Generative AI and Information Literacy

“How do I know if I can trust the information produced by an AI tool?” 

A lot of people have concerns about this question. Generative AI tools produce a lot of information and often present the information confidently or authoritatively. The question of whether you can trust this information is fundamentally about information literacy. 

The Association for College and Research Libraries' (ACRL) Framework for Information Literacy for Higher Education defines Information Literacy as “the set of integrated abilities encompassing the reflective discovery of information, the understanding of how information is produced and valued, and the use of information in creating new knowledge and participating ethically in communities of learning.”

This definition is supported by the following framework of six interconnected concepts that point to knowledge practices and dispositions related to information literacy: 

• Authority is Constructed and Contextual
• Information Creation as a Process
• Information Has Value
• Research as Inquiry
• Scholarship as Conversation
• Searching as Strategic Exploration

Furthermore, Critical Information Literacy is “a theory and practice that considers the sociopolitical dimensions of information and production of knowledge, and critiques the ways in which systems of power shape the creation, distribution, and reception of information” (Drabinski and Tewell, 2019).

Information produced by Gen-AI tools is fundamentally different from information produced by human authors in important ways. With human-created information, we understand that “information creation is a process,” where a human has made choices of how to (or how not to) research, create, revise, and communicate ideas. Human-created information leaves a trail that can be fact-checked, or checked for other biases, motivations, or contexts.

Information created by Gen-AI tools, on the other hand, is the immediate output of predictive algorithms responding to a specific prompt. The information output can change based on how a prompt is phrased, even if the idea behind the prompt is the same. Many times, the information produced is accurate, or at least partially accurate. 

However, there are many ways in which the information produced by a Gen-AI tool can be wrong, even if it seems correct:

• Information could be factually incorrect 
• Information could be incomplete or inaccurate
• Information could be misleading (e.g. misrepresenting or misappropriating selective information without context)
• Information could be made up (these are called “hallucinations” - where information is simply generated based on predictions of what information should come next)
• Information could be based on a misinterpretation of your prompt
• Information could be wrong because your prompt contained information that was also wrong

Where does the information from a Gen-AI tool come from?

Gen-AI tools like Open AI’s GPT series or Google’s Gemini are examples of Large Language Models (LLMs). These LLMs are proprietary, and are trained on vast amounts of data using algorithms that are often invisible to us, the users.

Because the inner workings of these tools are invisible, it is difficult to understand how we can effectively determine the authority of information created by LLMs. Information created by humans “reflect their creators’ expertise and credibility, and are evaluated based on the information need and the context in which the information will be used” (ACRL Framework for Information Literacy, “Authority is Constructed and Contextual”). However, LLMs operate much like a “black box” - where users “can feed the system an input and receive an output, but you cannot examine the system’s code or the logic that produced the output” (Bagchi, 2023). In other words, the exact mechanisms that LLMs use to produce information are hidden.

How do I evaluate information from a Gen-AI tool?

Lateral reading is a strategy used by professional fact checkers where you open up multiple new tabs and windows on your browser to verify information using outside sources. This is a useful strategy for evaluating any kind of online information.

When evaluating information created by a Gen-AI tool, it is important to take several additional steps. This is because LLMs may be drawing from factual and false source materials indiscriminately, and also because LLMs are responding to specific prompts, and those prompts may also contain misinformation or assumptions.

Evaluating ZotGPT using Lateral Reading

This short video demonstrates lateral reading in response to a prompt asking UCI’s ZotGPT about the role of the anteater in UCI’s history.

On Information Literacy

Association of College and Research Libraries. “Framework for Information Literacy for Higher Education." Accessed June 17, 2024. https://www.ala.org/acrl/standards/ilframework.

James, Amy B., and Ellen Hampton Filgo. "Where Does ChatGPT Fit into the Framework for Information Literacy? The Possibilities and Problems of AI in Library Instruction" College & Research Libraries News, Volume 84 Number 9 (2023): 334-341. https://doi.org/10.5860/crln.84.9.334.

On LLMs and Misinformation

Bagchi, Saurabh. “What Is a Black Box? A Computer Scientist Explains What It Means When the Inner Workings of AIs Are Hidden.” The Conversation, May 22, 2023. http://theconversation.com/what-is-a-black-box-a-computer-scientist-explains-what-it-means-when-the-inner-workings-of-ais-are-hidden-203888.

Park, Peter S., Simon Goldstein, Aidan O’Gara, Michael Chen, and Dan Hendrycks. “AI Deception: A Survey of Examples, Risks, and Potential Solutions.” arXiv (2023). https://doi.org/10.48550/arXiv.2308.14752.

Walters, William H., and Esther Isabelle Wilder. “Fabrication and Errors in the Bibliographic Citations Generated by ChatGPT.” Scientific Reports 13, no. 1 (2023): 14045. https://doi.org/10.1038/s41598-023-41032-5.

White, Jeremy. “See How Easily A.I. Chatbots Can Be Taught to Spew Disinformation.” The New York Times, May 19, 2024. https://www.nytimes.com/interactive/2024/05/19/technology/biased-ai-chatbots.html.

• How is information that is created by an LLM/Gen-AI tool different from human-created information?
• How is prompting an LLM/Gen-AI tool different from “traditional” searching for information in a database or search engine?
• How might evaluating information from an LLM/Gen-AI tool be similar to or different from evaluating other kinds of web-based information? 
• Consider each of the six frames of the Framework for Information Literacy in Higher Education. What does that frame mean to you, and what do you think it means in relation to AI-generated information? (For example, what does "Authority is Constructed and Contextual" mean to you if you think about scholarly articles? Social media? An answer to a prompt in ZotGPT?)