AI Scoping is the Secret to Quality Results

Previously I gave an overview of the key elements that I use for guiding AI chatbots productively: context, scope, and goals . In this second part I'll zooming in on scope - helping you to understand how defining the limitations of your inputs is critical for generating quality output.

With the current focus on trying to create "master" prompts that generate perfect outputs from generative chatbots, narrowing scope gets far less attention than defining context or stating goals. That's because of all three tools - context, scope, and goals - scope is the one that lands primarily in the lap of the user.

Let's review what each of these terms means in this context:

• Goals: These are the specific objectives or outcomes that you are aiming to achieve in the conversation. What do you want the AI to do or produce?
• Context: This is the overall setting or background for the interaction. What is the conversation about? Who is having it? What are the underlying assumptions or premises?
• Scope: This refers to the boundaries or limits of the conversation. What information should be included, and what is excluded? think of it as the 'walls' that guide the conversation in a particular direction.

Because AI chatbots like ChatGPT have inhaled so much information and are designed to consider the connections between all the elements they contain, scope may be the most difficult of the three elements to define clearly.

It also tends to overlap a bit with both goals and context. The difference is one of action vs. output. I limit scope to better satisfy goals. I narrow scope to better define context. Understanding this distinction is crucial:

• Scope is about input: Scope defines what information is relevant or irrelevant to the task at hand. It's like setting the rules for what can and cannot be considered.
• Context is about the range of information: In what areas of knowledge and
• Goals are about output: They guide what you want the final result to be, whether it's an answer to a question, a piece of writing in a specific style, or something else.

By layering these three elements together properly, I can more quickly and effectively complete my work using AI tools.

Your own relationship to scope may be modified by what you're looking to get the AI to generate for you. The more you want to use AI for outputting specific materials designed to satisfy particular criteria, the more you'll want to tightly define scope to make sure the results fall into the particular parameters that mean the results satisfy your specific needs. In this way, you're using scope to try and more effectively reach your goals.

On the other hand, when you're creatively brainstorming, you'll appreciate what the AI can create for you when the aperture of your scope is wide open. In that way, it can spark your imagination and deliver results that are unique and inspiring.

To see just how defining scope can help your work, let's look at the way you can it in ChatGPT (a text tool) and Midjourney (an Image generation tool).

Midjourney

In a visual tool like Midjourney, scope is not only necessary to good outputt, but it's possible to come at it from multiple directions with far more direct impact in terms of output. What is the visual style you want your image to be in? Should it represent a particular time period? An aspect ratio? A particular time of day? A unique lighting style? With properly defined scope in your initial prompt, you are bringing in the walls around your subject and hopefully getting something that is bringing more clarity to your ideas.

You can also feed it an initial image with the hope that it will use elements of that to define the scope even further. Say I want an image of my cat in a top hat, for example. I can start by giving it a link to an image of my actual cat. That's limiting the scope of its output to cats that look like the one in the photo. (And yes, I've actually done this...)

Once the initial input has been made, Midjourney will give you 4 outputs. Choosing one of these images to feed back into the LLM limits your scope so that when it makes modifications it can try and satisfy your needs even more clearly.

The act of choosing a single image is a clear reduction in scope even as the reasons you may be picking a particular image are to satisfy your goals.

ChatGPT

When facing the blank prompt, your initial instinct will be to focus in on context and goals. What do I want and what is it for?? In this context Scope is primarily about dynamically editing your piece. "Use my voice," for example, is a scoping request.

Let's look at a specific example. Imagine you're writing a blog post about environmental conservation, and you need a concise explanation of the greenhouse effect for a general audience. Your goal is clear: a simple and engaging explanation. Here's how you might define the scope:

• Context: "You are an environmental scientist. Write a blog post for a general audience about environmental conservation."
• Goal: "Give me a concise explanation of the greenhouse effect that's easy to understand."
• Scope: "Please avoid using scientific jargon or complex terms. Stay focused on greenhouse gasses. Keep the explanation to around 100 words."

Interestingly, in the current version of ChatGPT some scoping are built into the interface rather than happening through the prompt. You can save scoping (and context) information that allows you to return to individual sessions and not to repeat your requests over and over again.

When you're considering using scope for writing, imagine it's a chisel that lets you carve away some of the material and better define the limitations of the chatbot's output. In this case, your goals are how much liquid you plan to drink, and scope is the size of the glass that you want the AI to use.

In practical terms, it may be explaining to the chatbot the number of characters/words you want it to output or the complexity of language you want it to use. In the example above, the scope was defined both by the simplicity of language and the word count.

You can say, “Make sure to always use words that are simple and direct,” or go the other way and ask for "occasional florid swirls of language."

Like everything in ChatGPT, scoping is iterative. After each output, you can identify areas that need refinement and scope the next request accordingly. By carefully defining and redefining the scope in line with your goals, you can direct the conversation to a result that truly meets your needs.

Each back-and-forth should shape the focus for the next exchange. Scope begets scope. Stay nimble and be prepared to zoom in or out as needed.

Scoping towards Goals

As you define (and redefine) scope, it ultimately connects to your core goals. Knowing what you want to accomplish at any given moment lets you tailor scope accordingly.

When ideating, keep prompts short and output broad. If revising, keep limiting scope through defining style and structure. When finalizing, use focused scope for improvements.

Scope takes practice

Like any skill, effectively scoping conversations improves over time. At first, you can err towards over-scoping - tighten prompts and outputs gradually as you observe results.

Pay attention to when AI loses focus or regurgitates text. Dial it in until you hit the sweet spot of concise, productive results.

With proper scoping, you’ll be amazed at what creative sparks emerge. Creativity thrives with clearly bounded spaces to explore. Use scope purposefully and artfully to unlock your AI’s hidden potential.