Educators have expressed concerns about students outsourcing the labor of learning to generative artificial intelligence (AI) chatbots such as ChatGPT rather than engaging with course learning. Here we offer eight strategies to help instructors think through how generative AI chatbots can best become part of the teaching and learning space and enhance student learning.
We order the strategies sequentially to guide instructors through a series of possible pedagogical choices. This article addresses AI chatbot text generation and student writing as our central case, but the strategies apply to AI content generation more generally.
1. Orient your pedagogical choices around your course learning outcomes
Rather than considering writing and generative AI in general or in the abstract, focus on the learning outcomes of the specific courses you teach.
- What do you want students to learn and do by the end of your course, or a class session, or an assignment?
- Will writing help students make progress toward those learning goals, and can students demonstrate what they have learned through writing?
- Can students learn, and demonstrate their learning, through other activities and assignments as well?
Articulating learning outcomes at different scales for students can help you decide whether writing activities and assignments offer the best, or only, way to give your students the practice they need to make progress, and the medium in which to demonstrate their progress.
2. Diversify your course activities and assignments
If writing is not necessary for students to meet course learning goals, consider what other activities and assignments students might do, and what accommodations and other steps you should take to make course learning accessible for all students.
Examples include requesting students to:
- Develop and distinguish between different interpretations of a text through oral presentations
- Develop and refine an argument through class discussion or debate
- Communicate findings through live graphing or diagraming
- Tell stories visually through physical posters
- Work collaboratively in groups on quizzes or exams
3. Test generative AI yourself
When you decide to use writing as a preferred way for students to make progress toward learning goals, you might spend some time using freely accessible AI chatbots (e.g., ChatGPT) to see what kinds of outputs the chatbots generate when you enter your class prompts. You will probably need to repeat this exercise at least once a quarter, as tools continue developing.
Many students may also choose to use more sophisticated paid versions; you might request departmental or other funding to try those out as well.
Some factors to consider:
- What course readings and other sources does the chatbot currently have access to?
- Are the chatbot’s assertions and citations accurate?
- Do the chatbot’s sentences meet grammatical, stylistic, and other criteria?
- Does the chatbot generate interesting or novel ideas?
- Do certain prompts generate more interesting or sophisticated outputs than others?
- Do certain prompts generate flawed outputs that might prove useful to course learning?
4. Incorporate live writing in class
Consider how you can incorporate informal and low-stakes writing activities in class to foster creative exploration, engagement with course materials, critical thinking, and learning. These “writing-to-learn” activities uncover for students the uses and values of writing beyond summative evaluation as a demonstration of learning (e.g., submission of work to receive grades). Live writing can be solo or collaborative.
Examples include writing in class to:
- Access and bring forward prior knowledge
- Generate questions about course readings and key or unfamiliar concepts
- Identify assumptions in course materials
- Develop research questions or a research plan
- Translate a specialist concept to a non-specialist audience
- Connect course topics to personal experiences
5. Encourage writing outside of class
For writing that you would like students to do outside of class and on their own, unaided by other people or generative AI, consider your pedagogical rationale. What do you want students in your course to learn and why does it matter? How does their learning depend on engaging and sometimes struggling with ideas and words themselves, rather than outsourcing the work to other people or generative AI?
Examples of generative and engaging writing include:
- Connecting writing activities and assignments to students’ experiences and interests, to demonstrate relevance and meaning to their lives and goals
- Encouraging students to set their own writing goals and assume ownership of their learning and progress
- Communicating the value of writing as a process and not just a product
- Scaffolding writing through pre-writing, drafting, revising, and reflecting
- Recognizing student progress and promoting a growth mindset that emphasizes practice and persistence
- Situating writing within a supportive course community that includes formative feedback and check-ins with instructors and TAs, peer discussion, and feedback, workshopping, and sharing of progress and insights
6. Cultivate intellectual transparency and responsibility
To guide students to specific types of writing aid that do not interfere with course learning, and might even enhance it, you can make explicit for students what you allow or encourage. Just as instructors can communicate types of acceptable or productive peer collaboration in a course, instructors can also make explicit which uses of generative AI are acceptable or even encouraged, and request that students cite such uses.
