You are currently viewing Brainstorm with a Bot: Designing a Custom AI Tool to Help Students Find Their Arguments

Brainstorm with a Bot: Designing a Custom AI Tool to Help Students Find Their Arguments

When generative AI tools first became widely available, conversations in higher education centered on whether students should be allowed to use them and under what conditions. At the same time, it became increasingly clear that many students were already using AI for assignments. Rather than attempting to prohibit or ignore that use, I became interested in whether I could guide it more intentionally.

At the time, I was teaching first-year composition and introductory speech. A common challenge in both courses was helping students move from having no topic at all to identifying a topic of interest and developing a specific, arguable claim. Students must make this decision before they can effectively research, draft, or revise.

I wanted to provide a way for students to explore possible topics and arguments while ensuring that the intellectual work of evaluating, selecting, and defending those ideas remained their own. By brainstorming with AI, students could explore a wider range of possibilities than they might have considered alone. They could test possibilities early, reducing the likelihood of discovering later that a topic is too broad or narrow, lacks sufficient evidence, or cannot sustain a compelling argument.

Building the bot

To build the bot, I used PlayLab, a web-based platform my institution adopted for creating custom AI tools for educational use. The platform is publicly available online and includes a no-cost tier. Instructors can also create similar tools on other platforms, including custom GPTs in ChatGPT and Gems in Google Gemini.

I set up a bot for each course by providing relevant materials, such as the syllabus and assignment instructions. I then gave the bots directions to guide students from broad interests toward a focused claim through structured questioning in stages. As students narrowed their focus, the bots posed follow-up questions that encouraged them to evaluate the strengths and weaknesses of potential positions, clarify assumptions, consider evidence, and refine their claims. For instance:

Help the student to identify a topic:

  • Help students to identify an issue of interest
  • Encourage evaluation of suggestions.
  • Encourage consideration of multiple perspectives.

Help the student refine their topic:

  • Ask questions that help identify and refine a claim.
  • Ask students to explain and justify their reasoning.

Help the student think about objections:

  • Present reasonable objections and require a response.
  • Challenge assumptions and potential weaknesses.
  • Require students to defend their position.

Guardrails:

  • Do not generate thesis statements.
  • Do not write outlines, paragraphs, introductions, conclusions, or any other part of the assignment.
  • Do not select or endorse a single position.

These design choices emerged through iteration. Early on, a student asked whether she could use a bot-generated thesis statement in her essay, a question that made clear the need for stronger constraints. I adjusted the bots to preserve student ownership of claims rather than to produce them.

As additional patterns emerged, I continued to modify the bots. When students were not engaging deeply with alternative perspectives, I added a stage in which the bots were instructed to challenge students’ claims and argue against their positions. Rather than simply helping students generate ideas, the bots now required them to defend those ideas.

The assignment, Brainstorm with a Bot, asked students to reflect on their experience using the tool. I reminded students that AI outputs reflect training data and may contain bias, and I encouraged them to evaluate the bot’s suggestions critically.

Student reflections

Across three semesters, approximately 360 students have completed the assignment and submitted reflections on their experience using the bot.

Students consistently reported that the bot helped them transform vague ideas into specific, arguable claims. Many appreciated the follow-up questions that prompted them to clarify and refine their thinking. Others noted that it was faster and less overwhelming than starting with open-ended web searches.

Students also identified limitations. Some found the suggestions to be too broad or lacking originality. Others observed occasional bias or one-sided framing, requiring them to filter or reject the bot’s recommendations. In this sense, the bot became an object of analysis that supported critical thinking and AI literacy.

Nearly all students emphasized that the bot was a starting point and that final decisions about topics and arguments required their own judgement, experience, and values.

Suggestions

Three insights emerged from the experience. First, AI integration does not have to be all or nothing. Focus on one high-impact stage of a single assignment, the point where students must find a topic and a claim, keeping the work manageable while addressing a challenge that many students face. Second, guardrails are essential to the design. Without appropriate constraints, a general-purpose chatbot tends to generate content that students should produce themselves. The value of the bot depended largely on the boundaries I established and on its ability to support student thinking while allowing them to maintain agency. Third, designing AI for teaching is an iterative process. The bots I use today are substantively different from the ones I first deployed and continue to evolve as students interact with them and provide feedback.

The goal of Brainstorm with a Bot was to create conditions in which students engage more deeply in thinking before committing to a claim. Its value was not in generating ideas for students, but in helping them to discover, test, and refine ideas of their own.


Jane S. Nazzal is a professor of English at Mount San Antonio College.

The post Brainstorm with a Bot: Designing a Custom AI Tool to Help Students Find Their Arguments appeared first on The Teaching Professor.