-By Xipeng Wang
***
As a lecturer teaching an AI course, I understand the importance of learning to use GenAI tools and leveraging them to facilitate and accelerate learning. At the start of this year’s class, I shared my GenAI policy and statement with the students as follows:
As we are in a class focused on learning AI techniques, I strongly encourage students to use GenAI to facilitate and accelerate their learning. However, I advise students not to directly copy questions into GenAI tools and simply copy the answers back. Additionally, for quizzes, exams, and group discussions, please refrain from using GenAI tools.
I hope to convey two key points: first, I encourage you to use the tool to enhance your learning as much as possible. Second, avoid relying on it for everything. However, this has caused some confusion among students. Some of their questions include:
- When should I use GenAI tools, and when shouldn’t I?
- Will I face penalties for using them on assignments not explicitly mentioned?
- Why can’t I use them for quizzes and exams?
- If I don’t use GenAI tools but others do for quizzes, will our grades be the same?
- …
I thought my points were clear, but apparently they weren’t. This made me reflect on how to improve them. After much consideration and several iterations, I’ve identified three key recommendations:
- Provide transparency
- Offer autonomy
- Establish clear purposes
I want to keep this blog short, so let’s get straight to the point!!!
Provide transparency
I believe most students care about their grades (though in the GenAI era, I don’t think grades are an ideal way to evaluate a student’s abilities, especially since many educators don’t include clear GenAI guidelines in their syllabi—but that’s a separate discussion). As the course instructor, you need to be very clear about where using GenAI is prohibited, where it is encouraged, and where it is discouraged.
Another aspect of transparency is showing how you use GenAI in the course. This demonstrates the effort you put into the course by leveraging GenAI, which could make it easier to convince students not to use GenAI in certain situations. For example, if your entire course material and examples are generated with just a single prompt, how can you persuade students not to use GenAI to get an A+? Being transparent about how and where you use GenAI sets a good example, and you can share this approach with others.
Actionable items:
- Be explicit about when using GenAI will result in penalties.
- Be clear about how and where you incorporate GenAI into the course.
Offer Autonomy
In my class, I give students full autonomy to decide when to use GenAI. Different people have different learning goals: some may just want to pass the course, while others want to explore new concepts. Some students may already be familiar with the topic and feel comfortable with their knowledge, while others might use GenAI to deepen their understanding. You can offer suggestions and recommendations, but make it clear that they are free to make their own choices. This may seem easy, but you need to clearly communicate this, as students often treat recommendations as requirements.
Actionable items:
- Make it clear that students can and should decide when to use or not use GenAI based on their own needs and experience.
Establish Clear Purpose
With transparency and autonomy in place, many students might still be tempted to use GenAI constantly, as it saves them time and effort. This could make some instructors’ recommendations ineffective. To counter this, we need to help students establish clear purposes for when we recommend certain actions. For example, I tell students not to use GenAI during quizzes, exams, and group discussions. But why? If we don’t provide clear reasons, they’ll be more inclined to use GenAI all the time.
Actionable items:
- When recommending that students avoid GenAI for specific tasks, provide a clear purpose for why they should do so.
Updated GenAI Statement
Finally, I’m sharing my updated version of my GenAI statement. It’s certainly not perfect, but I’m quite satisfied with it after incorporating the three points mentioned above. I plan to use this in the AI course I will teach next semester, gather feedback from students, and continue to refine it!
- Provide transparency
- Offer autonomy
- Establish clear purposes
As we are in a class learning AI techniques, I strongly encourage students to use GenAI to facilitate and accelerate their learning. However, I want to be very transparent about using GenAI in the class.
You will not be penalized for using GenAI-generated results for your assignments and projects. You don’t need to tell me whether or not you’ve used it for your work. That said, I would appreciate hearing about how you use it if you’re interested in sharing.
There are several instances where we encourage you not to use GenAI. Again, you won’t be punished for using it, but understand that relying on it to get direct answers will not help you learn.
If you put in good effort in this class, you will receive good grades and learn things that will be beneficial for your future. So, don’t be discouraged if you spend more time on the course but receive the same grade as others.
In my classes, all lecture slides are either created by me or borrowed from other professors or researchers, with proper citations of the original publications. I do use GenAI to help me understand certain concepts and to find better ways to explain them to students.
All assignments are designed by me, and I also generate the answers myself. I do use GenAI to help write comments, such as docstrings for code.
Here is a list of areas where I don’t recommend using GenAI:
– Quizzes and exams: These are designed as interview-style questions (some of which I use when interviewing candidates). Memorizing and understanding these will be helpful in future job interviews. You can use GenAI to prepare quizzes and exams.
– Class discussions: These simulate discussions with colleagues. In the workplace, I guess most people won’t enjoy conversing with a robot, even an intelligent one. Although this may seem a bit strict, I’d prefer if you avoid referencing anything generated by it.
– Paper reading: This activity helps develop critical thinking. GenAI often provides superficial, general ideas. To become a good researcher or outstanding in your job, you need independent and critical thinking, which requires practice. Currently, GenAI won’t help much with this.
Notes:
After finishing my blog, I used GenAI to help generate a title, and it definitely offered better options than what I had in mind. I also rely on GenAI to proofread my writing, fixing typos and grammar mistakes along the way.
This document is also a GenAI-generated blog. While it has many good points, it’s too lengthy and remains too broad, leaving educators unsure of the practical steps to take. By focusing on just one or two key points, a blog post would offer more actionable insights, making it easier for educators to incorporate these ideas into their courses. In the GenAI era, I find myself gravitating toward shorter articles, as I tend to assume longer ones are more likely AI-generated (just my biased opinion).
Authored by Xipeng Wang who is a lecturer in the Computer and Information Science Department at the University of Michigan-Dearborn. He earned his Ph.D. in Computer Science from the University of Michigan. Currently, he teaches two courses: Introduction to Artificial Intelligence and Deep Learning.
Image generated by U-M GPT DALL-E 3 Model