Hey Google, speak human to me! A quantitative experiment of human-likeness and personalization in voice assistant interaction

Abstract: Voice assistants are a new customer touchpoint for businesses, and the market for voice shopping is expected to grow substantially in upcoming years. Despite high societal and business interest, academic marketing research on consumer interaction with voice assistants is scarce. There is a critical need for application developers to understand how to best design the interaction with the voice assistants, and for brands to understand the impact of being advertised through the device. This study contributes to research by studying voice assistant interactions as a service encounter and as an advertising context. Specifically, it focuses on consumer interaction with voice assistant in a low-involvement purchase situation. Through an audio experiment using a quantitative survey method, the study compares a series of two manipulations: making the voice appear more human-like, and the dialogue more personalized. 203 answers were collected using Amazon Mechanical Turk. Results strikingly show the importance of voice human-likeness for a positive perception of the device as well as improved brand attitudes. An explanatory path was found in how users anthropomorphized the device to a greater extent and experienced a sense of social presence during the interaction. However, dialogue personalization showed no significance for either dependent variable. Finally, attitudes towards the voice assistant were found to mediate the attitudes towards the brand. Such results will be useful for voice app developers and marketers alike as they establish the benefits of a strong focus on voice human-likeness. Likewise, this thesis is a first building stone to the nascent field of voice assistant research.