Teaching with AI

Examples include asking students to respond to or expand on an in-class discussion/debate with references to specific comments; asking students to visit local sites such as Shark Valley  or other local sites rich in vegetation and describe and assess the invasive species situation; or have students attend a live performance, describe and reflect on their experiences, and analyze a portion of it using the dramatic/musical/artistic techniques learned in class. 

The more required citations, the more challenging it is for ChatGPT. Also, asking students to use textual evidence to build their academic arguments is a skill we want them to develop. Note, longer essays and papers requiring a relatively high number of citations will make it easier to identify questionable submissions as the AI will likely cite things that novice learners would be unlikely to use.  

ChatGPT does not pull any citations after 2021. This also allows you to ask students to consider how current knowledge is the result of foundational work. 

For example, ask students to respond to questions from the perspective of someone living in Miami, as an FIU student, or from the culture of their family of origin. Have them reflect on how their lived experience impacts their understanding of the topic. 

When the overall grade in a course relies on only a few high-stakes assignments, students are more likely to panic and,in some cases, turn to inappropriate behaviors. This phenomenon is outlined in James Lang’s Cheating Lessons. Doing this also provides the opportunity for the faculty to gauge overall class understanding earlier and redirect if needed. 

Several teaching structures (e.g., flipped classrooms, hybrid, active learning) require students to do work in the classroom. It allows students access to the expert (the faculty) when they are struggling with the content. There are many ways to reduce the grading load for larger classes such as group quizzes, personal response devices, and shorter writing assignments. CAT can help you figure out what might work for you or connect you with other faculty at FIU that are using some of these techniques. 

Having an example of their natural writing style may be helpful as a comparison later in the semester. Students with work that appears to be drastically different from what you have on record may indicate a closer examination or conversation is warranted.   

Incorporating the AI responses into your assignments can help students recognize the limitations of these products and can help students develop critical skills. You may consider providing students with an AI response to a prompt and ask them to critique it for correctness, style, or substance. For shorter written work, ask students to compare their work to responses from the AI (Hint: plan to have the AI responses prior to class as AI servers can be down or at capacity without warning). 

Asking students to perform behaviors that reflect their subject area understanding in class or recorded in small group settings for online/hybrid classes can provide evidence of students meeting desired learning outcomes. Possibilities include presenting problem solutions complete with rationales for their approach,editing each other’s work using rubrics that align with your expectations (for formative stages of an assignment),assessing their own work using your rubrics (as a portion of their grade), or recording interviews or podcasts with experts (or for a different twist, novices) that reflect the knowledge of the student (i.e., the questions and the follow-up to the responses are graded).  

Submit sample prompts to ChatGPT to see what types of responses are produced. You may want to submit a variety of the kinds of prompts you give your students to get a feel for the nature of the responses ChatGPT creates. The language used may be very different from what you normally see in student writing. For example, ChatGPT responses can feel “canned” with standard writing structures and word choices. Students often use more expressive language and may not always use standard essay structures. Additionally, AI responses are on the low end of the taxonomical skill set. You are not likely to see language reflecting analysis, critique, interpretation, conjecture, or synthesis. 

AI products are asking students to submit their phone numbers and/or emails. This may result in their information being shared or sold. There is also the potential for fake AI products and scams. 

Outlining expectations and your reasons for them help students understand why you have designed your course and its assessments in the manner that you have. Don’t assume that their prior experiences will align with your expectations. Students will come to your class with experiences that range from “anything goes” to highly secure, proctored examinations.Use this link for syllabus language suggestions. 

You may be inclined to start by sharing what the consequences of misrepresenting work looks like in your discipline (e.g., loss of funding or credentials, legal actions, or even injuring others), but it may be more impactful if you help students understand the benefits of doing their own work. Here are some ideas: 

1. Have a discussion at the start of the semester about their fear of not “getting it right”. Share how challenging it can be for you to do some of your research and what you do when you feel stuck. 

2. Explain how you use what they submit to guide what you do in class or the attention that you give them.  

3. Create space for failure and confusion in your classroom. Help students understand that “getting it right” requires time, effort, and, often, failed attempts.  

4. Ask students to talk about what they think of dishonesty in general and then, what they think constitutes academic dishonesty and what does not. A class consensus on what academic dishonesty will look like in your classroom may take a bit of time, but it may be a useful deterrent. 

More frequent, lower-stakes assessments help minimize students’ perceived need to cheat (see Carnegie Mellon’s very useful resources). However, if we overload students with so many low-stakes assessments in all of their classes, they may not have sufficient time to complete all of their work and resort to the practices that we were trying to avoid. Here are some ideas: 

1. Use ungraded peer feedback (using your rubrics) for smaller assignment chunks. 

2. If you choose to have more frequent assessments, consider making them shorter and more focused. 

3. Practice is an important part of learning. But, if students are practicing without understanding, they are likely internalizing incorrect/incomplete understandings. Consider using tools that help students identify, learn about, and practice the content that they don’t yet fully understand and skip the content that they do. 

4. Provide students with assessment options. Some faculty at FIU have found success with allowing students to select which assignments they will complete to show their understanding. 

Here is an extensive list of almost 300 interactive techniques (link) created by a member of our Florida Consortium of Metropolitan Universities, Dr. Kevin Yee at the University of Central Florida. 

Teaching in the Age of Artificial Intelligence. A presentation by Dr. Matt Acevedo, Executive Director for Learning Innovation and Faculty Engagement, University of Miami. FIU Alum and Adjunct Teaching Professor. 

Watch Florida Atlantic University faculty discuss instruction in the world with ChatGPT.

Panel members: Donna Marion (dmarion@fau.edu, Psychology) and Jeff Galin (jgalin@fau.edu, writing center and professor) offer specific strategies faculty might implement. Anil Bilgihan (abilgihan@fau.edu, Hospitality Management) shares expertise and experience in EdTech and Kevin Wagner (kwagne15@fau.edu, Political Science) provides examples of how he uses ChatGPT in journalism.

Watch a "Fireside Chat" by Madeline Pumariega, President of Miami Dade College, Dr. Ernesto Lee, and Noelle Silver as they discuss AI in education.

For information about AI tools, links to numerous articles, a discussion of the limitations of AI tools, warnings about potential data issues/scams, and some do’s and don’ts about teaching in the age of AI, visit educational technology developer Doug Holton’s post using this link. For an abbreviated version, use this link.