Strategies to Mitigate Unauthorized GenAI Use – Learning and Teaching Consulting

Strategies to Mitigate Unauthorized GenAI Use

Many instructors have expressed concern that students are using GenAI tools to complete entire assignments, or to provide answers during tests. This is a valid concern, as participants in student panels have acknowledged using GenAI heavily, even after noticing a decline in their final assessment scores, simply because it’s so easy and available during busy or stressful times. Here are some measures faculty can take to help students understand the cost to their own achievement, in doing so.

Share your AI Policy Rationale

It’s important that your AI policy include a rationale that explains why you chose as you did, and how your policy centers student learning in your course. Students need to understand how your policies aid their learning and professional development if you want to secure student buy-in.

The line between using GenAI to support learning or as a substitute for learning is not always an easy one to draw. If you are going to integrate GenAI tools into your course, this will take careful thought and likely take additional instructional time to discuss your guidelines with your students and ensure that students are not outsourcing their work, but using the technology in service of learning. If you choose to exclude GenAI tools from your classrooms, your students will still need you to explain why and give them enough context to understand the impact on their skills and knowledge you are seeking to avoid.

Develop Honor Statements or Contracts

It is important to be proactive, supportive, and educational in your approach to academic integrity, in order to secure the most student cooperation. One way to do this is to co-create an academic integrity contract with students or provide honor statements for your class that clearly lay out what is acceptable and what is not. Appealing to students’ pride and sense of integrity can be a strong move to incline students to cooperate. This Contract Template from Vancouver Community College provides sample language of a co-created classroom agreement on the acceptable use of generative AI. The GenAI Honor Statement Matrix includes language to help you clearly communicate your expectations around AI use to students. LSA does not have a College-wide honor policy, as such, but individual instructors are always free to develop such for their own courses and learning communities.

Focus on Authentic Engagement 

The main appeal of AI for students is not necessarily that it can outperform them, but that it is an easy and tempting alternative to persisting through learning challenges. Designing assignments that engage and support students during the learning process will help mitigate students’ overreliance on AI. Strategies to promote student engagement can be summed up in the following approach: 

  • Scaffold instructional content and assignments
  • Promote authentic student responses
  • Design iterative learning tasks

Students perceive instructional tasks that are appropriately scaffolded, authentic, and iterative as a way to learn rather than a stressful academic performance with high stakes consequences. Inviting students to share personal experiences, evaluations, and self-reflections about course content helps students see how coursework is relevant to their personal goals and interests. Tasks designed in this way are approachable and student-centered, a combination that increases intrinsic motivation in students to engage fully with the learning process. The instructional strategies and tools recommended below align with this approach.

Scaffolding and Chunking Large Assignments

The first reason students may turn to AI to complete assignments is a lack of understanding. If students become stuck during the course of attempting an assignment, they may turn to AI for information and support. Scaffolding is a way to teach students the value of engaging with the learning process. Scaffolded assignments are chunked into smaller parts and assigned in order of difficulty with foundational tasks at the start of the assignment and higher order thinking components assigned later in the learning process. Students can more easily perceive possible steps to take when they have a question about a smaller assignment task, and are more likely to take those steps early on so that they don’t find themselves feeling stuck in an overwhelmingly large task without any clear way forward and running up against a deadline. 

Another reason students turn to AI is a lack of time. Time management is a skill that needs to be learned, and our students benefit when instructors intentionally plan assignments in a way that teaches time management skills. Breaking up an assignment into discrete tasks and setting individual deadlines for those tasks will provide students with a series of steps allowing them to make tangible progress on large assignments and projects. When students have the support of regular assignment deadlines and feedback, they are less likely to procrastinate and turn to AI as a result of poor time management. 

Finally, and arguably most importantly, scaffolding an assignment provides students with frequent opportunities for feedback. Feedback does not have to be voluminous; one or two concrete steps that the student can take to improve, at each stage, will keep students engaged by placing them in the driver’s seat during their learning process. When instructor feedback makes visible to students the ways they can improve at a skill, students are far more likely to engage their ability to learn. The joy of learning is the strongest motivator for students to stay engaged with the learning process that is, by nature, a struggle. Scaffolding tasks by cognitive demand and coaching students with actionable feedback as they complete each step of the assignment is one of the most powerful things instructors can do to help students develop metacognition. 

