Prompting Frameworks

Prompting frameworks are invaluable tools for structuring interactions with AI, guiding the model to produce relevant, effective, and often nuanced responses. Each framework offers a unique approach to structuring prompts, catering to a range of needs, from straightforward requests to iterative content enhancement. However, each also has distinct limitations and ideal use cases that are worth understanding to optimize AI-human collaboration.

The RTF (Role-Task-Format) framework organizes prompts by specifying the role the AI should take, the task it should complete, and the format for the response. This is a general-purpose framework, useful across various domains due to its simplicity and adaptability.

Strengths

Its simplicity is a significant advantage, making it suitable for users of all levels. The RTF framework reduces ambiguity by assigning a clear role, task, and format, which helps the AI generate responses in the desired tone and structure. This framework can be applied to both work-related tasks, such as generating summaries or proposals, and personal ones, like storytelling or casual explanations.

Limitations

While the RTF framework is versatile, it may lack the depth needed for complex, multi-step tasks. Because it is a “jack-of-all-trades,” it can sometimes produce surface-level responses where a more detailed framework might yield richer, more structured outputs.

Use Cases

Ideal for straightforward tasks such as generating descriptions, drafting summaries, or answering simple questions. For example: “Act like a historian. Give me a summary of the Renaissance period in paragraph format.”

Prompt Template: “Act like a [insert the role you want AI to take]. Give me a [insert task] in [insert format] format.”

The Chain of Thought framework enhances reasoning by prompting the AI to think through a problem in steps, making it particularly effective for analytical and problem-solving tasks. By guiding the AI to work sequentially, this approach fosters more logical, reasoned responses.

Strengths

This framework is powerful for improving reasoning, as it encourages the AI to tackle tasks step-by-step. This makes it particularly effective for tasks requiring logical progression, such as troubleshooting, analysis, or mathematical reasoning.

Limitations

While valuable for reasoning, Chain of Thought may not be suitable for quick, direct responses or for tasks where a step-by-step approach is unnecessary or over-complicates the prompt.

Use Cases

Useful for complex analysis, decision-making support, or problem-solving where multiple considerations must be accounted for. For example: “Explain the impact of climate change on marine ecosystems. Let’s think through it step-by-step.”

Prompt Template: “[insert your prompt instructions]. Let’s think through it step-by-step.”

The RISEN framework, which organizes prompts into Role, Instructions, Steps, End Goal, and Narrowing, provides a structured, goal-oriented approach. This framework is especially effective for larger, multi-phase tasks like writing projects, research, or complex planning tasks.

Strengths

RISEN is highly effective for in-depth, multi-step tasks. Its structure enables the AI to understand both the process and the final goal, enhancing focus and clarity. It’s especially useful for tasks that require a logical sequence and where specific constraints need to be respected.

Limitations

This framework can be overly detailed for simpler tasks and may require additional time to set up. Users may find it cumbersome if they are seeking a more spontaneous or flexible response.

Use Cases

Ideal for structured projects like research outlines, instructional content, and project management. For instance: “Role: Research assistant. Main Task: Create an outline for a blog on renewable energy. Steps: 1) Introduction, 2) Key types, etc. Goal: Provide an informative structure. Constraints: Focus on latest technologies.”

Prompt Template:

Role: [Insert the role you want AI to take.]

Main Task: [Insert the task you want AI to complete.]

Steps to complete task: [Insert numbered list of steps to follow.]

Goal: [Insert goal of the output]

Constraints: [Enter constraints].

RODES (Role, Objective, Details, Examples, Sense Check) is a framework for tasks where the user has specific output examples that can guide the AI. The Sense Check component allows the AI to confirm its understanding, making RODES ideal for high-stakes or quality-sensitive tasks.

Strengths

RODES is valuable for precision-based tasks. By offering examples, users can anchor the AI’s response, which is particularly useful when specific quality standards or formats are required. The Sense Check helps confirm alignment, reducing misinterpretations.

Limitations

This framework can be time-intensive, as it requires gathering examples and constructing detailed guidance. Additionally, it may be less adaptable for exploratory or creative prompts where fixed examples are less relevant.

Use Cases

Best suited for tasks that benefit from specific examples, such as producing content with a consistent style, tone, or format. Example: “Role: Technical writer. Objective: Draft a user manual introduction. Details: Focus on accessibility. Examples: Sample introductions provided below. Sense Check: Can you summarize your understanding?”

Prompt Template:

Role: [Insert desired role you want AI to take.]

Objective: [Insert objective you want AI to achieve.]

Details: [Insert any context or constraints for the AI.]

Examples: [Insert examples of good outputs].

Sense Check: Do you understand the objective and specific guidelines for this task?

The Chain of Density framework is designed for iterative refinement, guiding the AI to enhance content through successive improvements. By adding missing details in each iteration, this framework is highly effective for optimizing content quality, such as enhancing depth or comprehensiveness.

Strengths

This method encourages richer, more refined outputs by instructing the AI to evaluate and improve its response iteratively. It’s beneficial for tasks that require careful expansion, such as content editing, adding depth to arguments, or improving initial drafts.

Limitations

It can be time-consuming and may yield diminishing returns if overused. This framework is less practical for simple queries or tasks that do not benefit from iterative refinement.

Use Cases

Ideal for content that requires iterative quality improvements, such as enhancing arguments in an essay or refining product descriptions. Example: “Instructions: Improve the depth of this article on sustainable fashion. Recursion: Repeat steps to identify missing details and rewrite.”

Prompt Template:

Instructions: Here is [insert content you want to improve]. You will generate increasingly better versions of this content.

Recursion: Repeat the following steps 5 times for a total of 5 iterations:

Step 1. Identify 1-3 points missing from the initial output.

Step 2. Write a new, improved output of identical length that includes the missing points.

Benchmark: Here is more information on what makes a good XYZ: [insert info].

Additional guidelines: Follow these specific guidelines [insert guidelines].