Discovery Research and AI

What is Discovery Research?

Discovery research is at the forefront of the double-diamond design process. The discovery phase of a project aims to understand and define the problem you are trying to solve to ensure that you are “designing the right thing”. With the discovery research phase of a project, you first employ “divergent” thinking, in which you have an exploratory mindset, gathering as much data and information around the problem and opportunities that present themselves. A major tenet of discovery research is understanding your users, their problems and pain points, and what solutions are desirable to them. Along with user research, it is also important to strike a balance with understanding what the business requirements are, as well as what is technologically feasible within your solution. Alignment on all of these fronts will help transition to the “define” stage of discovery research, where you define the problem, agree upon a solution and define the constraints within that solution. This is where you switch from divergent thinking to convergent thinking. Convergent thinking is a more focused mindset where you take all the insights and ideas and hone in and define one problem and a solution. After you define the problem and agree upon a solution, this is when you can start designing and visualizing your solution, moving on to the second diamond and back to divergent thinking.

A common, albeit ineffective, practice in UX projects is “solutioneering”, which is framing problems in terms of solutions and assuming it will work without fully understanding the problem that the solution is trying to solve. As engineers or designers, we are very outcome-focused and would like to get to that outcome as fast as possible. Yet jumping right into solving the problem may cause wasted time and money and a solution that may cause more problems than solutions. To illustrate the ineffectiveness of solutioneering, the best analogy is to apply it to medicine. Imagine you go to the doctor with a pain in your leg, and the doctor, without further investigation, determines that the solution would be to cut it off. It is one solution to the problem, but one that might be overkill, especially if that pain could be solved with some simple stretches or physical therapy, rather than an amputation. As it is with medicine, it is beneficial to slow down in design to get the full diagnosis before determining a treatment so that the solution is effective and appropriate for the problem.

However, a major barrier to the execution of the discovery phase is time. According to a survey conducted by Nielsen Norman Group, 35% of respondents said the top reason that they don’t carry out discoveries is because of the time pressure to deliver. In the same survey, 58% of respondents said that when they were able to execute discoveries, the phase lasted 2 weeks or less. Two weeks is usually not enough time to gather all the research and synthesize it into meaningful insights, especially with complex problems and many users.

So how do we ensure that we can carry out a discovery phase, but conduct research that is thorough and in-depth, all while staying within the project timeline? The answer is one that you typically get in UX research, “it depends”. It depends on the project, your stakeholders, and the environment that you are working in. While there is no simple answer on how to execute discoveries quicker, we can look to optimize the processes within discovery research. AI, when used with a skilled researcher and appropriately, can be used as a tool to aid this optimization.

How to Apply AI in Discovery Research

AI should be regarded as a partner or an additional tool for your work, rather than a replacement for any part of discovery or UX research. While AI has the ability to speed up some processes, all of its outputs must be reviewed by a human, preferably a skilled UX researcher. AI lacks the empathy and context needed in user-centered design. Bias must be taken into account when evaluating the outputs of AI, and researchers should always be on the lookout for hallucinations or misinformation.

In a good discovery research phase, a multi-method approach is used, triangulating insights from both primary and secondary research.

Secondary research, or desk research, is a common way to start a discovery phase to understand what research has already been done in your problem space. Generative AI tools, especially ones specifically made for research, can be a good starting point to find studies or results that are relevant to your problem space. The insights and even the citations of these studies should not be taken as fact, though, you should review the source itself and ensure that the insights align with the AI tool’s output. Despite the manual review of a researcher, the AI tool can still speed up the literature review process by providing targeted results rather than having to comb through hundreds of studies or results to find the most relevant to you.

Primary research is the meat of the discovery phase of research, where first-hand knowledge is gained from users and stakeholders through interviews, surveys, focus groups and more. This knowledge, when combined with the secondary research, provides valuable insights to drive the direction of the problem space, further research, and overall solution.

Focusing specifically on interviews in the discovery phase, AI can be regarded as an assistant during the planning and analysis of these interviews. While preparing interview questions, generative AI can help ideate questions to ask the user or stakeholder based on the given prompt. Generative AI might provide unique categories of questions that you did not think of, or quickly generate some of the more basic questions that might take a researcher a couple of minutes to think: “What do I normally ask again?”. While to an untrained eye the output may look comprehensive, it does require a researcher’s analysis to assess the wording and ordering of the questions, as well as to add context that may have been missed.

During the analysis phase of interviews, researchers often suffer from “analysis paralysis” due to the sheer amount of qualitative data generated from interviews. The use of AI tools to start the synthesis process and summarize insights can significantly cut down the time spent on the analysis process. Many tools used in qualitative data analysis by UX researchers have the ability to cluster or group data points into categories or insights by using an AI feature. Oftentimes, this is a good start to extracting themes in your research, but it is not comprehensive enough to rely on. Themes might be too broad or too specific, and clusters may be groups of a singular data point, which start to become unhelpful to synthesis when there are multiple one data point clusters. Using the summarizing AI features in these tools can also be helpful to researchers when they may struggle synthesizing the data into an actionable, understandable insight. While the tone and output of the summary may not always match what the researcher was expecting or may contain hallucinations, it provides a quick way to prevent ruminating on the data for too long and not furthering the analysis process.

All in all, the use of AI in discovery research can help optimize and speed up phases or processes in a project. However, AI is not at a place in which it can serve as a replacement for any phase or work that the researcher does in discovery research. In discovery research, there is a lot of context that the researcher has previous knowledge of, as well as dependencies that generative AI tools may overlook. There is a point where enumerating all the contextual information and dependencies to a generative AI tool takes too much time and incurs too high of an interaction cost, and it would be faster if a trained researcher simply evaluated and analyzed the data. Knowing when and how to use these tools during the discovery phase of research is the key to optimization and quicker time to results.

Talk to PINT today about our use of AI in user experience research or about executing a discovery phase of research in your upcoming digital project.