Research Guides: Generative AI and Learning: Prompting with Purpose

Prompting Techniques

By the end of this section, you should be able to apply prompt engineering techniques and develop a clear strategy to improve the quality and relevance of AI-generated outputs.

When interacting with an AI tool, your goal isn’t just to get an answer—it’s to get a response that is relevant, clear and useful. A prompting strategy helps you avoid trial and error by giving you a structured way to communicate exactly what you need. Think of it like giving directions; the clearer your instructions, the better the outcome.

Recall that prompt engineering is the process of designing effective inputs (prompts) to guide generative AI tools toward producing accurate, relevant, and ethical outputs. In short, learning how to prompt improves your ability to communicate more effectively with the AI tool and ensure you get an output that is relevant and reliable.

The prompt engineering techniques below will help you guide AI tools more effectively. Click the hotspots to explore simple but powerful examples of each one.

Prompt Engineering Techniques
Use this guide for a complete description of this image.

Prompting Strategies

Turning Prompting Techniques into Prompting Strategies

A single good technique can improve your prompt. But combining several techniques into a structured approach or strategy can make interactions with AI more intentional, consistent, and effective. Below are three frameworks that incorporate multiple prompting techniques to help you craft better prompts and avoid common pitfalls, such as vague or unclear questions that lead to irrelevant or biased results.

All three frameworks emphasize the importance of reflection and evaluation. They encourage you to critically assess AI responses and fact-check claims using credible sources. This not only strengthens the reliability of the output but also helps you refine your approach. With practice, prompting becomes less guesswork and more of a purposeful, repeatable process, where you deliberately guide the AI toward meaningful, relevant results.

Prompting Frameworks

CLEAR stands for Concise, Logical, Explicit, Adaptive, and Reflective

The CLEAR framework emphasizes clarity, critical thinking and iterative refinement, making it a valuable tool for improving the quality of the prompt and the accuracy/usefulness of the response. By breaking down complex tasks into manageable steps, evaluating the output and refining your prompts, CLEAR helps you make informed decisions about the quality of the outputs.

The The CLEAR Path: A Framework for Enhancing Information Literacy through Prompt Engineering by Leo S. Lo under a CC BY-NC 4.0 license.
Image generated using Napkin.ai

Follow these guidelines when using the CLEAR framework

C - Concise – Keep your language clear and direct. Use simple, specific wording to avoid confusion.
- Check: Does your prompt avoid unnecessary words?
- Example: "Summarize the main points from an open-access or publicly available article on climate change."
L - Logical – Structure your prompt in a sequence that’s easy to follow. Guide the AI step-by-step.
- Check: Does your prompt follow a logical sequence?
- Example: "First, identify the key arguments in the article. Next, summarize each argument in less than 100 words. Finally, conclude with the overall message of the article."
E - Explicit – Don't assume the AI knows what you mean.
- Check: Does your request include enough detail or examples for added context?
- Example: "Explain the causes of climate change. Include at least three causes and provide a brief explanation for each."
A - Adaptive – Refine your prompt as needed.
- Check: Does the AI's response meet your expectations?
- Example: If the initial summary is too long, "shorten it to focus only on the most critical points."
R - Reflective – Evaluate the output and reflect on how to improve future prompts.
- Check: What worked well and what needs to be revised?
- Example: Review the summary for accuracy and clarity. Follow any links that are provided to ensure the sources are real. If any important points are missing, revise the prompt to include those details.

Activity: Try out the full prompt in Copilot

Instructions:

Open the USask version of Copilot and paste the prompt below into the 'Ask anything' field.
Review the AI's response using the CLEAR checklist.
Revise the prompt if needed, and try again to improve the results.

Full Prompt to Copy and Paste (reorganized for clarity):

Find a publicly available article that explains the causes of climate change. Start by identifying the main points or arguments the article makes about what's causing climate change. Then, write a short summary of each point using your own words, keeping each summary under 100 words. After that, explain the overall message or conclusion of the article in a short paragraph. Finally, list at least three specific causes of climate change mentioned in the article, and give a brief explanation of each one. Use clear and simple language, as if you're explaining it to someone who's new to the topic. Make sure your writing is easy to follow and helps the reader understand the key ideas.

Review the AI's summary for accuracy and clarity. Follow any links that are provided to ensure the sources are real and that the information is accurate. If any important points are missing, revise the prompt to include those details. If the initial summary is too long, ask the AI to shorten it by focusing only on the most critical points.

CLEAR Checklist
Use this checklist to evaluate the AI's output and refine your prompts.

