Brainstorming with an LLM Student Activity

3 students brainstorming with a robot on the laptop screen.

My LLM brainstorming student activity was accepted to SIGCSE TS 2026's Special Session ACM Generative AI Task Force Special Session: Teaching with Generative AI: Tools You Can Use Today. This special session asked the SIGCSE community to apply to present how they use generative AI (GenAI) in their teaching. Those accepted would then present what they do during the session.

[Also posted on Medium.]

This blog post is a companion piece to my presentation, so that people have more to reference than just my slides. It also includes all the nitty-gritty details people can use and copy-paste.

Context

But first context, because context matters! This is for my CompSci 216: Everything Data (here is the link to the Spring 2026 offering; later links will also point to this semester's instance). It is a post-CS2 elective data science course with an average of 111 students over the last five semesters (Spring 2024 to Spring 2026). We are on the 15-week semester system, and it has a 4-5 student group project that is semester-long. The project has 4 milestones and 5 deliverables. The first milestone, the Initial Plan, is due in the 5th week of the semester, after we spend about 2.5 weeks forming groups. So, as far as the students are concerned, it is "early." The Initial Plan serves two purposes: (1) to get the students thinking about what they want their data science project to be, and (2) to get them to think through how they will collaborate for the semester. The deliverable is a brainstorming reflection and a collaboration plan.

Activity - Website Part

On the Initial Plan web page, I have directions and requirements for the assignment. The brainstorm is part 1 after the general directions. It is worth 40% of the milestone, and the students have two options for the brainstorm. One is a mind map, and the other is using a large language model (LLM) chatbot of their choice. I give them two because some students prefer to limit their use of GenAI, and I wanted to support that.


The LLM instructions tell them how to access DukeGPT services and obtain a paid ChatGPT account through Duke. It then gives guidance on how to prompt the LLM (including a prompt template), requires them to do 2-3 rounds of prompting in the same chat window, and to add their chat log as an appendix to their milestone submission. I'll include the full text on the website at the end of this section.


I also provide an example chatgpt.com chat log of me using the template and prompting a few rounds with it. After the brainstorm, I have the students also answer reflection questions, with one question specific to the kind of brainstorm they did. The point of the reflection is to have them think critically about their process. Brainstorming is a skill that students should practice, and this is a low-stakes way for them to do so in my class.


Here is the full text on the website:

For the discussion with an LLM, do the following:


  1. Tell the LLM you are brainstorming for data science projects, what your group’s interests are that could be potential sources of data, and that you need to find the data yourself.

  2. Ask it what ideas it has for your project.

  3. Tell it what ideas you liked, didn’t like, why a suggestion isn’t a good one, etc.

  4. Do at least 2-3 rounds of steps 2 and 3 with the LLM.

  5. Put your chat as an appendix in this submission.


Here is a template you can use as your prompt:


Context: We are intro to data science Duke students, and we want you to help us brainstorm for a semester-long data science project. Here is the course website (https://sites.duke.edu/compsci216sp2026/) read through it to understand our context.


We are interested in the following topics: TOPIC_1, TOPIC_2, and TOPIC_3. There are COUNT of us in the group, and generally the project should require enough work for that many people, sometimes 1 research question per person, but not necessarily.


Ask us questions to help us come up with ideas or give us multiple ideas on what our projects can do, suggest where we could find data sets for that idea, and try to connect our interests together if possible. Start by asking us clarifying questions to help us brainstorm better.


Here is an example of using the template using chatgpt.com. Note, you must have an active ChatGPT Edu account through Duke to see this.


After your brainstorm (regardless of if you used the mind map or an LLM), reflect by answering the following questions:

  1. Why did you choose the method you used?

  2. What patterns do you see in what you found interesting?

  3. What research topics or questions did your group generate from this brainstorming? Which of these ideas can you see your group potentially pursuing?

  4. Do you feel like more brainstorming is needed before you find a topic?

  5. If you used

    1. The mindmap: Did you find your brainstorming narrowing or diverging as you discuss ideas to write down?

    2. LLM: Did it help you find a project your group is interested in? Or did it just generate text that all seems reasonable?


