When emotions matter: the contextual effectiveness of emotional expressions in AI agents

Purpose This paper aims to examine how artificial intelligence (AI) agents’ emotional expressions (positive or neutral) interact with service task types (subjective or objective) to shape consumer service satisfaction, and explore the mediating roles of mind perception, including perceived agency and perceived experience, in this process. Design/methodology/approach This work uses four studies, including a secondary data analysis of relevant Weibo posts and three scenario experiments involving a total of 587 participants, to underline the interaction effect between AI agents’ emotional expressions and service tasks on service satisfaction, and demonstrate the dual-path mediating effect of mind perception. Findings Across four studies, the authors find that the effectiveness of AI agents’ emotional expressions depends on service task characteristics. The secondary data analysis of Weibo posts first documents consumers’ differential responses to AI emotional expressions across subjective and objective service contexts. Three scenario experiments then establish that positive emotional expressions are more effective for subjective tasks, whereas neutral emotional expressions are more effective for objective tasks. Further analyses show that these effects operate through distinct psychological pathways: positive emotions enhance perceived experience in subjective tasks, whereas neutral emotions strengthen perceived agency in objective tasks, ultimately increasing consumer satisfaction. Research limitations/implications This study, while strong in internal validity, faces external validity limitations. It focuses on AI communication in commercial service texts, neglecting other domains like education and health care, and overlooking the nonverbal emotional expressions of embodied AI. In addition, the research primarily uses service satisfaction as the sole dependent variable, thus missing a broader spectrum of consumer responses such as repurchase intentions and word-of-mouth. Practical implications This research informs managers that AI service agents should be designed as social actors with mind perception rather than as purely functional tools. Firms can leverage emotional expressions to shape consumers’ social evaluations of AI. But emotional design should align with service task types: positive emotional cues are more effective for subjective, experiential tasks, whereas neutral styles better support objective, data-driven tasks. Originality/value This study contributes to the emotional AI literature by uncovering critical situational factors that shape the effectiveness of AI emotional expressions and advances the authors’ understanding of the underlying mechanisms through a dual-path mediation framework of mind perception. The findings also provide practical implications for the strategic deployment of emotional AI in service design.