AI Reference Interview Activity

What is AI prompting ?

AI prompting is the science and possibly art of giving AI a prompt or question that produces the desired outcome. This deceptively simple explanation can cover everything from the one to two sentence prompts you might use to ask a casual question, all the way to machine-coded queries that are hundreds of lines long and not even readable by humans!

Because AI’s work off of a probability model, the information they receive in the question or prompt has a huge impact on the output they create. Prompts open can open unexpected doors, both good and bad, in how you interact with these tools.

AI Prompting exercise

This activity has the goals of:

• introducing you to PRIMO research assistant, UIdaho Library’s new AI research tool
• guiding you through exploring how to develop a prompting strategy that helps you with your specific search

You’ll need:

• an active UIdaho library account (all students, staff and faculty have this!)

• to visit PRIMO Research Assistant, just look for the “…” at the top of your library search screen, then look for Primo Research Assistant.

Prompting practice

1. Go to Primo Research Assistant, make sure you are logged into your library account or just login when it asks you to.
2. Put in your search query in whatever format makes sense to you, for some this might be a few key words, or others it might be their actual research question or a similar normally worded sentence
3. Review the results, what do you notice? Are the sources cited ones you had come across previously? How do they look with respect to an evaluation framework, for instance CRAAP - which reviews sources to see if they are: current, relevant, accurate, authoritative, and intent/purpose.
4. Revise and shape your search query:
   • For instance try adding some of the following phrases and see how it shapes your results:
     • literature review OR meta synthesis
     • case study OR experiment
     • adding phrases like “impact” OR “outcome”
     • qualitative OR quantitative analysis
   • Use Research Assistant’s summary and input to explore and define key research question criteria such as:
     • relevant dates/dates of interest
     • populations of interest
     • geographies/locations of interest
     • existing theories
5. Conversational AI is also useful to explore your existing/starting hypothesis and understand what perspectives you might need to add to your existing frame to have a complete picture.

Pro-Tip: You can also search Google Scholar for AI served articles, this can be handy for tracking down citations that come out of large language models like ChatGPT.

Getting Good Results with AI

What results you get from any generative AI are hugely dependent on how you phrase and construct your questions. Because AI works in a probabilistic manner, which means it is determining which words are most likely to occur following previous words and weighting how those results are shown based on hundreds of thousands of virtual training hours, how you phrase, order, and logically construct your question makes a big impact on the type of results you get back.

Check out the CLEAR Framework for more on developing strong AI prompts.