AI-chatbots’ positive emotional expression may backfire: A perspective of emotional intensity

As emotional artificial intelligence (AI) chatbots become increasingly prevalent in customer service settings, understanding how their emotional expressions shape customer responses has become ever more important. Based on expectancy violations theory (EVT) and feelings-as-information theory (FIT), this study explores how positive emotional intensity (PEI) expressed by AI-chatbots influences customer satisfaction. Through three scenario-based experiments, this study reveals an inverted U-shaped relationship between AI-expressed PEI and customer satisfaction and demonstrates that this effect is mediated by perceived emotional expectancy violations. Moreover, customers’ own PEI significantly moderates this inverted U-shaped relationship. Specifically, when customers are in low PEI states, AI-expressed PEI exhibits an inverted U-shaped effect on their satisfaction. In contrast, when they are already in a moderate or high PEI state, AI-expressed PEI positively influences their satisfaction. These findings not only advance our understanding of human–AI emotional interaction, but also offer actionable insights for the effective deployment of emotional AI-chatbots in customer service settings.