What Is a Prompt Framework? Top 7 Prompt Frameworks That Will Help You Heighten the Accuracy of ChatGPT’s Answers

When we attempt to answer the common question “What is a prompt framework?“, we always turn to the old adage: Why reinvent the wheel when we already have effective tools and structures that guide us toward success?

A prompt framework is a set of guidelines or methods used to craft effective prompts for AI models like ChatGPT—i.e. the figurative ‘wheel’ that’s already proven and tested to heighten the accuracy of ChatGPT’s answers. These frameworks help users communicate more precisely with AI, to ensure the responses generated are accurate, relevant, and aligned with the user’s needs.

And what are the benefits of applying a framework when you prompt?
•  Improves Quality: Ensures responses are clear, relevant, and precise.
•  Increases Efficiency: Streamlines the prompt creation process, saving time.
•  Versatility: It is likely suitable for a wide range of applications, from content creation to technical support and beyond.

This article discusses the key elements of an effective prompt and 7 popular prompt frameworks you can apply—and which among them is our absolute favorite.

Key Elements in Prompt Frameworks

You would have heard many times that “clarity and specificity are key in prompt engineering“, but how exactly do we write clear and specific prompts?

An effective prompt includes the following elements. Later in this article, you will see how prompt frameworks encapsulate them:

1. Role Definition: Assigning a specific role to the AI to shape its response style.
2. Instructions: Providing clear directives on what the AI needs to do.
3. Context or Background: Setting up the scenario or providing relevant information.
4. Objective or Purpose: Clearly stating the goal of the prompt.
5. Expected Response Format: Specifying how the response should be structured (e.g., bullet points, paragraphs).
6. Tone and Style: Indicating the desired tone (formal, casual) and style.
7. Examples or Specifics: Providing examples to guide the response.
8. Clarifications or Boundaries: Setting rules to keep the response on track.
9. Reasoning and Justification: Requesting logical explanations or supporting evidence.
10. Desired Outcome or Conclusion: Indicating the end goal of the interaction.

Top 7 Prompt Frameworks

These are the top 7 prompt frameworks based on our research:

1. RICE (Role, Instructions, Context/Constraints, Examples)

• Role: Defines the perspective or persona ChatGPT should take (e.g., teacher, analyst). This will guide the style and tone of the response.
• Instructions: Provides specific directions on what ChatGPT needs to do, ensuring clarity in the task.
• Context: Sets the scenario or background information to help the AI understand the relevance of the task.
• Constraints: Establishes boundaries like word limits, tone, or style requirements to keep the response focused and appropriate.
• Examples: Includes samples of the expected outputs to guide the AI’s response style and substance.

See the following example prompt:

(All the examples below will be labeled with [XXX] to clearly mark the elements of the prompt framework involved. These labels are for illustration purposes only and should be excluded from the actual prompts sent to ChatGPT.)

[Role] Imagine you are a travel blogger who has been traveling full-time for 5 years. [Instruction] Provide a detailed guide on the top 5 destinations in Europe, explaining why you recommend them. [Example] For example, highlight budget-friendly locations, hidden gems, or places famous for specific attractions. [Constraint] Keep the guide under 500 words. 

2. FOCUS (Format, Objective, Context, User Needs, and Specifics)

• Format: Specifies the desired response format, e.g., list, paragraph, table.
• Objective: Clearly defines the main goal or purpose of the prompt. Tells ChatGPT what to achieve.
• Context: Provides background information that sets the stage for the prompt to ensure relevance.
• User Needs: Outlines the specific needs of the users. In this context, “users” refer to the target audience or individuals who will benefit from ChatGPT’s responses as a result of this prompt.
• Specifics: Details any other particular instructions or constraints.

Example prompt:

[Format] Provide a bullet-point list [Objective] explaining the top 5 cybersecurity threats businesses face in 2024. [Context] This list will be part of a blog post discussing cybersecurity threats, targeting business owners who may not have technical knowledge. [User Needs] Highlight practical steps companies can take to mitigate these risks in layman's terms. [Specifics] The response should be concise and tailored for non-technical business owners.

3. CORE (Context, Objective, Requirements, Examples)

• Context: Includes relevant background information to help ChatGPT understand the scenario.
• Objective: Define the specific goal of the prompt.
• Requirements: Outline any specific requirements or constraints, such as the preferred format, length, or tone of the response.
• Examples: Provide examples of the desired output or response style.

Example prompt:

[Context] I’ve just adopted a one-year-old cat who is very shy and hides most of the time. There are no other pets in the house, and I’ve tried giving her space, but she still seems scared. [Objective] I need advice on how to help my new cat feel comfortable and adjust to her new home. [Requirements] The response should be concise, about 200 words, and presented in a step-by-step format. Include practical tips, [Examples] like creating safe spaces or gradually introducing the cat to new rooms.

4. PAS (Problem, Action, Solution)

• Problem: Identifies the issue or question that needs addressing.
• Action: Describes the steps or actions required to address the problem. Include the instructions to guide the AI toward a practical response.
• Solution: Specifies the expected outcome or resolution to the problem and guides ChatGPT on what its response should focus on.

