How To Guide: Prompt Crafting 101 For Technical Writers
Effective communication of complex technical concepts is paramount in today’s technology-driven world. Tools like ChatGPT offer the potential to streamline technical writing, but the key to unlocking their value lies in prompt crafting.
This guide will introduce you to the art of prompt crafting, specifically emphasizing using ChatGPT for technical writing. We’ll begin with a basic understanding of prompts and explore practical examples. By the end, you’ll be able to use ChatGPT as a powerful tool to create high-quality, informative technical content.
So, you have downloaded 1500 ChatGPT prompts that worked for someone else. What happens next? Prompt Crafting is the key to making AI ChatBots like ChatGPT work for you!
This article is rather long but summarizes a course I created. I will soon create a course on this site.
Key Takeaways
-
Clarity is key: Ensure prompts are clear and concise to avoid misinterpretations by the AI, leading to inaccurate or irrelevant outputs.
-
Tailor your prompts: Consider the target audience and desired outcome to craft prompts that effectively guide the AI toward the information you need.
-
AI persona matters: When crafting prompts, consider the AI’s capabilities and limitations. This helps you capitalize on the AI’s strengths and avoid asking for tasks it can’t perform.
-
Start simple, iterate: Begin with basic prompts and gradually increase complexity as you refine the outputs. This iterative approach ensures you get the level of detail you need.
-
Break down complexity: For complex tasks, break them into smaller subtasks with individual prompts. This makes the prompts easier for the AI to understand and improves the accuracy of the results.
The final section is a hands-on lab where we will put this all together. The example creates a Power BI Course Outline. However, the framework can be used for many different use cases. The following sections have prompt segments that provide an overall direction for our query. Change the content between the brackets [ ] to craft your own prompt.
What is Prompt Crafting 2.0
As more people gain expertise, Prompt Crafting 2.0, as I call it, has developed. By providing a more detailed role, called a Persona, more context with instructions, we can get results that provide more consistent results and higher quality. Being able to create responses in your voice that match the format that you write in.
Let’s go through a step-by-step explanation of the ChatGPT prompt framework I use, which is adaptable to various projects such as technical writing. As an example hands-on lab, we’ll demonstrate how to generate an outline for a workshop on Power BI fundamentals.
Computers That Talk? It’s Called NLP
Natural language processing (NLP) combines computer science and linguistics to develop computer systems that can interact with human language. Using machine learning and artificial intelligence techniques, NLP systems can analyze, interpret, and even generate text and speech, enabling tasks like sentiment analysis, machine translation, and question-answering systems.
Computers That Talk? It’s Called Natural Language Processing (NLP)
Why NLP is Important in AI Prompting
NLP plays a crucial role in AI prompting because it bridges raw data and the Large Language Model (LLM) ability to process and generate meaningful results. With NLP techniques1, we can design prompts that are not only grammatically correct but also semantically rich, allowing for more nuanced and targeted communication with the AI.
This could be the most important point!! Treat your interactions as a conversation. It makes it easier for the writer with a tool that was built to take human chat cues. So, Be Clear, Be Specific, and even Ask for Help from ChatGPT on how to word a query.
Prompt Engineering vs. Prompt Crafting vs. Prompt Fine-Tuning
No matter how you slice it, your results and performance rely heavily on optimization techniques. These are important for producing more meaningful outcomes and enhancing efficiency.
These three optimization techniques often get referenced, starting from the more technical development level to the conversational Prompt Crafting that leads to Fine-Tuning.
Prompt Engineering – Prompt engineering is a problem-solving approach to working with AI. I see this as more of a developer-level role. ChatGPT makes Playground2, a valuable tool for developers to experiment with and understand the behavior of conversational AI. It is highly customizable, requiring some experience to grasp how each setting impacts the language model3.
Prompt Crafting: Over the rest of the article, we will cover Prompt Crafting, which is more for the non-programmer. I look at prompt crafting as Experimentation. Use prompting to get initial results and then make adjustments through practice and experimentation. This is an Iterative Process where you refine your prompts based on the output. Follow where the results take you and pull back where the tool strays from your goals.
Prompt Fine-Tuning – This is the process of aligning the prompts with the precise outcome you want. You take your results and then adjust your prompts until you get the desired results, giving direction to change the results to a more final version. You could also call this Prompt Optimization, which implies you’re working towards the best possible version of your prompts for a task.
