AI Considerations for Teaching and Learning

Generative AI applications like ChatGPT can assist students during the earliest stages of the writing process, auto-generating text for learners who are either stymied by writer’s block or stuck during the brainstorming process. When prompted, these systems can produce reams of raw content (of varying levels of quality and accuracy) that students can further evaluate, interrogate, and research. Using AI tools in this fashion can help students during the difficult preliminary stages of the composition process, creating a workable path toward their own inquiries and investigations (Gero, 2022; Krause, 2022; Weissman, 2023).

AI chatbots can also condense and summarize longer texts with only moderate error, potentially aiding students during the reading and research process. They may help in clarifying and explaining daunting or challenging texts in simple, digestible language. This function might potentially help learners (especially English language learners) gain a deeper comprehension of dense academic materials by making obscure prose and concepts more approachable and accessible (Anson and Straume, 2022; Warner, 2022). ​

When prompted, generative AI applications like ChatGPT can provide direct and immediate automated assistance for students struggling with grammar, mechanics, and syntax. They can identify, explain, and even correct basic grammatical mistakes. Additionally, AI language systems can function as a fluent conversation partner for informal language practice. This might be of particular benefit for English language learners and multilingual students who are still learning the basic mechanics of writing in English (Warner, 2022).

Generative AI applications also provide an avenue for discussing various facets of rhetoric, authorship, and academic integrity with students. They can function as the focal point of a broader conversation about the ethical questions posed by AI language systems, especially as their continued use and development alters our understanding of plagiarism and cheating. Using generative AI, we can help students develop their own style, skill, and voice as authors, particularly when we ask them to review and discuss their work in contrast to machine-generated texts (Fyfe, 2022; Grobe, 2023; Anson and Straume, 2022).

Although they draw from vast datasets of text, AI-powered chatbots remain limited to the information available to them at the time of their training. In other words, they cannot access or consult with external sources of information, nor can they self-correct or fill knowledge gaps with correct information. For example, ChatGPT often punctuates its responses with obvious fabrications, failing to maintain accuracy when tasked with generating knowledge outside its dataset. Users can also prompt ChatGPT to churn out obvious misinformation and nonsense, making it generate “garbage output” that is presented credibly and uncritically. In particular, it struggles when prompted to generate text about current events or recent developments, particularly on anything that has occurred after 2021 (Fyfe, 2022; Schatten, 2022; Grobe, 2023).

In addition to a penchant for generating misinformation, generative AI tools are incapable of conducting research and substantiating claims with credible evidence. When prompted to conduct research or cite secondary sources, for example, ChatGPT often fabricates research references and riddles the text with plausible sounding but entirely false or made-up claims, quotations, and scholars (Fyfe, 2022; Krause, 2022).

Generative AI’s understanding of American academic integrity and copyright standards is virtually nonexistent. Texts generated by language models have consistently committed frequent and flagrant acts of plagiarism, from direct, word-for-word plagiarism to misrepresenting others’ ideas as their own (Tutella, 2023).

Generative AI applications can sometimes employ biased or discriminatory language, repeat extreme or controversial viewpoints, or slip into explicit racism, sexism, homophobia, transphobia, and so on. Even when safeguards are added to filter out some of the more extreme or discriminatory positions, AI language systems are still prone to generating text that reinforces certain stereotypes, biases, and belief systems (Hutson, 2021).

Generative AI applications use predictive algorithms to generate text based on user input. Despite their relative fluency and adaptability, they cannot comprehend the meaning behind their words or exhibit human-like levels of critical thinking. This disconnect sometimes leads to text that sounds stilted, makes insubstantial claims, and lacks the subtle intricacy of human expression. AI can also commit rhetorical errors with relative frequency, pepper its texts with meaningless filler phrases, and over-rely on certain writing formulas. For instance, ChatGPT has a strong preference for five paragraph essays with short, three-sentence paragraphs and often overuses single word modifiers and transitions (Grober, 2023; University of Central Florida, 2023).

ChatGPT and other tools are third-party software systems that can track, collect, and share data from their registered users. For example, OpenAI’s privacy policy claims that it reserves the right to harvest and share user data but does not clarify for whom and for what purpose. This brings up potential concerns regarding privacy and security (Caines, 2023), and some students may be uncomfortable using or creating accounts with generative AI applications. Therefore, exercise caution around requiring students to use AI applications and consider alternatives for supporting student learning and meeting your instructional goals.

