Archives of Design Research

Copyright : This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/), which permits unrestricted educational and non-commercial use, provided the original work is properly cited.

Owing to its advanced information technology (IT) development, the South Korean market is a beta test for new technologies. The cumulative number of smart speakers sold by South Korea’s top three mobile carriers, Naver and Kakao, doubled in a year to 4.12 million units as of March 2019 (Lee, 2019). The smart speaker has drawn attention from the IT industry because of its considerable potential as a control tower for home automation; hence, many global companies are developing and distributing smart speaker technology. The virtual personal assistant (VPA) is embedded as a standard in every smart speaker using a voice user interface (VUI). It is “a software agent that provides professional administrative, technical, or social assistance to a human user”(Saad et al., 2017, p. 12518). Owing to the increase in the smart speaker penetration rate, the VPA in smart speakers is more pervasive and widely used than ever. From the perspective of advancements in AI, the VPA will gain more importance in the future. From smart home automation to self-driving, the VUI will play a vital role in human-machine communication. It is highly likely that speakers, robots, and cars with a VPA will have the biggest AI technology-related impact on us.

Research on smart speakers has increased with their increased usage. A VPA study pointed out that parasocial relationships enhance users’ satisfaction post VPA adoption (Han & Yang, 2018). Lau and his colleagues (2018) conducted a qualitative study to examine smart speaker adoption and non-adoption motivations by posing questions to 17 smart speaker users and 17 non-users. The results indicated “efficiency” and “early adopter identity” as the primary adoption motivations. Kowalczuk (2018) found that perceived enjoyment more strongly influenced users’ behavioral intention than other factors based on the structural equation modeling. Park et al. (2018) also used the TAM and confirmed that people considered platform-wise benefits more important than device-wise benefits. Consumers perceived smart speakers as an outlet connecting home automation. As these early studies share a common focus on users’ motivation to purchase smart speakers, they can help us understand the mechanism of the adoption process. Apart from the adoption reasons, there is little information on the psychological factors and interaction-related factors affecting users’ satisfaction while interacting with VPAs in daily routines. There are studies exploring smart speaker users’ interactions with smart speakers and VPAs. The Alexa users’ daily usage behavior and interaction pattern were observed more comprehensively using both history logs and in-home contextual interviews (Sciuto et al., 2018). Lopatovska and colleagues (2018) also analyzed user interaction using Alexa users’ online daily diary responses. Meanwhile, in the conext of human-computer interaction, parasocial interaction could occur in situations where human-VPA communication takes place, as discussed by Han and Yang (2018), who focused on the parasocial relationship theory. Purington and her colleagues (2017) analyzed user reviews of the Amazon Echo and the Alexa, and found that people personified the device and agent. Another user review analysis revealed this personification phenomenon along with usage behavior patterns (Gao et al., 2018). A study discussed how the elderly perceived the voice assistant of Echo in terms of anthropomorphism, and its link to their loneliness (Pradhan et al., 2019). Few studies have reported the possible relationship between IPA use and loneliness. All of these qualitative approaches and quantitative approaches using review data revealed usage patterns in general and interesting phenomena such as personification in VPA usage experience. However, these concepts have not yet been examined statistically using responses from users. Therefore, the current study focused on the effects of interactions with VPAs.

All statistical analyses were performed using IBM SPSS Statistics 24.0. To investigate the research question, correlation analysis was carried out, and the results are summarized in Table 1. According to the correlation analysis results, satisfaction was significantly positively related with parasocial interacton (r = .73, p < .01), and assistant/helper type (r = .66, p < .01), colleague type (r = .65, p < .01), friend type (r =.65, p < .01), and family member type (r = .62, p < .01). Meanwhile, satisfaction was not significantly correlated with loneliness.