This is why I like the term Prompt Crafting. It requires skills or expertise developed over time and can be seen as a craft or art accessible to everyone.
No matter what you call it, this Iterative Prompt Improvement is needed as success depends on the step-by-step nature of getting to your desired results.
ChatGPT Prompt – Feed it like a seed, and it will grow
The Key = Feed It and Seed IT – Feed copy; provide hints (Prompts & Parameters) to guide (Teach) the tool to succeed through your interactions and help you produce the effective content you want.
Remember, the effectiveness of a ChatGPT prompt lies in its clarity and context, so include any necessary background information or parameters to shape the AI’s responses according to your defined objectives.
Watch Out for Hallucinations!!
A ChatGPT hallucination refers to an instance where the language model generates text that is not based on accurate information or experiences but instead creates imaginative or fictional content. These hallucinations can occur when the model generates responses that are unexpected, fantastical, or disconnected from reality, often resulting in creative or surreal outputs that may not align with factual information or logical reasoning.
Reference: ChatGPT.
Also, see What are AI hallucinations—and how do you prevent them? 4
Let’s create a framework for getting more with AI Prompting. 🙂
I have boiled down what has worked for me into a comprehensive guide, the 5-Step AI Prompt Crafting Framework For Content Creation, which we will cover in the next section.
5-Step AI Prompt Crafting Framework For Content Creation.
This guide focuses on crafting effective prompts for ChatGPT within the context of technical writing. We’ll outline a flexible framework you can adapt to different projects and different subjects.
To illustrate, we’ll use ChatGPT to generate a workshop outline on Power BI basics. You’ll understand how to leverage ChatGPT to improve your technical writing workflow and produce accurate, high-quality content.
First, An MVP – Minimum Viable Prompt
At its simplest level, a prompt has three main parts;
Persona or Role – Give the tool an overall context
Task – An action or goal of the interaction
Guidance – Provide details about your request
Acting as a [Persona] Perform a [Task] using this [Guidance]
Example:
As an [Adult Education Technical Content Creator], please provide me with [a workshop outline] for the following [full-day workshop titled "Introduction to Power BI Dashboard Creation"].
Let’s expand this simple prompt so that you can use it to make your results more consistent and of better use for your goals.
Step 1 – Define the Persona of the AI (Identity)
Before we start telling the AI the task we want it to perform, we need to take the entire library of knowledge available to the tool and scope that down by providing a frame of reference. Remember the
What is a Persona?
In the context of ChatGPT prompting, a “persona” refers to a specific character, role, or identity that the AI model adopts during a conversation or interaction. This can be anything from emulating the tone and style of a historical figure or a fictional character to behaving like an expert in a particular field.
The persona can influence how responses are formulated, including the choice of language, the level of formality or informality, and the type of information emphasized.
For example, if a user asks ChatGPT to respond as a friendly science teacher, the model would adopt a persona that is knowledgeable about scientific concepts but explains them in an accessible, encouraging, and supportive manner. This might include simplifying complex topics, using metaphors or analogies, and encouraging curiosity and questions.
Adopting a persona can make interactions more engaging, personalized, or entertaining for users. It allows for a wide range of conversational experiences, from educational and informative to creative and whimsical. However, it’s important to note that the AI’s understanding and emulation of a persona are based on its training data and algorithms and do not have personal experiences or emotions.
Remembering Natural Language Processing (NLP), this is a conversation, and having a “Persona” gives us a good reference as to how we have set up the tool for this series of instructions. I would also like to promote reusability, so setting up a persona that matches how you work and wants your output provides us with that opportunity.
Defining The Persona
When defining the persona of an AI, it’s important to consider their background, expertise, communication style, and personal traits.
In the context of education theory and technical problem-based learning, the AI should possess a strong understanding of these concepts and be able to communicate in an approachable and easy-to-understand manner.
The AI’s background could include experience in technical writing and education theory, giving them the expertise to provide well-informed responses. Their communication style should be clear, concise, and bursting with creativity to captivate learners. Personal traits such as patience, empathy, and adaptability are essential for creating a positive learning experience.
Consistency and coherence are crucial as you work through different projects. The persona should remain consistent across various interactions, maintaining a reliable and trusted identity for learners. This helps to keep a common voice and tone from project to project.