Questions regarding the intellectual ownership of AI-generated texts remain contentious, with no foreseeable resolution in sight. Users should err on the side of caution when it comes to claiming outright authorship of anything created by ChatGPT or other AI language systems (McKendrick, 2022). There is also a lack of clear guidance on how to cite material generated by AI in any of the major citation styles such as MLA and APA.

As with many free applications, we should consider that generative AI applications could someday require payment or offer “premium” services to paying users, bringing about potential equity and access issues. OpenAI has already introduced ChatGPT Plus, a pay-to-use service.

Communicate in your syllabus about expectations for students’ course-related uses of AI (Wheeler, Palmer, and Aneece, 2019). Set tone, routines, and guidelines early in the term and engage students in openly discussing the opportunities and limitations of AI, as well as what represents misuse of AI in your course or for specific assignments and activities. See Support Academic Integrity below for more guidance.

Apply the Transparency in Learning and Teaching (TILT) framework to communicate the purpose, task, and criteria for success for each of your course activities, assignments, and assessments. Are AI-related skills relevant? How will they help students meet learning outcomes? Use this exercise to explicitly state why, when, and how AI can or should be used or prohibited in your course (Winkelmes et al., 2016). Visit http://www.tilthighered.com for TILT templates, examples, and resources.

Provide a rubric—or work with students to co-develop a rubric—for all assignments and assessments. To support student understanding of assignment expectations, you could even have them use the rubric to evaluate AI-generated responses to the assignment prompt.

Avoid transactional engagements with students when addressing AI—do not simply provide policies about AI technologies in your course or recite assignment instructions. Instead, adopt transformative approaches that involve students in the review, feedback, and decision-making around your policies and assessment criteria. For example, you might craft a discussion exercise in which students examine the grading rubric, ask questions, offer feedback, and practice applying it to sample work. They can then reflect upon the rationale for the inclusion or exclusion of AI tools in the given assignment (Balloo et al., 2018).

Discuss with students the current and emerging roles of AI in real-world professional and practice settings and incorporate these connections into assignments when relevant. How might AI-related skills benefit students in their studies or careers beyond your course? Speak candidly about the ethical considerations of organizations and professionals using AI to inform decision-making, policy setting, and other aspects of work. Conversations can stay within the confines of your field of study or extend to broader contexts with which students have contact, including education, finance, government, healthcare, and law (Villasenor, 2023).

Openly communicating the university’s policy for academic integrity in your syllabus will help to level set expectations for your course as well as for students’ academic careers at the university. Facilitate a conversation around university expectations at the beginning of the term and, as needed, during your course.

• For potential language to include in your syllabus, see this sample syllabus statement for academic integrity and misconduct. 
• Ohio State currently has a definition of academic integrity.  Additionally, students are required to abide by the university’s Code of Student Conduct, which includes academic misconduct. It’s always best practice to include the university’s policy on academic misconduct in your syllabus.

The university does not currently have an official policy on the use of generative AI chatbots or other AI tools. However, if a policy specific to AI is adopted, it would be best practice to include that policy in your syllabus. You may consider other sample syllabus statements concerning generative AI applications, such as these from Princeton University and Northwestern University.

Discuss with students how the use of AI fits within the university’s policies on academic integrity and misconduct. For example, students may have the misconception that since a chatbot is not a real person, they can use the text it generates without plagiarizing or committing misconduct. Prompt them to consider the nuances of using AI-generated works and related ethical considerations to help them understand what is permitted and prohibited at the university and in your course.

Beyond university policy, you may have additional expectations for academic integrity in your course. While your course policy does not need to read like an official university definition, it should be clearly communicated and use inclusive, student-centered language. Include a statement in your syllabus and facilitate a conversation around your expectations at the start of term. Your course policies might address whether students can use specific AI applications such as ChatGPT, the conditions under which they are permitted to use them (e.g., generating ideas, developing outlines, creating rough drafts, revising and editing), how they should reference their use of AI, and who students should contact if they have questions about using AI.

If you decide to use any AI-detection tools in your course, clearly communicate how and when you will use them via multiple outlets (e.g., syllabus statements, policy discussions, Carmen announcements, in-class reminders, assignment instructions). Explain your rationale for using them to help students see the broader context of the affordances, limitations, and ethical implications of using AI.

There may be specific assignments or assessments in your course for which you definitely want students to use or not use AI applications. Beyond course-specific policies and syllabus statements, remind students of these expectations before relevant activities, assignments, and assessments. Include explicit recommendations or restrictions in prompts and directions, and clarify them in open discussions with students.

