Voice Assistant

International Journal of Information Management xxx (xxxx) xxx Contents lists available at ScienceDirect International Journal of Information Management journal homepage: www.elsevier.com/locate/ijinfomgt Research Article Psychological determinants of users’ adoption and word-of-mouth recommendations of smart voice assistants Anubhav Mishra a,*, Anuja Shukla b, Sujeet Kumar Sharma c a Jaipuria Institute of Management, Lucknow, Vineet Khand, Lucknow, UP 226010, India b upGrad Education Pvt. Ltd., India c Information Systems & Analytics Area, Indian Institute of Management Tiruchirappalli, India ARTICLE INFO ABSTRACT Keywords: Individuals are increasingly using Smart Voice Assistants (SVA) to derive functional, hedonic, and symbolic Smart voice assistants benefits. SVA adoption is in the nascent stage and we have little knowledge about what drives SVA usage. Technology adoption Following the ‘attitude shapes behavior’ approach, this study examines the role of hedonic and utilitarian atti­ Artificial intelligence tudes on SVA usage and word-of-mouth (WOM) recommendations. The study also investigates five antecedents Hedonic attitude (playfulness, escapism, anthropomorphism, visual appeal, and social presence) to both attitudes. Through an Utilitarian attitude online survey of 360 respondents, the study suggests that playfulness and escapism positively influence hedonic Anthropomorphism attitude. On the other hand, anthropomorphism, visual appeal, and social presence determine utilitarian attitude. Further, utilitarian attitude has a stronger impact (vs. hedonic attitude) on SVA usage and WOM recommen­ dations. The findings reveal that individuals who perceive SVA as a symbol of high prestige (vs. low prestige) are less likely to relate playfulness to hedonic gratifications. In contrast, people attributing prestige to SVA are more likely to use it as an escape route from everyday life. These findings contribute to the growing literature on SVA adoption and offer insightful recommendations to various stakeholders to increase the likelihood of SVA adoption and generating favorable WOM recommendations. 1. Introduction to COVID-19 because most users prefer ‘talking’ over ‘touching’ to avoid physical touch and maintain social distancing. Mirror, mirror on the wall, who’s the fairest of them all? Smart speakers and voice assistants have drastically changed the In the famous story of Snow White, the evil queen regularly performs consumer decision journey, especially how users perform voice searches a ‘voice search’ about the fairest person on the earth. The magical mirror and make a purchase using SVA (Dwivedi & Ismagilova, 2021; Kush­ answers her questions similar to the Smart Voice Assistants (SVA) of waha & Kar, 2020). Many users use SVA as part of their daily routine for today’s world. SVA are Artificial Intelligence (AI)-based technology doing a voice search, multitasking, and shopping (Google Insights, including software and hardware (e.g., Alexa, Google Home, and Siri) 2017). The increasing internet penetration across the world (approxi­ that can communicate with users and respond to their queries or com­ mately 54%) promises a massive market for the new technologies, yet mands (Verma, Sharma, Deb, & Maitra, 2021). SVA are the latest en­ the use of voice assistants is more on the phone (58%) than on a smart trants to the ever-changing and growing field of human-technology speaker (22.9%, PWC, 2019). Therefore, the adoption of SVA has sub­ interactions (Bag, Pretorius, Gupta, & Dwivedi, 2021; Balakrishnan & stantial implications for marketers who need more profound insights Dwivedi, 2021a, 2021b; Belk, 2017; Grover, Kar, & Dwivedi, 2020; into the users’ usage of SVA to tailor their offerings (Dwivedi, Ismagi­ Kushwaha, Kar, & Dwivedi, 2021). SVA sales hit a record high in 2019 lova, Rana, & Raman, 2020; Dwivedi, Hughes, et al., 2021). (146.9 million units) and the growth continued in 2020 despite the COVID-19 impact on supply chain and retail distribution (Swoboda, Extant research on SVA adoption is mostly limited to existing the­ 2020). Furthermore, SVA usage is expected to substantially increase due ories and frameworks of general technology adoption, such as Tech­ nology Acceptance Model (TAM, Davis, 1989; Chatterjee & Kar, 2020; Fernandes & Oliveira, 2021), Unified Theory of Acceptance and Use of * Corresponding author. E-mail addresses: [email protected], [email protected] (A. Mishra), [email protected] (A. Shukla), [email protected] (S.K. Sharma). https://doi.org/10.1016/j.ijinfomgt.2021.102413 Received 19 April 2021; Received in revised form 7 August 2021; Accepted 7 August 2021 0268-4012/© 2021 Elsevier Ltd. All rights reserved. Please cite this article as: Anubhav Mishra, International Journal of Information Management, https://doi.org/10.1016/j.ijinfomgt.2021.102413

