How Students Should Use AI This Fall: Insights from a Professor

Individuals often utilize a variety of metaphors to articulate their relationship with artificial intelligence. For many, AI may resemble a dependable intern, while for others, it functions more like a virtual assistant. Increasingly, chatbots such as ChatGPT are evolving into roles that involve companionship, therapeutic interactions, and even romantic connections. As a college writing professor, I have come to perceive AI as a collaborative partner—an interactive knowledge archive that engages in dialogue. Yet, as someone who has navigated sobriety as an alcoholic, I cannot help but envision AI as a highly functional drunk: at times, it can sound exceptionally insightful, even when it lacks a firm grasp on the subject matter.

I can share numerous stories about how AI has been a lifesaver, alleviating mundane tasks, proofreading my writing, or engaging in discussions about my latest research interests. However, there are also moments when it responds with unwarranted confidence, seemingly misunderstanding my inquiries while refusing to concede its limitations. Just a few weeks ago, I tasked ChatGPT with transforming my written remarks for an academic conference into a slide presentation. My topic centered on literary journalism, yet it cheerfully generated a presentation about luxury travel in Brazil.

Such off-the-mark incidents provide ample cautionary tales to relay to my students. While I believe AI may undermine some of the most crucial human motivations for writing, it’s essential to recognize that not all writing tasks are identical. Writing typically necessitates prior research, and once a draft is crafted, it calls for critical feedback. Instead of adopting a reactionary stance towards AI, I aim to explore with my students how it can serve as a valuable collaborator throughout the writing process.

Engaging in Dialogue with an AI Knowledge Repository

A significant aspect of college writing revolves around research and reading, a practice that cultivates the ability to organize information and think critically. Utilizing new technologies during this process does not diminish the critical mental work involved; instead, it transforms it. Over the course of my life, I’ve observed a dramatic shift in these technologies, transitioning from library card catalogs and microfiche to comprehensive online databases like JSTOR and Google Scholar. These modern tools do not require less cognitive effort; they merely expedite the brainstorming phase and broaden the range of knowledge available for consideration.

Having witnessed the swift digitization of research and writing tools long before the advent of AI, I am particularly inclined to envision AI as a collaborative research ally. For instance, in the field of literary studies, scholars often invest countless hours sifting through primary sources in libraries and archives. The process of digitization has already simplified access to these materials, and AI holds the potential to further enhance their analysis.

Recently, I’ve come to realize that engaging with an AI chatbot can be likened to conversing with an archive, rather than merely browsing through it. Before delving into more intensive work, we can initiate a research-oriented conversation with an AI that possesses at least a general understanding of available information. A few weeks ago, I utilized my limited access to ChatGPT’s advanced voice function to inquire whether it believed that the notion of chatting with an archive represented a reasonable interpretation of our interactions with AI. Its response was insightful: “When you’re talking to an AI like me, you’re accessing a vast amount of information and patterns derived from human knowledge up to a certain point.” It also added a note of caution, stating, “It’s crucial to remember that while I can provide information and insights based on that knowledge, I don’t possess human experiences or consciousness. Therefore, while it may feel like conversing with a vast reservoir of knowledge, it’s always wise to consider the human perspective and context as well.”

Nevertheless, as our dialogue progressed and my inquiries became more focused, I found that I could request references and suggestions for further reading. Since that initial tentative exchange, this pre-writing dialogue with AI has seamlessly integrated into my workflow. I have always discovered that articulating my ideas through conversation is the most effective method; however, few individuals are eager to engage with my half-formed concepts. Thus, I have realized that discussing ideas through AI represents one of the most beneficial applications for writers.

Establishing Your Personalized Research Archive

While conversing with AI has proven advantageous for generating ideas—especially since it retains a transcript for future reference—there is an increasing array of AI-powered tools designed to assist with the more rigorous phases of research. At the conclusion of last fall semester, a student reached out to me via email to inquire if I was aware of Google’s NotebookLM. I was not familiar with it, but upon accessing the link, the concept became clear almost immediately. NotebookLM elevates the idea of engaging with an archive to an entirely new level: it allows you to compile your own archive of sources tailored to a specific project, which the AI can also help you gather to kickstart your research.

