Since its public rollout in late 2022, OpenAI’s ChatGPT has been the subject of much consternation and discussion in higher education. Academic libraries have not been spared the stress of acclimating to this disruptive technology and the far-reaching implications it will have for Information Literacy (IL) instruction. While many in the academic library community immediately recognized the urgency of this developing situation, clear guidance for how we, as individual library units, might respond has remained somewhat elusive. Given that, we decided to proactively chart our own course and compose a document that would address this. The result is this document that outlines our considered approach to teaching and integrating ChatGPT into our academic ecosystem in a way that centers the ACRL Framework for Information Literacy for Higher Education.1 With this document we aim to shed light on our rationale, specifically as it relates to each frame within the ACRL Framework, with the understanding that the document can and should evolve over time.
All mentions of ChatGPT refer to the free version (3.5) available as of June 2023.
What is generative AI?
Below are some brief definitions from various sources, in no particular order:
IBM Research
“Generative AI refers to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on.”
Tech Republic
“In simple terms, generative AI is a subfield of artificial intelligence in which computer algorithms are used to generate outputs that resemble human-created content, be it text, images, graphics, music, computer code or otherwise.”
Center for Democracy and Technology
“Generative AI systems use machine learning to produce new content (e.g., text or images) based on large amounts of training data. That data is typically examples of the type of content the system will produce (such as enormous amounts of text for systems like ChatGPT that will produce text responses, or hundreds of millions of images for DALL·E, which produces images in response to prompts).”
Microsoft
“Generative AI refers to a category of AI that uses systems called neural networks to analyze data, find patterns and use these patterns to generate or create a new output, such as text, photo, video, code, data, and more.”
What is AI literacy?
Long and Magerko (2020) define AI literacy as “...a set of competencies that enables individuals to critically evaluate AI technologies; communicate and collaborate effectively with AI; and use AI as a tool online, at home, and in the workplace” (2).
AI Literacy in Higher Education
Like news, media and digital literacies, AI literacy is part of the larger domain of Information Literacy (IL). While instruction librarians may not be experts in each of these literacies, we strive, both in pedagogy and praxis, to engage students in acquiring and honing relevant skills needed to apply these literacies in their academic, work and personal lives.
Generative AI tools such as ChatGPT are predicted to become ubiquitous; indeed, such tools have already been integrated into many widely used platforms and search engines. The need for AI literacy to prepare students for this reality is great. Although the frenetic pace of ChatGPT’s availability, adoption and integration has certainly catalyzed a great deal of technostress, the librarians aim to meet that challenge to the best of our ability by teaching various digital-age literacies through the use of emerging technologies.
Higher education is one of the places students can be taught how to use generative AI tools equitably, ethically and responsibly. Library instruction in all its forms can present many rich and unique opportunities to engage students with AI literacy; this is a crucial component of digital citizenship which, in turn, is critical to a student’s development into an active, metaliterate participant in today’s society.
Generative AI Tools and the ACRL Framework
The ACRL Framework for Information Literacy in Higher Education provides a foundation upon which the instructional librarians conceive, plan, design, implement and evaluate all library instruction. ChatGPT's purpose and function intersect, to varying degrees, with each frame. These intersections provide potential opportunities for the librarians to engage students with AI literacy.
Information Creation as a Process
Considering how Large Language Models (LLMs) work it is worth noting that ChatGPT in fact generates text, rather than creates information. Even so, understanding how LLMs work directly connects to the “underlying processes of creation” referenced in this frame. Additionally, it could be argued that ChatGPT’s output fits within the “range of information formats and modes of delivery.” Viewed through this lens, one can see abundant opportunities for students to develop many of the Knowledge Practices (KPs) associated with this frame.
Research as Inquiry
This is perhaps the strongest element of ChatGPT’s intersection with the Framework. ChatGPT excels at simulated brainstorming, generating keywords, drafting and refining research questions, simplifying complex concepts and organizing information – all of which are tasks that closely align with many of the KPs associated with this frame. Students very unfamiliar with a topic could stand to benefit from ChatGPT usage for this stage of research.
Searching as Strategic Exploration
Competency within this frame is adjacent to – and highly informed by – Research as Inquiry, a frame with which ChatGPT has a strong intersection. A student’s mastery of one of this frame’s KPs in particular – the appropriate use of different types of search language such as controlled vocabulary, keywords and natural language – could potentially be enhanced by ChatGPT usage, since it excels at generating these types of search language.
Authority is Constructed and Contextual
This is perhaps the most challenging element of ChatGPT’s intersection with the Framework. ChatGPT presents its output in an authoritative-sounding way, whether it turns out to be correct or not. Although OpenAI has become more transparent in communicating the need to verify its output, the onus is on the user to check output veracity. Additionally, because of how it works, ChatGPT is typically unable to accurately cite sources; when pressed to do so it often “hallucinates” non-existent sources. Noting and discerning these anomalies can be confusing to students with emergent IL skills. Accurate information requires credible, reliable sources, so on this count, ChatGPT must further evolve before it can intersect with this frame in a constructive way.