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx Technology (UTAUT, Venkatesh, Morris, Davis, & Davis, 2003; Vimal­ and personalized service to enhance users’ attitudes and likelihood for kumar, Sharma, Singh, & Dwivedi, 2021) and Theory of Planned favorable WOM recommendations. Behavior (TPB, Ajzen, 1991). SVA technology appears to serve a different purpose to other technologies such as mobile banking as SVA The remaining sections of this paper are structured as follows: The are used in more personal or private settings for immediate personal following Section 2 reviews existing research on the role of hedonic and gratifications (Lee, Lee, & Sheehan, 2020). Hence, future research on utilitarian attitudes in technology usage and WOM behavior, and ante­ SVA need to look beyond the concepts examined using the traditional cedents to both attitudes. Section 3 describes the conceptual model and adoption models. Recent research suggests that attitude is a critical presents hypotheses. Section 4 explains the research methodology, fol­ antecedent to technology usage and adoption behavior (Dwivedi, Rana, lowed by results in Section 5. Next Section 6 presents discussion, theo­ Jeyaraj, Clement, & Williams, 2019). Furthermore, identification of retical contributions, implications for practice, and limitations. Finally, factors that attribute to a favorable attitude toward technology is an the paper concludes with Section 7. underexplored research area in the context of SVA. There are specific calls for research to explore how brands can leverage emerging tech­ 2. Literature review nologies like SVA to influence consumer engagement behaviors (Dwi­ vedi & Ismagilova, 2021, p. 8). Similarly, McLean and Osei-Frimpong Rapid technological advancements have led to the development of (2019) call for future research on factors that influence SVA’s utilitarian many innovative products that fulfill the need of consumers of modern benefits. era. Every decade sees a radical change in how humans interact with technology. The journey of human-computer interactions encompasses SVA work as standalone devices delivering a spectrum of functional, the evolution from desktop to the Internet, touch-screens, SVA, and to hedonic, and symbolic benefits to users (Mishra & Shukla, 2020). SVA blockchain technology (Hughes et al., 2019; Lee et al., 2020). The topic technologies considerably differ from existing technologies as they offer of technology adoption has attracted many academicians and market almost humanlike interaction experiences (Fernandes & Oliveira, 2021). practitioners, who have used behavioral and cognitive theories to pro­ SVA have design concepts that predispose users to personify and interact pose various models and frameworks (e.g., Balakrishnan & Dwivedi, like conversing with a human being (Foehr & Germelmann, 2020). SVA 2021a; Hernandez-Ortega & Ferreira, 2021). It is important that the provide users an assurance of someone being present (social presence) in existing theoretical models should be updated to meet the challenges of their vicinity, who can empathize and follow their commands. More­ changing consumer behavior and preferences due to advances in tech­ over, users regard SVA as friends and communicate with SVA to deal nology (Dwivedi et al., 2019; Williams, Rana, & Dwivedi, 2015). with their loneliness or to have fun (Park, Kwak, Lee, & Ahn, 2018). Thus, users personify SVA and this behavior of assigning humanlike 2.1. Motivations and technology adoption traits to objects is known as anthropomorphism (Epley, Waytz, & Cacioppo, 2007). Users’ interactions with smart objects lead to devel­ Most of the research on technology adoption stems from behavioral oping humanlike relationships, which remains an unexplored research sciences and offers frameworks specific to the utilitarian perspective. area in the context of SVA adoption (Novak & Hoffman, 2019). Hence, For example, TAM builds on the Theory of Planned Behavior to inves­ this research specifically focuses on the anthropomorphism and psy­ tigate how perceived usefulness and ease-of-use determine attitude to­ chological perspective of gratifications derived from SVA devices. ward technology and subsequent behavioral intentions to use the technology (Davis, 1989). UTAUT is another widely used model to Users’ intentions to use a product, service, or