Fact-Checking Generative AI and Avoiding Plagiarism

For a detailed understanding of Unity’s stance on the use of generative AI in your studies, please read through Unity’s Policy on Generative AI use by Students in Distance Education.

Does this sound familiar?

You’re chatting with a friend and they mention that they’ve started using Generative AI like ChatGPT to help write their essays for school. It’s saving them a ton of time and improving their writing. A few days later, it’s Sunday night and you have a long academic report due. You are only halfway done writing it and you are worried you won’t finish on time — you are STRESSED.

You remember your friend’s experience using ChatGPT and you think maybe you could use it to help you write the final few paragraphs of your report. As you enter a simple prompt, you discover that ChatGPT is incredibly easy to use and you throw a few AI generated paragraphs at the end of your essay, then hit submit.

A week later, you receive your grade — it is disappointingly low and comes with a message from your instructor:

“While we permit the use of generative AI in our courses, it is considered plagiarism if you do not properly cite the AI tool you use and indicate which pieces of this assignment were generated by AI. You’ll also want to fact-check anything written by AI, because it may be inaccurate, or even false. I saw some inaccuracies in your writing that you definitely should have double-checked.”

Why do we need to Fact-Check Generative Artficial Intelligence?

AI tools like ChatGPT can be powerful, but they aren’t perfect. Generative AI tools use algorithms or formulas that gather information from a number of sources like private and public data, data created by AI programmers, and vast amounts of text from the internet.

But be careful! Sometimes generative AI tools can provide answers with inaccurate information!

This is because:

They are susceptible to the biases and inaccuracies present in the data they use.
Sometimes they scramble to make up answers for prompts they do not understand or have knowledge of.

Just like you wouldn’t trust information posted by a random person on social media without double-checking it, you shouldn’t blindly trust AI either.

IMPORTANT: Generative AI tools like ChatGPT and Claude DO NOT note when they provide misinformation, they will provide answers regardless of the algorithm’s accuracy, therefore it is up to you to assess the reliability of information provided in generative AI outputs!

Some examples of inaccurate information generative AI has been found to provide:

Outdated Information: AI might not have the latest updates if newer information isn’t included in its training data.
Oversimplification of Complex Topics: AI may not have the capacity to interpret complex topics accurately and may provide oversimplified interpretations.
Fabricated Facts: If AI does not have access to specific details on a topic it will make an educated guess on the topic and make up an answer.
Misinterpreted Data: AI may not be able to complete complex data analyses and may misinterpret data, providing false claims.
Generalizations and Biases: Depending on its sources, AI may provide answers with biases from sources on the internet. AI does not have the social consciousness to evaluate and question these biases.
Incorrect Mathematical Calculations: AI can sometimes provide incorrect answers to mathematical problems, but will always attempt to provide an answer even if it does not know how to do the calculation.
Misquoted or Nonexistent References: AI can misquote or attribute quotes to the wrong sources and has been known to make up sources and citations entirely.
Fiction Presented as Fact: AI does not possess the cognitive ability to differentiate when information in its sources is based in reality or not, so it may present fictional information as if it were fact.

Best Practices for Fact-Checking AI

Step 1: Verify with Reliable Sources

When the AI gives you information, cross-check it with trusted sources:

Use Academic Databases: Check AI generated information against primary and peer-reviewed sources from academic search engines like JSTOR, PubMed, or Google Scholar.

Consult Reputable Websites: Look for information on well-known and reputable sites like official government websites (e.g., .gov), educational institutions (e.g., .edu), and major news outlets (e.g., BBC, The New York Times).

Step 2: Cross-Reference Multiple Sources

Don’t rely on just one source. Cross-reference information from AI with at least two or three reliable sources to confirm the accuracy.

If more than one source provides contradictory information to what is provided by AI, go ahead and change this information then cite these sources.

Step 3: Evaluate the Relevance and Credibility

Assess the credibility of your sources by considering:

The Author’s Credentials: Are they real and actually an expert in the field?
Source Authenticity: Is it a real publication in a peer-reviewed journal or a real website from a recognized institution?
Bias: Is the source objective, or does it have a particular bias?
Relevance/Timeliness: Is the date of the publication recent enough to trust its relevance?

Steps for Avoiding Plagiarism with Generative AI

Step 1: Understand What Constitutes Plagiarism

Plagiarism is using someone else’s words or ideas without giving proper credit. This includes copying text verbatim, paraphrasing too closely, or using ideas without acknowledgment.

Step 2: Always Cite Your Sources

When you use AI-generated content, you must cite it just like any other source. See the previous guide on how to do this properly in APA, MLA, or Chicago style.

Step 3: Paraphrase Properly

If you use AI to help you write, make sure you:

Understand the Content: Fully grasp what the AI is saying.
Use Your Own Words: Rewrite the information in your unique style and structure.
Credit the AI: Even when paraphrasing, give credit to the AI tool.

Step 4: Use Plagiarism Detection Tools

Because generative AI is compiling information from several sources, you want to make sure that it is not pulling directly quoted text from sources that are not properly cited. To avoid unintentional plagiarism, try the following:

Before submitting your work, run it through plagiarism detection tools like Grammarly, Copyscape or other online tools to ensure there is no directly quoted text that is not properly cited.
- NOTE: Please be aware that online plagiarism detection tools can be helpful but they are not 100% assurances that plagiarized text has not been included in your work.
Whether or not you use a plagiarism checker, we also advise you to thoroughly review your work before submission. If you find any facts, statistics, or ideas that are not supported with citations, you should investigate this information and find appropriate citations to include.

Step 5: Keep Records of Your AI Interactions

Save transcripts of your AI chats and include them in an appendix if necessary. This transparency can help you avoid plagiarism accusations.

Examples

Example 1: Fact-Checking in Action

AI Output: “The Great Wall of China can be seen from space.”

Fact-Check: According to NASA, this is a myth. The Great Wall is not visible from space with the naked eye.

Example 2: Avoiding Plagiarism

AI Output: “The mitochondrion is the powerhouse of the cell.”

Your Work: “Mitochondria are often referred to as the cell’s powerhouses because they generate most of the cell’s supply of ATP (Adenosine Triphosphate), used as a source of chemical energy.”

Tips: Be Skeptical and Curious!

DO NOT JUST COPY AND PASTE!
Always question the information and dig deeper.
If something looks or sounds off in an AI output, it probably needs more scrutiny.
Generative AI tools are imperfect and using them properly will require discernment.
Be sure you are properly citing information and text generated by generative AI.

This resource was developed with assistance from ChatGPT-4. Illustrations by Storyset.

Info For