This research aims to explore why high-fidelity digital human avatars induce a stronger adverse response from customers than an artificial static robot interface condition following a service failure.
A total of three studies were performed. Studies 1 and 2 (n = 322, n = 282) involved a between-subjects design that was set up in the context of educational administration and aimed at investigating the effects of the main variable, the mediation mechanism and the moderation effects of the size of the failure. Study 3 (n = 205), on the other hand, adopted a post-performance only design involving adults in the banking sector.
The findings suggest that service failure by a high-fidelity digital human avatar creates a much larger drop in trust than a similar failure by a static chatbot. This effect is partially mediated by a higher experience of expectancy violation, and this effect is amplified in the case of a more serious service failure. The results were replicated in the retail banking context, and the indirect effect of perceived expectation violation was still significant while holding discomfort, perceived deception and institutional blame constant.
This research questions the dominant notion of the more human, the better in AI service design by examining the unique susceptibility of customer trust. This research recognizes service failure as an important boundary condition to the normally positive effect of designing a high fidelity digital human avatar. This research proves that if there is failure on the part of the high fidelity digital human avatar, then the higher expectations created by the avatar may create an erosion of trust.
