Search results

Past research has emphasised the potential for conversational artificial intelligence (AI) to disrupt services. Conversely, the literature recognises customer expectations as fundamental to service quality and customer satisfaction. However, the understanding of users’ expectations for conversational AI services is currently limited. Building upon previous research that has underscored the importance of users’ expertise, this study aims to provide valuable insights into the expectations of users with varying levels of expertise.

Forty-five semi-structured interviews were conducted, on three populations: experts, quasi-experts and non-experts from various countries including Japan, France and the USA. This includes 10 experts and 11 quasi-experts, as in professionals in conversational AI and related domains. And 25 non-experts, as in individuals without professional or advanced academic training in AI.

Findings suggest that users’ expectations depend on their expertise, how much they value human contact and why they are using these services. For instance, the higher the expertise the less anthropomorphism was stated to matter compared to technical characteristics, which could be due to a disenchantment effect. Other results include expectations shared by all users such as a need for more ethics including public interest.

The study provides insights into a key yet relatively unexplored area: it defines three major expectations categories (anthropomorphic, technical and ethical) and the associated expectations of each user groups based on expertise. To the best of the authors’ knowledge, it also highlights expectations never detected before as such in the literature such as explainability.