AI is a hot topic over the past year. Major advances in Large Language Models (LLMs) havepeople wondering if there is an existential threat to their careers. AI will likely push every industry to places never thought imaginable, for better or worse. I am mostly optimistic about what this means and hopeful for the opportunities AI can bring to the table. I am also skeptical about using it and cautious and intentional when I do.
This is not meant to bash the potential of AI nor glorify AI as some utopian solution to present problems. Rather, I’m sharing my perspective as a practical recommendation on using AI as a design practitioner, specifically using AI for research. While the use of AI is widely debated (even here at Atomic Object), I believe there are places we should and shouldn’t be using it, and it’s important to be clear about that.
Definitions
Before digging into my opinion about using (or not using) AI for doing research, I want to first define what those two terms mean, at least in this context.
Broadly speaking, Design Research — aka User Research or User Experience (UX) Research — is an approach to understanding real-world problems and framing solutions to solve them. Researchers do this by gaining a deep understanding of people who encounter the problem and empathizing with their experience. Not always, but often, the problem is centered around a business, and the goal of research is to identify a product or service that could solve that business’s problem.
Artificial Intelligence (AI) is the brainpower, so to speak, of a machine or a computer rather than the wisdom of a human or an animal. Today, we use AI more than most of us realize for basic, everyday tasks. In this context, I’m referring more to generative AI (like ChatGPT), which can help us take disparate ideas and information and make something new.
When You Should Use AI for Research
There’s no doubt that AI can help speed up manual, antiquated processes of working. Research is no exception. I’m all for leveraging this technology for the busy work. That way, we can focus on truly understanding the problem and delivering solutions.
There are several ways we can use AI to speed things up for us. They include the ability to:
- Extract useful information from an interview with a customer or subject matter expert
- Translate what someone said from a recording into a transcript
- Organize research observations in a sensical manner
- Group information together into similar clusters
- Summarize actionable to-dos from the feedback you heard
- Inspire you to find creative ways to solve the problems you’re uncovering in your research
I have recently used some tools to replace the repetitive, mundane tasks that were ultimately just a timely means to a valuable end.
Otter.ai
Otter.ai has helped me get reliable transcripts of meetings and interviews, so I don’t have to spend hours transcribing them myself or go off of half-complete notes. I can also generate a good summary based on the feedback, which is helpful to quickly turn around early, high-level research findings to a client (instead of rushing to get out an executive summary).
Miro AI
Miro is a helpful tool for clustering different interview observations to uncover themes and patterns. There are new, although relatively basic, features for grouping individual observations by keyword or sentiment, which is a helpful place to start the research synthesis process.
ChatGPT
This is the tool I use least often and most cautiously, but it’s hard to argue against its use in the right situation. After I’d facilitated interviews, gone through the transcripts in Otter, and clustered observations in Miro, I could define the problems we needed to solve. I understood the task at hand. ChatGPT was pretty remarkable at summarizing what all the interviewees said and helping me outline their feedback into sensible action items. I could also pull out the specific details of a process that someone walked me through in an interview, upload it to ChatGPT, and have it outline the steps of a workflow (something that can take a lot of time and thought to get started).
Once we’ve processed the information, I feel it’s safe to experiment with AI to help uncover opportunities and inspiration. At that point, we’ll know right from wrong when a robot generates an idea, but we still have to gain deep empathy for the people we’re designing for, no matter what.
When You Shouldn’t Use AI for Research
You can’t skip getting a good grasp of the information you glean from the research process. You still need to absorb that information and understand the problem, but AI can speed up the process of defining that problem and help you generate concepts faster.
There are several ways I think we should avoid using AI during research:
- To identify what you should be researching in the first place
- To summarize an entire interview without listening to it ourselves
- To identify the actual problem(s) you need to solve
- To create design solutions without having a firm understanding of the problem at hand
Don’t rely too much on AI tools – you need to become familiar with the research topic(s), almost to the point of becoming a subject matter expert. No tool can wave a magic wand and turn you into an expert overnight. You still need to do the work and understand the content, but you can speed up that process. Then if you do use a generative AI tool, you are the judge of a good (or bad) output.
Find a Balance
We should leverage AI to speed up the research process where it makes sense. This is especially true for small teams, as research is highly valuable in the product development process, but can be hard to do with a small team or a low budget. Using tools to make research more time-effective will allow us to focus on understanding what our customers are experiencing, and that isn’t something we should skimp out on. Then we’ll know what needs to be done to solve their problems, getting solutions in their hands even faster.