.png)
Caitlin K. Kirby, Michigan State University
Min Zhuang, Michigan State University
Imari Cheyne Tetu, Michigan State University
Stephen Thomas, Michigan State University
Keywords: Generative AI, Course Design, Teaching Philosophy
Key Statement: We offer a framework to help navigate a response to generative AI in higher education classrooms through a series of guiding questions.
Introduction
Generative AI in higher education has been widely discussed following the record-breaking uptake of ChatGPT in November 2022 (Dwivedi et al., 2023; Gordon, 2023). Generative AI models function based on their training with large amounts of data to recognize context and meaning to suggest the likely next word or series of words based on a query (Deng & Lin, 2023). The paid version of ChatGPT has been continually upgraded and can also create and read images, transcribe voice conversations, and create code in multiple languages (OpenAI, n.d.).
Relevance
According to a global survey in five countries, nearly 74% of university students indicated that they planned to use ChatGPT for academic tasks (Ibrahim et al., 2023). Concerns about academic dishonesty, the replacement of human labor by generative AI, and the ethics around generative AI use are paramount in discussions around generative AI in higher education (Ibrahim et al., 2023; Sallam, 2023).
Students can use generative AI for academic tasks such as brainstorming, drafting, or improving writing; summarizing, explaining, or synthesizing texts; and data analysis (Sallam, 2023). Generative AI tools can identify grammatical issues and provide tailored suggestions to edit students' writing. Students can seek personalized clarification on challenging concepts. Generative AI tools can refine research topics, support data analysis, and format references. One important limitation of generative AI is its “hallucinations,” where it makes up facts or references (Deng & Lin, 2023), a concerning feature for engaging with it academically.
Framework for Developing Your Approach to Generative AI
AI can complete so many cognitive tasks that it can be overwhelming to comprehensively address generative AI in your courses. We have created a set of guiding questions to serve as a framework for developing your approach to generative AI. Although the capabilities of generative AI tools will continue to expand, our guiding questions are designed to maintain relevance despite shifts in technologies.
We offer two initial check-in questions to educators. Throughout this piece, discussion/reflection questions are highlighted with green text. They are also found in the Discussion Questions section for easy access.
What are the gaps in your understanding about generative AI’s applications in your discipline or courses? How can you address those gaps?
To what extent are you interested in integrating generative AI into your classroom?
These questions can help frame the next steps in your inquiry on generative AI. For those who want to improve their understanding of generative AI’s capabilities and outputs, consider which topics feel most important for you to learn about first, such as: how generative AI performs on common assessments in your discipline; what tools are available to students; and how others are modifying course materials to discourage the use of AI or integrating AI into assessments.
Generative AI and Your Discipline
After a sense of initial impressions, our guiding framework asks you to think broadly about your discipline and the skills students acquire in your program. Based on your understanding of generative AI’s current capabilities, consider which of those skills may become more or less important in a world with widely available generative AI. For example, in technical writing, editing for concision may be achievable with generative AI; however, understanding context enough to ensure that the writing is understandable will likely still require human skills. This is a theoretical exercise without one right answer, but your response reveals key foci as an instructor.
It is important to explain to students that they will need to possess certain skills in their future careers, particularly when requiring students to complete tasks without AI. This transparency helps students contextualize their coursework in the context of mastery of skills, which reduces the likelihood of students engaging in academic dishonesty (Krou et al., 2021; Yang et al., 2013) or simply failing to put in necessary, foundational work. As you think about your discipline relative to generative AI, consider:
What fundamental skills do students need in your courses or disciplines?
How does this shift in a world with widely available generative AI?
Even if some tasks can be accomplished with generative AI, are the concurrent skills necessary to prepare students for more advanced tasks?
How do you communicate these skills and their relationship to your course content to students?
Generative AI and Your Assessments
It is worth considering how easily generative AI could complete your course assessments. Based on current widely available generative AI capabilities, we outline considerations of where the work is performed, what is being assessed, assessment complexity, and assessment context (Table 1).
