Enhancing AI conversational agents for effective advice-giving: exploring factors affecting user-agent interactions

As conversational agents (CAs) become more ubiquitous in digital interactions, it becomes essential to understand whether their design influences users. This study investigates the CA communication style in terms of language formality, avatar outfit formality, and advice alignment to see how these factors influence psychological reactance and intention to follow advice. Based on politeness theory, psychological reactance theory, and responsibility attribution theory, three empirical models test some key determinants affecting CA effectiveness in advice-taking situations. Model 1 examines the effect of language and avatar formality on advice adherence. Results indicate that formal language has a significant positive effect on adivce adherence, whereas the formality of the avatar outfit has no effect. Mediation tests indicate that formal language increases perceived negative politeness, which in turn has an impact on reactance and adivce adherence. In Model 2, the effect of language directness and advice alignment on adivce adherence is examined. Outcomes show that aligned advice with user’s previous preferences reduces psychological reactance, thereby increasing adivce adherence. Language directness, nonetheless, does not independently affect adivce adherence, which attests to the importance of alignment in overcoming resistance to AI suggestions. In Model 3, the focus turns to how responsibility attribution and the success or failure of an advice received from a CA impact future adoption of a chatbot. Users put less responsibility on a CA for the outcome when they received advice that is aligned with their previous preference. While the success of the dvice directly affects intention to use the CA positively. In summary, this research advances human-machine interaction literature by clarifying how verbal and nonverbal cues influence user behavior in AI-driven advice settings. Practical implications include strategies for designing persuasive, user-centered CAs that optimize engagement and adherence while managing user resistance and responsibility perceptions.