When writing revision response letters, I most easily fall into three bad habits: opening with pleasantries, writing "in summary" in every paragraph, and not giving abbreviations when terms first appear. Every time I finish, I have to go back and delete them myself.
Later, I turned these habits into a checklist and gave it to MuchSci's personalized answers.
My eight preferences
- **Do not write "in summary" or "in conclusion"** and other such closing words.
- Terms and abbreviations: give the full name and abbreviation on first use, and use only the abbreviation thereafter.
- Do not put a concluding sentence at the start of a paragraph; put the conclusion at the end.
- Do not use the first-person plural ("we believe" → "this study").
- Do not use comparative adjectives (change "better" and "significantly improved" to specific values or delete them).
- Number reviewer comments one by one, with each response corresponding to one comment.
- Flag statements you are unsure about; better to leave them to ask yourself than to write vague sentences.
- Use the target journal's template for citation format, not the default.
The eight rules aren't complicated, but they are "my" way of writing. The official description of personalized answers is: it continuously updates its understanding of you through interaction and uses it in later answers, so responses better fit your current research needs. My experience with it is—it narrows things down first, then answers: it brings the preferences, direction, and stage I have mentioned into this answer.
How I provide the checklist
I didn't write it all in one place; I came up with it while writing and revising:
- When I see "in summary," I say on the spot, "Don't use this word again."
- When a term doesn't include an abbreviation, I add one on the spot.
- I made the most changes in the first two rounds; starting from the third round, I basically no longer need to delete anything it produces.
This is how the checklist takes effect: it is not a one-time configuration but ongoing correction. Personalized answers accumulate every correction and use them directly in the next answer.
Three scenarios where I use it
- Revision response letters: respond to reviewer comments one by one, keep the tone restrained, and avoid pleasantries. This is where I use it most.
- Submission emails and cover letters: explain the manuscript's contribution in one paragraph and do not write "we are very honored."
- Point-by-point revisions based on advisor feedback: list the advisor's comments one by one and give a revision for each.
What the three scenarios have in common is that they all need to sound like I wrote them. No matter how polished the tool's wording is, if it doesn't sound like me, it will have to be rewritten before I submit it.
The three accompanying features
- For newly added references to cite in a revision, use MuchSci's Paper Search Assistant to verify DOI and citation information.
- The manuscript itself and the reviewer comments are archived in MuchSci's Personal Knowledge Base, ready to access whenever I write a response.
- For parts of the finalized response letter that have separate formatting requirements, use MuchSci's Office Assistant Agent to produce an editable document.
Put the three together: MuchSci's personalized answers handle "writing like me," MuchSci's Personal Knowledge Base handles the source material, and MuchSci's Paper Search Assistant and MuchSci's Office Assistant Agent handle citations and the final document.
I suggest writing the preference checklist from the start. Spend twenty minutes listing your bad habits, and you'll save yourself from having to go back and revise every time later. This is also the most easily overlooked way to use a graduate-student AI assistant: it is not just a tool that answers questions; it can be "taught" by me to fit my style.