Thinking through the “what” and “why” of your generative AI policy is critical; clearly communicating the policy and its rationale to your students throughout the semester is key to making that policy successful. Torrey Trust, Professor of Learning Technology, describes her iterative process for developing, refining, and engaging students with her generative AI policy.
What initially prompted you to develop this policy/statement?
After ChatGPT came out, I realized that the university’s academic integrity policy was not sufficient at the time to help students figure out when to, and when not to, use GenAI tools for academic purposes. Students need guidance, they need specifics, and they need their instructors to be transparent…otherwise, they are left guessing (and if they guess wrong, it can mean failing an assignment or course!). I initially added 2 paragraphs in my syllabus, which I thought were detailed enough to let students know when GenAI tools were allowed and when they were not allowed. However, a couple students clearly used GenAI tools to write their 2-3 sentence reflections for class, and I realized that my policy needed to be a lot more clear, specific, and value-driven. Also, as I have been studying GenAI in education since the launch of ChatGPT, I have been thinking a lot about the impact of these tools on thinking, learning, and communication from a motivation and learning standpoint. So, I started to expand my AI syllabus policy.
How did you design your AI statement?
I began by identifying specific ways students might use GenAI tools for my class and thinking about what I would allow and what I would not allow. I did not focus on tools, because those change and there are too many to keep track of. I created a table for examples of allowable uses (e.g., generating ideas, improving writing quality, assisting with studying). As I worked on this table, I added a “Things to Keep in Mind” section which focused on ethical issues surrounding AI, as well as insights regarding theories of motivation and learning.
For example, I allow students to use GenAI tools to help with their studying – like creating a study guide or getting directions for assignments – but in the “Things to Keep in Mind” section I warn students that GenAI tools, even newer ones, are known for making up (hallucinating) information, and they need to double-check and verify the accuracy, credibility, and reliability of what GenAI tools produce. Hallucinations are not going away anytime soon, but students put way too much trust in these tools. They often use them instead of search engines to find information!
In another example, I tell students that it’s okay to use GenAI tools to help make information easier to understand (like summarizing a text or providing a podcast of an article), but if you look at cognitive load theory, offloading the productive struggle of learning can backfire. So, I recommend first reading through and trying to figure out the information/text, then going to GenAI for help; rather than going to GenAI first.
Next, I added a table for non-allowable uses. On this table, I added the MOST important feature of the policy which is WHY. Telling students “AI is banned” or “AI is not allowed on this assignment” is not enough to convince students not to use it. You need to tell them why. For example, I tell students that they should not use GenAI tools to automatically summarize complex texts for them…because: 1) they are offloading their thinking to GenAI tools; 2) they are relying on GenAI tools, which are guessing machines, to make sense of a text even though these tools can’t “think”; and 3) often when students do this, they end up uploading copyrighted text to the GenAI tool without permission from the author and that data can then be collected by the GenAI company to use for training purposes.
Finally, I shared my generative AI policy with educators around the country and accepted or considered their feedback as I refined my policy.
How do you get students to engage with and understand the policy?
I invite my students to annotate my syllabus in a collaborative document during my first class of the semester. Every student must add at least 2 comments. The AI policy is toward the end of my very long syllabus but often has a few questions/comments and that allows me to go over it with the whole class and address their concerns. One student even commented last year how much she appreciated having the policy. Another student said they were unaware of the UMass Amherst’s Center for Teaching and Learning list of indicators for identifying AI-generated content (which I use as my guide), and they said thank you for linking that resource. Then, a few more students asked clarifying questions like what’s the difference between using GenAI tools to make something easier to understand and using GenAI tools to summarize a text. So, I thought more deeply about those, and we talked about it in class. I also often make further revisions to my syllabus policy after talking with students.
What impacts have you seen since adopting this policy? Is there anything you are considering changing?
So far, I have not seen a high instance of using GenAI tools to do academic work in my classes. However, I attributed most of that to my assignment design. My assignments are built upon design-based learning practices. Students design 3D digital models, create global collaboration project websites, build interactive branching Google Forms. GenAI tools might help them with the creative thinking part but not with doing the work.
I did find that one assignment led to about 10% of the class using GenAI tools – it was a one-page reflection, toward the end of the semester. Students were exhausted and overwhelmed with all their courseloads and workloads. The assignment was not as hands-on, engaging as the other ones, and I started to notice very generic AI-generated text. So, I brought it up with the entire class. I reminded them about the purpose of the assignment (check out the Transparency in Learning and Teaching framework if you have not yet done so – this works wonders in improving student motivation). I told students it was quite easy to identify AI-generated work and if they felt they needed to turn to AI to get work done, to see me instead for an alternative assignment. I also revised a future reflection to shorten the length of it and make it even more personalized. I did not see any cheating after that assignment.
Any advice to instructors as they consider a statement of their own?
Be specific, clear, and transparent when you write your statement. Include the “WHY” (why is this allowed or not allowed). Give students time in class to read and discuss the statement, and ideally, add to/edit it. Also, review the statement before big assignments. This is a proven way to reduce cheating (Vahid, Downey, Pang & Gordon, 2023). Review your statement every semester because GenAI tools change so much, it’s important to be willing to change your statement too. And, try to learn about the GenAI tools that are available to students – not just ChatGPT; but ones being promoted to “go to students’ lectures and take notes for them” and “do all their readings for them.” Keep those in mind as you write your policy.
