AI Persona Design and Perceived Difficulty of Legal Advice Tasks
This study demonstrates that varying the occupational persona of a legal advice chatbot significantly influences users' perceived task difficulty, with a mediator-like persona rated as least difficult and suggesting that designing AI with high agreeableness and low neuroticism could enhance the understandability of complex legal advice.
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
Imagine asking a computer for help with a complicated legal problem, like figuring out who is responsible when a flowerpot falls from a balcony or whether a worker is truly an employee. For many people, the idea of getting advice from a machine feels cold, confusing, or even intimidating. This is especially true in fields like law, where the stakes are high and the language can be dense. Researchers have long known that the way a person interacts with technology changes how they feel about it. Just as we react differently to a stern teacher versus a friendly neighbor, we likely react differently to a computer that sounds strict compared to one that sounds helpful. This field of study looks at how the "personality" we give to a computer program—its tone, its style, and the role it plays—shapes our trust and our ability to understand complex information. The question is not just whether the computer gives the right answer, but whether the way it presents that answer makes the task feel manageable or overwhelming.
A team of researchers at Sigmund Freud University in Vienna set out to test this idea with a small, focused experiment involving legal advice. They wanted to see if changing the "job title" and personality style of a chatbot could change how difficult people found the task of understanding legal scenarios. They created three different versions of the same legal advice bot, each programmed with a distinct set of personality traits. One bot was designed to sound like a judge: structured, rule-focused, and somewhat distant. Another was styled as a lawyer: strategic, persuasive, and ready for a debate. The third was built to act like a mediator: empathetic, focused on finding common ground, and eager to listen. These personalities were not random; they were carefully constructed using a standard psychological framework that measures five key traits: how open a person is to new ideas, how organized they are, how outgoing they are, how cooperative they are, and how emotionally sensitive they are. The researchers gave these traits specific values to create the distinct "voices" of the judge, the lawyer, and the mediator.
Sixteen adults, mostly students with backgrounds in psychology, volunteered to take part in the study. Each person was randomly assigned to talk to just one of the three bots. They then worked through three different legal situations, such as a dispute over a bicycle sale or a question about privacy rights at work. After reading the scenario and chatting with their assigned bot to get an explanation, the participants had to rate how difficult they found the task. The researchers also measured the participants' own personalities to see if a person's natural style matched better with a specific type of bot. The goal was not to prove which bot was the best for everyone, but to see if the different styles created noticeable differences in how hard the work felt.
The results showed a clear and striking difference between the three groups. People who spoke with the mediator bot found the legal tasks significantly easier than those who spoke with the other two. On a scale of difficulty, the mediator group gave the lowest scores, indicating they felt the tasks were straightforward. The group talking to the lawyer found the tasks moderately difficult, while the group talking to the judge found them the most challenging. The gap between the mediator and the judge was particularly large, suggesting that the warm, cooperative style of the mediator made the complex legal information feel much more accessible. In contrast, the strict, rule-bound style of the judge persona seemed to make the same information feel heavier and harder to grasp.
The study also looked at whether a person's own personality changed how they reacted to these bots. Because the groups were so small, the researchers could not draw firm conclusions about these connections, but they noticed some interesting patterns. For instance, people who were naturally very organized and detail-oriented seemed to find the judge's strict style particularly difficult to navigate. Meanwhile, the mediator's style seemed to work well across the board, making the tasks feel easier regardless of the user's own personality. This suggests that a calm, cooperative approach might be a safer bet for helping people understand complex advice, whereas a more authoritative or confrontational style might create unnecessary barriers.
Despite these promising findings, the researchers are careful to note that this was a small pilot study, not a final answer. The group of participants was small and drawn from a specific university, so the results cannot yet be applied to the general public. The study also did not measure whether the advice given was actually correct or if the participants truly understood the legal concepts, only that they felt the tasks were easier or harder. The researchers acknowledge that the "mediator" bot was a mix of many positive traits, so it is not yet clear if the ease came from the friendly tone, the specific job title, or a combination of both. They also found that while people trusted the computer to be competent and useful, they still trusted human experts more, especially for important decisions. This suggests that AI should be seen as a tool to help people access information and prepare questions, rather than a replacement for professional judgment.
The main takeaway from this work is that the design of a conversational AI matters deeply. How a computer speaks and what role it plays can change a user's experience from frustrating to manageable. The study suggests that giving a legal advice bot a mediator-like personality—characterized by patience, empathy, and a focus on cooperation—could make complex legal advice feel much less daunting. However, the researchers emphasize that this is just the beginning. To truly understand how to design these tools, larger and more rigorous studies are needed to test these ideas with a wider range of people and to ensure that making a task feel easier does not accidentally make people overconfident in their understanding. For now, the evidence points to a simple but powerful idea: when asking a machine for help with serious matters, a friendly and balanced voice may be the key to unlocking understanding.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.