The Effects of Expectation Violation of AI Speakers on Expectation and Satisfaction: Through the Anthropomorphic Moderation Effect

This study investigated how interactions with artificial intelligence (AI) speakers impact users' expectations and satisfaction, with a particular focus on expectation violation (EV)s. The findings revealed that both changes in expectations and satisfaction are influenced by these violations. Specifically, when expectations were exceeded (positive EV), users experienced increased satisfaction, whereas unmet expectations (negative EV) led to disappointment, consistent with previous studies on Expectancy Violation Theory (EVT). Furthermore, our aim was to explore the role of anthropomorphism in shaping users' expectations and their confirmation of AI speakers' functions. As expected, interactions with AI speakers that surpassed expectations and exhibited more human -like traits resulted in the most positive changes in expectations and satisfaction. Conversely, encounters with AI speakers failing to meet expectations and displaying rigid, machine -like behaviors led to the most negative shifts in both metrics. However, irrespective of anthropomorphism levels, both positive and negative EVs consistently influenced changes in expectations and satisfaction within each group. Therefore, it could be argued that meeting or exceeding user expectations is a critical factor for fostering a positive user experience, regardless of the degree of anthropomorphism.