Generative AI at SSU: Citations and Ethics of ChatGPT

Why Should You Cite AI?

References tell your reader where your information came from and how you used it in your work. If you use content created by a tool like ChatGPT, including it in your works cited - as you would with any other source - is the responsible thing to do. If you use AI like ChatGPT to help write or structure your paper, even if you do not otherwise quote or paraphrase its content, you will likely wish to acknowledge your use of it in some manner. This provides transparency to your reader.

It is always better to err on the side of caution when it comes to being fully transparent! Remember, the unauthorized use of AI on your coursework may be treated like plagiarism.

Tips for Citing AI Responsibly

• Save a transcript of your chat. Make it available to or retrievable by your reader, possibly by including it as an appendix to your work or as an online supplement.
• Describe the prompt that generated the specific ChatGPT response.
• Include the date when the response was generated or date of access. This is important as these tools will update regularly.
• Acknowledge how you used the tool. You should do this even if you only use ChatGPT to help plan your paper or generate ideas and don't include any of its generated content.

from Salt Lake Community College.

Chatbots and You: AI as a Personal Communication

When citing AI, both APA and MLA recommend that you treat it as a personal communication instead of a book, article, or encyclopedia entry, even if the AI generates an essay in response to your prompt.

At their foundation, these programs are chatbots that return a statistically likely answer to your prompt, not vetted research tools.

AI Hallucinations and Misinformation

Again, AI is not a perfect tool, and it is prone to something that developers call "hallucinations."

These hallucinations may also be known as misinformation. Similar to a friend confidently misremembering who won the last season of Survivor or a history professor mixing up dates during a lecture, misinformation is usually unintentional but convincing. Often, you will not realize that you were misinformed until you come across the correct information in a different source.

Developers are working to increase AI's accuracy and reliability, but the frequency of mistakes is still high. Current studies suggest it feeds users false information in 15-25% of its responses. Because AI pulls from a broad variety of sources to craft an answer, many unvetted, and does not fact-check itself to the standards expected in academic and professional settings, it is important to fact-check the information that it gives you.

Additionally, some malicious groups have already begun using AI to generate purposefully misleading stories, also known as "fake news." This deliberately false information is known as disinformation.

To learn more about how you can identify misinformation and disinformation, check out our research guide on fake news.

Can the Sources Cited by AI Be Trusted?

Sometimes, when a user prompts AI to generate a list of resources or cite the sources for the information it generates in response to a prompt, AI cites sources that do not exist or contain no link to the content of the response. This is another form of AI hallucination and misinformation.

To avoid any confusion or unintentional plagiarism, always double-check the sources AI provides.

1. Make sure the source exists, that the details like the author, publisher, title, and date are correct using a site like GALILEO.
2. Skim the source to confirm that it includes the referenced content.
3. Finally, make sure to cite these original sources in your reference list too.

If you do not know how to do this, bring your list of sources to the library or email a librarian - we would be happy to help!

MLA Reference Format

"Description of chat" prompt. Name of AI tool, version of AI tool, Company, Date of chat, Link to source.

Example Using ChatGPT

“In 200 words, describe the symbolism of the green light in The Great Gatsby” follow-up prompt to list sources. ChatGPT, 13 Feb. version, OpenAI, 9 Mar. 2023, chat.openai.com/chat.

----------

MLA In-Text Format

("The first three words of the description of the chat in the full reference")

Examples Using ChatGPT

(“In 200 words”)

----------

For more information, see the MLA Style Center's article on citing generative AI (March 17, 2023).

APA Reference Format

Developer of the model. (Year of the version you used). Name of the AI model (Version number) [Additional description - for AI, this is used to detail the program as described by the developers]. Link to source.

Example Using ChatGPT

OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat

----------

APA In-Text Format

Parenthetical citation: (Author/developer of the model, year of the version you used)

Narrative citation: Author/developer of the model (Year of the version you used)

Examples Using ChatGPT

Parenthetical citation: (OpenAI, 2023)

Narrative citation: OpenAI (2023)

----------

For more information, see this post about AI citation styles on the APA Style Blog (April 7, 2023).

Because tools like ChatGPT are so new, many style guides do not yet have a preferred citation style for them. Please speak to your professor or contact Student Success & Outreach Librarian Elle Davis at davise@savannahstate.edu for further guidance.