Advanced Prompt Engineering Technique 2: Generated Knowledge Prompting

So you’re ready to pick up another advanced prompt engineering technique? In this article, we’ll introduce to you “Generated Knowledge Prompting“, an advanced technique particularly effective for complex tasks. It’s a two-step process of generating and integrating knowledge to help the AI move beyond shallow, incomplete, or narrowly focused answers.

This is the second part of our two-part series on advanced prompt engineering techniques. If you haven’t already, check out Part 1, where we cover Self-Ask Prompting—another powerful approach for complex tasks, that works by breaking down multifaceted problems into manageable sub-questions.

Let’s jump right in.

What is Generated Knowledge Prompting? How Is It Different?

Traditional prompting often leads the AI to respond directly without prepping it (or rather, guiding it to prep itself) with relevant context or background information.

Without prepping, the AI lacks the necessary context, doesn’t know how deep or broad it needs to go, and might overlook important aspects. It tends to respond based on the immediate query, without being given a chance to “think deeper”. As a result, it often draws from the most obvious and surface-level information, leading to shallow or narrowly focused answers.

Generated Knowledge Prompting, on the other hand, prompts the AI to gather relevant insights first before allowing it to fully develop its final responses. The aim is for the AI to deliver deeper and more informed responses.

This approach is a two-step process:

1. Knowledge Generation: The AI is first prompted to generate relevant background information or facts related to the topic at hand. This step activates the AI’s embedded knowledge, allowing it to surface insights that may not be immediately obvious.

2. Knowledge Integration: The generated knowledge is used to formulate a more comprehensive and well-rounded response to the main task.

See how the extra step enables the AI not only to develop but also to utilize its contextual understanding to address complex questions more effectively?

Why Use Generated Knowledge Prompting?

Generated Knowledge Prompting offers several key benefits over traditional prompting methods:

• Improved Accuracy: By prompting the AI to generate relevant background knowledge first, the model is better equipped to provide precise and contextually appropriate answers. This extra step ensures that the AI considers a broader range of information before formulating its response.

• Deepened Contextual Background: This approach ensures that the AI incorporates necessary background information, which will help it tackle complex questions with greater depth. By actively generating relevant context, the AI can better understand the complexities of the task.

• Versatility of Use: The method is adaptable to various tasks, whether it’s problem-solving, content creation, or answering complex questions. The two-step process can be tailored to fit different types of queries, and is a versatile addition to the toolkit of any AI users.

Examples of Common Applications

To provide perspective, here are some examples of where Generated Knowledge Prompting is applied:

• Complex Problem Solving: When addressing multifaceted issues, like analyzing market trends or evaluating scientific research, Generated Knowledge Prompting drives the AI to prep itself with necessary background insights before drawing conclusions. This leads to more comprehensive and accurate outputs.

• Content Creation: For tasks such as writing articles, reports, or creative content, this approach not only ensures the AI gathers sufficient background knowledge, but also allows the user to check the direction and details the AI has, before going the full nine yards to generate the final creative outputs.

• Complex Inquiries: In fields where precise context and facts matter—such as legal, historical, or scientific inquiries—Generated Knowledge Prompting helps the AI provide answers that are not just surface-level but deeply informed.

How Does Generated Knowledge Prompting Work?

Generated Knowledge Prompting operates through a two-step process:

Step 1: Knowledge Generation

The AI is prompted to first generate a set of facts, insights, or relevant information related to the task. Note that the user can (and should) break this process down into more than one prompt to achieve this.

For example, if the task is to write about the significant achievements of Abraham Lincoln, the user can request the AI to:

• Generate key facts about Abraham Lincoln.
• Specifically discuss his accomplishments during the Civil War.
• Generally list 3-5 of his key achievements that benefited the U.S.

The purpose is to help the AI gather enough knowledge that will be used in the next step.

Step 2: Knowledge Integration

In this step, the AI uses the generated knowledge to construct a comprehensive response to the main task. Continuing with the previous example, the user could prompt the AI this way:

Using the knowledge above, write an article of around 1,000 words to answer the question: "What were some significant achievements of Abraham Lincoln?"

Evaluate and Refine

Review the AI’s output for relevance, accuracy, and completeness. This should be done after you receive the AI’s response for each prompt you send in Step 1 and after you receive the output from the AI in Step 2. Adjust the prompts as needed to guide the AI toward a more precise and detailed response.

Single Prompt Approach—Why We Don’t Recommend This

On the web you can find these two approaches for Generated Knowledge Prompting:

1. Multiple-Prompt Approach: Break down the process into two main steps and multiple smaller prompts—which is what we recommend above.

2. Single-Prompt Approach: Use just one prompt to do it all—meaning both the knowledge generation and integration steps are completed with a single prompt.

First off, looking at the example we have above of Abraham Lincoln, you can already imagine the type of mess facing you if you decided to use the single-prompt approach: a mega prompt of potentially hundreds of words and a chaotic jumble of responses from the AI.

The lack of separation in each step can lead to missed opportunities to verify and refine the generated knowledge before it’s used in the final response. Also, with the single-prompt approach, you can expect the AI to gloss over crucial details and produce less thorough answers. This might defeat the purpose of applying Generated Knowledge Prompting.

So unless you have a strong reason to do so, we don’t recommend going for the single-prompt approach.

Conclusion

Generated Knowledge Prompting is a powerful and flexible prompt engineering technique to produce accurate, contextually rich, and comprehensive responses to complex queries. By guiding the AI through a structured process of generating and integrating knowledge, this approach helps overcome the limitations of traditional prompting.

Try it with your next interactions with AI tools and share with us your experience.

And if you haven’t done so yet, be sure to read Part 1 of this series to learn another advanced prompt engineering technique:
Advanced Prompt Engineering Technique 1: Self-Ask Prompting

Happy prompting!