Given the limitations previously discussed, evaluating the output of generative artificial intelligence (GenAI) is a crucial skill for anyone using GenAI in any context, but especially for those using GenAI in post-secondary. Failure to evaluate GenAI output could result in a vague essay at best, and plagiarism at worst.
These tips are based off the SIFT fact-checking model developed by digital literacy expert Mike Caufield, and adapted to specifically address evaluating GenAI output.
GenAI can make up sources. It is therefore crucial to verify that a source actually exists and that the information attributed to the source is correct. As an NWP student, you have access to licensed databases that you can use (alongside Google Scholar or individual journal websites) to confirm that a particular article exists and analyze the information in the article for yourself.
Find other sources to compare and verify the claims made by the text GenAI created. As an NWP student, you have access to tons of peer-reviewed and scholarly articles on a multitude of topics via our Library databases.
If you are required to find recent information, make sure you check how recent the sources GenAI is using are, and remember that programs like ChatGPT are often limited (e.g., as of July 2024, ChatGPT 3.5 -free version- can offer no knowledge or input regarding information later than January 2022)
The video below by UBC Learn (2024) describes some considerations and approaches when evaluating the content created by GenAI.
Developed by the University of Alberta, this webpage offers useful tips on verifying the information provided by GenAI.
Stop. Investigate. Find. Trace. This webpage arranges the process of evaluating GenAI output into a four-step process.
The material in this guide is based on material from Using generative AI by Deakin University (2023). Content has been adapted for the NWP Learning Commons in July 2024. This work is licensed under a Creative Commons BY NC SA 4.0 International License.
All icons on these pages are from The Noun Project. See individual icons for creator attribution.