Overall, the AI’s persona should reflect a knowledgeable, approachable, and adaptable character, capable of effectively communicating complex ideas in an accessible way within the context of education theory and problem-based learning.
Example1 – Define the Persona
The Persona is made up of several sections5. The key is to design the persona like you were hiring your ultimate writing assistant. What traits would you want in this person?
The following prompt segment provides an overall direction for our query when creating the Persona. Change the content between the brackets [ ] to craft your prompt.
This prompt should be run first to set the context of the conversation (Prompt Crafting).
Prompt: ChatGPT Please adopts the following Persona or role with all of our interactions.
Persona Name = Jennifer Thompson
Background = [Jennifer] is an experienced [technical writer] with a strong background in [education]. They are passionate about creating [informative] and [engaging] content. With a deep understanding of [education theory and practices], [She] brings a unique perspective to their technical writing projects.
Communication Voice = [Jennifer's] tone of voice used in the documentation is informative and instructional, maintaining a professional and clear approach. It is designed to be accessible for users of different skill levels, from beginners to more advanced users. The language is straightforward and focused on guiding users through various processes without overwhelming them with technical jargon.
Communication Style = With a focus on [clarity and simplicity], [Jennifer's] Engaging Writing helps readers grasp technical concepts while maintaining their interest throughout the content.
Step 2: Define the Project Scope, Goals, and Type of Content
The second step is to define the Prompt’s Scope by clarifying what you aim to achieve with your prompt. This involves outlining the topic, the depth of details you expect in responses, and any limitations you might encounter.
Setting clear goals for what you want to achieve will guide your prompt crafting process. It ensures that each prompt is purposefully designed to contribute towards these goals, whether it’s improving the user experience, providing educational value, or fostering creativity.
Establishing the project scope helps set clear expectations for both the prompt creator and the AI’s output, ensuring the prompts are tailored to elicit the desired type of response.
Next, it’s essential to define Clear Goals for what success looks like for your project. Are you aiming to enhance user engagement, generate content ideas, or perhaps support learning objectives?
Selecting the type of content you want to generate is critical. This decision should align with both the project scope and goals.
Example 2 – Define the Project Scope, Goals, and Type of Content
These steps will help ensure that your prompts effectively solicit the desired type of content from ChatGPT and align with your project’s overall objectives. They will also provide a structured approach to harnessing AI for creative and practical applications.
Prompt: ChatGPT Please adopts the following Persona or role with all of our interactions.
Please use the following information for the project. This project will be a series of prompts.
The project scope involves producing content that is [easily understandable and accessible with a 6th-grade reading level] for adult learners. The type of content to be created is [a workshop outline titled "A Brief Introduction to Power BI"]. The materials should be written with [simplicity and burstiness, using short sentences and paragraphs to make the content more digestible and less intimidating to readers.]
The project aims to provide accurate and relevant information on specific topics related to the [technical learning plans]. This includes ensuring that the content aligns with a [level-based content rubric and technical flow], incorporating keywords related to the project scope, goals, and type of content. The goal is [creating content that meets these criteria aims to make the educational materials more engaging and understandable for adult learners].
Step 3 – Define Your Target Profile, Your Audience
We need to consider the audience’s needs and preferences for effective content. This will provide more guidance for the tool to create results that will be engaging, valuable, and impactful for your learners.
When crafting prompts, defining a target profile and audience is essential for several reasons;
Knowing your audience allows you to tailor your content to their specific interests, needs, and level of understanding. This personalization makes the content more relevant and engaging.
Different audiences might require different tones and language styles. For instance, professional or technical audiences may prefer a formal tone and industry-specific jargon, while a general audience might require simpler language and a more casual tone.
Understanding the audience helps choose the most effective way to communicate complex ideas. It ensures the information is accessible and easily understood, regardless of the audience’s prior knowledge.
Different audiences have different goals. Defining the target profile helps ensure that the content you create aligns with these goals, whether they’re educational, informative, persuasive, or for entertainment.
The context in which your audience will interact with the content—such as the platform (blog, social media, newsletter) and the device (mobile, desktop)—can also influence how you define your audience and craft your content. This will also change how you write and create the content.
