Unlocking the service attractiveness of AI assistants: does multi-modal anthropomorphic interaction dynamically manipulate users' mindset metrics?

The rapid rise of in-vehicle AI assistants is providing users with a more engaging and intelligent service experience. However, what drives the attractiveness of these multi-modal anthropomorphic AI agents and how they influence users' mindset metrics (attitudes, perceptions, and intentions), remains unclear. This study aims to identify the service attractiveness components of AI assistants using big data text analysis techniques, and reveal their manipulative effects on users' mindset metrics (satisfaction, perceived service quality and brand liking) in a time-considered dynamic perspective. To this end, 4,584 valid users' reviews from Chinese car evaluation websites have been collected and analyzed. Conclusions show that: (1) the service attractiveness of AI assistants consists of Functional attribute (service process and service outcome), Relational attribute (sociability and friendliness) and Physical attribute (human-likeness and multi-modal); (2) All three attributes can significantly and positively manipulate users' mindset metrics; (3) Car usage time exerts a differential impact on the positive manipulation of the three attributes of service attractiveness. In addition, the potential causes of the differential impact have been explored by constructing structural topic models to identify real-time users' concern about service attractiveness in review texts. Our research provides new insights into the service attractiveness enhancement of AI assistants.