Information Today, Inc. Corporate Site KMWorld CRM Media Streaming Media Faulkner Speech Technology DBTA/Unisphere
Other ITI Websites
American Library Directory Boardwalk Empire Database Trends and Applications DestinationCRM Faulkner Information Services Fulltext Sources Online InfoToday Europe KMWorld Literary Market Place Plexus Publishing Smart Customer Service Speech Technology Streaming Media Streaming Media Europe Streaming Media Producer Unisphere Research

Vendors: For commercial reprints in print or digital form, contact LaShawn Fugate (

Magazines > Computers in Libraries > September 2023

Back Index Forward
Vol. 43 No. 7 — September 2023

Helping Students Navigate Research With AI Tools
by Chad Mairn and Shelbey Rosengarten

Is it a tsunami or an earthquake? Are we wading through the heavy gravitational pull back to the ocean’s center, or are we caught in the aftershocks of a permanent seismic shift? You won’t find the answers to this by plugging the questions into ChatGPT or Bing; we’ve tried, and while we got passable answers involving metaphors, they still weren’t satisfactory. They were too broad and too vague and filled with all the faults that people have called out about these fascinating tools. It remains up to us to reflect and apply the general knowledge to our own situations and conflicts and create meaning out of what we are experiencing.

One recurring theme across several discussions is how AI will catalyze an emphasis on skill-building in higher education. Rather than solely knowing stuff, students will need to know how to do stuff. Both are inherently necessary to a well-educated mind. However, it’s good to revisit and articulate the types of skills we want students to build. Most educational frameworks (such as Bloom’s taxonomy) delineate skills such as evaluating the pros and cons of a situation or analyzing processes. The research process requires students to actively develop and use these skills, so it’s been fertile ground for a partnership between librarians and instructors. There’s plenty of information readily available. How do we sift through it without drowning in it? How do we refine that search, evaluate what we find, and integrate it into an artifact that shows our understanding?

Although ChatGPT generates a human-like response quickly, it can free up time for thoughtful reflection before we decide to use that response or not. However, this technology is still not a useful problem-solving tool because it does not truly know anything; instead, it is predicting a sequence of words from a “dictionary” with billions of words in it. Going further, reinforcement learning from human feedback is an important aspect of generative AI since humans are currently providing the necessary feedback to help the large language model (LLM) learn and become more astute with its responses. This is a good enough reason for students to become AI literate—so that they’re better prepared for a world in which this technology becomes a ubiquitous component in the workplace, for personal use, and for lifelong learning. Librarians and instructors can work together to help students navigate the tsunami.

Finding Information Without Getting Lost in the Search

Composition instructors see all kinds of students with all levels of competence, not just in writing, but in research, clarifying ideas, critical thinking, evaluating, judging, and synthesizing bits and pieces of information into a deeper understanding and a more solid knowledgebase. Trying to lead a class of a dozen different levels through finding, understanding, assessing, and internalizing many sorts of resources is chaos even on a good day. This is one reason that over the last decade or 2, college instructors and librarians have increasingly prompted their students to take advantage of learning resources.

In almost any course, librarians can model a research journey for the students. It’s not just a simple task anymore to go look something up, since there are so many types of sources, and within each type, hundreds or thousands of sources themselves. Such a journey can touch on different facets or phases. These may include the following:

Information literacy—What are the different types of sources, who are their audiences, and what are their respective purposes? What is broad and what is narrow within a field? This also includes distinguishing between a general reference on subjects and unique takes from experts of various stripes.

Exploring a range within a subject area—Given that not all sources are written for all audiences and that discovery is half the joy of learning, librarians can lead students to a virtual shelf where they may discover related sources.

Knowing which databases fit which situations—What a freshman composition student needs is very different from that of an upper-division business administration student, a paralegal in training, or an RN student mired in a busy semester of clinicals.

Finding information and refining prompts to find better information—Many times, students don’t know how to commit to an idea because they are unsure of what is out there. They may start very broadly, but there are technical tools and search strategies available to help.

Media literacy—Students need to become sensitive to the different ways information is presented and how it may shape someone’s understanding as well as emphasize (or even embellish) certain ideas and perspectives due to a range of viewpoints.

Modeling a Research Journey

Students may find search engines themselves to be overwhelming, not to mention riddled with ads and distractions. Libraries work within a subset of vetted databases but still return hundreds (if not thousands) of sources—not all of which will be relevant or intelligible to a general audience. ChatGPT may appear to sidestep these problems by giving the information in a few paragraphs. It may not always be right, as widely noted. This past semester, Rosengarten experimented with using ChatGPT to generate short lists of sources. Its entries were spotty, and as a GPT-3.5-powered tool, source dates only went through 2021. This was an exercise to test its limits for finding resources. The short summaries of sources were so repetitive as to be nearly interchangeable. It also generated some hallucinations in the form of nonexistent sources made up of a mix-and-match blend of language, author names, and publications. In other words, it’s not yet a reliable tool for finding sources themselves, let alone for integrating the information into a finished product. So, what might libraries have available for patrons and students who want to refine their searches? Can a chatbot model provide a good research journey of sorts? Can it show a student what to do with it?

Click or Scan to view AgentGPT tasks.There are many plugins for GPT-4 that are designed to add various functionalities to applications such as ChatGPT. For example, Auto-GPT is an autonomous GPT-4 experiment that allows users to define tasks for the chatbot to accomplish. For a test that Mairn conducted, an Open-AI API key was added to AgentGPT, a web-based service to help assemble, configure, and deploy autonomous AI agents inside a web browser, with nothing to download or install. A Librarian-GPT agent was created to help break down the researcher’s goal (i.e., the prompt) into multiple tasks, which was to write a research paper on how AI will transform society.

