Extending Office Hours: How AI Tools Can Support Test Prep
Students gather to study inside the Michigan Union. VP of Communications, Regents of University of Michigan
How this helps
Are your students stressing over studying for their final exams? Cramming into the earliest hours of the morning, coffee in hand? With the help of artificial intelligence tools, there are a variety of ways that you can support them, even during those 3 a.m. study sessions.
Start with a Guiding Framework
Rhonda Holton sees value in generative AI for students.
As the assistant director for technology-informed pedagogy at the Center for Research on Learning and Teaching in Engineering, Holton largely works as a consultant, using her expertise to advise instructors and directing departmental workshops on how to effectively incorporate technology into courses.
“I have instructors who have told me, ‘I would love to teach my course as though I am in a liberal arts college with only 20 students, but that’s not realistic; I have 250 students,’” Holton said. “The greatest potential for GenAI is how we can take what we already know as being good educational practices, like robust feedback, and close that gap between the ideal learning experience and just the realities of where we’re actually at.”
One resource Holton frequently provides instructors is a technology integration framework called TPACK, which stands for technological, pedagogical, and content knowledge. This framework identifies three types of knowledge instructors need to account for to effectively combine education and technology.
TPACK’s Three Types of Knowledge:
- Content Knowledge: “What are you teaching and what is your own knowledge of the subject?”
- Pedagogical Knowledge: “How do your students learn best and what instructional strategies do you need to meet their needs and the requirements of the lesson plan?”
- Technological Knowledge: “What digital tools are available to you, which do you know well enough to use, and which would be most appropriate for the lesson at hand?”
This framework outlines how these types of knowledge intersect and can help you, as the instructor, map out the role of technology in your courses.
U-M Tools
The University of Michigan also provides a variety of secure, vetted tools for instructor use.
Maizey: You can start by enabling Maizey in your course’s Canvas page. This allows you to upload your course material to create tutor-like study instruments for your students, such as study guides, module reviews, practice quizzes, and more.
Gemini Notebook: Similar to Maizey, Google’s personalized research and thinking assistant can use the materials you upload to create a variety of study assets, including infographics, flashcards, and even a podcast-style audio review that can help students navigate the lessons.
MConnector: Another option is to enable the MConnector tool. With this application, you can locate and create useful content to complete supported tasks.
Nick Gaspar, distance learning director and adjunct lecturer for the School of Nursing at UM-Flint, utilizes these tools himself and with his students. His first consideration is ensuring students understand what is acceptable, and what is not, when using AI tools. Establishing that expectation, and building trust and transparency, is an important factor when using AI with students.
“[If you] go to some other tool, Claude, ChatGPT, and then copy and paste your content in there, you’ve lost total control of where the content is going,” Gaspar said. “If you’re trying to help students with test prep, turn on the Canvas connector and then you have a built-in mechanism that you, the faculty member, can control.”
Need a step-by-step process? Gaspar helped outline one:
- Enable the tool(s) in Canvas. Both Maizey and MConnector can be connected to Canvas, though they support different goals. Learn more about their capabilities.
- Identify relevant course content to inform and provide context for the model (UM GPT). You can upload as little as one module of course content or as much as is desired.
- Test the tools and outputs. Behave like a student and ask the questions your students would ask. Determine whether the output is accurate and the model is talking to your students as you intended.
- Inform your students. Consider including an activity to help your students practice interacting with the tool. Model the behavior you want your students to follow.
AI Tutors
Those seeking a more comprehensive support tool can craft an AI tutor using the above tools and their own course materials.
Seeing a need for students to have more concrete and substantial studying options, Pete Bodary, a clinical associate professor and associate dean of undergraduate education at U-M’s School of Kinesiology, was inspired to explore if AI could assist as a study aid for learners.
Bodary and his former research assistant, Giselle Aronoff, developed what they dubbed “Wolverine Tutor,” a Socratic AI tutor. They built this tool using only Bodary’s course materials with the aim of providing guidance, not answers. This scaffolded support tool prompts students and then evaluates and identifies potential gaps in their understanding of a particular concept. The tutor provides more support if the student demonstrates a need, but backs off and provides a mastery check once the student exhibits deeper understanding.
“Our primary emphasis in all of this has been to cultivate productive struggles,” Aronoff said. “This isn’t functioning by any means as an answer bot.”
Bodary said that one of the AI tutor’s most impactful benefits is its ability to offer feedback and support when it’s most convenient to the student. It’s a solution to the large-class-size problem that limits many professors’ availability to focus on any one student.
“I teach courses that are 50 to 80 students and, if they were all coming to office hours, I don’t know if I could manage that,” Bodary said. “This is a way we can get them something that is much more like one-on-one feedback where it’s totally scalable.”
Final Considerations
Ultimately, integrating AI test prep tools into your courses is about finding the right balance.
- To make the right decisions for your teaching environment, understand the capabilities of the technology. “Wariness is understandable and in some cases very warranted, but don’t let that wariness stop you from understanding what the technology is and what it’s capable of,” Holton says. “Hesitancy is fine, but don’t allow that to turn into ignorance.”
- Keep learning objectives specific and digestible: Bodary recognizes the limitations of the tool regarding its adoption of learning objectives. “The Socratic tutor works well when learning objectives are clear and finite, and it works much less well when the learning objectives are super broad and vague.”
- Build trust with your students: “Learning cannot happen without trust between the instructor and students,” Gaspar says. “Bring up a clear and open dialogue [of values and expectations] … if we’re not talking about how the technology is being used, we don’t know. And be aware that not all students are going to interact with it. The AI tool should not replace traditional study material, but rather enhance it.”
Resources
GenAI Workshop Series — A series of hour-long workshops available on-demand that cover the basics of GenAI and its applications in academic and everyday settings. Those who complete the courses, which are slotted into beginner, intermediate, and advanced levels, are eligible to earn GenAI badges.
Teaching with GenAI — A self-paced course for U-M instructors and staff seeking skills on integrating GenAI tools in their teaching practice. Those completing all eight modules receive a Teaching with GenAI badge.
Additional ITS and U-M resources:
- GenAI 101 An Overview of Generative AI (self-paced microlearning course)
- Generative AI Prompt Literacy (self-paced course)
- U-M Generative AI Website
- ITS AI Services and Roadmap at U-M
- Generative AI on the Academic Technology@Michigan website
- Getting Started with Generative Artificial Intelligence: U-M Instructor Guide
- Course Policies & Syllabi Statements
- Learning with GenAI