Information Has Value
There is much fertile ground within this frame to teach students the value of information. While information can have value to us as seekers and consumers, it can be challenging to realize the information we produce has value to others, both individual people as well as corporate entities. OpenAI, the company that created ChatGPT, openly states that they “…may use the data you provide us to improve our models” (Schade). Additionally, the biases within which LLMs operate influence their output, thereby contributing to the continued systemic marginalization of underrepresented groups. As with the previous frame, ChatGPT does intersect with it but in a way that illustrates the inequity of information’s value to various groups and individuals as opposed to corporate parties.
Scholarship as Conversation
The reconciliation of hypocognition with cultivated information-seeking behavior (ISB) is a perpetually evolving consequence of the human condition; indeed, students with emergent IL skills may struggle with this to a greater degree than the seasoned researcher. A potential use of ChatGPT within this frame is as a means of inquiry into various perspectives and ideologies on a topic. As students acquire greater awareness of various perspectives, they become more empowered to strategically explore those perspectives via more conventional methods of ISB (e.g., open web searches, library database searches, etc.).
Tools, by their very nature, are meant to evolve over time, both in design and application of use. When critically applied and evaluated, ChatGPT’s output has the potential to be beneficial to students during the research and discovery process. However, it should be emphatically stressed that ChatGPT alone should not be employed as a substitute for scholarly exploration or analysis; its output may sound authoritative but must be checked for accuracy. Rather, generative AI should be utilized mindfully as one of many possible starting points. Otherwise, we risk undermining the cultivation of divergent critical thinking skills so important to students’ development as agile and independent thinkers.
Ethics of Use
The ethical use of information, data and scholarship is a core principle of the ACRL Framework. Students, faculty and librarians all bear responsibility in differing ways for engaging this principle with IL. If a librarian should choose to use generative AI tools during interactions with student, staff and faculty, part of that responsibility means we will strive, to the degree it’s feasible and appropriate, to address ethical issues such as academic integrity, bias, critical thinking, privacy concerns and copyright. SFCC Library acknowledges that the emerging field of generative AI brings up many ethical considerations; additionally, we recognize the need for transparency about our use of such tools in assisting them.
Application of Use
The SFCC Library embraces a community of practice that recognizes the expertise, experience and professional individuality of our librarians. The decision to utilize, or not utilize, generative AI tools (e.g., ChatGPT) during an instructional interaction with a student is an exercise in professional judgement; one librarian’s use/non-use of such AI is not representative of a particular stance held by the SFCC Library.
References
Association of College & Research Libraries. (2016). Framework for Information Literacy for Higher Education. https://www.ala.org/acrl/standards/ilframework
Hughes, O. (2023). Generative AI defined: How it works, benefits and dangers. Tech Republic. https://www.techrepublic.com/article/what-is-generative-ai/#what
IBM Research. (2023). What is generative AI? https://research.ibm.com/blog/what-is-generative-AI
Long, D. & Magerko, B. (2020). What is AI literacy? Competencies and design considerations. CHI '20: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. 1-16. https://aiunplugged.lmc.gatech.edu/wp-content/uploads/sites/36/2020/08/CHI-2020-AI-Literacy-Paper-Camera-Ready.pdf
Schade, M. (n.d.). How your data is used to improve model performance. OpenAI. https://help.openai.com/en/articles/5722486-how-your-data-is-used-to-improve-model-performance
Microsoft. (2023). AI Explained. https://news.microsoft.com/2023/04/04/ai-explained/
Quay-de la Vallee, H. (2023). Generative AI systems in education – uses and misuses. Center for Democracy and Technology. https://cdt.org/insights/generative-ai-systems-in-education-uses-and-misuses/
Additional sources that informed this document
Cox. C. & Tzoc, E. (2023). ChatGPT: Implications for academic libraries. College & Research Libraries News, 84(3). 99-102. https://crln.acrl.org/index.php/crlnews/article/view/25821/33752
Forrestal, V. (2023). Do we need librarians now that we have ChatGPT? Choice360 [Blog post]. https://www.choice360.org/libtech-insight/do-we-need-librarians-now-that-we-have-chatgpt/
Helper Systems. (2023). AI in higher education: The librarians’ perspectives. https://www.helpersystems.com/wp-content/uploads/2023/03/HS-AI-Survey-Whitepaper-3-7-22.pdf
Lo, L.S. (2023). My new favorite research partner is an AI. College & Research Libraries News, 84(6). 209-211. https://crln.acrl.org/index.php/crlnews/article/view/25932/33850
Papini, A. (2023). ChatGPT: A library perspective. Krupp Library, Bryant University. https://library.bryant.edu/chatgpt-library-perspective
UNESCO. (2023). ChatGPT and artificial intelligence in higher education. https://www.iesalc.unesco.org/wp-content/uploads/2023/04/ChatGPT-and-Artificial-Intelligence-in-higher-education-Quick-Start-guide_EN_FINAL.pdf