AgentGPT then created five tasks, designated as follows:

1. Gather and analyze data on the current and potential impact of AI on society.

2. Conduct a literature review on the different perspectives and opinions regarding this topic.

3. Synthesize findings, and create a comprehensive research paper outlining the transformational potential of AI on society.

4. Identify specific case studies or examples of how AI has already transformed society in various industries, and analyze the results for inclusion in the research paper.

5. Analyze the potential ethical implications of AI’s transformational impact on society, and provide recommendations for responsible development and deployment of AI technology.

Once AgentGPT finished those tasks, it displayed an image, a PDF, and other options to copy and share the results. In this case, we can conceive that AI has been learning from us, and reviewing the results of this concatenated set of tasks may be one way we can learn from AI. Additionally, AgentGPT’s output resembles the framework for information literacy standards and indicators in which students can use the tasks to again help guide their thinking throughout their research journey.

Teaching the Tools

Let’s tackle the next issue. How can we get students to feel more comfortable searching for themselves, rather than relying on a search bot? It’s best that they become savvy to the same techniques most information professionals have mastered. We all use Google often enough that it’s become a verb. In response to some of the criticism of its SEO-driven returns, the company created Google Scholar, which is now often brought up in many research sessions for all sorts of classes. This method has also evolved along with AI, especially once Microsoft partnered with OpenAI and incorporated ChatGPT into Bing.

Google Scholar for Libraries was developed to give easy access to scholarly articles from within a familiar Google-like search interface. It works with library vendors to gain access to their OpenURL link resolver to create an automated list of titles that the library subscribes to. Campus IP addresses and other technical information are also gathered to make sure that users who have appropriate credentials can find library resources that are behind authentication platforms. After initiating a search query, the user looks for their library link (e.g., Full Text @ SPC), where they are then redirected to their institution’s identity provider to log in.

Meanwhile, the browser Edge has integrated ChatGPT for finding information and also has a generative AI writing assistant. Bing, Microsoft’s search engine, has also integrated ChatGPT and is being touted as an answer engine. Many people will be motivated to use this tool since it tends to be an easier approach to finding and using information. However, the links that are normally displayed are not authoritative, so it would be a positive step forward to have libraries work with Bing (as they do with Google Scholar) to provide easy access to library resources as well as open web results. Google’s Bard may be able to implement a Google Scholar-like experience quicker than Bing, but time will tell.

Tome generates AI-powered presentations.Tome is another application that will build stories in seconds using AI. Mairn’s first attempt at using it was to ask Tome to build a presentation outlining the legislative process for information specialists, researchers, code revision, bill drafting, and committee and policy analysis for the National Conference of State Legislatures. It generated eight slides in about 30 seconds. There are options to add video narration, change themes, export to PDF, create various tiles, use prompts for generative art and video, and more.

Many instructors nationwide continue to experiment with when, where, and how to integrate AI into their coursework and planning. In addition to the source-generation exercise this past spring, Rosengarten has also tried a few iterations of other exercises. During a prewriting phase for a research project, she used ChatGPT to generate research questions, wanting it to model what good ones look like as well as those that are far too broad and general. Students uncovered angles they hadn’t originally encountered or thought of, used them to find sources, and revisited and refined the focus. Rosengarten has investigated and reported AI uses in different fields as a research and exploration project in one of her summer composition classes, and the range of applications is fascinating. Students have been reporting their own findings and seem intrigued about having a head start on how it will be used in their fields of study.

We have come to rely on internet search engines, and many libraries even model their interfaces on them. A deliberate use of AI to streamline a search for good information and resources would be a fine analog. Before we feel obsolete, though, we should differentiate between the straight information and the ability to critique, evaluate, and synthesize information to create new knowledge and ideas. AI can’t do these things, and they are an important suite of skills that we need our students to continue to develop.

Final Notes

Where might this all be headed as AI writing models evolve? Perhaps we will see style bots emerge? Imagine collecting everything you have ever written in school, on social media, and in emails, and all that text was added to an LLM. It seems possible that users would be able to prompt a chatbot to generate text based off of their personal style.

Society has been influenced by algorithms for a long time, but learning how to utilize and play with these types of tools can help enhance creative learning while potentially developing a personalized learning plan for those students who have certain strengths and weaknesses that they would like to focus on. Also, exposing students to the AI tools mentioned in this article—and hundreds more being developed each week—will help them to be more adept and comfortable with these types of technologies when entering the workplace of the future. However, it is important to provide clear expectations about the use of generative AI tools in classrooms because utilizing them inappropriately can have detrimental consequences. But these tools can also help us discover things that we did not know existed before. We are truly living in challenging and remarkable times.

Shelbey RosengartenChad MairnChad Mairn (L) is a librarian, teacher, author, and self-described geek who frequently shares his enthusiasm for all things technology as a speaker at library and technology conferences. He is an information services librarian, assistant professor, and founder of the Innovation Lab at St. Petersburg College. While he was an undergraduate studying humanities at the University of South Florida (USF), Mairn was awarded a Library of Congress Fellowship, helping archive personal papers and other items in the Leonard Bernstein Collection. During his library and information science graduate work, also at USF, Mairn became a technology liaison between the Bill and Melinda Gates Foundation and Florida public libraries.

Shelbey Rosengarten
(R) is an associate professor of communications at St. Petersburg College. For more than 20 years, she has taught courses in composition, literature, and technical and professional writing, experimenting with instructional technologies all along the way. Rosengarten spends a lot of time exploring the great outdoors, trading screen time for green time.