The multiple linear regressions analysis was performed (R² =.60, F_{(6, 527)} = 131.42, p < .001). As shown in Table 2, parasocial interaction satisfaction positively affected satisfaction (β = .53, p < .001). Only the assistant/helper type positively influenced satisfaction (β =.29, p < .001) among the personification types. Colleague type, friend type, and family member type were not related to satisfaction. However, loneliness negatively affected satisfaction (β = -.11, p < .001). There was no multicollinearity among the independent variables.

First, people with a high level of parasocial interaction exhibited a higher level of satisfaction. This result was in agreement with Han and Yang’s (2018) research findings. In their research, parasocial relationships increased satisfaction with consumers’ IPA use. It can be assumed that when users perceive the VPA as a social being more strongly, they receive stronger satisfaction from the interaction experience. Audio modality can fully generate a sense of humanness.

Second, personification phenomena were confirmed in the case of assistants/helpers. The four different types of classification indicate different levels of intimacy with VPAs and also reflect the different social roles VPAs are perceived to assume. Contrary to previous qualitative research findings (Purington et al., 2017; Gao et al., 2018), treating VPAs as their friends or family members did not affect users’ satisfaction level. Smart speaker manufacturers have been steadily promoting their devices as “intelligent assistants” and “competent secretaries” for a long time. The operation systems are also programmed to respond to commands and gently make suggestions. Although they sound friendly, their speech patterns remains unnatural. This might be why other personification types did not have an impact on satisfaction.

Third, people who lack social connections evaluate satisfaction from human–machine communication negatively. What they really seek might be social interaction with a real person and not with a device. When people feel lonely, communicating with a synthetic voice probably makes the interaction experience more intimidating. Moreover, it may make their loneliness more salient. If the same study were conducted with elderly individuals who experience considerable loneliness, we might have obtained opposite results.

These findings provide several implications for designing VUIs. First, the current technology is not capable of initiating and maintaining conversations seemlessly. To narrow the existing gap between users’ expectations and the actual technology (Luger & Sellen, 2016), an AI-based VPA should be able to initiate and lead the conversation on their own. When VPAs can have a more natural and fluent conversation with the device, people might feel a stronger social attachment. Smart speaker manufacturers offer speech options such as a male voice or celebrity voices in addition to the default female voice. Different personalities can be generated through voices by making changes in volume, pitch, pitch rate, and speech rate (Nass & Brave, 2005). By implementing the personality that users prefer the most, the devices could become much more anthropomorphic and human-like.

Second, treating VPAs as colleagues, friends, or even family members had no observable effect on user satisfaction. Therefore, current positioning of VPAs as “intelligent assistants” and “competent secretaries” seems appropriate. When people perceived the VPA as a smart helper, they experienced more satisfaction. To develop a smarter assistant, interaction designers can learn from the protocols of business relations, providing more efficient and optimal experiences for users.

Third, people who felt lonely reported lower levels of satisfaction with the technology. Even though users personify the devices, intimacy does not directly derive from this personification. It was reported that consumers express their feelings to devices (Choi, 2019), and when in a situation of increased loneliness, people longed for more emotional conversations (An, 2019). Thus, how to make the devices more humanized is a key issue to consider. It is possible that friendly conversations and intimate responses could increase user satisfaction, which in turn would decrease loneliness. Furthermore, if a voice can be designed to detect and respond to users’ emotions carefully, it would generate more positive interaction experiences as well.

There are limitations in this study. First, even though each variable was discussed as an important concept in previous studies, to the best of the authors’ knowledge, no study has investigated the combined effect of all the variables. However, these variables can all reflect user psychological states that occur from interactions with devices. All these concepts can be understood in terms of user social relations and desires. In future research, a theoretical framework should be applied that can explain concepts in a more sophisticated way. Second, there was an assumption that the independent variables might have influenced the dependent variable; however, it is not possible to fully confirm casuality using only this cross-sectional data. This is a limitation that every study using cross-sectional data has; thus, results should be carefully interpreted. Third, as this study included users in their 20s to 40s as major target consumers, future research needs to be conducted with consumers in their 50s and older in terms of loneliness-related issues.