Using Generative Artificial Intelligence for Research

In research, the planning phase is key - it sets the stage for the entire project.

Taking time to think deeply about your topic and learning to incorporate generative AI tools can help you streamline exploration and analysis. For information on getting started with the research process, check out the Research Guides by subject and research skill.

In the next page of this guide, you will learn how to construct effective prompts for text-based and image-based outputs. 

Understanding the scope, details and limitations of your assignment will affect how you approach your research.

Read through your assignment guidelines, and consider the following:

• What is the project format? (i.e. written essay, oral presentation, poster presentation, media production, etc.)
• How many sources will you need?
• What types of sources can you use? (i.e. scholarly journals, books, newspapers, websites, primary or secondary sources, etc.)
• Will you need visuals and graphics?

Clarify any doubts you have about your assignment with your instructor, and be sure to verify if your instructor allows the use of generative AI to support your assignment.

Although generative AI chatbots can be very useful in various phases of the research process, it is important to be aware that chatbots are not the same as search engines. Not all chatbots have real-time access to the internet, and while search engines like library databases or Google connect users to existing texts, generative chatbots create new text based on complex LLMs and the information they have been trained on. Chatbots that do have real-time access to the internet will often include internet search results along with the chat response. However, all generative AI tools will have distinct limitations that users should be aware of:

• Hallucinations or fictitious information
  • LLMs are prone to generating fictitious information presented as being factual or accurate. This can include citations and publications.
• Biased information
  • Chatbots learn from the training data they are fed, which could contain biases that can influence the chatbot's responses. 
• Establishing credibility
  • Generative AI tools are not databases of knowledge, but rather a tool that synthesizes and reproduces information they have been trained on. This makes it difficult to properly validate or establish credibility their outputs.