In creative work that I claim as my own, here are the guidelines I use toward LLM technology.
The general guideline is — AI is only acceptable if it prompts me to reconsider, broaden, or refine work that I already did. Any AI use which directly affects any meaning in the final product I claim as my own is not acceptable. No AI language is ported wholesale into my writing — I type every last character.
Acceptable
- Grammar/readability edits: "Point out verbosity in the following passage." "Make line edits to the following passage."
- Source selection: "Point me to 10 scholarly secondary sources on the Meiji Revolution's Impact on Sino-Japanese relations."
- Trivial Calculation: "Show your work, find the mean wheat output between 1600-1700 in this giant table."
- Counter-argument creation: "Ideate 3 potential counterarguments to the next passage."
Unacceptable
- Summary of sources: "Summarize this article" — alters my analysis of secondary sources, you want to read human thoughts. If I cite something then I read every page at least of the chapter I cite.
- Direct edits of writing: "Make the following passage less verbose." — Pointing out errors lets me react to faults and address them, direct edits are likely to be accepted and alter my voice.
- Strengthening arguments: "What are ways to further argue this point?" — Altering my active stance. Counterarguments let me wrestle with alternate viewpoints maintaining my stance or adapting my stance, direct strengthening leans toward an alteration of what I am saying.
- Outlining: "Create a 5 paragraph structure for the following argument with these pieces of evidence." — Structure is substance and style. I take ownership for the skeleton of everything I write.