Whether you choose to create a mind map or use an LLM, use this exercise to brainstorm project ideas that your group collectively believes are interesting, relevant, and worthwhile to your time in this course.

Activity - In-Class Part

After students have their groups and a few days before the initial plan is due, I discuss the initial plan milestone and its rationale in class. See the slides from my talk in the special session for examples. The main rationale I give them is to get the groups started thinking about what data analysis project they will do and how they will organize themselves. I also explain that this is something they can add to their portfolio for job interviews, but that requires a frank discussion now about how they will work as a group and how they can publicly share that work.


After the rationale, I then show them what I expect from the brainstorm. I show them an example mind map and the prompt. I also go to the actual chat log example and walk through it a little. I especially highlight that my responses to the LLM are not one-word answers but longer statements that answer the questions it generates and provide explicit guidance and directions for the LLM to follow.

Conclusion

Overall, I think the students appreciate having a sanctioned way to use GenAI in their learning. I've also been slowly incorporating how students can use GenAI into my courses, including how to disclose that use. Changing and updating is a slow process, but given that GenAI is now part of students' learning environments, we should find ways to address it, just like any other technology that could heavily impact their learning. I know that for many it's a struggle and can be draining, but I believe we can do this like the thoughtful and careful teachers that we are, one step at a time.


If you have questions or thoughts on any of this, feel free to leave a comment or send me an email!

Comments

Post a Comment

Popular posts from this blog

How I track my to-do list

This is also on medium . I use Trello  to track all the things I need to do. Trello, at its core, is just a to-do list management tool. It is versatile to many organization strategies and is great for a single person or a team to organize the tasks that need doing. However, this versatility can be very daunting as a novice because you are just starting to figure out a system to keep track of what you need to do. And a novice rarely does well in an environment where there are many ways to do the same thing, a.k.a. a versatile system. A beginner does not have enough experience to determine what technique is better than another. Moreover, a novice finds it challenging to keep track of the many ways to do the same thing. So, if you are a novice, here are my suggestions for how to use Trello to track your personal to-do list (how I use it to organize my research team(s) and teaching team(s) I'll leave for a different blog post). First, I'll go over the fundamental way I use Tre...
Read more

A Computer Science Concept that Helps Me Manage Overwhelm

Image
Spring 2025 was rough for me. I was teaching a new course alongside one of my usual courses, which had over 100 students. On top of that, I was learning the ropes for the spring term responsibilities of my new role as Associate Director of Undergraduate Studies in the department. [ Also posted on Medium. ] In one of my lower moments of feeling weighed down by it all, I asked Microsoft Co-Pilot to create an emoji to express it all. This is what I ended up with. And it now lives in a few of my Slack workspaces to give some levity during those moments of overwhelm that never seem to entirely stop happening. That aside, overwhelm is real, and I feel my peers and I are especially prone to it in academia. And a computer science concept that helps me be realistic about what I can actually do is priority heaps. So I thought I’d share in case you also need a useful metaphor, and to ask you all if you have a metaphor of your own that helps when in overwhelm. What’s a priority heap, you ask? It’s...
Read more

Webinar: How to Create and Use Formative Assessments at Scale

Codio  asked if I'd be interested in sharing my expertise on scalable formative assessments through a webinar  on July 14th, 1PM EDT ( slides ). I liked the idea given our future for the next academic year. Moreover, it was an opportunity for me to consolidate my thoughts on the scalable formative assessment techniques I use and reflect on how I currently do it, as well as how I can be better. To help with my thought process, I wrote this blog post. Then this post could also provide all the relevant resources and references in one place, like a companion piece for the webinar. For context, when I'm referring to things like lecture, I'm talking about in-person live lectures because that's where my experience lies. I do write about ideas for remote teaching since that is the world many are likely facing. For full disclosure, I am getting an honorarium from Codio for doing this webinar. However, I do not use Codio nor have plans to use it in the future. They also included ...
Read more