Many newer AI tools feature enhanced capabilities like RAG (Retrieval Augmented Generation) and Reasoning. RAG improves accuracy by retrieving real-time information from external sources, such as research papers or databases, rather than relying solely on pre-trained data and reasoning allows users to see how the AI interprets their prompts, which can help refine queries for more accurate or relevant results. This is particularly helpful for users seeking to understand how the AI arrives at its conclusions.

The PROBE Framework complements these capabilities by guiding users to critically evaluate AI responses. The 'R' in PROBE stands for 'Reason,' encouraging you to ask the AI to explain its reasoning process. This is particularly useful with advanced AI systems designed for research, coding, and academic writing, as they can break down responses into logical steps.

By combining PROBE with advanced AI tools, you can achieve more accurate, relevant, and transparent results in your AI interactions.

PROBE stands for Pose, Reason, Obtain, Break, and Evaluate

Image generated using Napkin.ai

Follow these steps when using the PROBE Framework

Pose the Question – Clearly define what you want the AI to answer.
- Example: "Explain the impact of social media on mental health."
Request Reasons – Ask the AI to explain its thinking.
- Example: "Provide a logical explanation for each point you make, showing how specific aspects of social media use lead to particular mental health outcomes."
Obtain Supporting Evidence – Seek examples, references, or sources.
- Example: "Present your explanation as a series of cause-and-effect relationships, with each point supported by at least one research finding or statistic."
Break it Down – Divide complex topics into smaller, manageable parts.
- Example: "First, discuss the positive impacts of social media on mental health. Then, explain the negative impacts. Finally, describe any factors that might influence whether the impact is positive or negative."
Evaluate the Output – Critically assess the AI’s response.
- Example: Cross-check the cited research or statistics against reputable sources, look for over-generalizations or unsupported claims, and identify any potential biases in the AI's explanation.

Activity: Try out the full prompt in Copilot

Instructions:

Open the USask version of Copilot (NSID required)
Paste the prompt below into the 'Ask anything' field.
Evaluate the output for accuracy, reliability and bias.
Revise the prompt if needed, and try again to improve the results.

Full Prompt to Copy and Paste

Explain the impact of social media on mental health. Provide a logical explanation for each point, showing how specific aspects of social media use lead to particular mental health outcomes. Present your explanation as a series of cause-and-effect relationships, with each point supported by at least one research finding or statistic. First, discuss the positive impacts of social media on mental health. Then, explain the negative impacts. Finally, describe any factors that might influence whether the impact is positive or negative.

Evaluate the Output: After receiving the AI's response, critically assess its accuracy by cross-checking cited research, identifying any over-generalizations, and evaluating potential biases in the explanation.

⚠️ Be mindful of the type of license attached to the material you use or enter into an AI model. While AI models can learn from open resources and the public domain, copyrighted or sensitive materials cannot and should not be used without proper permission from the content's creator.

This framework uses the acronym PROMPT and was developed with the assistance of Perplexity.ai to emphasize the importance of evaluating an AI's output for accuracy and relevance, as well as refining the prompt as needed.

PROMPT stands for Purpose, Role, Outcome, Method, Preferences, Test

Image generated using Napkin.ai

Steps to follow when using the 'PROMPT' framework

Purpose: Clearly define the goal of your prompt. What do you want the AI to achieve?
- Example: "Generate a concise summary of a recent scientific discovery in climate change."
Role: Assign the AI a specific role or persona to provide contextual guidance. Who should the AI act as?
- Example: "Act as a science journalist explaining key findings to a general audience."
Outcome: State the goal or result you expect from the AI’s response. What should the AI’s response include?
- Example: "Summarize the study's key findings and explain their significance in everyday terms."
Method: Outline the approach the AI should follow. How should the AI process the information?
- Example: "Identify the main discovery, describe its impact, and provide an example of real-world implications."
Preferences: Set constraints like length, format, or tone. What specific requirements should the response follow?
- Example: "Keep the summary under 150 words and use clear, engaging language."
Test: Evaluate the response for accuracy. Did it meet your expectations? If not, how can you improve the prompt?
- Example: "Review for clarity and accuracy. Verify the information with reputable sources. If the response is too technical, too long, or too short, refine the prompt for a more accessible explanation."

Activity: Try out the full prompt in Copilot

Instructions:

Open the USask version of Copilot and paste the prompt below into the 'Ask anything' field.
Review the AI's response for clarity and accuracy.
Revise the prompt if needed, and try again to improve the results.

Full Prompt to Copy and Paste

Generate a concise summary of a recent scientific discovery in climate change. Act as a science journalist explaining the key findings to a general audience. Summarize the study’s main discovery, describe its impact, and provide an example of real-world implications. Keep the summary under 150 words and use clear, engaging language.

Test: Review the AI's response for clarity and accuracy. Verify the information with reputable sources. If the response is too technical, too long or too short, refine the prompt for a more accessible and useful explanation.