Should AI express emotion? Mitigating the negative effect of AI emotional expression on usage intention

Integrating emotional expressiveness into artificial intelligence (AI) systems has emerged to play a pivotal role in boosting user participation and human–robot interaction, propelled by recent advances in AI and deep learning. Nevertheless, the potential adverse implications of AI emotional expressions in specific operational settings remain an under-explored research area. This study introduces derived anxiety as the psychological mechanism underlying the negative impact of emotional expression on usage intention. Empirical data from four experiments across diverse industry contexts refined the theoretical framework for identifying key drivers of customer resistance to AI emotional expressions. Findings indicate that AI emotional expressions significantly heighten users' derived anxiety, thereby diminishing usage intention in specific contexts. Importantly, this effect is moderated by three critical factors: (1) relationship type, (2)relationship norm orientation, and (3) appearance anthropomorphism. These insights offer essential guidance to the principled design of AI systems, especially in establishing appropriate emotional expression protocols for customer service applications. The study contributes to both theoretical understanding and practical implementation, providing a framework for the responsible integration of affective AI while optimizing interaction effectiveness.