In my role as an instructional designer for a small university, I use AI on a daily basis. I use it to aid in my procedures for online course design and I strive to incorporate it into student learning activities. I am quick to have AI create module objectives or a course map straight from an instructor’s syllabus, and I can write a prompt to create rich AI patient simulations for budding counselors or future nurses. I can even find ways to implement AI-based in-class tutoring. 

In my role as an adjunct biostatistics professor, however, I found myself falling a step behind. While I am quick to propose AI-based assignments for professors building a new course, I learned the hard way that I wasn’t as astute applying those principles to my personal course load. I’m not sure if it is easier for me to think creatively about content that isn’t mine or if veteran teachers somehow get stuck in the way we think our content is “supposed” to be delivered.  

Either way, I entered this semester of Biostatistics at risk of becoming (gasp!) outdated. The instructional designer in me could see that I was still looking at my content like it was 1999.  In order to prove that students have a functional knowledge of statistics, they should be able to read a research article and complete a worksheet answering questions about the results and the implications. The problem is that it isn’t 1999 – it isn’t even 2022! In the two years since I developed these materials, AI has flipped traditional assessments on their ear, and I’ve slowly seen it infiltrate my students’ submissions. Throughout the Fall 2023 and Spring 2024 semesters, I watched more and more submissions have a robotic tone to the language and a somewhat different flavor to the reflective analysis. Despite my warnings about the inaccuracy of AI and the possible academic integrity ramifications, I knew it was a battle I was losing.

The more I thought about my students’ AI temptation, the more I started to understand it.  I realized that for a majority of my students, doing their biostatistics homework was akin to me making a batch of cookies. Just because I can make a mean snickerdoodle, don’t be fooled into thinking that I actually enjoy the process of baking. That’s why, even though I know that mixing the batter by hand and carefully folding in each of my ingredients will produce a somewhat superior cookie, more often than not you will find me dumping everything into the electric mixer and letting it do the work for me. Sometimes, I choose speed and ease over quality. For my students, ChatGPT was their electric mixer. While it may not have been able to analyze the article with the same insight or accuracy, it was fast and it was easy. Whether you are a chaotic doctoral student or a freshman struggling with that unique blend of overconfidence and insecurity, I can see how the AI upside might sometimes outweigh the risks. 

So, this semester I decided to ride the wave. Instead of fearing that my students would use AI to complete their article analysis, I required it. From repeatedly uploading the article and worksheet to ChatGPT, I already knew there were a few questions it habitually got wrong and others where it fell short of a mastery level answer. Don’t get me wrong, ChatGPT would absolutely pass my class – probably never earning lower than a B, but that wasn’t the point. The point was to allow students to discover ChatGPT’s strengths and weaknesses for themselves. In a controlled environment, I allowed my students the ability to safely test the limits of AI in research analysis and instead of having to hide it, be required to reflect on it. In the assignment directions I asked them to read the article and attempt to answer the questions on their own. The next step was to upload the assignment to ChatGPT and screenshot the answers it generated.  After comparing their answers to ChatGPT’s, I asked them to write a final answer for each question. Their submission included their first draft, the ChatGPT responses, their final answers and a reflection on the process.

As you might expect, the results of the first assignment varied. There were students who were empowered by the scope of AI and students who were confused by it. There were students who blindly fell for “hallucinated” conclusions and some who were able to take things ChatGPT found and strengthen their original answers. The assignment went all sorts of sideways – and I found it wonderful. Every student took something important away from it and every student applied something they learned to the assignment the following week. By the end of the first six weeks, most students found their AI comfort zone as well as a healthy distrust of AI generated answers.  

We all know that AI isn’t going away. Our students will be expected to thrive in a world where using AI is expected and avoiding its pitfalls equates survival. Suddenly the tried-and-true assignments of our past are irrelevant. They were born in a world that no longer exists. What we need to do now is find a balance between the past and the future.  We need to figure out how to apply the fundamentals of our craft to a landscape that keeps changing. For me, that meant meeting my students where they were. I wasn’t going to stop them from using AI, so the alternative was to teach them to use it responsibly. Regardless of your discipline, I encourage you to try this approach. If you are an education professor, don’t wonder if students are using AI to write their lesson plans; ask them to use AI then critique and perfect the product. How about having business students use AI to create a business plan or a marketing strategy that they then dissect, improve, and implement? Whatever your area of expertise, think of creative ways to advance your field or customize your content. Take it from me, if you fear your students are using AI under the table - don’t wait - flip the tables. Not only am I an instructional designer, but I’m from Buffalo, so I’m disappointed that I didn’t think to jump through that table sooner!