Blogs

By Paul Dagnall

Has AI ruined your class? Many instructors have run into this situation reporting concerns such as:

• Students use AI to write their papers.
• Project work is mostly completed by AI.
• AI is writing the students’ code.
• Take-home tests are a joke due to AI.
• AI shows the students how to solve problems.
• Students show no critical thinking due to reliance on AI.

Does this sound familiar? While we do not like to start from the position that students are using AI to cheat, it’s obvious that many are (Wiley, 2024), (Nam, 2023). 

 

What does not work

Before we explore optimal solutions, let’s cover a few things that are frequently recommended (Watson & Bowen, 2024) but tend to not adequately address the problem. In other words, we do not want you to spend time re-creating assignments only to find that students can still exploit them with AI!

 

Forbidding AI is Not a Complete Solution

You can forbid the use of AI on some or all of your assessments. Policies like this must be included in your syllabus and addressed on day one of your class. The honest students will obey this rule. The more rascally students, of course, will not. Often you can spot AI-generated work and you can confront the student - not a comfortable activity. Though more enterprising students do know how to adjust the text into their own words.

However, we do strongly encourage the use of “AI-free moments” in your courses. These take thoughtful planning and can be very beneficial. 

 

AI Detection Software Doesn’t Work

MIT cites high rates of false positives when using detection software, as well as the ease with which AI-generated text can be modified to evade detection (MIT, 2025). This is the basis for why UD does not recommend AI detection software. 

 

Requiring Personalization Does Little to Discourage AI

Some believe that requiring that assignments be tied to the students’ personal interests, experiences, or in-class discussions will dissuade them from using AI, and this may help a bit; however, the more rascally students just need to add a sentence or two to their AI prompt and their personalization will be included in the output. The same goes for reflections. Asking a student to write a personal reflection, in no way, discourages the use of AI. 

We’ve received numerous complaints about students using AI to write personal reflections.

Progress Stage Check-Ins Do Not Reduce the Lure of AI

When ChatGPT arrived many educators thought that asking students to first create an outline or plan, then a draft, and lastly a final version would create smaller, iterative steps where the temptation to use AI wouldn’t make sense. Instead this created three occasions to use AI. 

This approach combined with instructor-to-student feedback between each stage can certainly still lead to a good learning experience, but alone, it just doesn’t strongly reduce the lure of using AI to do the majority of the work.

Embarking on a Teaching Revolution When You Don’t Have Time

AI definitely should prompt a massive redesign of a substantial amount of higher ed courses. We support this! But if you don’t have the time to plan, design, and build it properly, there could be serious problems if you're unable to get all the work done before the term starts.

These aforementioned approaches do have merits but by themselves, they are insufficient to outright stop students from using AI when you don’t want them to. There are definitely situations where our instructional designers will still advocate for one of these methods, but we would advise caution and careful guardrails around those activities. 

Assessments that Resist AI-Exploitation

If you’ve been burned by students using AI to exploit your course assessments. The following examples are good approaches to adopt. We certainly suggest you do set up an appointment with an instructional designer to talk through your vision. 

 

Embrace AI Rather Than Fighting It

The best move, if your course objectives allow it, is to adjust your assessments to expect a measure of AI usage. This takes different forms. Think about how few of your own professional tasks can be dominated by AI. Sure, AI is a helpful tool, but the critical elements of your job are beyond what AI can do. Why is that? How can such tasks be infused into your teaching? Aim to create tasks that align with your learning outcomes where AI should be used.

In-Class Performances

Direct students on how to prepare and put them on the stage. Single or group presentations, debates, scenario role-play, debates, or in-class problem solving put students in position to perform within your parameters. The drawback here is that it takes a lot of class time, but these are quick to assess and can be coupled with prior work used to prepare.

Video Assignments with Rules

Include instructions to disallow reading of prepared statements. Require posting a video to a Canvas Discussion where questions must be asked and answered. This is good practice for Zoom or recorded interviews. Students may use AI to prep and they could refer to notes, but they need to be looking at the camera and speaking naturally. 

Interview Exams

This was covered in The Cheat-Proof Assessment, so please take a look! In short, interview exams are oral exams where the student meets you one-on-one. You ask them questions face-to-face. They know stuff or they don’t. There’s no fooling you. These are less time-consuming than you might think due to how quickly you can grade them.

Observed Experiential Learning

UD truly believes in the value of experiential learning in the higher ed experience. This can take whatever form is compatible with your discipline and course learning objectives. As the instructor you seek to assess student preparation, participation in immersive experiences, reflection, and application of knowledge. The Office of Experiential Learning can help you find community partners and provide specific guidance on how EL can be included in your course. 

In-Class Testing

The old school approach with a modern twist. UD supports a product called Respondus LockDown Browser, which forces each students’ computer to only allow the Canvas testing screen. Another option is to use Gradescope bubble sheets which can be imported into your Canvas site. Both of those allow for the benefits of auto-grading. Still there may be scenarios where it makes the most sense to use paper. 

Next Steps

If you’re reading this you probably need to update your course, and you likely do not have time for a complete redesign. Our hope is that these ideas provide a way to address the most glaring areas of your course that have become problematic. 

 

If your class has been ruined by AI or you’re not sure what to do next please set up an appointment with an instructional designer. We’re here to help and we definitely do not want you to waste time with adjustments that will not solve your problem. 

 

References

Wiley. (2024, July 29). AI Has Hurt Academic Integrity in College Courses but Can Also Enhance Learning, Say Instructors, Students. https://newsroom.wiley.com/press-releases/press-release-details/2024/AI-Has-Hurt-Academic-Integrity-in-College-Courses-but-Can-Also-Enhance-Learning-Say-Instructors-Students/default.aspx

Nam, J. (2023, November 22). 56% of college students have used AI on assignments or exams. https://www.bestcolleges.com/research/most-college-students-have-used-ai-survey/

MIT Sloan Teaching & Learning Technologies. (2024, December 19). AI Detectors Don't Work. Here's What to Do Instead. https://mitsloanedtech.mit.edu/ai/teach/ai-detectors-don’t-work/

Watson, C. E., & Bowen, J. A. (2024, April 30). Teaching with AI: A practical guide to a new era of human learning.