Answered By Library Staff
Last Updated: Aug 29, 2023     Views: 51

The answer is yes and no. 

While some text-based generative AIs like ChatGPT can help you brainstorm your research topic or make sure you're not missing something about a topic you've selected, you need to be aware that to quote from AI or to even use what it generates in your own work might present issues. You cannot trust that everything an AI says is true, even if it does present sources. Oftentimes, the sources it cites do not exist or, if the source is real, it might not be the most academic source available or have interpreted the work correctly. You will need to verify and think critically about the source or claims from the AI. We recommend the SIFT method for doing so.

Here is an example of a search using Perplexity, where it gives an answer, but the sources itself could be good leads (that you would need to verify using the SIFT method, much like you would if using a Wikipedia article) for the research question: 

screenshot of link example


Also, quoting from an AI or copying and pasting from it could present plagiarism or academic integrity issues for your work. Be sure to check with your instructor for what is allowed in your assignment before using the work of an AI. 

Using AI for brainstorming or summarizing  

But, in terms of how AI might help you in your research, it could be a good brainstorming tool that gets you to narrow in on your research topic or summarize a topic in a new way so you can better understand. It can, for example, give you a list of keywords or search terms related to your topic in order to create better searches. AIs like ChatGPT are not a search engine, so the previous ways you would phrase a search might not yield results that will work for you. Outputs from AI are usually in conversation form, building off previous prompts. Sometimes, however, it can summarize a topic better than, say, a Wikipedia page, being better at condensing the issue or rephrasing it multiple times and therefore potentially being (but not always!) better at introducing the topic. Take this "Age of Aquarius" example (transcript attached). While Wikipedia might be where you turn to to start verifying aspects of this topic, the response the AI gave is a lot more approachable and easy to understand (and a lot shorter than the Wikipedia article in this case!): 

shows side-by-side wikipedia article and AI resoponse

(open image in new window to expand view or see transcript below)

Below are some suggestions for how to to get ChatGPT to work as effectively as possible, based on the tips mentioned in this video below for educators and librarians

Prompt Tips for Interacting with Generative AI: 

Open image in new tab to expand. Transcription of interaction is attached as a Word file. Chat conversation is meant to illustrate the brainstorming abilities of AI, as well as help the researcher shape their topic and glean keywords and areas of research focus. 

  1. Provide context for what you want the AI to generate,
    • What do you want the answer it generates to include? (A list? Pro/Con arguments about a topic? 5 paragraphs?) What are 5 Questions 
    • Give the AI a role or persona or POV. Christian 
    • The AI will generally match your language and terminology. "Might have" used in prompt and "might have" used in response. 
    • "What are 5 questions about Islam that a Christian might have?" example and answer
  2. Ask follow up questions. 
    • It's a conversation (not a search engine), so you can ask it to rephrase something, clarify a point, refine something. Or, you can edit your prompt/question or have it regenerate a response. Use of #3 in example
    • It's OK to ask it to answer questions step-by-step or to break things down by components. 
  3. Get the AI to solve a problem, or clarify your task or objective: 
    • "Summarize the following text for me..." 
    • "What is the theme of X work...." 
    • "... in words a child could understand." 
    • "Debate or reflect on the issue of..." 
    • task clarification
  4. Minimize hallucinations and wordcount by structuring prompts with language like:
    • “if you don’t know the answer please respond ‘I do not know’”
    • "limit your response to around 500 words." 
    • asking AI to limit to 300 words or less