I have often thought of the first few months of ChatGPT’s public launch in 2022 as the Wild West for online teachers. As I have described in another article, I was getting so much AI-generated content that trying to sort through it all was taking up to 50% of my time as an online teacher. I also made some mistakes — falsely accusing students who were really innocent, missing students’ work who were guilty because they told their chatbot to write like they were in 8th grade — but in the absence of clear policies and guidelines, dealing with AI felt like whack-a-mole.
Things have changed a lot in the last number of years, but I would argue that most of the ways we have used to try to prevent AI cheating have been unsuccessful:
- Just banning the use of AI doesn’t work.
- AI-detection software is problematic, because it can lead to false accusations, broken relationships, and unfair consequences for students who were just trying to use a grammar checker.
- AI agents can do entire online courses without student involvement, and AI is advanced enough now to mimic a student’s writer’s voice or make mistakes on purpose.
Even requiring students to hand-write all of their assignments does not work, because even this can be faked by AI and it creates additional accessibility issues for students who struggle with handwriting.
So what is the one thing that we have used that actually seems to work? How have I reduced my incidents of cheating to virtually zero in the last two years? (Ironically, I had what I believe was my first incident of AI cheating in a long, long time while writing this article!) There are four steps:
Step 1: Directly Teach the Ethics of AI Use
First, all of the courses that I teach now start with three videos on academic integrity and AI. Students have to watch these videos before they start the course. This provides a helpful framework for them — having AI complete your homework is like getting a sibling to complete your homework for you. It is wrong to do this, and it will not help you to learn and grow.
Just saying, “Don’t do it!” won’t be enough for some students, though, particularly if they are under intense pressure to perform or are in a time crunch. They have to want to do the right thing for you.
If you want to take this first step, you can do that right now. StudyForge believes this topic is so important that we’ve made these videos freely available on YouTube, so anyone can access and benefit from them.
Step 2: Build a Relationship with the Students
Second, I meet with students one-on-one before they start a course. The primary purpose of this is to start building a relationship with them — I want them to know me and know that I care about them. Secondarily, I believe this helps them to want to do well in the course. We have all had teachers that we recognized as good people who had cared about us, so we wanted to do good work to please them. Be that teacher.
As part of this meeting, I talk to them directly about AI use, saying they are allowed to use AI in some ways, but they are not allowed to get it to do their work for them. This way, they know that I am watching for AI-created assignments, but they know to look for ways that they are allowed to use it with integrity as part of their learning process.
Step 3: Clearly State How They Are Allowed to Use AI
Our team at StudyForge developed a framework that has been very useful in communicating guidelines for AI use. There are some assignments we do not want students anywhere near AI for; there are other assignments where it can be a useful tool. We use a six-level system that outlines the various ways AI could be used in particular assignments.
It ranges from Level 0A — don’t use AI for anything — to Level 5 — AI use is required. If you are interested, here is the full set of guidelines:
Level 0a)
Level 0b)
Level 1
Level 2
Level 3
Level 4
Level 5
These instructions are posted both in the assignment sheet and in the submission area so that there is no confusion for students.
Step 4: Students Self-Disclose Their AI Use
Next, we include an AI disclosure component to our assignments. This includes a drop-down menu where students can say how they did or did not use AI, as well as a place for them to share their chatbot history through a link or screenshot. Here are the options we provide in our drop-down menu:
- I did not use generative AI for any portion of this assignment. All of the work is my own.
- I used generative AI to help me generate ideas and structure for this assignment. I have cited the AI program as a source and provided a link to the chat history or a screenshot of the conversation.
- I used generative AI to suggest improvements for this assignment. I implemented all the suggestions myself. I have cited the AI program as a source and provided a link to the chat history or a screenshot of the conversation.
- I used generative AI according to the specific instructions for this assignment provided by my teacher. I have cited the AI program as a source and provided a link to the chat history or a screenshot of the conversation.
- Other (Provide explanation)
Could students lie? Certainly they could lie, but it would take a special kind of gall to watch a video about how you are not allowed to cheat with AI, be told in a face-to-face video call not to cheat, then cheat, then tell your teacher that you did not cheat. In my experience, most students are not like this.
Most students want to do the right thing most of the time. Most often, I have found that it is desperation caused by anxiety to meet deadlines that causes students to use AI in unethical ways. If you have a student cheat in spite of all that you have done to warn them against it, the pressure they are feeling might be a good place to start your conversation with them.
Interestingly, I have also found that the vast majority of students in my courses are not using AI at all, even when they are allowed to. I have even had students ask to be excused from the “AI-required” component of assignments because of their moral objections to using AI.
What I believe our team has stumbled on, however, is a helpful formula to reduce AI cheating. Rather than expecting teachers to police AI, we let the students’ own consciences do the work by:
- Teaching them the ethical reasons not to cheat using AI
- Building relationships so that they want to do well for us
- Clearly outlining ways AI can and cannot be used in each assignment
- Asking them to self-disclose their AI use
This little bit of accountability has worked wonders for my own teaching practice. As I only have a small online teaching load right now, I have an admittedly small sample size to work with, but the difference has been night-and-day, making me think that maybe the Wild West isn’t so wild anymore.
About the Author
