Skip to Main Content

Quinebaug Valley Campus Library

Quinebaug Valley Artificial Intelligence

Prompt Engineering

Getting information from AI is like all other research, it is a process, not just a one search and then finished.  There are good prompts and queries that will generate good information.  There are also bad prompts that will generate no, false, or misleading information.  As AI continues to evolve the way we communicate with it will evolve.  For now try using the CLEAR Framework to get the most out of your prompt.  Collecting information with AI is like research with other technology.  To answer questions you will likely spend a lot of time refining your queries and using multiple queries to gather the information you need.  

CLEAR Framework

C - Clear (& Concise)

Like searching in a database we need to go through the process of identifying keywords.  This helps us to determine what the most important part of our question are.  Present the most important parts of our question first, and make sure that we leave out any unnecessary information.  The kind of information could cause the AI to become confused.

 

L – Logical

Because of the way that AI understands information, it is important that queries are structured with a logical progression.  Structure the prompt in a way that allows the AI to easily make connections between the concepts.  Make sure that your prompt makes sense to you.  Try reading it outload.  If something does not sound right, try reprioritizing or rewording.

 

E – Explicit

This consideration involves detailing exactly the kind of information you are seeking and the format that you want that output to be presented in.  Do not just ask about science fiction movies but ask about climate change in post-apocalyptic movies.  Make sure to tell the AI how many much information you want to be given, if the response be in a formal or informal voice, if you want the AI to omit certain data points, whether this is an outline or a summary, etc.

 

A - Adaptive

Never settle for the first response from AI.  Be willing to adapt your question based on the information you receive from the AI.  Try multiple variations of your research question and compare the different responses to see the differences or similarities in the outputs.  If you have too much information or if the information provided is too broad add more details.  Read the AI generated materials to see if anything seems interesting, is stated in a way you would not have considered, if data is presented that should be excluded, etc. 

 

R - Reflective

Always think about the information you are given by an AI and how it fits into your research or task.  Are there any elements of the response that do not make sense?  Does anything in the output seem like it might be outdated?  Double check the information with multiple sources.  Ask the AI to provide citations for the information provided.  Determine if the AI has missed or ignored aspects of the prompt or the research.  Are there any further prompts that would help answer the research question?