Get better responses from ChatGPT with example-based thinking
Crafting great examples is the key for designers to make use of ChatGPT.
Learning to craft great examples will become a highly valuable skill for designers, especially if you want to use ChatGPT with your design process.
I previously talked about how the value of design may start to increasingly come from thinking like a designer, especially when AI can help you generate your mockups and prototypes.
However, trying and learning everything you need to know about design thinking takes time and effort. You might find yourself asking, where do I even begin?
When researching how to craft an ideal ChatGPT prompt, I stumbled upon the answer: Example-based thinking. Learning that one skill doesn't just help you communicate more effectively; it's also crucial to getting better responses from ChatGPT.
Here's why that happens.
Examples, not analogies, help bridge gaps in knowledge
Have you ever had a team member suggest, "We should design something like X"? Phrases like, "We should design our search page like Google," or "Our e-commerce homepage should look like Amazon," are often common but not that useful.
Your team uses these phrases because they're trying to bridge their knowledge gaps around design. They may not know the specifics of the design pattern, but they've seen it used somewhere.
You may have even used these techniques if you've ever had to explain what you do as a Designer.
After all, isn't it more accessible to talk about a bad user experience someone encountered and how you fix those problems rather than a detailed explanation between a UX Designer and a Graphic Designer?
However, there’s a difference between these two scenarios. In the designer's case, we're providing a detailed example that people can relate to (like the last time they used their bank's horrible website) to help people understand the specifics of what our job entails.
In our team's case, though, they're using a very vague analogy: after all, what does it mean to design a search page "like Google"? Does that mean there's only one text field and much whitespace? Or something vague, like "it looks good and simple"? It’s to take action on such a vague descriptor.
This is why learning to craft specific, strong examples is such a valuable skill: it helps to narrow down your focus and provides actionable next steps.
In addition, that same skill is critical to writing a great ChatGPT prompt.
Example-based thinking and how it helps your ChatGPT prompt
ChatGPT has helped me in several design tasks, such as creating documentation, organizing my research, and summarizing points like meeting minutes or the 1st draft of my research.
However, one of the things I've noticed is that the quality of the response primarily depends on the quality of my prompt. In other words, if I input a garbage prompt, it outputs a garbage response.
As I mentioned previously, the first mistake to avoid with generative AI tools like ChatGPT is treating them like a search engine. Your prompts don't need to be limited to a single sentence or question: the longer and more detailed your response, the better it usually turns out.
However, when I learned about Jeff Su's ChatGPT prompt model, I truly started getting the quality responses I needed.
It consists of six elements, in order of descending importance:
Task
Context
Exemplar
Persona
Format
Tone
However, the thing to focus on here is the third point: exemplar, a fancy way of saying example. Examples can constrain generative AI tools like ChatGPT to produce more targeted and relevant results.
To use our previous example about explaining UX to our parents, adding a strong example to our ChatGPT prompt can provide us with much better and more relevant results.
Adding that single example allows you to probably get exactly what you were looking for, rather than having to mess around and try several different phrasings with ChatGPT.
This is why learning to craft these detailed examples is so crucial for designers. So how exactly do you start learning this skill? It starts by narrowing your focus: whether in discussions with teams or crafting prompts for ChatGPT, learning to focus on the details can make a difference.
Think atomically about Design
"Atomic Design" by Brad Frost may be an older concept, but it's an excellent reference point for adopting example-based thinking.
While designers often have access to existing design systems or pattern libraries, you should focus on how to break down designs for example-based thinking.
When someone suggests, "We should design our search page like Google," your first instinct should be to dissect that statement into something actionable.
What exact part of the design is appealing to the stakeholders? Is it:
The page layout or template of search being front and center?
A feature or module, like having Gmail, Google Play, and more easily accessible from this page?
A pattern, like the search bar having autocomplete results?
Identifying the specific details of what elements resonate with stakeholders can clarify vague statements. That focus on the details of the statement is the first part of creating a great example to work with.
After this comes seeking inspiration from different sources to compare and contrast how organizations approach similar tasks. Strangely enough, one of the best approaches I've found for this came from a fence repair company.
Create a scrapbook of Design element references
I had no idea the difference between Shadow Box or Privacy Scalloped fences when I needed to repair my fence. Most people wouldn't, so the company brought a scrapbook full of fence photos to help me figure out what I wanted.
Contractors often have stylebooks with photos and descriptions, allowing customers to see past industry jargon and get a visual indicator of what each of these words means.
A design scrapbook works similarly. If stakeholders appreciate Amazon's search bar design, are there other designs that could be references?
Finding and maintaining examples of specific design elements for reference can not only clear up any misunderstandings: but it can also help you with your ideation process.
If you told ChatGPT that you wanted to create an e-commerce page "like Amazon," you would likely get a very generic response.
But looking at a scrapbook of design inspiration, and knowing that you want an e-commerce website with a center-constrained grid, cards as primary design elements, a search bar as the primary navigation tool, and more, can help you define your example in a way that allows you to leverage ChatGPT.
Doing these two steps can allow you to harness ChatGPT's power effectively.
Create one strong example, and ChatGPT can fill in the rest
One of the most valuable use cases that I've found for ChatGPT is to 'fill in the rest of the work.' For example, if I document one row or column of a table, ChatGPT can fill in the rest of the rows if I give it enough guidance.
Whether generating a table of colors for design documentation, creating bullet point summaries of meeting minutes, organizing (and summarizing) user research findings, or more, ChatGPT can save me immense time and effort on otherwise tedious tasks.
But to do that, you need to be able to craft a strong example for it to reference. This is why example-based thinking will be a valuable skill to learn for designers.
We come across more than a few vague examples from team members daily. By learning to drill into the specifics and provide specific examples, we can constrain ChatGPT to give us relevant and targeted results to help us work more effectively.
So, if you're looking to level up your thinking skills as a designer, especially working with generative AI, begin by focusing on how to craft strong examples. Doing so helps you better communicate with your team and can also allow you to communicate effectively with ChatGPT.
Kai Wong is a Senior Product Designer and Data and Design newsletter writer. His book, Data-Informed UX Design, provides 21 small changes you can make to your design process to leverage the power of data and design.