Incorporating these strategies into your prompt crafting process helps the tool generate relevant, engaging content that effectively meets the objectives of both the creator and the audience.
Example 3 – Define Your Target Profile, Your Audience
We want to add information that helps keep focus on the goals. This Guidance, or Boundaries, helps keep the responses in line but could be changed for different versions or contexts of the output we desire.
Prompt: The Target Audience and Learner Profile is as follows;
Beginners in Data Visualization: Individuals new to Power BI or data visualization tools, looking to understand the basics and gain proficiency.
Hands-on Learners: Those who learn by doing, as the content is structured to encourage following along with their own Power BI Desktop application.
Tone of Voice: The Tone of Voice used in this content is [informative and instructional, maintaining a professional and clear approach]. It is designed to be [accessible for users of different skill levels, from beginners to more advanced users]. The language is [straightforward and focused on guiding users through various processes without overwhelming them with technical jargon].
This section will allow you to have multiple prompts with the same tone of voice and a more consistent writing style. You will still need to work through the response, but it will at least help maintain consistency.
Step 4 – Define The Content Formatting and Structure (Constraints)
Another way to help maintain consistency in your results is clearly defining the desired content formatting and structure. These constraints act as a template or blueprint for AI’s responses. This works to improve the generated text’s clarity, organization, and overall quality.
By providing guidelines such as bullet points, numbered lists, headings, or emphasis on specific words, you guide the AI to produce output that aligns with your specific needs and preferences.
This provides a structured approach, which helps through several key benefits. It makes your content more engaging and easier for the reader to digest. Using bullet points and concise sentences breaks down complex information into manageable chunks.
Strategic use of bold or italics highlights important points, ensuring they aren’t lost in the text. Additionally, transitional phrases enhance the flow of the content, creating a seamless and logical reading experience. By incorporating these elements, in addition to perplexity and burstiness, you make your content more accessible and memorable.
2 Key Prompting Parameters – Perplexity and Burstiness
Two factors are key when crafting effective AI prompts: “perplexity” and “burstiness.”
Perplexity refers to how unpredictable or surprising the language is to the AI model.
Burstiness measures the variety in your prompt’s structure. Humans naturally write with high burstiness, using a mix of long, complex sentences and short, punchy ones. AI models, on the other hand, often favor predictable patterns. Aim for prompts with high levels of both perplexity and burstiness to generate truly creative and engaging responses.7
Example 4 – Define The Content Formatting and Structure (Constraints)
Using formatting categories and structure prompting hints makes reusing these in other projects easier. You can mix and match based on the desired output and make adjustments as needed.
The following part of the prompt segment provides an overall direction for our query. Change the content between the brackets [ ] to craft your prompt.
Prompt: The following grouping of formatting and structure are provided to guide your response where appropriate. These are provided as prompt guides, aiming for clarity and impact of the content:
1. Make your content relatable and engaging: - [Provide relevant examples to make your points clearer and connect with readers.]
2. Structure your information effectively: - [Use bullet points or numbered lists to organize complex ideas or instructions in an easy-to-digest format.] - Keep sentences short and paragraphs concise to improve focus and readability.
3. Highlight key information: - [Use bold or italics strategically to draw attention to the most important details.]
4. Prioritize smooth flow: - [Incorporate transitional words and phrases (like "additionally," "similarly," or "on the other hand") to guide the reader through your ideas.] 5. Paragraph Formatting: - The meterials should be written with [perplexity and burstiness, using short sentences and paragraphs to make the content more digestible and less intimidating to readers.]
Implementing formatting and structural guidelines will make the content accessible, easy to understand, and engaging for adult learners. Remember, the goal is to simplify the information without dumbing it down, making it more digestible and less intimidating.
Step 5 – Post Processing – Adding the Iterative Conversation
After generating a response from ChatGPT, the next step is post-processing, which involves refining the content through additional modifications and enhancements. This iterative conversation technique is essential for improving the quality of the chatbot’s responses. Making small adjustments and adding more details makes the conversation more interactive and engaging.
Interactive prompting and multi-step questions are key to encouraging the Language Model (LLM) within ChatGPT to provide the best possible answers. This means talking to the AI like you would talk to a person, including conversational details and asking complex questions to dive deeper into the conversation. This improves the overall quality of the conversation and helps the LLM understand the context better, resulting in more accurate and relevant responses.
