We are happy to share some pedagogical considerations for generative AI and ChatGPT in the classroom. We recognize that not all will apply to every course but hope to showcase a range of considerations that may resonate with academic staff and teaching assistants across disciplines here at CEU. To set up a consultation or share your own reflections, contact us email@example.com.
Learn more about...
- Communicating with Students Around the Use of Generative AI
- Redesigning Assessments and Writing Assignments
- Incorporating AI into Your Courses
It is important to communicate with students about your expectations surrounding the use of generative AI. These should be communicated in writing, either on the syllabus or a set of written instructions, to avoid potential misunderstandings.
First, you may wish to decide if and how much generative AI use is acceptable in your course. Is it allowed for brainstorming? Translation or checking grammar? Debugging code? For a specific assignment? A recent JISC report summarizes the ways that students may be using generative AI as a study and writing tool.
Once you have determined how much AI use is appropriate, the following resources can help you develop a syllabus statement or set of assignment instructions:
- Should you Add an AI Policy to Your Syllabus: What to consider when drafting your course policy on students’ use of tools like ChatGPT (Gannon, 2023)
- Example Syllabi Statements: A crowdsourced collection of 100+ syllabus policies for generative AI (Eaton, 2023)
- Can I Use AI? Tool: A form-based tool that can help you develop a table for how students can or cannot use AI for your assignments (Watkins, 2023)
If you are allowing the partial or full use of generative AI, it is also beneficial to communicate clearly how you expect students to acknowledge or cite its use in their work. This guide provides acknowledgement and citation examples (Monash University, 2023), and the APA, MLA, and Chicago manuals have released guidelines on their websites.
Finally, it helps to be transparent with students about why AI is or isn’t allowed or recommended for certain tasks. Rather than starting with questions about students’ own integrity, which may create an atmosphere of mistrust, it may be more productive to frame conversations in a way that create a climate of encouragement and that drives dialogue, collaboration, and inquiry.
If you choose to modify assignments, keep in mind that different formats can advantage or disadvantage different students, and that some formats may pose greater barriers than others. For example, in-class, handwritten or oral exams may disadvantage students who are studying in a foreign language or students with disabilities. Asking students to create videos at home might require different sets of skills and equipment than on-site work. There is likely to be no “AI-proof” type of assignment, and it is important to consider multiple goals — such as the learning purpose of the assignment and minimizing potential barriers — in addition to reducing the risk of cheating. That said, the scholarship suggests several strategies worth considering:
Revisit course learning outcomes. Returning to your learning outcomes can help you reflect on the educational purpose of assignments and opportunities to adjust the format. For example, if the purpose of the assignment is to become familiar with literature, would an annotated bibliography be a good alternative to an essay? If the purpose is a close analysis of one text or solving a logic problem, it may be that an essay is the best format, but sometimes a shorter or more focused approach may be worth considering. It is also sound pedagogical practice to communicate the purpose of an assignment to students, as we learn more from an activity if we know why we are doing it.
Focus on what motivates students. There is evidence that robust and constructive feedback, authenticity, alignment with students’ interests and goals, and building relationships can all boost motivation (CAST, 2012), which makes it more likely for students to see the value of completing assessments.
Consider ‘authentic’ assessment. This is an approach that requires students to use the same competencies or combinations of knowledge, skills, and attitudes that they need to apply in professional life situations, seeking to not only measure but also improve learning (Gulikers, Bastiaens, and Kirschner, 2004). Generally, authentic assessments are realistic; require judgement and innovation; ask students to do the subject; replicate or simulate the contexts in which people are tested in the workplace, in civic life, and in personal life; assess the ability to use a repertoire of knowledge and skills efficiently and effectively to negotiate a complex task; and allow appropriate opportunities for students to rehearse, practice, consult resources, and get feedback on and refine performances and products (Fink, 2003).
Emphasize process over product for writing assignments. Assignments that are broken down into parts or drafts over time and graded with an eye on process — requiring and incentivizing the many steps along the way rather than only valuing the final product — have proven quite effective in inviting students to lean into their own work. This emphasizes the student's ability to explain why and how they made choices they did. For example, you might consider devoting some class time to the early writing, creating, and sharing of ideas. To foster metacognition, have students write occasionally about the process they are engaged in and, at the end, write a “process” paper in which they narrate how they got to their final submission.
Additionally, here are some small adjustments that can be implemented relatively quickly, for which there is evidence that students are more likely to complete the assignment themselves.
- Provide clear instructions, sufficient time to complete assignments, and some flexibility with deadlines. Pace larger assignments over the term, focusing on quality over quantity of work. If an assignment is due all at once and counts for a large portion of the grade, the temptation to cut corners will be greater.
- Provide multiple opportunities for self-, peer-, and instructor- feedback throughout the semester (e.g., individual conferences, self-assessment rubrics). Reassure students that you read closely and engage with their ideas.
- Be clear about assessment criteria, whether or not you use a rubric. ChatGPT is relatively good at writing seemingly coherent and readable syntax, which can be distracting. Being clear about the other aspects of an assignment (e.g., deep knowledge of theories or sources, close analysis, logical argumentation, and evaluation), can focus attention on what matters for the assignment. Consider clarifying how and to what degree you would like specificity, share examples of strong work and work that needs improvement, and consider asking for exact page numbers in sources.
- Require students to include materials that are only available in your classroom sessions or labs, or sources you have assigned in your course.
- Build in more on-demand, in-class writing that requires students to engage material with a greater degree of immediacy.
As the AI landscape continues to evolve, instructors and students can build AI literacy by incorporating it meaningfully into their coursework to practice skills and foster meta-conversations.
Like any assignment, it is worthwhile to consider the educational purpose of activities that incorporate AI. Will the assignment strengthen students’ writing skills? Their ability to analyze an article for credibility or spot logical fallacies in a piece of writing? Their ability to correct AI output or use AI for discipline-specific tasks in the workplace? Here are some concrete exercises that may help you get started:
Discuss. Facilitate a discussion with students about an article related to generative AI. This could be discipline-specific or about higher education and the writing process more generally. Here are some essays we have found useful for starting conversations:
- The End of the Take Home Essay (Corey Robin, 2023)
- You are Not a Parrot (Emily Bender, 2023)
- Artificial Intelligence Implications for Academic Cheating: Expanding the Dimensions of Responsible Human-AI Collaboration with ChatGPT and Bard (Jo Ann Oravec, 2023)
- Speculative Futures on ChatGPT and Generative Artificial Intelligence (AI): A Collective Reflection from the Educational Landscape (Aras Bozkurt, 2023)
Analyze. Generate a piece of AI text and ask students to analyze or critique it. These sentence starters (Mills, 2023) can help students frame their critique around clarity, bias, accuracy, and relevance and focus.
Compare. Model a generative AI tool alongside another discipline-specific tool and compare their performance as a group. Examples might include translation of archival materials or coding exercises.
Experiment. Allow students to experiment with AI for different tasks, such as brainstorming or data analysis, and ask them to document and reflect on the process.
Research. Encourage students to choose AI-related research topics for an assignment or presentation. Students may wish to explore ethical considerations of AI, such as the environmental impact of GPT, bias in AI-generated images, intellectual property issues, or content moderation practices.
More assignment ideas can be found through the TextGenEd project. Please note that at this time, it is not advisable to require students to create their own personal accounts or to input any student data into generative AI tools due to data privacy considerations.
This page is a collaboration between the Elkana Center and the Center for Academic Writing. It acknowledges the contributions of Kaitlin Lucas, Sanjay Kumar, Michael Kozakowski, Tamara Kamatovic, and Irene Lubbe.