technology greatly explain technology use and adoption (Venkatesh et al., 2003; Williams depend on the word-of-mouth (WOM) recommendations received from et al., 2015). The updated version UTAUT2 incorporates hedonic others (Mishra, Maheswarappa, & Colby, 2018). WOM plays a critical motivation to understand users’ motives to adopt a technology (Ven­ role toward the success of products and services via network effect. A katesh, Thong, & Xu, 2012). Similarly, TAM model has been updated to positive WOM motivates users to try and use new technology and can a third version (TAM3) that includes elements of experience, enjoyment, turn users into advocates (Tamilmani, Rana, Nunkoo, Raghavan, & and playfulness (Pillai, Sivathanu, & Dwivedi, 2020; Venkatesh & Bala, Dwivedi, 2020; Vilpponen, Winter, & Sundqvist, 2006). Individuals’ 2008). WOM behavior depends on their experiences and gratifications derived from the use of products or services (Berger, 2014). Prior research Extant research in technology adoption broadly uses two types of confirms the positive influence of hedonic motivations (e.g., joy and fun motivations - utilitarian-motivation and hedonic-motivation (Agarwal & of using technology) in adopting technology like mobile banking Karahanna, 2000; Lee et al., 2020). Utilitarian motives include the (Alalwan, Dwivedi, & Rana, 2017). However, the impact of hedonic functional aspects of technology such as performance, usefulness, and motives on WOM behavior, which can significantly influence further ease of use (e.g., TAM, Davis, 1989). In contrast, the second category technology adoption, is yet to be ascertained. Therefore, the main ob­ focuses on hedonic intentions such as enjoyment, immersion, and flow jectives of this research are: (e.g., UTAUT2, Tamilmani, Rana, Prakasam, & Dwivedi, 2019; Ven­ katesh et al., 2012). Moreover, SVA may lead to the state of cognitive • To examine the influence of hedonic and utilitarian attitudes on SVA absorption because they offer temporal escapism and heightened usage and WOM recommendations. enjoyment by addressing users’ curiosity (Liao, Vitak, Kumar, Zimmer, & Kritikos, 2019). A recent meta-analysis by Tamilmani et al. (2019) • To identify the effects of psychological antecedents to hedonic and reaffirms the crucial role of hedonic motivation in technology adoption. utilitarian attitudes. In fact, researchers have integrated the TAM model with flow theory to ascertain the role of immersion and interaction in human-technology The present study contributes to the extensive research on technol­ interactions such as online gaming (Hsu & Lu, 2004). Flow theory is a ogy adoption, especially in the area of emergent smart technologies. The psychological concept that explains individuals’ mental state when they findings indicate the importance of utilitarian and hedonic attitudes are completely immersed in an activity with extreme focus and enjoy­ toward users’ intentions to adopt SVA and share WOM recommenda­ ment (Nakamura & Csikszentmihalyi, 2009). For example, users can be tions. The results suggest that SVA provide many benefits to users such in a flow state while playing online games or online impulsive buying as playfulness, escapism, social presence, and, most importantly, as a (Wu, Chiu, & Chen, 2020). Therefore, flow state significantly influences humanlike companion. SVA usage depends on their esthetic appeal and users’ attitudes and behavioral intentions. figurative personification created by users. The findings provide actionable strategies to the various stakeholders who can tailor their offerings specific to users interacting with SVA. For example, manu­ facturers and marketers must emphasize esthetic, playful experience, 2

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx 2.2. SVA and multisensory experiences Germelmann, 2020). Hence, social influence can impact users’ SVA buying and usage intentions. One stream of research treats computers as SVA significantly differ from traditional technology like computers, social actors (CASA) to describe or explain the social interactions with laptop, or mobile banking in terms of usage situation, sensory experi­ computers (Kim & Sundar, 2012). In addition, people build relationships ences, and gratifications (Hoffman & Novak, 2018). Users generally do and emotional bonds with SVA during the regular interactions and not use SVA for increasing productivity in an organization, but they use discuss the experiences with others in the form of WOM behavior (Hur, SVA as a personal device in a relatively more private location (e.g., Koo, & Hofmann, 