Write Discussion And Conclusion

If information is missing, make the minimum reasonable assumptions and label them. Do not invent contributions, managerial implications, or stronger claims than the findings can support.

Router

Resolve the route in this order: task -> section -> output.

Task

• discussion mode: write or revise the main discussion logic
• contribution mode: articulate theoretical or methodological contribution
• implications mode: write practical, policy, or broader implications
• limitations mode: write limitations and future research directions
• conclusion mode: write the final closing paragraph or conclusion section
• QC mode: diagnose why the back end feels weak, repetitive, scattered, or overclaimed

Section

• opening of discussion
• theoretical implications
• methodological contribution
• practical implications
• limitations
• future research
• final conclusion

Output

• one subsection
• one integrated discussion block
• a diagnostic report
• a closing paragraph

Default Routing

• wording such as too repetitive, just restates results, afterthought, doesn't close -> QC mode
• wording such as contribution, theoretical implications, so what -> contribution mode
• wording such as practical implications, managerial implications, policy implications -> implications mode
• wording such as limitations, future research -> limitations mode
• wording such as conclusion, ending, final paragraph -> conclusion mode
• otherwise -> discussion mode

Core Rules

• A discussion section must interpret, not merely repeat.
• Return to the paper's opening promises. The back end should answer the same what will we learn? question the front end raised.
• Narrow overbroad opening promises when the findings or design support a smaller claim. Do not carry an inflated front-end statement unchanged into the back end.
• Organize discussion around contribution, not around a second narration of every coefficient.
• Write discussion paragraphs as mini-pyramids: finding or implication -> interpretation -> contribution or boundary -> bridge.
• Explain what supported findings, nulls, and surprises mean for the larger conversation.
• Distinguish theoretical implications from practical implications; do not merely restate one in different words.
• Acknowledge limitations candidly, but do not turn the section into a self-destruction exercise.
• Make future research directions grow from the study's actual findings and limits, not from generic filler.
• Close at a higher level than the raw results. A good ending returns to the research question or larger problem and shows how the conversation has moved.
• Prefer active, concrete prose over abstract uplift language and management platitudes.
• If the user is really asking for introduction, theory, or results drafting, route to the appropriate specialized skill instead of forcing a discussion-shaped answer.

Outputs

• discussion mode: return a coherent interpretation block that reconnects findings to the paper's question and contribution
• contribution mode: return one or more paragraphs that explain what the paper changes in the literature
• implications mode: return a practical or policy implications block that is specific and not merely decorative
• limitations mode: return a limits-and-next-steps block with credible boundaries and plausible future directions
• conclusion mode: return a closing paragraph that elevates rather than repeats
• QC mode: return findings first, then the minimum repairs needed

Reference Loading

Use progressive loading rather than preloading the whole folder.

• Always load discussion-core.md for substantive drafting, revision, or QC.
• Load paragraph-and-prose.md for paragraph-level logic, interpretive movement, and language cleanup in discussion or conclusion sections.
• Load contribution-and-implications.md when writing theory, methods, or practical implications.
• Load limitations-and-future.md when writing boundaries, caveats, or future research.
• Load conclusion-closure.md when writing the final paragraph or conclusion section.
• Load back-end-qc.md when the discussion feels repetitive, scattered, shallow, or overclaimed.

Assets

Load these templates when drafting to maintain stable output quality:

• Use assets/discussion-block-shell.md when drafting the main discussion paragraphs.
• Use assets/implications-shell.md when writing theoretical or practical implications.
• Use assets/limitations-shell.md when writing limitations and future research.

Execution

Discussion Drafting

1. Restate the research question and main answer in compressed form.
2. Move immediately from what was found to what it changes in the conversation.
3. Interpret nulls, reversals, or surprising findings rather than hiding them.
4. Tie each major implication back to the paper's opening claims.
5. Keep the section moving; do not re-run the results table.

Contribution And Implications

1. Identify the contribution promised in the introduction.
2. State how the findings address that promise.
3. Say whether the paper fills an incompleteness, exposes an inadequacy, or pushes against a prevailing assumption.
4. Separate theory implications from practical implications.
5. Keep claims bounded to what the findings actually support.

Limitations And Future Research

1. State the limit clearly.
2. Explain what interpretive boundary it creates.
3. If possible, note why the limit does not invalidate the whole paper.
4. Derive future research from the actual limit or unexpected finding.
5. Avoid generic future studies should... filler.

Conclusion

1. Return to the paper's question or larger problem.
2. Summarize the answer at a higher level than the results section.
3. Show how the conversation has moved.
4. End with a sentence that feels final, not administrative.

Boundaries

• Do not merely summarize every result again.
• Do not inflate contribution because the author is tired or worried the paper feels small.
• Do not turn practical implications into vague management platitudes.
• Do not use limitations as a dumping ground for every imaginable weakness.
• Do not end with a generic call for more research if a stronger closing statement is available.

Gotchas

这里记录实战中只有踩过坑才知道的经验。请在明天补充你遇到的真实案例。