AI Study Buddy

The instructional practices shared in this article are ideas for exploration, not requirements for any instructor. They were developed by Northwestern IT Teaching and Learning Technologies in partnership with the Provost’s Generative AI Advisory Committee. Please note:

• Accessing Copilot via your Northwestern credentials is the recommended path for accessing a generative AI tool. Questions about whether a risk assessment has been performed or an institutional contract exists for a specific AI tool can be directed to the Northwestern IT Information Security Office (security@northwestern.edu). Procurement of new AI tools should follow university processes and policies regarding licensing and third-party risk assessments.
• Output from large language models (LLMs) can include false or incorrect information. Verifying accuracy via other sources is a critical practice for instructors, students, and staff to engage in when using LLMs.

In winter 2024, Northwestern IT announced that access to Microsoft’s large language model (Copilot) would be provided at no cost to students, faculty, and staff. Access to this resource, which includes use of OpenAI’s GPT large language model (LLM) means several key things for the Northwestern community. First, students and instructors have access to the most updated version of a large language model. Second, having a generative AI resource that is provided at no cost to students and faculty means we can explore opportunities for generative AI to support student study needs. As part of Northwestern’s “AI and Teaching” series, McCormick School of Engineering Adjunct Assistant Professor Kate Compton spoke of the great potential of a ChatGPT-like model to serve as “the world’s greatest tutor” to boost learning outcomes. In the video, she advocates for shifting from a content generation or “service” approach to an “interview” mentality, where students explain things to the generative AI chatbot they are interacting with.

Video Resource

"Asking the Right Questions": 

Kate Compton’s advice for turning generative AI chatbot into a tutor from the series, "AI's Implications for Teaching and Learning," sponsored by the Provost’s Generative AI Advisory Committee.

AI-Supported Retrieval Practice

Using a large language model like Copilot as a study tool rather than a source of content generation offers interaction with generative AI that centers the learning process. In particular, the study practice of retrieval can significantly improve test performance and understanding and comprehension of complex, educationally relevant material. To learn more about retrieval practice, check out “Enhancing Learning by Retrieval: Enriching Free Recall with Elaborative Prompting” and "Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping."  

Step One: Have a Conversation With Copilot 

The following table contains three types of questions to try with Microsoft’s Copilot chatbot. Each of these questions encourages students to think critically about the material they are studying and requires them to recall information, put things into their own words, and go beyond the surface-level in their responses to the chatbot.

Step Two: Verify and Integrate

Because large language models use pattern prediction to create their responses, “hallucinations” and incorrect information can be present in the output. Using the “interview” method during prompting helps reduce this likelihood, but learning how to verify outputs is a critical step in using generative AI.

Fortunately, verifying the content provides another opportunity for students to continue their study process. Provide students with these ideas for verifying generative AI output.

• What sources did Copilot cite during the exchange? Do you need to check against another source to verify the information that was given?
  • Use this Libraries resource for ideas on how to evaluate a source.
• Verify against your course materials and texts.
  • Confirm that information is correct by checking it against course materials, lecture notes, and other resources from an instructor. Connecting your Copilot practice with course materials will help you make connections between ideas and move information into your long-term memory.

Having a self-testing conversation with Copilot can help students access the study benefits of practice testing, which is a low or no-stakes practice that students can engage in on their own. Following up self-testing with verification of the information covered in the chat by checking lecture notes, rereading course materials, or reviewing previous assessments can be part of distributed (spread out over time) or interleaved (mixing various kinds of problems or materials) study practice. For more details on learning techniques check out, “Improving Students’ Learning with Effective Learning Techniques: Promising Directions from Cognitive and Educational Psychology." For more study strategies, share Strategies for Success from Academic Support and Learning Advancement (ASLA) with students.

Sharing AI Prompts with Students

To explore using Copilot as a study tutor, try some of the prompts and see how they perform. Tweak the prompts by providing more details to the chatbot about the behavior you want to see during the student study process. Once you have some prompts that seem helpful, you can share them with your students. When sharing with students, remind them that using Copilot is optional and that they have other study resources available to them at Northwestern from ASLA and the Writing Place.

1. To help students understand how AI as a study tool might be relevant to them, share the following statement (or edit it to your uses) with your students: “AI tools like Copilot can offer us a way to more deeply engage with classroom content, helping us to prepare for assignments and exams. You should approach Copilot as a tutor who will ask questions to help you study. This is a two-step process, which begins with a conversation with Copilot and continues with you verifying and integrating what you practiced with course materials and lecture notes.”
2. Share one or more of the example prompts provided in the table. We recommend sharing them in the immediate lead-up to a major assessment or during a unit with particularly difficult concepts.
3. Share the output verification questions to help students interact critically with information provided by an LLM.
4. Reflect: As an optional assignment, ask students to share which study prompt they used and how the experience went for them. Did they need to refine their prompt? What was helpful (or not) about what Copilot did? Would they use this as a study tool again?

As a pedagogical tool, AI like Microsoft's Copilot has the potential to provide students with a more interactive and engaged study experience. The proposed questions above ask the chatbot not to generate content, but rather to support the learning process overall, especially the process of recall/retrieval and writing comprehensively about a topic.

Continue Exploring this Topic

• Generative AI and Teaching for Fall 2024
• Generative AI on Your Terms: Data and Privacy 101
• Breaking Down the Writing Process with AI