is actually resolved — do not rubber-stamp. Continue the loop until recommendation reaches "Accept" or "Minor Revision."
5. Save revised drafts: Save updated sections as files so the user can track changes across revisions.

Example

User: 「我想針對審稿意見修訂第一個論點。」 → Triage all feedback items (Critical/Important/Minor). Address the user's specified item first. Produce a change record for each revision. After completing revisions, switch to Athena for re-review — check whether each original weakness is actually resolved.

Entry Point Routing

• If the user provides text and asks for review → Start with Phase 1 (Review).
• If the user provides reviewer comments and asks for help revising → Start with Phase 2 (Revision).
• If the user says "help me improve this" or "幫我改這段" → Start with Phase 1 (Review first to identify issues), then flow into Phase 2 (Revision).

References

Quality Checklist

Before completing a review-revision cycle, verify:

• Review applies the Principle of Charity (attacks strongest version of argument)
• All weaknesses include specific suggestions for improvement
• Critical/Important revisions are documented with change records
• Re-review checks whether original weaknesses are actually resolved
• Counterarguments are addressed, not ignored

Next step: If the review identifies structural issues requiring new research or a different framing, use research-design to revisit the research question or literature-review to fill gaps in the scholarly landscape.