While AI-generated output can be useful, it is not always accurate, unbiased, or considered a credible source of information without verification. AI tools, such as ChatGPT and other generative chatbots, apply probability to create information using mathematical models to “guess” at likely words to use based on the tool’s training data. Therefore, the AI does not understand the meaning or context of the text it creates and may generate misinformation, fabricated information, and citations that do not exist. It is important to evaluate all AI-generated outputs critically, just as you would when evaluating information from any other source to ensure you are using trustworthy and credible information.
Lateral Reading is a technique that can be helpful in determining if AI-generated output is reliable. Lateral Reading is carried out by applying fact-checking techniques to evaluate what the AI has provided based on the prompt. See the video below for more information on how to carry out lateral reading.
Video: Stanford History Education Group
When evaluating AI-generated output, instead of asking “who’s behind this information?” we must ask “who can confirm this information?” In the video above, lateral reading was applied to an online source with an organization name, logo, URL, and authors whose identities and motivations can be researched and fact-checked from other credible sources. But AI-generated output often has no identifiers. Instead, its’ output can be a composite of multiple unidentifiable sources. Due to this, you must assess the claims made in AI-generated output based on their validity rather than their sources. Since AI-generated output is drawn commonly from multiple unknown sources, that could be both factual and false, you will find it useful to break it up into smaller components of information that can be evaluated independent of each other.
When working with generative AI, it’s helpful to remember that all generative AI chatbots use probability to create their outputs. However, it is dependent on the chatbot model whether sources are unidentified or identified inaccurately. See the AI Fact-Checking Steps box on this page for more on how to fact-check an output from ChatGPT or similar tool.
Alternatively, rather than using generative AI platforms, which are designed to provide coherent, but not accurate, responses, to find information, you can use tools designed to help with research. See the University Library’s AI Tools for Research and Education page or the many information seeking tools provided by the University Library.
Content from University of Maryland Fact-Checking AI guide, creativecommons.org/licenses/by-nc/4.0/, Southern New Hampshire University AI guide, and The University of British Columbia AI guide.
Content from: University of Maryland Libraries Fact-Checking AI guide creativecommons.org/licenses/by-nc/4.0/
Here's further detail on how to fact-check an output from ChatGPT or similar tool:
Break down the information. Take a look at the response and see if you can isolate specific, searchable claims. This is called fractionation.
Next, it’s time for lateral reading. Open a new tab and look for supporting pieces of information. Here are some good sources to start with:
a. When searching for specific pieces of information: Google results or Wikipedia
b. When seeing if something exists: Google Scholar, University Library's OneSearch, or Wikipedia.
c. Tip: Some things to watch out for – is the AI putting correct information in the wrong context? Is it attributing a fake article to a real author?
Next, think deeper about what assumptions are being made.
a. What did your prompt assume?
b. What did the AI tool assume?
c. Who would know things about this topic? Would they have a different perspective than what the AI tool is offering? Where could you check to find out?
Finally, make a judgment call. What is true, what is misleading, and what is factually incorrect? Can you re-prompt the AI to try and fix some of these errors? Can you dive deeper into one of the sources you found while fact-checking? Remember, you’re repeating this process for each of the claims the AI made – go back to your list from the first step and keep going!
University of Maryland Libraries Fact-Checking AI guide creativecommons.org/licenses/by-nc/4.0/
Report ADA Problems with Library Services and Resources. Resources on this page may require Document Viewers
