Using Large Language Models: How to Write Better Prompts

Different types of prompts are key to getting the most out of large language models, as they shape everything from tone to level of detail in the responses. Think of prompts as both a question and a set of directions—how you word them influences how the model “understands” what you need. Changing up the way you ask questions can help the model give you just what you’re looking for—whether that’s a quick overview, a detailed analysis, or a burst of creative ideas. Open-ended prompts, for example, encourage broader, more imaginative answers, which are really helpful when you want to brainstorm or dive into new ideas. If you ask the model something like, “Describe a futuristic city,” it’s going to give you a big-picture, creative answer. But if you say, “Describe a futuristic city with flying cars and eco-friendly skyscrapers,” you’re guiding it to focus on specific features, narrowing the output while still allowing for creativity.

This approach also applies in more technical or professional settings. When you need structured information, using a clear, specific prompt can help. For example, asking, “List the principles of machine learning in bullet points,” will lead the model to organize the information in a straightforward, easy-to-read format. Asking, “Explain the principles of machine learning” invites the model to go into more detail, adding context and depth for those who want to understand more. The tone of the response also changes with the way you word your prompt—casual prompts usually get friendly, conversational replies, while more formal prompts lead to responses that feel polished and professional. By experimenting with different prompt types, you can tailor the model’s responses to fit whatever tone or depth you need for a project.

What is the Question Refinement Prompt Pattern ?

The question refinement pattern helps clarify questions by breaking them into specific, targeted inquiries, leading to more focused and informative responses. This is accomplished by asking the large language model to suggest a better version of the question you ask each time you ask a question.

Example of a Prompt Using the Question Refinement Pattern

Original Question: What should I do to improve my health?

Refined Prompt: From now on, whenever, I ask a question, suggest a better version of the question to use instead and ask me if I would like to use it instead. – What Should I do to improve my health?

What is the Cognitive Verifier Prompt Pattern?

The cognitive verifier prompt pattern is a structured approach used in prompt engineering, particularly for language models like ChatGPT. It enhances the model’s ability to produce accurate and comprehensive responses by encouraging it to break down complex questions into simpler, manageable parts. Here’s how it works:

Key Rules for the Cognitive Verifier Prompt Pattern:

1. Clarify the Question: When asked a question, begin by clarifying what is being asked to ensure understanding.
2. Generate Sub-Questions: Create a series of additional questions that delve deeper into the topic. This helps to gather all necessary information for a thorough answer.
3. Answer Sub-Questions: Respond to each sub-question individually, providing concise and informative answers.
4. Synthesize Final Answer: Combine the answers from the sub-questions to formulate a comprehensive final response to the original question.

Example of a Prompt Using the Cognitive Verifier Prompt Pattern

Original Question: Can you explain the importance of emotional intelligence in leadership?

Refined Prompt: Clarification: To provide a thorough answer, I need to understand the specific aspects of emotional intelligence and leadership you’re interested in. Are you looking for definitions, examples, or the impact on team performance? Generate Sub-Questions: What is emotional intelligence, and what are its key components? How does emotional intelligence differ from traditional intelligence (IQ)? In what ways can emotional intelligence influence leadership styles? Can you provide examples of leaders who effectively use emotional intelligence? What are the benefits of emotional intelligence for team dynamics and performance? Answer Sub-Questions: Emotional Intelligence Definition: Emotional intelligence (EI) refers to the ability to recognize, understand, manage, and influence emotions in oneself and others. Key components include self-awareness, self-regulation, motivation, empathy, and social skills. Difference from IQ: While IQ measures cognitive abilities and intelligence, EI focuses on emotional and social skills that are crucial for effective interpersonal interactions. Influence on Leadership Styles: Leaders with high EI can adapt their leadership style to fit the needs of their team, fostering a more inclusive and supportive environment. Examples of Effective Leaders: Notable leaders like Oprah Winfrey and Satya Nadella exemplify high emotional intelligence by showing empathy, building strong relationships, and creating a positive workplace culture. Benefits for Teams: Teams led by emotionally intelligent leaders tend to have better communication, increased morale, and improved performance, as these leaders can address conflicts and motivate their team members effectively. Synthesize Final Answer: Emotional intelligence is crucial in leadership as it encompasses the ability to manage one’s emotions and understand others’ feelings, which can lead to effective communication, stronger relationships, and enhanced team performance. Leaders like Oprah Winfrey and Satya Nadella illustrate how high emotional intelligence can transform team dynamics and create a positive work environment, ultimately benefiting the organization as a whole. By following this cognitive verifier prompt pattern, the response not only becomes more structured and clear but also provides a richer understanding of the topic.

