Teaching and Generative AI

State your policy on your syllabus and assignments

Students are navigating a wide range of instructor expectations and policies around GenAI. It’s critical to clearly convey your policy and expectations on your syllabus. 

• You should also include a statement about your GenAI expectations on each assignment, and discuss your expectations with students in class. This is especially important if you allow different uses of GenAI for different assignment types, but even if you have a consistent policy, it is helpful to remind students on each assignment. 

A committee of faculty formed by Dean Wade created a resource to help instructors develop syllabus language that meets the needs of their classroom. The resource is designed to provide sample language that you can modify to the specifics of your course.

• The AI Assessment Scale is a peer-reviewed framework that can help you decide what role (if any) AI should play in an assessment.
  • If you use this scale, keep in mind that you still can—and should—add your own specific guidance. For example, you might find some but not all examples in the “AI Collaboration” category acceptable.
  • You should also remember (and clearly communicate to students) that these levels are independent. For example, permitting some level 3 “AI Collaboration” uses of AI does not mean you must allow level 2’s planning activities. 

Lead a conversation about GenAI and expectations early in the semester

In this conversation, you should explain your policies and give students an opportunity to ask questions about them. You may also choose to ask students to help shape these policies, discuss GenAI’s impact on your field, discuss GenAI’s impact on student learning, or explore ethical and environmental concerns about GenAI use. 

These conversations can help build student buy-in and cultivate a sense among students that their peers disapprove of, and will not use, inappropriate forms of assistance on their coursework. (This is among the most well-established factors affecting student cheating behaviors.)

Consult this resource from the Stanford Teaching Commons for detailed guidance on setting norms and commitments around Gen AI in your class. 

We generally discourage using or relying on GenAI detectors. While accuracy has improved, even low false positive rates create a lot of false-positive cases across many students. 

Even if false positives are somewhat rare, they can lead to serious damage to the student-faculty relationship. We have heard from multiple students who described being hurt, angry, indignant, and disengaging from a class after a false accusation that they had used GenAI. 

To address potential or suspected inappropriate use of GenAI: 

• Run your assignment prompts through GenAI tools so you know what typical GenAI output looks like for your assignment.
• If you have suspicions about a piece of work, start with what aroused your suspicions. Is it different from the student’s other work? Does it include topics, authors, or ideas beyond the scope of the class?
• Does a piece of work include fabricated citations or sources? This is an academic integrity violation, regardless of whether you can establish that the student used GenAI.
• If you have an academic integrity case, review the Rutgers–New Brunswick academic integrity policy and process. It is critical to submit all academic integrity cases through this formal process. (You may want to adjudicate academic integrity violations yourself. This is possible in many situations, but you must file a report first. If it is the type of case that can be handled by the instructor, academic integrity staff will let you know and then assist you throughout the process.)

GenAI’s impact on teaching and learning is evolving, raising big questions with no definitive answers. There's also a lot of exploration and innovation happening; we know there are vibrantongoing conversations happening within all departments and disciplines. 

Here are a few suggestions that we know can help promote learning in your course, regardless of discipline: 

Be transparent about assignment purpose, task, and criteria 

When students don't see the value in doing the work, they're more likely to use AI (or other assistance) in impermissible ways. You can address this by transparently connecting assignments to students’ learning processes and goals. 

To make your assignments more transparent, focus on explaining an assignment’s value to students’ learning; how it connects to course learning goals; and how the skills and knowledge that students develop will help them in their academic, personal, or professional lives. 

Look at the “Example assignments” section of this Transparency in Learning and Teaching (TILT) resource for sample transparent assignments from STEM, social science, and humanities disciplines. 

Shift focus to mastery and process 

Students are more likely to use impermissible tools if they’re focused on performance and outcomes, rather than mastery of skills and course topics. There are many ways to encourage a focus on process and mastery. Any one of these strategies can have a positive impact: 

• Scaffold major assignments through stages like idea generation, annotated bibliographies, and drafts. 
  • Consider devoting in-class time to working on or practicing these scaffolded steps. This provides students with a lower-stakes setting to practice and develop skills, while easily accessing your assistance. In-class work is also an opportunity to guide students’ use or avoidance of AI tools, and can help you to become familiar with your students’ voices and levels of understanding. 
• Use ungraded, relatively simple Classroom Assessment Techniques (CATs) to formatively assess student learning in real time. 
  • This CMU resource on Classroom Assessment Techniques explains how to implement some useful CATs, including application cards, minute papers, and the muddiest point. These provide low-stakes opportunities to develop and demonstrate learning without the pressure to perform perfectly. 
  • CATs enable you to make informed instructional decisions, so you can address difficulties or misconceptions before students feel the need to resort to an AI tool for help.
  • CATs also enable students to make informed decisions about how to approach their learning (e.g., I thought I understood this until I tried to apply it; I need to practice this more before the exam!). In addition to focusing students on mastery and process over grade outcomes, this builds students’ sense that they can improve through effort without resorting to impermissible use of AI tools.
  • CATs can be implemented in any modality and any class size. (Yes, CATs can scale up to very large classes!). In a large class, just responding to a sampling of student work lets you give whole-class feedback on common errors. 
• Add “wraparound” practices around major assignments to focus students on process, skills, and plans for improvement.
  • Discuss upcoming major assignments in a way that normalizes productive struggle as a part of the learning process. After major assignments, encourage metacognition and reflection through exam or assignment wrappers that focus students on process, skills, and plans for improvement.
  • These approaches build students’ sense of self-efficacy: their sense that they can succeed in your class through their own effort, without resorting to impermissible use of AI. If you allow some use of AI, these approaches enable students to understand how they learn and where AI assistance would help or harm their learning.