Approaches to AI and Academic Integrity
Anna Mills
English Instructor, College of Marin
A presentation for College of the Siskiyous, January 24, 2025
What to expect: One teacher’s thinking and current set of practices to discourage unlabeled AI use that interferes with learning
Let’s acknowledge some reservations around focusing on AI and academic integrity
Fundamentally, if we assess a student’s learning by reading text, we need to know if that text was put together by the student
(A slide by Dr. Tricia Bertram Gallant)
Ignoring the impact of GenAI Tools on your Course Learning Outcomes undermines:
This slide is from “Crafting Your GenAI & AI Policy:
A Guide for Instructors” by academic integrity expert Dr. Tricia Bertram-Gallant, shared under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International license.
AI Policy Development
“You should be able to expect clear guidance from your instructor… guidelines should make clear which specific systems or tools are appropriate for any given assignment.”
--Kathryn Conrad in A Blueprint for an AI Bill of Rights for Education, Critical AI (Duke University Press)
AI is a lot to process for all of us…
AI policy can offer
Students want and deserve clear policy surrounding AI. And that may be a tall order. But we don’t have to figure this out on our own once and for all.
Seek student input or collaboration on AI policy formation
Activity: discuss Joss Fong’s Vox.com video interviews with teachers and students
“AI can do your homework. Now what?
Students and teachers grapple with the rise of the chatbots.”
Of course, ultimately the teacher needs a policy that they are comfortable will help students meet learning goals, a policy consistent with department, program, and institutional policy
Are you aware of any college, program, or department policies or guidance on AI use? Describe where to find any and/or what they say.
Go to the Mentimeter poll at https://www.menti.com/al5gfof99qtc(in the chat).�
Or go to Menti.com and enter 4818 7735 to rate each option.
It’s hard to get specific enough about all possible uses of AI. That will always be a work in progress, but it’s worth working on.
Consider: There are many possible uses beyond auto-generating the whole assignment (i.e. AI for brainstorming, AI for feedback, AI for help with organization, grammar, or genre conventions)
How can we make a policy that accounts for that variety and the variety of ways such uses could impact learning for different assignments?
The AI Assessment Scale developed by Leon Furze and colleagues is helpful; I’d like to share my adapted version.
An AI Assessment Scale (the short version)
Note: This is the short version. See also the version that explains each one.
This scale is adapted by Anna Mills from the Perkins, Furze, Roe, and McVaugh (2024). The AI Assessment Scale. CC BY NC SA. I have incorporated aspects of the earlier version by the same authors and made my own edits.
Sample policies
–From Crafting Your GenAI & AI Policy: A Guide for Instructors by Tricia Bertram Gallant of UC San Diego
A large collection: Classroom Policies for AI Generative Tools, curated by Lance Eaton
Templates and worksheets
A “Tools and Rules” Section for each assignment?
The idea and phrase come from ESL instructor Julie Carey of Cañada College. Here’s some sample language I use:
Students will want to know what to expect from you if you suspect AI misuse. Recommendations to consider:
�From the MLA/CCCC Task Force on Writing and AI working paper on policy development
Purposeful, meaningful, supported writing instruction
Known best practices in writing instruction may be the most effective way to reduce incidence of AI misuse.
Why is writing in college important? Not for the product. For the thinking process.
“A fundamental tenet of Writing Across the Curriculum is that writing is a mode of learning. Students develop understanding and insights through the act of writing. Rather than writing simply being a matter of presenting existing information or furnishing products for the purpose of testing or grading, writing is a fundamental means to create deep learning and foster cognitive development.”
First and foremost, emphasize purpose and engagement
Build relationships and community as the context for reading and writing�
Presentation by Anna Mills, licensed CC BY NC 4.0.
Teach and support the writing process
Accountability
I focus on the preventative, supportive practices just described. But I don’t find them to be adequate.
“Trusting students not to cheat isn’t fair to the students who don’t cheat no matter what….For all other human behaviors, we understand that people need help sticking to their goals or keeping their promises.” –Dr. Tricia Bertram Gallant, “Crafting Your GenAI & AI Policy Guide”
In-class and oral assessments
These don’t have to be the sole mode of assessment to help reduce AI misuse.
Used judiciously, these may allow us to get to know student voice so we inquire when there’s a discrepancy.
Of course, many students have anxiety about performance and about talking to teachers, and their spoken style may differ widely from the way they write.
Are there kinds of academic writing assignments it can’t generate?
I don’t know of any.
It can produce something passable for most prompts with persistent prompting.
Feed it a reading, snippets on current events, bits of personal experience, a transcript or an image (with Bing)—whatever the prompt demands.
It will generate a personal essay, a piece about something timely and hyper-local, a simulated reflection on the writing process, a close analysis of a text, or a response to its own outputs.
Can we distinguish AI text from student writing?
