Information Literacy & Library Research: AI Prompt Engineering

AI Prompt Engineering

Generative AI, such as ChatGPT or Google Gemini, can be helpful in many ways. But AI output is only as good as your input. To get good results, you need to ask it good questions or give it good instructions so it knows exactly what you want from it. And then you need to evaluate the responses and adjust accordingly.

Never just copy and paste straight from AI, but use it as a tool to help you improve your own workflow and boost your creativity. It’s better used to help you make the process easier, and not to replace the process altogether. Never outsource your thinking!

CLEAR Framework for Prompt Engineering

The CLEAR Framework for Prompt Engineer (see Figure X) provides directions for writing effective AI prompts. CLEAR stands for Concise, Logical, Explicit, Adaptive, and Reflective

Concise: Focus on keeping your prompts short, sweet, and to the point. Only include exactly what you need to say to convey your point.

• Example: Brainstorm topics for a college research project on Fake News.

Logical: AI uses logic, so make sure your prompts are structured in a logical flow to be coherent and easy to follow, or the AI engine will get lost.

• Example: Take me through the steps of the academic research process, starting with planning and ending with dissemination, including library research.

Explicit: Be clear and precise with the exact details needed to give you the output you want. Provide precise details about output format, content, or scope. Be sure to give AI the context of what you need.

• Example: help me narrow down the topic of “fake news, confirmation bias, and sharing” for a research poster.

Adaptive: As you create your prompts, learn from the results and refine them until you get the results you want. If your first prompt is too vague, add more specifics, etc.

• Example: help me narrow down the topic of “fake news, confirmation bias, and sharing” into a research poster with a 5-minute presentation covering the poster for my college information literacy class.

Reflective: Continually review and evaluate prompts, improving them until you get the results that will be useful to you.

References

Lo, L. S. (2023). The CLEAR path: A framework for enhancing information literacy through prompt engineering. The Journal of Academic Librarianship, 49(4), 102720. https://doi.org/10.1016/j.acalib.2023.102720

Advanced AI Prompting Techniques

After perfecting your CLEAR prompting, here are additional techniques to create effective AI prompts. 

Examples: Using examples gives the AI a chance to see the pattern in what you’d like created, and will allow it to give a better output.

Stepwise: Have AI work through your request one step at a time, getting confirmation before proceeding to the next step. This is important because AI might fixate on the last part of your prompt, forgetting the beginning. This is called Recency Bias. If you force the AI to go one step at a time, that helps prevent it from getting lost in the process.

• Example:
  • I need to write an email address authentication code in Python. Go step by step and don’t move on until I respond with “next.”

Prompt in Conversation: Go back and forth, refining your prompt until the AI fully understands what you’d like. This relates to the adaptive step of CLEAR (see above).

• Example:
  • Your prompt: help me narrow down the topic of “fake news, confirmation bias, and sharing” 
  • AI gives several approaches
  • Follow up prompt: Can you narrow it down for a 5-7 minute presentation using approach 2?
  • etc.

Q&A: Present the problem and tell the AI to ask relevant questions until it has enough details to give a good enough answer.

• Avoids assumptions, problem-solving approach.
• Example:
  • I need to refine a topic for my college English 2010 paper on Mental Health. Before giving recommendations, please ask me relevant questions, one at a time, about my assignment requirements and topic ideas before you give me recommendations 

Pros and Cons: Ask AI to give the pros and cons of all the options listed.

• Helps to avoid bias, reveals options and ideas you might not have considered.
• Gives you control over the process as it allows for review and course correction.
• Example:
  • I’m writing a college paper about recycling. Please analyze the pros and cons of recycling on the environment. Consider factors like water usage, government policies, and economics.

Roleplay: Ask AI to take on a specific role or use certain expertise, to give AI context for advice.

• Focuses on specific domain expertise, provides different perspectives
• Example:
  • Think like a College English Writing instructor and give me feedback on the following research question. In your feedback, focus on the strength of the research question and how researchable the topic is: 
    • Are dogs good for mental health and can they help with depression?
    • Do college students who attend weekly campus therapy dog events report lower levels of perceived stress compared to students who do not?
      • How would you rewrite the question if yes/no questions aren't allowed?