Using the Carmen course template can help you set clear expectations for each assignment in your course. With the assignment template, utilize the bullet points and icons in the Academic Integrity section to communicate your expectations (i.e., prohibit, permit, permit and encourage, require) for using generative AI applications. Use the other sections (Directions, Resources) to provide more detailed information. For example, if you are requiring or recommending that students use a specific AI application for an assignment, include information about the tool and additional guidance for technology requirements, troubleshooting, account setup, and student privacy. You might also remind students of how to provide proper attribution and use your preferred citation style for all resources they use, including generative AI applications.

Creating assignments that cannot be completed solely by AI-assisted technology can help to minimize student reliance on it. For instance, instead of having students write a policy brief, ask students to create a presentation outlining their research findings and policy recommendations for a specific intended audience. You can still be flexible in terms of presentation tools and format (e.g., recorded voiceover PowerPoint, Flip, Adobe Express) to give students agency of choice. Having students articulate their learning in multiple formats helps them reframe their learning and gives them rhetorical flexibility to communicate in a range of contexts (Selfe, 2007).

Explore the Ohio State Toolset and Additional Tools as you consider the variety of technologies and formats students can use to present their work.

When given an input, generative AI applications produce an output. But learning is a process that is not just about generating a product, and this is an important distinction for students who feel the pressure of high-stakes assignments and grades. Incorporate scaffolded assignments that build on one another and include different types of tasks (e.g., proposal, outline, literature synthesis, rough draft, peer review, final draft). By breaking down a larger assignment into smaller chunks, and having more frequent lower-stakes assignments, students experience the process while recognizing the connection between components. Scaffolded assignments also provide ample opportunities for feedback and revision, which allow students to refine their thinking and learning (Bean, 2011).

New AI chatbots may sound more human than earlier versions. Nonetheless, having students include personal reflections or connect to their lived experience for your assignments will bring in a human element that chatbots cannot adequately replicate. Students might reflect on personal and professional experiences, their growth in your course, and why course content is valuable to them. You can also ask students to describe their writing process and reflect upon the steps they took. Creating intentional opportunities for students to reflect on their learning strategies can help them become more successful, self-directed learners (Ambrose et al., 2010).

Generative AI applications like ChatGPT are trained on an enormous amount of data, but they have limitations that you can leverage. For example, most of the data ChatGPT was trained on is from before 2022 (at present). Therefore, you could ask students to relate their learning to a current event for which ChatGPT may not yet have enough information to properly generate a response. Guiding students to connect and apply their learning to current events can help them see the value of their learning, improve their engagement and motivation, and apply their learning to relevant real-world contexts (Ambrose et al., 2010).

Students who see value in and feel connected to what they are learning may be less likely to rely on AI-generated support to complete an assignment. Consider how you can integrate real-world content (such as case studies) and authentic tasks (such as project proposals or practice client sessions), to increase students’ motivation on assignments. As with connecting activities to current events, engaging students in tangible real-world tasks allows them to translate key course concepts into meaningful practice.

Students may be more likely to resort to generative AI tools if their workload is heavy or their assignments and deadlines feel unmanageable. Take stock of the amount, length, and sequence of assignments in your course to mitigate any unnecessary pressure on students.  It can be helpful to estimate the workload in comparison to your course’s number of credit hours.

Analyze data for any summative assessments (such as exams, essays, projects, or presentation) in your course that aligned to instruction that leveraged or limited AI. What insights on your approach to AI can you glean from students’ performance?

Consider data and observations from formative assessments related to AI, including quizzes, in-class practice activities, and student reflections. What insights on your approach to AI can you glean from students’ engagement and performance?

Plan how you will collect student feedback about the use (or non-use) of AI in your course. You can survey students about your instructional approaches, any barriers they encountered in your course, and what elements most supported their learning and success. You might also revisit any informal written reflections students submitted about their use of AI during the term.

The Small Group Instructional Diagnosis (SGID) is a service offered by the Drake Institute for Teaching and Learning that provides instructors with valuable student feedback through a focus-group style evaluation conducted by an instructional consultant. Information is collected on supports and barriers to student learning, as well as ideas for positive change. Your consultant can gather student feedback on specific areas of interest, such as your approach to AI in your instruction.

Your personal reflections on the teaching and learning process, particularly around new uses of AI and specific AI-related learning activities, will be helpful in determining what to maintain or change for the next iteration of your course.