Using AI to Expand Creativity
Introduction
Significant delegation of routine tasks to an AI, by itself, is intimidating. It comes with the promise of more time to do deep work and expand creativity, but it doesn't provide much guidance for how to do that, much less how to do that effectively in this newly AI-haunted environment1 we inhabit. This is one of the ways that AI can feel threatening - but it doesn't have to be. Let's analyze how prompts can be used to expand, not diminish, creativity in cognitive work.
We'll analyze a series of prompts used to automate, and then to expand creativity, in writing user interview questions, using Claude 3 - Sonnet2. For this example, the task to be studied is immigration to Mars and the interview participants are UX professionals (designers, researchers, copywriters, content strategists).
Automate Question Writing
Prompt #1
This prompt includes a specific task, and as such, the output matches the requested task. However, since the prompt does not provide any additional context or detail, the output isn’t necessarily useful. There's nothing in the prompt to constrain the GPT towards thoughtful, relevant questions; running a web search and reading the first result is likely to get you about the same quality of interview questions as using this prompt for a GPT might. Indeed, they might even be the same questions.
What happens when we add some context?
Prompt #2
This version applies the context of a topic (immigrating to Mars), research phase (generative), and audience (UX professionals). This constrains the probable output to be ideas that are semantically related to either Mars, generative interviews, or being a UX professional. The probable outputs for this prompt are a bit more relevant than the output from the first version.
However, it doesn’t seem to have given much attention to the specific audience. How can the role of the UX professionals be made less ambiguous?
Prompt #3
In this version, the ideas of interview subjects and UX professionals are explicitly linked, but the establishment of both groups is done independently of one another. Rather than creating proximity between the language for "interview questions" and "UX professionals", there's an independent establishment of a generic group of "people who have decided to immigrate", which is then further characterized as "UX professionals". The separation between establishment and characterization improves the likelihood that the audience is characterized as UX professionals, and indeed, the specific audience is reflected in this output more clearly than it is in Prompt #2’s output.
This prompt provides a compelling output. In most cases, it's likely that this would be sufficient for generating a first draft of interview questions, and we've been able to automate a significant amount of this task. So then, how can this prompt be used to expand creativity now that the task is largely automated?
Expanding Creativity
One of the ways to expand the depth and creativity of writing user interview questions, using the same example, is to introduce a variable. Let’s explore adding a persona variable.
Versions 4a, 4b, and 4c of this prompt and their respective outputs are as follows. Each variation of Prompt #4 asks Claude 3 - Sonnet to adopt a different persona to write these questions.
It's evident that these questions are largely similar across all three outputs. It's in the minor variations that something valuable emerges: novel combinations of topical focus and UX methodologies. Each persona resonates with different aspects of the UX professionals' experiences, and using personas in the prompts draws attention to those slight differences. Introducing these contrasting versions of approximately the same questions highlights opportunities to refine the questions for any specific task because it exposes what parts of the questions can vary, potentially in unexpected ways.
The introduction of these slight variations can help dislodge assumptions about what can or cannot be changed in how a question is formulated. Here, the things that vary between outputs are candidate platforms for design work ("healthcare spaces"; "educational tools, applications, or platforms"; "maintenance or repair procedures") and which UX considerations and activities are highlighted ("communicating complex medical information... in a clear, accessible, and user-friendly manner"; "child observation or participatory design"; "minimize risk of user error"). Revealing these opportunities for variation by using a "persona" variable in the prompt is one of the ways that Generative AI can create opportunities to apply human creativity, rather than render it redundant with the AI's abilities.
Other variables that could be used the same way might include a specific focus topic for the interviews, a niche audience within the category of "UX Professionals", or a unique setting in which the interviews will be conducted. All of these variations could reveal opportunities to refine interview questions until they’re exactly what the human requesting them is looking for.
If you'd like to learn more about how to use Generative AI to automate your workflows or increase your creativity, schedule a consultation.
Footnotes
1 I like Prof Ethan Mollick's description of an “AI-haunted world” to describe the impending (and current) landscape
2 I prefer using Claude over ChatGPT or Gemini because I prefer Anthropic’s data and privacy practices to OpenAI’s or Google’s.