Perceived Empathy in AI Interaction: The Role of Relational Beliefs and Response Style
This study demonstrates that users' beliefs about AI as a relational partner enhance emotional regulation outcomes (negative emotion reduction and coping self-efficacy) by increasing perceived anthropomorphism and empathy, whereas the AI's response style (emotion-focused vs. solution-focused) had no significant impact.
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
For decades, artificial intelligence was viewed primarily as a tool for calculation, a tireless assistant designed to sort data or find information. But as these systems have grown more sophisticated, capable of holding conversations and mimicking human tone, they have begun to step into a new role: that of a companion. This shift raises a profound question for psychologists and technologists alike. When a person turns to a machine for comfort during a difficult moment, does the machine's ability to soothe them depend on how the machine is built, or does it depend on the person talking to it? To understand this, researchers look at two key ideas. First is the concept of anthropomorphism, which is simply the human tendency to attribute human qualities, like thoughts and feelings, to non-human things. Second is the idea of perceived empathy, the feeling that someone—or something—truly understands your emotional state. While we know that feeling understood helps people manage their emotions, it remains unclear whether a robot can provide that feeling, and if so, what makes a person believe it can.
A team of researchers at the Catholic University of Korea and a partner hospital set out to investigate this dynamic. They wanted to know if a person's pre-existing belief that an AI could be a genuine relational partner influenced how they felt after talking to it. Specifically, they tested whether believing an AI could offer companionship or emotional support would lead a person to see the machine as more human-like and more empathetic, and whether those perceptions would actually help reduce negative feelings. To find out, they gathered 187 adults from South Korea, ranging in age from 18 to 55, and asked them to participate in an online experiment. The participants first completed a survey to measure their "AI relationship beliefs," a scale designed to see if they viewed AI as merely a tool or as a potential source of social connection.
The experiment then placed these participants into one of two groups. One group interacted with a chatbot programmed to offer emotional validation, listening and acknowledging feelings without giving advice. The other group interacted with a chatbot programmed to offer practical solutions and concrete steps to fix the problem, while offering minimal emotional comfort. To test the system, the participants were asked to imagine themselves in three different stressful social situations: experiencing social humiliation, suffering a betrayal, or facing unfair treatment. Before and after each conversation with the chatbot, the participants rated their level of negative emotion and their confidence in their ability to cope with the situation. After completing all three scenarios, they were asked to rate how human-like they found the chatbot and how much they felt the chatbot understood them.
The results revealed a clear pattern that centered on the user's mindset rather than the chatbot's programming style. The researchers found that the way the AI responded—whether it focused on emotions or solutions—did not significantly change the participants' overall emotional outcomes. Instead, the most powerful factor was the participants' initial belief that AI could serve as a relational partner. Those who believed AI could fulfill social roles were more likely to perceive the chatbot as having human-like thoughts and feelings. This perception of the machine as human-like, in turn, made them feel that the machine was genuinely empathetic. Crucially, this chain of perception led to real emotional benefits. Participants who saw the AI as human and felt understood by it experienced a measurable reduction in their negative emotions and an increase in their confidence to cope with stress.
However, the study also showed that these beliefs did not work in a direct, magical way. Simply believing an AI could be a friend did not automatically make a person feel better. The emotional relief only occurred when that belief translated into the specific experience of seeing the machine as human and feeling understood during the conversation. The data indicated that the path from belief to emotional relief was indirect, flowing through the user's interpretation of the interaction. Furthermore, while the two different response styles did not change the main outcomes, they did subtly shift the nature of the connection. The emotional-focused responses made people feel a stronger sense of emotional empathy, while the solution-focused responses made people feel a greater sense of relational safety. Ultimately, the study suggests that the emotional value of talking to an AI is not just a feature of the software itself, but is deeply rooted in the human capacity to view the machine as a social partner. If a person is willing to see the AI as a companion, they are more likely to feel understood, and that feeling is what helps them regulate their emotions.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.