How Chatbots’ Conversation Skills Influence Users’ Satisfaction: Evidence from a Serial Mediation Model and Brain Activity

Abstract

Although chatbots are increasingly adopted, users often report lower satisfaction compared to humans, partly due to their lack of social and emotional intelligence. This study examines how chatbots’ conversation content (generic vs. tailored) and style (non-varied vs. varied) jointly influence users’ satisfaction and through what psychological mechanisms. Across two studies, tailored responses enhanced satisfaction more than generic ones, especially when delivered in a varied style. Serial mediation analyses revealed that tailored responses improved satisfaction by increasing perceived warmth or competence, which in turn elevated social presence. EEG evidence showed tailored-varied responses elicited stronger neural engagement, reflecting subconscious affective and motivational processing. These results demonstrate chatbots’ conversational skills shape both users’ conscious evaluations and subconscious responses. This research advances understanding of human-chatbot interaction by identifying a moderated serial mediation mechanism and offers design guidance: chatbots should adopt socially intelligent strategies that foster warmth, competence, and presence to optimize user satisfaction.