Let’s Discuss: How Should We Handle AI in One-Shots?

Whenever I teach a one-shot, there’s an AI-shaped elephant in the room. The rapid advancement of AI technologies has changed so much about how students approach completing assignments, which shapes the dynamic in the one-shot classroom. I worry that students, never the most enthusiastic one-shot attendees, now feel that library research skills are less relevant than ever.

On the surface, it appears that consumer-focused AI tools like ChatGPT are useful for student research. They can help students quickly locate sources, learn about a topic, summarize difficult peer-reviewed articles, and cite their sources—all tasks that are relevant to the learning outcomes of my one-shots. However, as I’ve discussed previously on ACRLog, I believe that higher education should be a place of particular caution when it comes to AI. The hazards of AI, such as copyright infringement, hallucinations, bias, lack of privacy, and the potential erosion of critical thinking skills, are particularly dangerous in an educational context.

Moreover, as Allison and DeRewal (2024) have pointed out, the missions of AI companies are directly opposed to our missions as libraries and universities. The goal of products like ChatGPT and Comet—and even AI tools built into educational products, like Primo Research Assistant—is to make tasks as easy as possible for the consumer. Our goal is to educate students, which cannot happen by making things as easy as possible. We intentionally build friction into our assignments and activities in order to help students develop the skills necessary to overcome inevitable difficulties in the research process. ChatGPT’s goal is to eliminate challenge; our goal is to strategically introduce it, because students can’t learn without it.

Here are the research-related AI applications I’m most worried about in regards to undergraduates I teach:

• Using AI tools to find sources (“Find 5 peer-reviewed articles about climate change.”)
• Using AI tools to learn information (“What are the main causes of climate change?”)
• Using AI tools to summarize articles (“What are the main ideas of this article?”)
• Using AI tools to generate citations (“Please format this list of sources into an MLA Works Cited page.”)

(In general, I’m also worried about students using AI tools to generate the text of their papers. However, that falls outside the scope of one-shot instruction, so I will not address it here.) 

With all of this swimming around in my head, it can be overwhelming to try to productively address AI issues in a 50-minute one-shot, especially when I have so much else to teach. Whenever I feel lost about how to tackle an AI-related issue, I return to three touchstones: clarity, transparency, and guidance. I believe that undergraduate students today deeply value these qualities in the classroom. So when I tell them a proactive, firm “no” (clarity), I immediately follow up with transparency (“here’s why the answer is no”) and guidance (“here’s what you can do instead”). 

I want to use the remainder of this post to share three of the strategies I’ve been using to address AI in one-shots—but I am just one librarian, liaising with the social sciences at a regional comprehensive university in the midwestern United States. My context may be very different from yours. I would love for folks to jump into the conversation and share strategies, thoughts, and questions in the comments below. 

Strategy #1: Address AI upfront

I always start my one-shots by talking about AI for a few minutes. I don’t want students zoning out the entire session, thinking “I’ll just use ChatGPT for this,” so I try to capture their attention by posing the question: Why don’t you just use ChatGPT to research this paper? Why are you here at the library today? Per my transparency touchstone, I’m explicit about my intentions with having this conversation. I know this is something some of you might be wondering, and I often get questions like these from students, so I wanted to take a couple of minutes to talk about this before we dive in.

I state that in general, when students ask me this question, I do not recommend the use of AI for completing research assignments. I share specific examples of times I’ve witnessed AI-assisted research going wrong: Cornerstone students combing the stacks for books that were never written, law students tearing their hair out as they search for Supreme Court cases that don’t exist. I try to get vivid and lean into the storytelling. 

I don’t want that to be you, I say. So today, we’re going to walk through how you can research this paper without running into any of those issues. Sound good? 

Strategy #2: Demonstrate better open web searching

Let’s be honest: Using Google is miserable. Half the websites are paywalled, and the other half are sponsored links that relentlessly spam you with pop-up ads. No wonder students are increasingly turning to Google’s AI overview and AI chatbots to find information. The interfaces are far cleaner and easier to navigate; this not only provides a better user experience, it also appears more professional and, therefore, reliable. 

But as information professionals, we know these AI tools aren’t actually more reliable, considering the frequency and magnitude of hallucinations. To provide an alternative when students want to search the web for information on a topic, I show them how to limit their search results by site domain (or even a particular website, if relevant to the course). I generally recommend .edu, but I sometimes show .org and .gov as well. I may also show how to use phrase searching or the intitle command in a basic Google search. Students are often surprised and excited to see how simple tricks can make Google searching so much easier. 

Strategy #3: Cultivate college-level research skills 

In most cases, it is currently illegal for a student at the University of Northern Iowa to download an article from one of our subscription databases and then upload it into a generative AI tool, such as ChatGPT. In some cases, it is not the student who would be liable for this action, but the library. 

Whenever I tell students this, it makes an impression. They’re especially shocked to hear that in some cases, their actions could lead to the library getting in trouble. I clarify that this is a result of the agreements we make with scholarly publishers in order to make their books, articles, and other materials available to UNI students. I explain that students often ask whether or not they can use ChatGPT to summarize sources for their research papers, and this is part of why their librarians and professors are telling them no.

However, I emphasize that I understand why students ask me this question—because, truly, I do. It’s difficult enough to understand peer-reviewed research when you’re new to academia and/or to a particular discipline; on top of that, many students are never actually taught how to read research, how to read in college. It’s very different from how we read for leisure or how they read to perform well on standardized tests in high school. Students may feel lost, like asking ChatGPT to summarize a difficult passage or even an entire article is the only way they can succeed in class. 

There’s only so much I can do in a one-shot; reading research is a skill, and real skill development takes guided practice over an extended period of time. Nonetheless, I try to plant a seed. I reassure students: It’s okay if reading peer-reviewed research is hard. I provide a few tips about strategic reading, such as starting with the abstract and conclusion, making multiple passes, and annotating key points. I love this guide from Brown University and often share it with students. I also take a few minutes to explain citation chasing. Citation chasing is another college-level research skill that seems second nature to us, but it won’t occur to most freshmen until somebody takes the time to teach them about it. You don’t need to ask a generative AI tool for “5 sources about climate change” because once you find one good article or book chapter, there are tons of additional sources right there at the end. 

—

I am in no way an expert on AI or one-shot instruction; this is only my second semester of teaching. I don’t come to this with the intention of sharing expert knowledge, but with the hope of starting a conversation. I would love for the comments section to become a place for folks to share experiences and brainstorm strategies around addressing AI in one-shots. Here are some questions to get started: 

• How have you been thinking about AI in your one-shots? 
• What strategies have you used to address AI in your one-shots? What has worked well, and what has not worked as well?
• What kinds of conversations have you had (or not had) with your faculty about classroom AI policies, AI usage in research assignments, and so on?
• Any other thoughts or questions that this post prompted for you!