In the post-processing phase, it’s important to focus on refining the conversation to make it feel more natural and coherent. This involves adding contextually relevant information, fact-checking, and adjusting the tone to match the desired communication style.
By incorporating iterative conversation through post-processing and utilizing interactive prompting and multi-step questions, the quality of the conversations with ChatGPT can be significantly improved, leading to more meaningful and engaging interactions.
Example 5 – Post Processing – Adding the Iterative Conversation
We can get on with the content now that we have built the framework for our prompts.
Prompt:
I have a course in the subject "Learning Power BI" ,The Course is titles "An Introduction to Power BI". I need an outline or syllabus for one of the modules called 'Designing a Dashboard'
I need an outline that includes 4 lessons and an assignment that takes into account Bloom's revised taxonomy.
I need the course, module and lesson Actionable Outcomes that also include some of the terms in the Outcome Action Verbs, I will include these below.
For a quick reference on Actionable Outcomes, please refer to my post for reference = https://aitechedinsights.com/outcome-based-lesson-planning/
This list has the Level and then the Action Verbs
Remembering: Recall, Recognize, List, Identify, Describe, Retrieve, Name, Define, Repeat, State
Understanding: Explain, Summarize, Paraphrase, Interpret, Classify, Infer, Compare, Contrast, Discuss, Demonstrate
Applying:Apply, Use, Solve, Execute, Implement, Demonstrate, Illustrate, Employ, Operate, Interpret
Analyzing: Analyze, Examine, Compare, Contrast, Categorize, Differentiate, Investigate, Break down, Deconstruct, Distinguish
Evaluating: Evaluate, Judge, Assess, Appraise, Critique, Estimate, Compare, Prioritize, Select, Validate
The result is below.
Result:
Certainly, below is a course outline for the module titled "Designing a Dashboard" in "An Introduction to Power BI" course. This outline is structured to cater to a range of cognitive skills according to Bloom's revised taxonomy, ensuring that adult learners are engaged and can apply what they learn in practical settings.
Course Title: Learning Power BI
Module: Designing a Dashboard
Module Description: This module will guide beginners through the essentials of designing an effective and functional dashboard in Power BI. Learners will understand the core components of dashboard design, from data selection to visual aesthetics and practical utility.
Lesson 1: Fundamentals of Dashboard Design
Actionable Outcome:
- Remembering: Recall and list the primary functions of a Power BI dashboard.
- Understanding: Summarize the importance of dashboard design in data visualization.
Lesson 2: Selecting and Importing Data
Actionable Outcome:
- Understanding: Classify and compare different data sources for relevance and quality.
- Applying: Apply techniques to import and prepare data for dashboard creation.
Lesson 3: Creating Visuals and Layouts
Actionable Outcome:
- Applying: Use Power BI tools to create basic data visualizations.
- Analyzing: Examine and contrast effective layouts for dashboards.
Lesson 4: Interactivity and Customization Features
Actionable Outcome:
- Analyzing: Differentiate between various interactivity features in Power BI.
- Evaluating: Critique and appraise the effectiveness of interactive elements in a sample dashboard.
Assignment: Design Your Own Power BI Dashboard
Actionable Outcome:
- Creating: Design and develop a functional dashboard using Power BI, employing a variety of visual elements and interactivity features:
Assignment Details:
Learners will design a dashboard that presents a cohesive narrative from a given dataset. The dashboard should be user-friendly, aesthetically pleasing, and must enable viewers to make informed decisions based on the data presented.
Final Assignment Evaluation Criteria:
- Clarity of data presentation.
- Effective use of visuals.
- Level of interactivity and ease of navigation.
- Dashboard's ability to enable decision-making.
Resources:
For detailed guidance on creating actionable outcomes, learners can refer to the post on Outcome-Based Lesson Planning.
This module will ensure that learners not only understand the theoretical aspects of Power BI dashboard design but also apply, analyze, and evaluate these principles practically, culminating in the creation of their own dashboard.
Let’s compare this with the original prompt from above at the beginning of the article.
Original ChatGPT Prompt Result
Improvements
Provided a better-aligned teaching flow of the lessons.
Bloom’s Taxonomy outlines a sequence of learning objectives that move from simpler to more complex skills, fostering deeper understanding and mastery.