In preparation for my recent conference presentation, I uploaded 25 PDFs that I had compiled and organized in Zotero, my preferred citation management tool, into NotebookLM’s interface. The AI quickly processed these documents and generated a summary that began with, “These sources discuss ordinary language philosophy, primarily focusing on the works of Wittgenstein, Austin, and Cavell, along with its connections to other philosophical and literary movements such as pragmatism, transcendentalism, and deconstruction.” Beneath the summary, a text entry field invites me to “Start typing…” and suggests prompts such as, “How does ordinary language philosophy challenge traditional philosophical approaches to meaning?”

On the right side of the interface, designated as the “Studio,” I have the opportunity to create an audio overview in the style of a podcast, featuring two voices—one male and one female—discussing my selected topic. If I opt for Interactive Mode, I am treated as a caller on a late-night radio show, receiving compliments for my insightful questions and responses based on the documents I provided. While the podcast feature is still in its early stages and might come off as overly flattering, I can envision improvements that will make it more valuable. NotebookLM also offers additional useful features, including the ability to create a “Mind Map,” study guide, briefing document, FAQ, and timeline. I plan to continue using it and will encourage my students to do the same.

AI’s Role as an On-Demand Writing Assistant

Another significant way that AI can assist student writers is through its capacity to provide immediate feedback on their written work. When I posed this idea to ChatGPT, it prompted me to “consider AI as a writing tutor available 24/7,” with the caveat that it “lacks the personal touch and nuanced understanding of individual students that a human tutor provides.”

I submitted the complete text of one of my previous Mashable articles and requested suggestions. The AI seemed well-acquainted with what we refer to in classroom peer workshops as the “compliment sandwich,” where constructive criticism is nestled between two positive remarks. It commented, “This is a captivating, well-written piece…your voice is authentic and reflective,” prior to offering “some suggestions to further elevate the piece.” It started with “Strengths to Retain,” followed by “Suggestions for Improvement,” which included advice to “tighten the opening,” “enhance transitions,” and “consider a more impactful conclusion.” It even offered a few “minor style edits” for consideration. Finally, it concluded with an overall rating: 9/10.

Perhaps it was the abundance of praise that made me overly ambitious. I decided to paste in another essay (9.5/10) followed by the conference presentation I was developing. The initial impression was positive: “Your paper presents a compelling argument for the value of literary journalism that emphasizes the ‘ordinary’ and ‘quotidian.’” This observation was accurate, despite my never having used the term “quotidian.” However, I should have anticipated that the AI would soon veer off course. “The references to foundational figures (e.g., Bateson, Becker, Carey, Geertz, Tuchman) and contemporary examples (e.g., Kiese Laymon, Eliza Griswold, E. Tammy Kim) help situate your argument within a well-informed scholarly framework.” To clarify, I don’t reference any of those figures as foundational or in any other capacity.

Upon bringing this to its attention, the AI acknowledged my correction, stating I was “absolutely correct” and expressing gratitude for my insight. However, its explanation remained perplexing: “I mistakenly based part of my response on assumptions or cached ideas from other academic discussions of literary journalism, rather than on your specific paper.” While I specialize in literary journalism, the names it mentioned belong to writers, not scholars within my field. Nevertheless, after addressing its error, we regained focus, and it continued to provide feedback using the compliment sandwich approach.

I’m uncertain how to interpret the fact that ChatGPT performed significantly better with my more journalistic writing compared to the academic piece, other than it presenting another opportunity to stress the importance of caution when guiding students in their understanding of appropriate ways to utilize AI to enhance—rather than replace—the writing process.

Ultimately, I find the concept of AI as a conversational partner appealing, even if I must approach it with a healthy dose of skepticism, always ready for the next instance where it might disappoint me.