Synthetic personas as support for UX research
This paper proposes a Design Science Research-based procedure for integrating synthetic personas and AI-generated journeys into UX research to accelerate early-stage analysis and hypothesis formulation, while emphasizing the critical necessity of human oversight to mitigate bias and ensure empirical validity.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Every product we use, from a smartphone app to a banking website, is built on a hidden assumption about who will use it and how they will move through it. Designers often create a mental picture of this user, but to make that picture useful for a whole team, they traditionally build a "persona." Think of a persona as a detailed character profile that summarizes the goals, habits, and frustrations of a specific type of person. Alongside this profile, teams map out a "journey," which is a step-by-step story of how that person tries to accomplish a task, where they might get stuck, and what they expect to happen next. These tools help teams agree on who they are designing for and where the experience might break down. However, creating these profiles and stories takes time, and once made, they often stay the same even as the product or the world changes. With the rise of artificial intelligence that can write stories and generate images, a new question has emerged: can machines create these user profiles and stories for us, and if so, how do we use them without being fooled by a convincing but fake story?
A team of researchers at the Federal University of Rio de Janeiro set out to answer this question by developing a new way to use artificial intelligence in design research. They did not simply ask a computer to invent a user; instead, they built a strict method to ensure that the computer's output remains a tool for thinking, not a replacement for real evidence. The researchers started by acknowledging a critical risk: an artificial intelligence can write a story about a user that sounds perfectly logical and detailed, yet it is entirely made up. If a design team accepts this story as fact, they might build a product based on a fantasy rather than reality. To solve this, the team created a step-by-step procedure that treats these AI-generated profiles as "hypotheses" or educated guesses that must be checked by humans.
The method they proposed begins with the human team clearly defining the specific problem they are trying to solve and the context of the product. Only after this context is set do they ask the AI to create a profile. The AI is given specific instructions about what matters for that particular situation, such as the user's main goal, what might stop them from succeeding, and what tools they are familiar with. The AI then generates a story of how this hypothetical person would move through the product. Crucially, the researchers insist that the human team must immediately review this story. They check to see if the story makes sense given the rules they set, if it avoids unfair stereotypes, and if it actually helps them ask better questions. The team also requires that every step be documented, creating a clear trail that links the original problem to the AI's guess and finally to the human team's decision. This ensures that no one forgets that the story came from a machine and not from a real person.
To test if this method worked, the researchers invited eleven experienced design professionals to try it out. The participants were asked to use the new procedure to create a user profile and a journey story for a project they were familiar with. The results were encouraging. The professionals reported that the process was very useful, with nearly all of them rating it highly on a scale of one to five. They found that the AI helped them get started quickly, providing a first draft of a story that they could then critique and improve. It acted as a spark for discussion, helping the team spot potential problems they might have missed if they started from a blank page. However, the participants also made it clear that the AI could not do the thinking for them. They emphasized that the generated stories needed careful editing and that the team had to remain in control to ensure the stories didn't become too generic or biased.
The study concludes that artificial intelligence can be a powerful partner in the early stages of design, but only if it is used with a clear set of rules. The AI can generate many different scenarios quickly, helping teams explore more possibilities than they could on their own. Yet, the researchers are firm that these machine-generated stories are never a substitute for talking to real people. The value of the method lies in its ability to organize questions and make assumptions visible, not in providing final answers. By treating AI output as a starting point for human review rather than a finished product, design teams can use these tools to prepare for real-world testing without losing sight of the complexity of actual human experience. The research suggests that when used responsibly, this approach helps teams move faster and think more deeply, provided they never forget that the story was written by a machine, not lived by a person.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.