A.3 - Using Generative AI Effectively

To learn to use Generative AI effectively, you must learn to prompt it effectively. An ineffective prompt costs just as much energy as an effective one.

To prompt effectively, consider using the CLEAR framework (Lo, 2023). The CLEAR framework can be a “scaffold for devising effective prompts for AI language models, combining precision and creativity, science, and art” (Lo, 2023, p. 208).  

Concise

Clear and precise prompts guide the AI model’s understanding, leading to a more relevant response.

Initial Prompt - "I’m working on this project with data and a report and I want it clear and organized and also need help figuring out what to include."

Improved Prompt - "Outline a data report, including key sections, analyses, and presentation suggestions."

Note: Limiting each prompt to one request or question is recommended to not overwhelm the tool. Remember, it is still just a machine!

Logical

Providing context within prompts helps the AI model create meaningful generated output.

Initial Prompt: “Tell me about programming.”

Improved Prompt: “Explain the concept of loops in programming languages and their importance in executing repetitive tasks.”

Note: To provide context, try defining the role you would like the AI tool to assume or describe your role or level of knowledge. For example:

• You are a first-year computer science student.

• Act as an expert in programming languages.

• Assume I am 10 years old.

• Explain the concept of loops in programming languages.

Explicit

Including a specific structure or style in your prompt may assist the AI model in meeting user expectations in terms of the format and length of the generated responses.

Initial Prompt: “What are the causes of climate change?”

Improved Prompt: “List and explain in 10 sentences the primary human-induced factors contributing to climate change, including the impact of carbon emissions and deforestation. Use lay terms.”

Note: To help you get more accurate results, consider specifying the following in your prompts:

• Background information related to the topic (what AI can base its response on)

• Length of response (number of sentences, paragraphs, words, or characters)

• Format (a list, paragraph, chart)

• Structure (essay, blog post, poem)

• Style (formal or informal)

• Target audience (who this is intended for)

• Examples (provide an example of what you are looking for)

----------

Westling (2023) sums up these details into the following AI prompt formula:

[Deliverable specifics] + [Voice] + [Objective] + [Ideal format]

Example AI Prompt: “Review my 1,000-word argumentative essay draft on climate policy (deliverable specifics) in a supportive but honest tone (voice). I want detailed feedback on how to strengthen my thesis statement and improve transitions between paragraphs (objective). Provide your advice in a numbered list with specific examples from my draft (ideal format).” 

Adaptive

Experiment with general and then more specific questions or statements to assist the Al model in generating a response that matches the desired output.

Initial Prompt: “How does technology affect society?”

Improved Prompt: “Discuss both the positive and negative impacts of technology on various aspects of society, including education, healthcare, and social interactions.”

Reflective

Evaluate and improve the accuracy, coherence, and utility of the generated responses. For example, after receiving a final set of responses, tally how many, what kind, and how the prompts were adapted to achieve the desired result. Identify what could be refined in your prompt and try again.

• • Rephrase your initial prompt.

  • Give AI more information.

  • Change your approach.

Note: Use the “chain of thought” technique to increase response accuracy, e.g. by using prompts like “Work on this problem step-by-step” or “Are you sure?” (Chen et al., 2023).

What Is Meta-prompting?

Meta-prompting is the practice of writing prompts about prompts—that is, crafting instructions whose primary purpose is to improve, refine, or guide the creation of other prompts. Instead of directly asking an AI to do a task (like “Write a poem about rivers”), a meta-prompt asks the AI to help you generate a better prompt or a set of prompts for that task (like “Generate three strong prompts that would produce creative poems about rivers”).

This approach stems from the idea that the quality of a prompt heavily influences the quality of an AI’s response. By focusing on the prompt itself, users can clarify their goals, structure their inputs more effectively, and gain richer, more useful outputs. In other words, meta-prompting adds a layer of design and reflection to the prompting process.

Why Meta-prompting Matters

• Higher-Quality Outputs – Well-designed prompts are more likely to elicit detailed, accurate, and creative responses.

• Efficiency – It saves time by frontloading the “thinking work” into the prompt instead of rewriting requests repeatedly.

• Learning Tool – It teaches you (and the AI) how to think critically about the instructions you’re giving.

• Scalability – In research, teaching, or content creation, meta-prompting allows you to produce a template for multiple similar tasks quickly.

Core Components of a Meta-prompt

When you build a meta-prompt, you’re typically doing one or more of the following:

• Clarifying the task (“Help me define exactly what my essay feedback prompt should ask for”).

• Specifying constraints (“Generate a prompt that limits the response to 300 words”).

• Defining style or tone (“Write a prompt that requests a friendly but professional voice”).

• Structuring outputs (“Ask for bullet points instead of paragraphs”).

Essentially, a meta-prompt acts as an instruction template for generating instructions.

----------

Example

• Direct Prompt: “Write an outline for an article about urban gardening.”

• Meta-prompt: “Generate three different prompts that would produce high-quality outlines for an article about urban gardening, each using a different organizational style.”

The meta-prompt doesn’t ask for the article outline itself; it asks for better ways of asking for that outline.

This chapter contains materials from:

•  Introduction to Academic Writing by Nancy Bray, University of Alberta, licensed under CC BY-NC-SA 4.0.