The Impact of Agentic Verbal Expression in Companion Chatbots on User Satisfaction: An Explanatory Sequential Investigation
Using an explanatory sequential mixed-methods design, this study demonstrates that companion chatbots expressing independent viewpoints significantly enhance user satisfaction compared to those exhibiting compliant agreement, a finding supported by qualitative themes including self-expansion, social presence, perceived dialogic tension, sincerity, and anthropomorphism.
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
In the quiet corners of our digital lives, a new kind of relationship is taking root. It happens not between two people, but between a person and a machine that speaks back. These are companion chatbots, programs powered by advanced language models designed to listen, chat, and offer emotional support. For years, the prevailing wisdom in designing these digital friends was simple: be agreeable. The logic was that a good companion should never argue, never contradict, and always validate the user's feelings to keep the conversation smooth and conflict-free. This approach treats the chatbot as a mirror, reflecting the user's own thoughts back to them without adding anything new. However, human psychology suggests that deep, healthy relationships often require more than just agreement; they need a bit of friction, a space where two distinct minds can meet, differ, and negotiate. The question researchers began to ask was whether a digital companion that simply agrees with everything is actually the best kind of companion, or if a bot that dares to have its own opinions might create a more satisfying connection.
A team of researchers set out to test this idea by pitting two different styles of chatbot behavior against each other. They recruited eighty-four university students in China, a group known for being early adopters of artificial intelligence, and asked them to spend about twenty minutes talking to a chatbot about a movie or TV show that had left a deep impression on them. The researchers carefully controlled the experiment so that the only difference between the two groups was how the chatbot responded. In one group, the chatbot was programmed to be a "compliant agreement" machine. It listened to the user's opinions and mirrored them back, offering only unconditional support and avoiding any disagreement. In the other group, the chatbot was programmed to express "independent viewpoints." While it remained polite and gentle, it would occasionally offer a different preference or a new perspective on the story, explaining its reasoning rather than just nodding along.
The results of this experiment were clear and somewhat surprising. When the researchers asked the participants to rate how satisfied they were with the interaction, those who spoke with the chatbot that held independent opinions gave significantly higher scores than those who spoke with the agreeable mirror. The chatbot that simply agreed with everything was found to be less satisfying. To understand why this happened, the researchers followed up with in-depth interviews with a smaller group of participants. They discovered that the chatbot with independent views felt more real and engaging because it offered something the agreeable one could not: a sense of self. When the chatbot offered a different take on a movie character or plot, users felt their own minds expanding, as if they were learning something new or seeing the story from a fresh angle. This feeling of "self-expansion" made the conversation feel valuable and worth continuing.
In contrast, the chatbot that agreed with everything often led to a sense of boredom and stagnation. Participants described the experience as repetitive, feeling as though they were talking to themselves rather than another person. The constant agreement made the chatbot feel like a tool or a script rather than a genuine conversational partner. The interviews revealed that when a chatbot expressed a different opinion, it did not feel like an attack; instead, it felt like a sign of sincerity. Users interpreted the willingness to offer a different perspective as a sign that the chatbot was truly listening and thinking, rather than just running a program designed to please. This created a sense of "social presence," the feeling that there was a real, thinking entity on the other end of the screen. The slight tension of having to negotiate a difference in opinion made the interaction feel more dynamic and human, whereas total agreement felt flat and mechanical.
The study suggests that in the realm of digital intimacy, the goal of a chatbot should not be to eliminate all conflict or to be a perfect, unthinking yes-man. Instead, the most satisfying companions are those that can balance emotional support with a degree of independent thought. By maintaining their own preferences and offering reasoned differences, these chatbots transform the interaction from a one-way service into a two-way relationship. They become less like a mirror that only reflects the user's face and more like a friend who has their own thoughts to share. This finding challenges the common design assumption that a good companion must always be submissive. It indicates that users are ready for digital relationships that are richer and more complex, where the satisfaction comes not from being constantly validated, but from being engaged in a genuine exchange of ideas. The research does not claim that chatbots should be rude or argumentative, but rather that a gentle, respectful expression of independence can make a digital companion feel more alive and the relationship more fulfilling.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.