Instructors might even consider reflective exercises or “methods” sections that ask students to detail the tools, queries, and outputs used in their writing process and how the results helped shape the content and form of the paper.
Example uses of generative AI students might cite include requests for:
- further information and context, such as definitions of terms, or descriptions of historical events
- additional resources to supplement course learning
- explanations pitched at varying difficulty levels
- additional ideas
- alternative theses or arguments
- questions, objections, or other feedback on their ideas and writing
In addition, just as relying on other people brings with it risks of being led astray, instructors should educate students about the need for evaluating generative AI outputs, rather than relying on them unreflectively.
7. Build generative AI into your course activities and assignments
Instructors can design activities and assignments that intentionally build in the use of generative AI to give students extra practice in developing higher-order skills. As generative AI tools develop, it might eventually become educators’ responsibility to weave generative AI into instruction and give students practice and guidance in using and learning from the tools effectively.
Examples of how students might use generative AI include:
- Summarizing long or dense material for further use in course activities
- Critically evaluating text the chatbot generates
- Comparing and contrasting r own responses to the chatbot’s responses
- Synthesizing ideas or text from multiple AI-generated responses
- Receiving and incorporating feedback on their work
- Practicing dialogue and role-play
- Translating texts from other languages that contribute to other course activities
Instructors can also guide students through critical discussion and evaluation of generative AI tools themselves. Examples include:
- Asking students to connect their personal lived experience to course materials and concepts and to contrast this with AI chatbot outputs, to illustrate the limitations of systems with no lived experiences to draw on
- Demystifying the technology, resisting anthropomorphism, and discussing emotional reactions to interacting with AI chatbots
- Fostering awareness of digital literacy as a skill set; that is, proficiency in locating, evaluating, and communicating information found through digital tools
- Addressing diverse representation, anti-bias, counter-stereotyping, and inclusion in the training and outputs of generative AI (and machine learning) systems
- Addressing intellectual and artistic property rights, and human rights, in the development and training of generative AI tools
- Presenting the selection of the right AI tools, and effective work with them, as skills in their own right
8. Craft course policies that align with your course design
Once you have worked through the steps above, you will be equipped to answer hard questions from your students about why and when they may, or may not, use generative AI for writing in your course. All of your answers will come down to your course learning goals for students, and how aided or unaided writing from students will help them make progress towards them. Your course policies around student use of generative AI will codify those answers and their pedagogical justifications.
Examples of course policies an instructor might develop include:
- a reminder that using generative AI to “substantially complete” an assignment or exam by entering the prompt and submitting the output as one’s own is disallowed per Stanford’s Honor Code
- banning use in writing assignments, but permitting cited use in other kinds of assignments
- banning any use in generating preliminary ideas, theses, outlines, or other content, but permitting cited use in generating additional ideas or getting feedback
- banning any use in generating ideas, outlines, or text for more novice students, but permitting cited use for more advanced students
- banning any use earlier in the course, but permitting cited use later in the course
- permitting cited use for the explanation of concepts and theories
- permitting cited use for literature reviews in advanced research papers
- permitting any cited use, along with evidence of student contributions to the writing process and product
- How to Best Use ChatGPT: A Basic Training Guide, Stanford UIT
- Teaching Strategies to Support the Honor Code and Student Learning, Stanford Teaching Commons
- Curricular Resources about AI for Teaching (CRAFT), Stanford GSE
- Newport, C. (2023). What Kind of Mind Does ChatGPT Have? The New Yorker.
- McMurtrie, B. (2023). How ChatGPT Could Help or Hurt Students with Disabilities. The Chronicle of Higher Education.