Providing Social Motivation to Produce Authentic Writing with Peer Review 

It is well known that LLMs are capable of producing polished writing samples. Instructors whose learning goals focus on developing students’ writing skills face a landscape where students have such easy access to GenAI tools that it becomes very tempting for students to simply have GenAI produce their essays. In addition to scaffolding writing assignments, one way to deter students from outsourcing their writing to GenAI is to include peer review. When designing for peer review, it works best if the author and reviewers are able to discuss peer review feedback in a writer’s workshop setting. When students are expected to discuss their feedback with peers, they feel accountable to their peers, making it less likely that they will turn to GenAI to provide drafts or even peer feedback itself. Finally, tools like Peerceptiv not only simplify the logistics of designing a peer review process, but hosts the review process itself, making it extremely time consuming for students to move in and out to outsource their peer review to GenAI tools. All of these steps add up to help make doing their own writing the easier path.

Authentic Discussions with Repeat Engagement

In traditional discussion boards, the instructor posts a question or prompt that garners student responses. This is usually one-way and one-time communication. Two-way communication, in which students hold a genuine, ongoing discussion with their peers and the instructor is far more interesting, and less likely to tempt students to use GenAI to produce. Shorter and more frequent posts and replies are also less tempting to outsource to AI tools as they are based on students’ recent experience, class discussion, and students’ own connections to prior knowledge and the outside world. LSA has several tools that can help with this.

Yellowdig, for example, uses a gameful learning approach to encourage repeat participation. Posts and replies to posts all earn students points, but the point total earned per post or reply is low enough that students must post and reply multiple times over the course of the point earning period to reach their point goal. This means that students must return to the conversation multiple times during the week, and re-engage with their peers. 

Harmonize encourages instructors to set multiple “milestone” due dates within a single discussion, requiring students to return several times and, once again, engage with their peers and whatever new ideas have emerged since they last visited. Harmonize also provides some useful dashboards for instructors, with visualizations of how active the discussion is and built-in tools to immediately reach out to and prompt students who have low engagement.

Bringing Students Back to the Text with Social Annotation Tools

A significant challenge of GenAI in teaching and learning is how easily students can upload a reading into an LLM and receive a summary of the text. They don’t even need to search for the information, or buy SparkNotes. For instructors whose learning goals are rooted in reading strategies, such as teaching students to decode challenging disciplinary texts, GenAI poses a significant barrier to helping students reach their learning goals. Tools that encourage students to engage with text-based materials can help with this. Social annotation tools such as Hypothesis and Perusall, in particular, encourage students to analyze the text collaboratively, as a group, and thus encourage authentic engagement.. Social annotation is a way to help ensure that students have engaged with the actual text and not just a summary. The time that it would take to outsource responses to peer comments and questions using an AI serves as a mild deterrent to AI use. In short, the social nature of the tool, where students are responding to their peers and building on each other’s insights, provides social motivation for students to engage authentically with a course text. 

Short Paper-based Formative Assessments

While writing long assignments by hand is not feasible or recommended, given how rarely students are taught or practice hand-writing, now, doing short, low-stakes assignments in class and on paper is one way to ensure students are doing their own intellectual work. Including some paper-based assignments provides a way to measure students’ actual learning progress and to recognize students’ own writing voice and style. Using this approach for a few scaffolded pieces of larger assignments engages students in the overall project, as well. To make managing such assignments easier for both students and instructors, LSA provides Gradescope as a tool to digitally “collect” paper-based student responses from students and provide prompt, meaningful feedback. The tool allows students to scan their work while still in class and upload it for immediate submission. Instructors can also use the grading features of the platform, such as responsive rubrics and limited writing recognition, to more easily grade work or assign participation points. Using Gradescope, collecting in-class work is logistically feasible, even for large-enrollment courses.

Scroll to Top