A Framework for Considering How We Might Use AI in Classroom Activities

Students can acquire knowledge, concepts, and information in a variety of ways. This could include reading a statistics book, attending a lecture or watching videos on a topic in their class, or reviewing example problems. While students generally need to practice acquisition, AI can help! The author of a textbook can define a term like “p-value”. That author may even give an alternate definition. However, some students may benefit from a definition expressed a bit differently, perhaps by ChatGPT or another AI tool. With a proper prompt, students can ask a chatbot to create alternate definitions, generate example problems for many topics, and help highlight important ideas in a textbook.

Statistics instructors can and should provide opportunities for students to identify connections between various concepts in their courses. Synthesis may involve combining topics across different acquisition tasks from the same course, like investigating the relationship between confidence intervals and hypothesis tests. Synthesis also includes connecting ideas across multiple courses, like combining ideas from introductory courses on inference and probability to discover the ubiquity of linear combinations in the undergraduate statistics program at most institutions. Some synthesis tasks can be completed solely by AI if they are well-known, but many synthesis tasks are difficult or even impossible for AI tools like ChatGPT. Instructors can and should work to identify tasks where AI is helpful as well as those where it isn’t. Students can benefit from learning when and why generative AI is not a useful tool.

In many courses, instructors seek to help students develop new ways of thinking: students learn about statistical significance and even grapple with the tenuous relationship between statistical and practical significance. At their best, courses will also ask students to develop their own new ideas. While a student can ask a chatbot or other AI to create a new idea, large language models have a limited capacity to engage in true genesis of ideas. Instructors should illustrate examples of genesis by investigating questions such as “how did significance testing become popular?” or “why have some journals considered abandoning the use of null hypothesis significance testing?” By engaging with these sorts of questions, students increase their own capacity for genesis, which AI does not typically replicate effectively.

In many courses, we ask our students to create some piece of work, like a statistical report, and then iterate on that work to hopefully create a final product that is exemplary. When students want to make changes to their project, such as making a report longer or removing passive voice, generative AI can rapidly provide suggestions. Not all suggestions will be ideal, but the speed of AI makes this tool a great “partner” in editing and refining the work. Asking students to use AI to iterate on their work can be an effective strategy that illustrates responsible use of this powerful tool!