Our intuition about what is AI text may help us initiate important conversations if we know the student’s writing well. But let’s be wary of our own unconscious bias and possible overconfidence. AI can imitate student writing styles.
Is this AI? Turnitin said 0%. Previous workshop participants largely guessed it wasn’t AI. But it is.
Robin Marantz Henig wrote an article in National Geographic about kids and their feelings about gender. She looks at two viewpoints: one from a scientist named Eric Vilain and one from a mom named Marlo Mack. Vilain thinks kids often mix up liking things typically done by the opposite gender (like a boy liking dolls) with feeling they are that gender. He says most kids who feel this way end up comfortable with their birth gender as they grow up. So, he suggests parents should tell kids it's okay to like whatever they want, but that doesn't change who they are. Mack shares her own story, where even after telling her child that boys can like 'girl' things, her child still felt she was a girl. The article seems balanced, but it feels like it leans more towards Mack's side. Henig uses Mack's story to show that what Vilain suggests might not always work. Readers might leave the article feeling more supportive of kids deciding how they feel about their own gender.
Studies on the extent to which instructors can differentiate AI from student writing suggest we are quite unreliable (drat!)
Is AI detection reliable? “In short, no”--OpenAI, after taking down its detector
Is AI detection biased against English language learners? Maybe
Note: free software exists that rephrases AI text to “humanize” it and get around AI detectors
The accuracy of detectors is constantly changing as AI and detectors change. Right now, Turnitin seems to be ahead.
“Turnitin turned out to be the most accurate and consistent one, with a 100% AI score even with the adversarial techniques.”
It even did pretty well on AI text that was edited or paraphrased. Researchers used “three adversarial techniques (edited through Grammarly, paraphrased through Quillbot, and 10%-20% editing by a human expert).. The study found that the four AI-detection tools showed inconsistent AI scores - from very high by Turnitin (almost perfect) to very low by Writer AI. ”
–AI vs AI: How effective are Turnitin, ZeroGPT, GPTZero, and Writer AI in detecting text generated by ChatGPT, Perplexity, and Gemini? (Malik and Amjad 2025)
“For those who decide to use AI detectors, please consider the following questions”
From the MLA/CCCC Task Force on Writing and AI working paper on policy development. (The Modern Language Association and the Conference on College Composition and Communication are the professional associations for writing, language, and literature faculty in American higher ed.)
• What steps have you taken to substantiate a positive detection?
• What other kinds of engagement with the student’s writing affirm your decision to assign a failing grade outside the AI detector’s claim that the text was AI generated?”
Process tracking
Process tracking means ask students to share their document history
“If a student actually does the writing themselves, they will…write things, move things around, add bits, delete bits; all the usual meandering manoeuvres of human writing. And all of that will appear in the version history, indelibly timestamped and tagged per user.”
Dave Sayers
“A simple hack to ChatGPT-proof assignments using Google Drive,” Times Higher Ed, May 25, 2023
Various apps and extensions allow replay and analysis of a document history, including cut/paste and time spent (check for FERPA compliance)
How much teacher labor does this involve? My approach involves some setup and then minimal labor.
I create a shared folder for the student to store their Google Doc (using a school Google account) so I have edit permission.
When they turn in an essay I use the Revision History Chrome extension to see some basic stats about their process across the top. These provide a basis for further inquiry and conversation if needed.
I can see the text of any large copy/paste (might be from another preferred writing app or from another place in the document.)
I can replay typing and edits.
Why I’m switching to Grammarly Authorship
The user can decide if the authorship report includes a replay of the whole drafting process. I probably won’t require students to share this. I pay more attention to time spent.
Grammarly authorship reports attempt to label text by origin
I ask students to reflect when they share an Authorship Report
Steps to try Grammarly Authorship
Is process tracking intrusive? This informal poll suggests opinion is divided.
We could make process history tracking part of the way students reflect on their learning process, as Nigel Robertson suggests.
Allow for alternative processes
My message to students: “Think of this as an online version of in-class writing that allows you more flexibility. It gives me a way to understand your writing process, both to encourage academic honesty and to encourage reflection on the writing process itself. If you have concerns or do not feel comfortable sharing your process in this way, please let me know and we can meet and work out another plan.”
Caveat: process tracking can discourage AI misuse, but it can’t completely prevent it.
Mentimeter Rating activity
I’ve just listed (too many?) strategies for discouraging misuse of AI.
Use the arrow to move to the next page in the same Mentimeter to answer at https://www.menti.com/al5gfof99qtc (in the chat).�
Or go to Menti.com and enter 4818 7735 to rate each option.
Questions or comments?�Thank you, and feel free to get in touch!
LinkedIn: anna-mills-oer
Slides, open for commenting: https://bit.ly/AISiskiyous
�This presentation is shared under a CC BY NC 4.0 license, except where otherwise noted.