What is the Audience Persona Prompt Pattern?

The audience persona prompt pattern is a structured approach to crafting prompts that specify the characteristics and preferences of the target audience for a piece of content or communication. This pattern helps ensure that the generated responses align with the audience’s needs, interests, and level of understanding. By defining the audience persona, you can guide the AI to create more tailored and effective content.

Explanation of the Audience Persona:

• Define who the audience is (e.g., age, profession, interests).
• Include their goals, challenges, and preferences.
• Specify the tone and style of communication that resonates with them

Example of Audience Persona Prompt Pattern

Original Question: Can you explain the importance of emotional intelligence in marketing?

Refined Prompt: Imagine you are a marketing consultant creating a blog post for small business owners aged 30-45 who are looking to improve their digital marketing strategies. They are often overwhelmed by technology and prefer straightforward, actionable advice. The tone should be friendly and encouraging, with practical examples to illustrate key points. What are five effective digital marketing strategies that you would recommend for this audience?

The Flipped Interaction Prompt Pattern

The flipped interaction prompt pattern is a structured approach to using conversational AI, where the user explicitly guides the AI’s questioning process. Instead of the AI providing information or responses immediately, the user requests that the AI ask questions to gather necessary details to achieve a specific goal (X). This method promotes a more interactive and tailored conversation, allowing the user to clarify their needs and the context effectively. The user also specifies when to stop asking questions (Y), ensuring the interaction remains focused and productive.

Example of The Flipped Interaction Prompt Pattern

Prompt: I would like you to ask me questions to help me develop a marketing strategy for my new product. You should ask questions until you have sufficient information about my target audience and marketing goals. Ask me the first question.” In this prompt: X: developing a marketing strategy for a new product. Y: until you have sufficient information about the target audience and marketing goals.

The Game Play Prompt Pattern

The game play prompt pattern is a structured way to request a game or interactive activity, typically involving a set of rules or guidelines. This pattern is useful for engaging users in a creative or problem-solving context, allowing them to explore a concept, theme, or scenario through game mechanics.

Example of The Game Play Prompt Pattern

Prompt: You should create a game for me around the theme of environmental conservation. These will be the fundamental rules of the game: Players take turns choosing actions to protect a specific ecosystem (e.g., forest, ocean, desert). Each action has a point value based on its impact (positive or negative) on the ecosystem. Players can collaborate to combine actions for greater effects. The game ends when a player reaches a predetermined point total, or after a set number of rounds, at which point the player with the most points wins.

The Ask for Input Prompt Pattern

The Ask for Input prompt pattern involves requesting specific information or context from the user to tailor the response accordingly. This pattern enhances interaction by ensuring that the AI can provide more relevant and personalized outputs based on the user’s needs.

Example of The Ask for Input Prompt Pattern

Prompt: From now on, I’m going to provide you with topics related to emotional intelligence in team performance. You will summarize the key points and suggest actionable strategies based on the information I provide. Ask me for the first topic. This way, the AI knows to wait for the user’s input (the first topic) before proceeding with the task.

The Outline Expansion Prompt Pattern

The Outline Expansion prompt pattern is a structured approach for generating content that allows users to start with a broad outline and then drill down into specific areas for deeper exploration. This method encourages a clear organization of ideas and facilitates focused discussions on selected topics.

How It Works:

1. Outline Generation: The initial prompt requests a bullet point outline based on the user’s input. This provides a high-level view of the main topics or ideas.
2. Selection for Expansion: After the outline is generated, the user selects a specific bullet point they want to explore further.
3. Detail Expansion: Once the user selects a bullet point, a new outline or detailed content is generated specifically for that point, allowing for in-depth analysis or discussion.

Example of The Outline Generation  Pattern

Prompt: Act as an outline expander. Generate a bullet point outline based on the topic of ‘Benefits of Emotional Intelligence in the Workplace.’ After creating the outline, ask me which bullet point I should expand on