Example prompt:

[Problem] A customer’s computer is overheating. [Action] List the possible causes and [Solution] outline the immediate actions they should take.

5. IDEA (Instruction, Details, Examples, Action)

• Instruction: Directs ChatGPT on the task to perform. Make sure you include clear and specific guidance.
• Details: Adds specific information or background to help ChatGPT understand the task.
• Examples: Provides examples to guide the style, depth, or tone of the response.
• Action: Specifies what action or type of response the AI should produce.

Example prompt:

[Instruction] Provide a step-by-step guide on how to set up a home for a new kitten.
[Details] Include necessary items like food, bedding, and toys, and discuss how to create a safe space for the kitten.
[Examples] For example, describe the best types of litter boxes and kitten-proofing techniques to avoid hazards.
[Action] End with tips on how to help the kitten adjust to its new environment.

6. SOCRATIC (Situation, Objective, Clarifications, Response, Action, Tone, Insights, Conclusion)

• Situation: Sets up the scenario or context that ChatGPT needs to address.
• Objective: Defines the goal or purpose of the interaction. This will guide ChatGPT toward a specific outcome.
• Clarifications: Requests ChatGPT to ask clarifying questions if necessary (we love this part!)
• Response: Provides ChatGPT with instructions on how to formulate its response by specifying the kind of analysis it should perform, the approach to take, or the aspects to focus on, etc.
• Action: Outlines the kind of actionable steps or recommendations that ChatGPT should provide.
• Tone: Specifies the desired tone, such as formal, casual, or instructional.
• Insights: Encourages ChatGPT to include key insights or analysis relevant to the task.
• Conclusion: Directs ChatGPT on how to end the response. For example, summarizing key points or providing a final recommendation.

Example prompt:

[Situation] My friend's cat has recently stopped using the litter box and is instead urinating around the house. 
[Objective] Prepare an e-mail that I can send to her to discuss why this behavior might be happening and suggest potential solutions. 
[Clarifications] If necessary, ask me to provide more information, e.g., recent changes in the cat’s environment, diet, or health that might contribute to this behavior. 
[Response] Provide an analysis of the possible reasons, including behavioral, medical, and environmental factors. 
[Action] Offer actionable steps my friend can take to encourage the cat to use the litter box again. 
[Tone] Use a compassionate and understanding tone, as this issue can be frustrating for my friend. 
[Insights] Include tips on monitoring the cat’s behavior and when to consult a veterinarian. 
[Conclusion] Summarize the key points at the end of the e-mail.

7. QUEST (Question, Understand, Example, Style)

• Question: Clearly states the inquiry or problem you want ChatGPT to address.
• Understand: Provide context or background information that ChatGPT needs to understand the question better.
• Examples: Offers examples of your desired output.
• Style: Specifies the desired tone or format for the response, such as formal, informal, bullet points, or essay style.

Example prompt:

[Question] Why does my new cat keep scratching the furniture even though I have provided a scratching post? 
[Understand] The cat is a one-year-old who has just joined our family, and we already have three other cats. 
[Examples] The reasons we're looking for could include, for example, the cat might prefer scratching the arm of a sofa due to its texture or positioning, or it might ignore the post if it’s placed in an inconvenient location. 
[Style] List your speculations and explain why you think so. Keep your explanations free from technical jargon.

Which Prompt Framework Do We Recommend?

We could easily recommend any of the above prompt frameworks, as they all contain the essential components needed to create an effective prompt. However, the weightage given to each component varies between frameworks, making each of them uniquely suited to different scenarios and user needs. This is what you need to consider when selecting a framework suitable for you.

Having said so, it’s beneficial to master one or two frameworks that you can rely on consistently, rather than frequently switching between different ones. Becoming familiar with a framework means that crafting effective prompts based on it becomes a breeze. This will save you time and guarantee the quality of your interactions with ChatGPT.

That’s why we recommend the RICE framework. RICE stands out as a general-purpose prompt framework that outshines others because of its:

• Adaptability
This framework is highly customizable and can be easily adapted to various needs, from simple research tasks to complex problem-solving. It is less restrictive compared to some other prompt frameworks.

• Structured Approach
RICE combines clear instructions with contextual guidance. The users are explicitly encouraged to specify the context, state the constraints, and provide references to minimize ambiguity. All these are what “clear and specific instructions”, as mentioned at the beginning of this article, is about.

• Ease of Use
RICE is easy to use for both beginners and advanced users. The steps are straightforward, and the framework doesn’t overwhelm users by requiring excessive information. The four core components are the essential minimum needed for an effective prompt.

In essence, the RICE framework simplifies prompt creation while optimizing the results of interactions with ChatGPT, in addition to being a framework that applies to almost all use cases.

If you want to pick up just one prompt framework and be good at it, we highly recommend you consider this framework.

Conclusion

In this article, we introduced 7 powerful prompt frameworks, each of which provides a unique approach to crafting effective prompts for ChatGPT, and offers different strengths suited to different scenarios and user needs.

We encourage you to experiment with these frameworks to discover which one aligns best with your specific tasks and style. However, if you’re looking to master just one framework, we recommend considering the RICE framework due to its versatility, structured approach, and ease of use.