Choosing Between AI Responses: How Valence, Arousal, and Dominance Shape User Preference

This study examines how emotional tone shapes user preference in human–large language model (LLM) interaction. Drawing on the Computers as Social Actors framework, we treat conversational AI as a social communicator whose affective cues influence user judgments. Using large-scale pairwise preference data from LMSYS Chatbot Arena, we model emotional tone through the Valence–Arousal–Dominance framework and analyze how positivity, activation, and assertiveness in prompts and responses relate to users’ choices. Regression results show that responses with higher valence are more likely to be selected, whereas higher arousal and dominance reduce selection probability in baseline specifications. Users also prefer responses that are more positive than competing alternatives and less aroused and less dominant than the original prompt. These effects vary across prompt-task categories, indicating that emotional tone is context dependent. The findings show that affective characteristics shape behavioral preference in LLM-mediated interaction and should inform future user-centered design of conversational AI systems.