Table 1. Assessment Characteristics and AI Potential
Characteristic | High(er) AI Potential | Low(er) AI Potential |
Where | Out of class | In class |
Online | Physical/tactile materials | |
What | Higher stakes assignments | Lower stakes assignments |
Final products with no iteration or view of the process | Scaffolded processes with iteration and feedback incorporated | |
Complexity | Individual work | Team work |
Assessments that are lower on Bloom’s taxonomy | Assessments that are higher on Bloom’s taxonomy | |
Text-only assessments | Multimodal assessments (e.g., incorporating video, voice, or visuals) | |
Do not require prioritization of various criteria within the assessment instructions | Require prioritization or more heavily weighting certain components in the instructions | |
Context | Assessments about basic factual or abstract knowledge | Assessments specific to in-class discussions or activities |
Assessments less related to future career or personal interests | Assessments directly related to students’ future careers or personal interests |
This table is modified from Monash University (2023).
After reviewing Table 1, consider:
How are your major assessments impacted by widely available generative AI? For those that are strongly impacted, how could you:
Integrate AI to further students’ learning and achievement of course objectives?
Modify the assessment to reduce the impact of generative AI?
Generative AI and Your Course Design
Considering the broader design of your courses, we present six course archetypes to demonstrate how AI might impact student learning and assessment of student work (Table 2). These simplified archetypes do not fully describe an individual course, but they provide examples for thinking through when and how assessment and learning is happening in your courses and how the process might be impacted by generative AI. In some cases, you might choose to enhance opportunities to interact with generative AI, while in others you might choose to limit students’ use of generative AI in order to focus on skill-building.
Table 2. Course Archetypes and AI Impact
Archetype Name | Distinguishing Features | Relationship to Generative AI |
Examiner (e.g., a biology course with lecture sessions and students assessed with only a mid-term and final exam) | Class focus is on delivering content | Low opportunity for students to use generative AI on assessments |
No graded homework assignments | Provides little preparation or incentive for students to engage with generative AI | |
Summative assessment via few high-stakes in-person exams | ||
Homeworker (e.g., a writing course where students complete several writing drafts through a semester and have a final paper at the end) | Class focus is on delivering content | High opportunity for students to use generative AI on required, graded assignments |
Skills developed on graded homework assignments | Consider modifying assessment features (Table 1) to reduce generative AI use | |
Summative assessment consists of project(s) or large written assignment(s) | ||
Experiencer (e.g., a science lab where student engage in experiments in class and submit lab reports after class) | Class focus is on experiential learning | Low opportunity for students to use generative AI for completion of experiential components |
Reports and reflective components can be generated with AI if students provide it with their description of the experience | ||
Project-Based Promoter (e.g., a general science course where students design a team experiment and submit components of a scaffolded report throughout the semester) | Class focus is on learning through team-based activities | Lower opportunity for students to use generative AI on multimodal project components |
Team-based project assignments with multiple iterations and feedback integration | Moderate opportunity for students to be able to integrate feedback into written components using generative AI if they provide their drafts and feedback | |
Flipper (e.g., a writing class where students read content before class, and spend class time working on their drafts) | Spends class developing skills individually | Opportunity for AI use depends on where graded assessments are taking place |
Content knowledge transfer happens out of class | Consider completing assessments in class or offering more credit for in-class assignments | |
Onliner (e.g., a computer science course where students follow along with lecture videos and submit coded assessments for evaluation) | Synchronous or asynchronous content delivery online | High opportunity for students to use generative AI in assessments |
| Consider modifying assessment features (Table 1) or whether proctoring may be appropriate to reduce generative AI opportunity | |
Students complete assessments on their own time | Difficult to address at the course level and may require new, as-yet-unknown approaches |
Generative AI and Ethical Considerations
Ethical considerations are key in navigating generative AI. The topic deserves more space than we can dedicate, but we offer suggestions for getting started. Access to generative AI tools may differ based on students’ technological skills, internet connectivity, and income, an extension of the digital divide and an ethical issue in and of itself. Many generative AI tools integrate the data from their user base into their training modules or otherwise take ownership of data that is put into their models, leading to concerns over data sharing and confidentiality, as well as bias in AI’s responses and using authors and artists’ works without permission or compensation (Sallam, 2023). Individual institutions often have their own policies on generative AI to consider. As you explore these ethical issues, consider:
What level of technology access is reasonable to assume for your students?
