Rationally or emotionally: how should voice user interfaces reply to users of different genders considering user experience?

Voice user interfaces (VUIs) have exploded in popularity over the past 3 years. However, there has been little research on the reply methods that VUIs can adopt to communicate with people. In this paper, we designed 2 studies with 20 participants to explore the influence of reply methods on user experience in 2 kinds of scenarios (applicational scenarios and giving a command) when using a VUI. We explored the performance of different reply methods (fact-only, rational, and emotional) at different times and in different scenarios. In addition, we examined whether there were gender differences when evaluating a reply and different preferences for different reply methods. A “Wizard of Oz” method was used in the experiments to simulate real scenarios for communication between the participants and the VUI. We divided a reply into three parts (fact + judgment + strategy) and constructed three kinds of reply methods. In the experiments, we used quantitative scoring (five aspects: affection, confidence, naturalness, social distance, and satisfaction), preference selection and an interview to measure the participants’ user experience. The results indicated that the participants were inclined to prefer the reply methods (rational and emotional) that offered judgments and strategies in our experiment script, and the emotional style received the highest evaluation. In addition, we found that male participants tended to have a higher evaluation of VUIs’ replies for all three reply methods in applicational scenarios and when giving a command than female participants in our studies. In general, these results may contribute to the design of VUI replies.