Published: Fri 12 December 2025
By Joseph Reagle
In praxis .
tags: teaching ai
This semester I required students to share the version histories of
their writing.
All assignments submitted to Canvas must have an
appendix with a link to your document’s version history (i.e., a link to
itself). Versioning and history is native to GDocs ,
Pages ,
HackMD and Wikipedia ; if
you use MS Word
you must use Northeastern’s Office 365 or keep it
in your OneDrive/Sharepoint
account. Ensure that your documents and their history are publicly
accessible from the start by sharing them with me with the
Edit permission. I might also ask you to speak about your work
with me. If you use AI tools for improving your work (e.g., ChatGPT for
feedback or GrammarlyGo and Quillbot for improving composition), include
a note or appendix describing your use , including important
prompts; failing to do so is misconduct.
For the 10–20% of students who forgot to provide a link to their
history, I assigned a 0 with a note stating the grade they will receive
upon completion.
This policy was a hassle, but I believe it lessened AI misconduct.
More importantly, reviewing students’ writing, especially with the
Google Docs Process Feedback
tool, is worthwhile and illuminating. I was surprised by how many
students wrote their larger (~1500 words) assignments in an hour or two
the night before it was due. Many wrote from the first sentence through
to the last, with a few edits along the way and at the end. I no longer
wonder why they don’t do well as I (or they) think they should. Their
work is not revised and doesn’t reflect earlier feedback. I suspect a
few of these students are typing in AI-generated prose; there are “humanize ”
plugins that do so automatically. The students who excelled spent around
four hours on these assignments, spread across a few days, and began
with an outline and notes about their sources.
Some students disclosed that they used AI tools. At the extreme, yet
within the bounds of my AI
policy , they used a chatbot for brainstorming, researching, and
outlining. They then wrote the sentences within the shape of the
detailed AI-generated outline. They might then use Grammarly for polish
or chatbots for feedback. Some students also wrote in their native
language and used chatbots to create the English version. Interestingly,
I could see the progression of a multilingual student writing in both
English and Spanish. I’m not yet sure what is happening with the
students who are not fluent in English.
Only one student linked to their session with a chatbot; their
version history and (minimal) prompts led me to comment:
While not misconduct, I am concerned about your learning given your
heavy use of AI in this assignment.
I think the lack of engagement with course sources is a reflection
that you prompted/directed AI to create a thesis, find and summarize
sources, and create an outline; you then you edited prose following its
suggestions.
Again, this isn’t misconduct under my course policies, but you are
wasting opportunities to learn how to do these things. For example,
there’s a difference from what you’ve done here and learning to find and
read these papers, or even asking for suggestions, but then discussing
those sources with the AI, asking for clarity about its findings,
methods, and implications.
I am uncomfortable with those students that use AI weakly and
extensively, especially for the brainstorming and outlining of a paper,
but I don’t think I can effectively police this. I’d rather have
honesty, and have their grade reflect if there’s a consequent lack of
quality.
During the course of the semester, I did identify students using AI
in violation of the AI policy (“query, don’t copy”). Those who
double-down received a misconduct referral for
additional dishonesty about the misuse. Finally, I suspected a few
students were misusing AI, but their version history showed otherwise.
This is the final benefit, especially for students, of transparency
about their work.
There are comments .