AI for OER Discovery: Building Good Prompts for an OER Search

Your prompt is what you enter into the "message" part of an AI chatbot. Better prompts = better outputs. 

• Tell Me Your Prompts and I Will Make Them True: The Alchemy of Prompt Engineering and Generative AI. Bozkurt, Aras (2024)

  This paper explores the emerging field of prompt engineering within generative AI, emphasizing its role as a critical intersection between art and science. Prompt engineering is identified as the key to unlocking the full potential of generative AI technologies by optimizing human-AI communication. Through a comprehensive analysis of the related literature, this study illustrates how prompt engineering transcends mere technical manipulation, requiring a blend of creativity, strategic thinking, and a deep understanding of generative AI capabilities.

Be clear and specific

Use complete sentences. AI is trained on natural language, so complete sentences will be the best way to translate your prompt into a desired output.

It helps to think of the exact results you want and write it out. Be specific about every little detail as it relates to your course or assignment requirements.

• Example: “Please list databases that contain open educational resources (OER) for undergraduate physics courses, including links and a brief description of each.”

Use keywords and phrases, especially some professional jargon to help guide AI in generating results specific to a chosen field. Being clear about what you want (e.g., type of resource, educational level, subject area) increases the likelihood of a helpful response.

• Example: “Find OER databases that offer nursing simulation scenarios for clinical instruction at the associate degree level.”

Follow-up

Most AI chatbots are tracking your inputs like a conversation and will be able to reference recent activity. If your first attempt does not yield the results you want, assess what may be missing and adapt your prompt accordingly. Use follow-up questions to narrow results or request alternative formats.

• Examples: “List OER platforms that provide resources for introductory psychology.”, "only include results that offer eBooks as a downloadable .pdf",  “narrow that list to those offering interactive media or video content.”, or "summarize the strengths and weaknesses of each platform”.

Check your own bias

The way you phrase a prompt affects the output. If you're not looking for persuasive/opinionated information, use neutral phrasing and don't ask leading questions. If you're asking about content regarding a controversial topic, but need both sides, ask for OERs that provide diverse sources. If you need research within a specific timeframe, include that as a parameter in your request. 

• Example: "Find OER content supporting the idea that Portugal's drug decriminalization policies could be applied in the United States with the same results." vs. "Locate at least 5 OER that contain information published in the last five years regarding drug decriminalization policies. Must include sources that provide comparative analysis, case studies, and primary source data from government and substance abuse treatment organizations."

Give context

Giving the AI context or a role to help deliver the most relevant results.

• Example: “I am an academic librarian helping a university professor locate OER for course design. Suggest OER repositories for a history professor seeking materials to provide students for their course on 20th-century African political and military history. Explain your process for evaluating the credibility of each database. Also, include a brief description for each repository regarding why the content you located is relevant to this request.”

Ask open-ended questions

To help determine the value of AI-recommended OER databases, you should ask questions. Chatbots don't stop at "yes" or "no" answers, even when you phrase your questions that way. However, by providing an open-ended question, you will get more information.

• Example: “What engineering-related resources are available on OpenStax, and how do they compare to those on MERLOT?”

Break down complex searches

If your course content covers a variety of broad topics, break your search into smaller, focused prompts that include subject-specific language. This will result in more manageable and thorough responses.

• Example: "Where can I find OER for STEM?" vs. “What are the most popular OER databases that offer open textbooks for undergraduate physics?”, “Which platforms host OER videos or labs for chemistry instruction?”, “Which STEM-focused OERs provide content aligned with standards like NGSS?”

Specify the file type format of your desired content

OER databases provide a variety of file types and types of sources for their content. You can always request specific kinds of content to best suit your needs.

You can ask for specific file types like .pdf, .csv, .txt., .wav, .png, etc.

You can ask for content that includes abstracts, data sets, literature reviews, illustrations, etc.

You can ask for content that includes primary sources, secondary sources, scholarly, peer-reviewed, grey literature, etc.  

• Example: "Locate three OER databases that provide primary source statistics for high school (grades 8-12) graduation rates in the United States from 1996-2004. I will be analyzing the data myself and prefer .csv files." 

General Guidelines for Success

Try new prompts and iterate on them to get the best results. 

Use a combination of the strategies above to yield the best results. 

Always verify every resource an AI suggests to make sure the content is real, accurate, current, and relevant to your needs.