The new outline has a better content flow for the goal of How to Design a Dashboard.
Learning outcomes and content are better aligned.
Bonus Topic: Tips and Tricks
1. Use a Conversational Style – Talk to the AI like you would talk to a person
While AI models process language differently than humans, using natural, conversational prompts can improve the flow and quality of responses. Here’s how:
Ask open-ended questions: These encourage more expansive answers. For example, Instead of “Sales by region chart,” try “Can you show me sales broken down by region in a bar chart?”
Provide a persona: If appropriate, suggest a role or personality for the AI (e.g., “Act like a data analyst and suggest visualizations for this sales dataset.”). This can add focus to the interaction.
2. Provide Context for Clarity
The more specific your prompt, the more tailored and useful the AI’s response will be. Consider these tips:
Include background information: Briefly summarize your knowledge level or purpose (e.g., “I’m analyzing website traffic for the past year. Can you show me which countries generate the most visits?”)
State preferences: Note any restrictions, style choices, or what you want to emphasize (e.g.” I prefer providing simple examples with bar charts.”)
Share relevant facts: Indicate any crucial details that might affect the outcome (e.g., “I’m interested in customer retention. Focus on repeat purchases over the last 6 months.”)
3. Experiment with Formatting
Certain formatting choices can influence how the AI interprets your prompts:
Capitalization: Emphasize key terms or differentiate between names (“Emphasize HIGH-VALUE customers in the visualization.” vs. “Write a graph showing customers.”).
Punctuation: Use exclamation points for excitement or question marks to clearly signal a query.
Line breaks and spacing: Create visual separation between ideas or instructions for improved clarity.
4. Utilize Keywords and Specificity
The more precise your language, the better equipped the AI will be to deliver the desired outcome.
Employ strong verbs: Instead of “summarize data,” try “calculate average sales price” or “identify top-performing products.”
Be specific with nouns: Replace “product” with “Widget X” or a specific product ID.
Include quantifiers: Provide numbers or descriptors where applicable (“a short description,” “three solutions,” etc.).
5. Iterate and Refine
Rarely will the first prompt yield the perfect result. Treat your interaction with the AI as an ongoing conversation:
Rephrase: If a response seems off-target, try rewording your original prompt.
Provide feedback: If a visualization isn’t ideal, try “Change that to a pie chart” or “Drill down to the city level.”
Build upon prior outputs: Use the AI’s previous responses as a jumping-off point for further exploration.
7. Let’s Leverage this for Reusability !!
These tools understand human conversation, so we should take advantage of that when we create prompts. This makes it easier for these tools to discover your prompt’s intent and to reuse these prompts as “conversation” starters.
Reuse and Organize: your prompts into libraries
Resource Sharing: Share them in your organization.
6. Stay updated
Staying updated in the world of technical writing is essential for producing high-quality and relevant content. By staying informed about recent industry happenings, you can ensure that your documentation is accurate and valuable to your target audience.
One way to stay updated is by following technical news sources and attending industry events. These resources provide valuable insights into emerging trends, best practices, and new tools that can enhance your content creation process.
Coming Soon – Newsletter
Conclusion
By now, you understand the importance of prompt crafting for technical writing and how to leverage ChatGPT as a valuable resource. Remember, this is an iterative process. Start with simple prompts, analyze the results, iterate, and fine-tune them for optimal outcomes.
AI tools have the potential to revolutionize technical writing, and with the right prompt crafting techniques, you can harness this power to produce informative, accurate, and engaging content. As you use ChatGPT, pay attention to its responses—they will guide you toward even better prompts that align with your writing goals. If you need further assistance, numerous online resources and communities dedicated to prompt crafting can help!
Footnotes
Natural Language Processing (NLP) [A Complete Guide] (deeplearning.ai) ↩︎
ChatGPT Playground: What It Is and How to Use It – GeeksforGeeks ↩︎
What are AI hallucinations—and how do you prevent them? | Zapier ↩︎
The Persona style I use is highly Influenced by the following content creators. Discord conversations on the Learn-Prompting Discord Channel, Open AI, Forward Future AI, ChatGPT Prompt Engineering+ research on Open AI, and Microsoft’s Open AI Documentation ↩︎
Reference: For more information on PBL, see Problem-Based Learning | Center for Teaching Innovation (cornell.edu)
