12d (edited) • 💬 General discussion
AI writing Scientific Research
Where's the line?
At what point does "AI helped me write this" become "AI wrote this"? And who's actually checking?
A few things I keep going back and forth on:
- The authorship question: If an AI drafts the literature review, structures the argument, and tightens the prose... is the human still the author, or just the editor? Journals are already scrambling to write policies on this, and most of them contradict each other.
- The quality question: Are AI-assisted theses actually worse? Or are we just assuming that because it feels wrong? A student who uses AI to organize scattered research and tighten weak arguments might produce something more rigorous than one working alone at 3am before a deadline. We don't have good data on this yet.
- The detection question: Detection tools are unreliable and getting more unreliable as models improve. So if we can't reliably catch it, is "stopping it" even a realistic goal, or should the conversation shift to disclosure and process instead?
What might actually work:
- Requiring a documented process (drafts, sources, prompts) instead of policing the final text
- Oral defenses that test whether the person actually understands what they submitted
- Norms around disclosure, similar to how we handle co-authorship or statistical consulting
None of this feels solved to me. Curious where you land:
- Should AI-assisted thesis work be disclosed like a methodology, or is that overkill?
- Is "did AI write this" even the right question, or should we be asking "does this person understand what they submitted"?
What do you guys think?
7
21 comments
AI writing Scientific Research
skool.com/cliefnotes
What we give away free beats most paid courses. Build durable AI systems with a Marine vet and Edinburgh researcher. 40+ lessons, growing.
Leaderboard (30-day)
1
+675
2
🔥
+644
3
+622
4
+523
5
+475
Powered by