How can you support access for those without it?
What are your institution’s policies about generative AI use for instructors? For students?
What is your discipline’s culture around generative AI use? Is it allowed but frowned upon? Or embraced wholeheartedly?
In terms of data and privacy, how do you feel about asking students to interact with generative AI?
How might bias in AI’s training data be apparent in your discipline?
What, if any, alternatives do you need to provide students for conscientious objectors, accessibility, or Universal Design for Learning perspectives?
Conclusion
As generative AI becomes integrated into workplaces, scholarly work, and students’ workflows, we have the opportunity to take a broad view of the role of generative AI in higher education classrooms. Our guiding questions are meant to serve as a starting point to consider, from each educator’s initial reaction and preferences around generative AI, how their discipline, course design, and assessments may be impacted, and to have a broad view of the ethics of generative AI use.
Discussion Questions
Detailed discussion questions are presented throughout the paper in green. In summary, they follow:
What are the gaps in your understanding about generative AI’s applications in your discipline or courses? How can you address those gaps? To what extent are you interested in integrating generative AI into your classroom?
What fundamental skills do students need in your courses or disciplines? How can these be supplemented with, taught using, or considered in light of AI?
How are your major assessments impacted by widely available generative AI?
What ethical considerations might you consider in your discipline and institution relative to the introduction of generative AI?
References
Deng, J., & Lin, Y. (2023). The benefits and challenges of ChatGPT: An overview. Frontiers in Computing and Intelligent Systems, 2, 81–83. https://doi.org/10.54097/fcis.v2i2.4465
Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., Baabdullah, A. M., Koohang, A., Raghavan, V., Ahuja, M., Albanna, H., Albashrawi, M. A., Al-Busaidi, A. S., Balakrishnan, B., Barlette, Y., Basu, S., Bose, I., Brooks, L., Buhalis, D.... & Wright, R. (2023). “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71, 102642. https://doi.org/10.1016/j.ijinfomgt.2023.102642
Gordon, C. (2023, February 2). ChatGPT is the fastest growing app in the history of web applications. Forbes. https://www.forbes.com/sites/cindygordon/2023/02/02/chatgpt-is-the-fastest-growing-ap-in-the history-of-web-applications/
Ibrahim, H., Liu, F., Asim, R., Battu, B., Benabderrahmane, S., Alhafni, B., Adnan, W., Alhanai, T., AlShebli, B., Baghdadi, R., Bélanger, J. J., Beretta, E., Celik, K., Chaqfeh, M., Daqaq, M. F., Bernoussi, Z. E., Fougnie, D., Garcia de Soto, B., Gandolfi, A., … Zaki, Y. (2023). Perception, performance, and detectability of conversational artificial intelligence across 32 university courses. Scientific Reports, 13(1), Article 1. https://doi.org/10.1038/s41598-023-38964-3
Krou, M. R., Fong, C. J., & Hoff, M. A. (2021). Achievement motivation and academic dishonesty: A meta analytic investigation. Educational Psychology Review, 33, 427–458. https://doi.org/10.1007/s10648-020-09557-7
Monash University. (2023). AI and assessment. Learning and Teaching: Teach HQ. https://www.monash.edu/learning-teaching/TeachHQ/Teaching practices/artificial-intelligence/generative-ai-and-assessment
OpenAI. (n.d.). Models. https://beta.openai.com/docs/models.
Sallam, M. (2023). ChatGPT utility in healthcare education, research, and practice: Systematic review on the promising perspectives and valid concerns. Healthcare (Basel, Switzerland), 11(6), 887. https://doi.org/10.3390/healthcare11060887
Yang, S., Huang, C.-L., & Chen, A.-S. (2013). An investigation of college students’ perceptions of academic dishonesty, reasons for dishonesty, achievement goals, and willingness to report dishonest behavior. Ethics & Behavior, 23(6), 501–522. https://doi.org/10.1080/10508422.2013.802651
About the Authors
