Advanced Prompt Engineering for Knowledge Harvesting: A Case Study Involving Indiana Dunes

When using generative AI platforms like ChatGPT, mastering Prompting can take you a long way. However, utilizing prompting for knowledge harvesting can open up new worlds to achieve success or advance your business goals.

In this blog post, I'll explain an advanced form of prompting known as Perspective Prompting, and explain how to harness it to harvest intuitive knowledge profiles. Whether you're a researcher or scientist, a real estate professional, a comic book collector or a startup founder, what you'll read in this blog post can be tailored for your industry or profession.

Beginning with a term you might not be familiar with--Perspective Prompting. It's the ability to instruct a tool like ChatGPT to produce information or responses from the perspective of a person, character, role, or persona. For example, you might use a perspective prompt like this: "Please provide me a list of the best national parks to visit from the perspective of a Park Ranger." ChatGPT is able to respond based on information it has learned from a large language model.

If you're not familiar with large language models, they are like big computer brains that mimic and predict human language by analyzing enormous amounts of text. Moving on from that, while perspective prompting can be very useful, it represents just the basics of prompt engineering.

The limitation of basic perspective prompting is that it produces responses lacking the nuances of human thought and real-life situational awareness. Take ChatGPT as an example: it draws from a vast database of information, yet it may struggle to fully grasp or reflect the full picture of fast-changing, real-life scenarios.

AI's limitations become even more apparent when we delve deeper into specialized scenarios. For instance, if we use the perspective prompt of a Park Ranger at Indiana Dunes and ask ChatGPT for a list of parks, the response, while accurate, only scratches the surface. Consider the following perspective prompt followed by a response from ChatGPT:

Perspective Prompt: "From the perspective of a park ranger who works at Indiana Dunes National Park, tell me how far the picnic tables are from the bathrooms, for a family with 2 teenagers and a toddler."

ChatGPT Response: As a park ranger at Indiana Dunes National Park, when assisting a family with two teenagers and a toddler, I would guide them to picnic areas where amenities are close enough to manage with the little one in tow. At our park, we strive to ensure that picnic areas are family-friendly, with restrooms usually no more than a short walk away."

As you can see in the example, ChatGPT may not always know the latest on-site developments at places like Indiana Dunes National Park, where certain areas could be closed for maintenance. These improvements at the park may affect accessibility to restrooms and nearby picnic spots. Such specifics often require local knowledge, like that of a Park Ranger. However, by talking to a Park Ranger or checking the Indiana Dunes website, you can gather information to create a data model that thinks like a Ranger on the ground.

Then, by consolidating that data or information and developing advanced prompt engineering, you can query the Ranger's harvested knowledge to analyze scenarios, answer specific questions about the park, simulate experiences at the park, or make predictions with on-the-ground accuracy.

Keep in mind, you don't have to talk directly to a Park Ranger, to train your data model. You can still train it by using natural language processing (NLP) to analyze and collect information from content or media that park rangers have shared. That information can then be fed into your structured data model. These could be articles, interviews, social media posts, public communications, podcast recordings, or any other form of unstructured data.

In my work, I'm particularly drawn to integrating cultural adaptability within the realm of prompt engineering and data model frameworks. I believe this is an opportunity for us to navigate a wide range of cultures, which empower 'machines' to understand and interact with the world more effectively.

Much like the beautiful skies above the Indiana Dunes, when it comes to prompt engineering and knowledge harvesting, there's practically no cap as to what we can do!

Are you interested in setting up your harvested knowledge profiles for your project or your business? I'd love to help you do that. Let's connect on LinkedIn and get a conversation started! (no pun intended)