AI Use in Graduate Studies – Where is the Line and Who’s Drawing It?
At the CEGEP level we’ve been wrestling with undergrad AI policy but the graduate and post-secondary question is even more complex and getting less attention.
I’ve been following the research on this. Hao et al. (2025) found that graduate students use AI at significantly higher rates than undergrads but are less likely to disclose it – the tacit assumption being that at the graduate level, intellectual integrity is a given. That assumption is increasingly wrong.
The specific challenge for grad students: using AI to summarize literature is qualitatively different from using it to draft analysis. Using it to check code logic is different from using it to write methods sections. Every graduate program needs to draw these distinctions explicitly, and most haven’t.
what are universities and colleges in your province doing? especially interested in Quebec CEGEP context and Ontario grad programs.
5 Replies
Join the discussion.
Log In to Reply
Citing Hao et al. is exactly right - that paper changed how our university approached the disclosure conversation. We revised our academic integrity policy after reviewing the data. The framework we landed on: AI as a cited tool. If you used it for any stage of your work, you cite it with specificity (which tool, what stage, what prompt category). Not disclosure as confession, but disclosure as scholarly transparency.
The graduate level requires a more sophisticated framework than detect-and-punish. Researchers use tools. The question is always: did the researcher think?
Mark's point about the supervisor relationship is key. i've seen two grad students in our department flagged by AI detection this year. in both cases the supervisor had no idea how to handle it - there was no protocol at the graduate level, just the undergraduate policy applied awkwardly. we genuinely need separate frameworks for grad work.
the line question in grad studies is also a disciplinary one. in qualitative humanities research, AI assistance with literature synthesis looks very different from AI assistance in empirical sciences where methods need to be your own. blanket policy doesn't work across programs.
grad school governance on this is 3-4 years behind undergrad. which was already behind. we are building the plane in the air.
Academic integrity in the AI era requires us to be more precise about what we're actually assessing. Is it the originality of the text? The quality of the thinking? The depth of the research? Turnitin tells you about text similarity. It tells you nothing about any of those things. Graduate programs especially need to decide what they're actually evaluating.