2015). Users show feelings of love, commitment, and home). SVA deliver a highly interactive experience where users can ask intimacy toward SVA as they do for fellow humans (Hernandez-Ortega SVA to do functional (e.g., setting an appointment or a reminder) or & Ferreira, 2021). One may easily notice how individuals proudly talk hedonic (e.g., telling a joke or playing a song) tasks (Foroudi, Gupta, about their emotional bonding with cars or watches. Moreover, SVA fill Sivarajah, & Broderick, 2018; Lopatovska et al., 2019). Furthermore, a conversational space and enable emotional venting, which are critical SVA provide a multisensory experience (esthetic and auditory) similar to antecedents to WOM behavior (Berger, 2014; Ismagilova, Rana, Slade, & emergent multisensory technologies like virtual reality and augmented Dwivedi, 2020). Therefore, we notice that users’ attitude toward SVA reality. Thus, we borrow from relevant research on multisensory tech­ and their usage depends on multiple factors. Hence, we consider the nologies to examine the importance of multisensory experiences (e.g., critical aspects of SVA usage like psychological gratifications to address playfulness and escapism) in SVA usage and WOM recommendations hedonic motivations, and anthropomorphism, to offer valuable insights (Mishra, Shukla, Rana, & Dwivedi, 2021). into users’ motivations to use SVA and WOM recommendations (Mishra & Shukla, 2020). 2.3. SVA and anthropomorphism 3. Conceptual model and research hypotheses Users consider SVA as humans like a friend or a companion (Poushneh, 2021). For ages, humans are anthropomorphizing objects A majority of human-technology interactions are gradually shifting and animals around them. Many examples exist in the mythological to SVA such as Alexa and Google Home. Many users use SVA for fun and literature, famous fables and stories (e.g., The Jungle Book and Alice in hedonic motives rather than only utilitarian benefits. Our study explores Wonderland), and movies (e.g., humanoids and R2D2 in Star Wars experiential and design factors that influence SVA usage and WOM movies). Similarly, brands utilize anthropomorphism to spread aware­ recommendations for SVA. To adequately address the gratifications ness and recognition (Guido & Peluso, 2015). Users can instantly derived from using SVA, the proposed model in our research (see Fig. 1) recognize and recall the associated brands, for example, Mickey Mouse builds upon two theories from the psychology area: flow theory and the (Disney), Cheetah (Cheeto), McDonald’s clown, and Michelin man. theory of anthropomorphism. The flow theory is used to capture the Therefore, we believe that the way users perceive SVA as humans will be immersive experiences of playfulness and escapism derived from using a critical factor for developing users’ attitudes and usage of SVA. SVA. The anthropomorphism concept explains users’ perceptions of SVA as humans and companions. We examine the role of various psycho­ 2.4. SVA and social influence logical antecedent variables that significantly influence users’ hedonic and utilitarian attitudes, which in turn affect SVA usage and WOM SVA deliver symbolic benefits of social prestige (McLean & recommendations. Osei-Frimpong, 2019). Many users consider SVA as a status symbol and use SVA to enhance their self-image among their social groups (Foehr & Fig. 1. Conceptual model. 3

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx 3.1. Hedonic attitude 3.3. Playfulness Users evaluate any product, service, or technology on two di­ Playfulness is defined as a degree of cognitive spontaneity in human- mensions: (1) hedonic or gratification value that is affective in nature, computer interactions (Webster & Martocchio, 1992). People interact and (2) utilitarian or instrumental value that is cognitive-intensive with technology to relish the experience. For example, users spend (Voss, Spangenberg, & Grohmann, 2003). Extant research suggests considerable time on mobile games to enjoy the thrills of racing, that attitude toward technology is a critical antecedent to technology adventurous journeys or missions, or building online empires. Users find usage and adoption behavior (Dwivedi et al., 2019). SVA offer users an element of playfulness in SVA interactions (Chatterjee, Kar, & Dwi­ enjoyment and entertainment by answering user’s questions in humor­ vedi, 2021; Lee et al., 2020). Most of the queries posted to SVA (e.g., ous ways. People find novelty in using personal assistants leading to Alexa) are functional in nature. Many users pose funny questions to SVA continuous usage (Gursoy, Chi, Lu, & Nunkoo, 2019). SVA provide an such as enquiring the marital status or asking to make a sandwich and element of playfulness and visually appealing esthetics, which satisfy get interesting humorous replies making the interaction amusing and the hedonic intrinsic experiential values (Mathwick, Malhotra, & Rig­ highly enjoyable. Playfulness stems from curiosity and enjoyment and don, 2001). A hedonic attitude increases the likelihood of trying new significantly impacts users’ attitude toward the use of technology (Ahn, technology such as eStores or mobile banking (Alalwan et al., 2017; Wu Ryu, & Han, 2007). Hence, we argue that users will actively use SVA for et al., 2020). Thus, we propose the following hypothesis: fun and enjoyment. Thus, we propose that: H1a. Hedonic attitude positively influences users’ SVA usage. H3. Playfulness positively influences users’ hedonic attitudes. When individuals derive psychological gratifications and satisfaction 3.4. Escapism from products and services, they are more likely to share their experi­ ences with others (Ismagilova et al., 2020). WOM behavior is a form of Escapism refers to people’s tendency to “get away from it all”. self-expression, where people share information to fulfill hedonic mo­ Escapism is defined as a state of psychological immersion in which a tives. Many people share their views related to experiential products like person is completely involved in the focal activity (Mathwick & Rigdon, much-awaited movies or highly rated restaurants immediately on online 2004). Individuals use escapism as a coping strategy to handle an un­ platforms (Mishra & Satish, 2016). Since SVA provide users an affective pleasant emotional state. Escapism offers individuals momentary satis­ experience, we expect users to display a similar WOM behavior. More­ faction and fulfillment of their needs. For example, people engage in over, SVA are considered as status symbols that enhance users’ recreational or leisure activities like shopping, consuming products (e. self-image among their social network. People display feelings of pride g., ice-cream), or binge-watching TV or web series as an escape route to by owning such advanced devices to influence others. Research suggests their routine life (Jones, Cronin, & Piacentini, 2018). that social influence and altruism are critical drivers for WOM activities (Berger, 2014). Thus, we propose the following hypothesis: Escapism as a concept has been widely used in research to explain users’ intensive engagement with virtual and highly interactive envi­ H1b. Hedonic attitude positively influences users’ WOM ronments like massively multiplayer online games (e.g., World of War­ recommendations. craft) and virtual worlds (e.g., Second Life, Kuo, Lutz, & Hiler, 2016). Multisensory technologies (e.g., augmented reality, virtual reality, and 3.2. Utilitarian attitude mixed reality) also offer an immersive experience leading to increased involvement with the environment. For example, the tourism industry Individuals use SVA to fulfill their functional needs, for example, uses virtual reality to connect with potential consumers and promote people regularly use Alexa for mundane tasks like setting up an alarm, activities like scuba diving in a highly immersive environment (Mishra scheduling meetings or appointments, and shopping. SVA reduce users’ et al., 2021). Such vivid experiences fulfill consumers’ desire for fantasy cognitive load by offering convenience. SVA are designed to learn and and novel experiences as evident from the growing usage of 3D visuals in adapt to the instrumental transactions and repeat them as desired by movies. In addition, many users take help of SVA to break the monot­ users at pre-configured intervals (McLean & Osei-Frimpong, 2019). onous routine by asking SVA to tell a joke or play entertaining music Many users use SVA extensively for seeking information on a variety of (Conner & Bradford, 2020). Thus, SVA perform the role of an avid topics like news, weather, products, sports scores, and cooking recipes. listener and entertainer who answers to the commands of its master. Recent research suggests that users will increasingly use SVA due to Hence, we propose the following hypothesis: convenience and competence offered by these personal digital assistants (Hu, Lu, Pan, Gong, & Yang, 2021). Thus, SVA usage becomes a habit for H4. Escapism positively influences users’ hedonic attitudes. many users to accomplish routine tasks. So, we hypothesize the following: 3.5. Anthropomorphism H2a. Utilitarian attitude positively influences users’ SVA usage. Anthropomorphism reflects human tendencies to label SVA as humans and to seek emotional support as a companion. Users consider Users with a favorable attitude toward SVA are more likely to use SVA as a child learning the tricks of communicating with the owner SVA to purchase and interact with customer support to resolve queries (Karimova & Goby, 2020). Recent research on personal assistants sug­ (Moriuchi, 2019). These consumers offer favorable recommendations of gests that anthropomorphism is a significant factor in human-SVA in­ firms that integrate SVA in their websites to enable direct communica­ teractions. For example, Hu et al. (2021) suggest that users evaluate SVA tion (Moriuchi, 2019). Moreover, consumers share novel, positive, and on two dimensions: warmth and competence. Warmth echoes the entertaining experiences online (Berger & Milkman, 2012). SVA in­ emotional benefits of SVA like benevolence, sociability, and caring. teractions deliver a highly interactive and entertaining experience to Users develop an emotional attachment with SVA due to persistent use. users. In a recent meta-analysis on eWOM behavior, Ismagilova et al. In fact, users enjoy communicating with SVA and humanize them as (2020) propose that altruism and homophily motivate people to share companions (Poushneh, 2021). WOM. We argue that users will share recommendations for SVA with others due to high levels of utilitarian value and convenience. Hence, we COVID-19 pandemic forced millions of people across the world to propose the following hypothesis: live in isolation under various restrictions on movement. Many people suffered a severe impact on mental well-being because of an unusually H2b. Utilitarian attitude positively influences users’ WOM long period of home confinement and lack of social gatherings and recommendations. celebrations (Lovett, 2020). In these tough times, SVA has helped people 4

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx to improve emotional and mental well-being and fight anxiety and the last decade, users have seen a drastic decrease in phones’ size with depression. People feel lonely when they do not have anyone to talk to, significant improvements in performance and features. Hence, SVA’s and in such cases, SVA played the role of users’ companion (or friend) to compact and symmetrical size can influence users’ perceptions of effi­ overcome people’s loneliness at home (Poushneh, 2021). In a ciency and utility of such devices. Thus, we hypothesize that: work-from-home culture, people start their home office by greeting SVA (Foehr & Germelmann, 2020). SVA’s critical role to fill the emotional H6. Visual appeal positively influences users’ utilitarian attitudes. void and loneliness of users is exemplified during the lockdown in 2020 when millions of people proposed to Alexa. Thus, SVA are emotionally 3.7. Social presence superior to other technologies in fulfilling hedonic needs and hence we posit the following hypothesis: Social presence reflects the sense or feeling of being with others while having a communication. In a technology-mediated interaction, H5a. Anthropomorphism positively influences users’ hedonic social presence is defined as “the degree to which users can feel others’ attitudes. presence in the result of interpersonal interactions during the commu­ nication process” (Walther, 1992, p. 54). Social presence encompasses According to Hu et al. (2021), competence (the second dimension to physical and psychological proximity to users (Roy, Singh, Hope, evaluate SVA) reflects the utility of SVA such as their capability to Nguyen, & Harrigan, 2019). Thus, the social presence of SVA can be complete the tasks efficiently as required by users. The UTAUT model explained as their ability to engage with users and to establish physical proposes two antecedents to users’ technology adoption - performance existence. Social presence is a critical factor that affects individuals’ expectancy and effort expectancy (Venkatesh et al., 2012). Performance affective and cognitive commitments toward their relationships with a expectancy of SVA measures relative performance compared to humans brand (Dabholkar, van Dolen, & de Ruyter, 2009). For example, when on parameters of accuracy, consistency, and errors. On the other hand, users interact with others on Twitter, a higher degree of social presence effort expectancy shows the degree of ease and learning curve required leads to better fulfillment of social connection needs (Han, Min, & Lee, to use SVA. Both antecedents influence users’ attitudes toward tech­ 2016). Likewise, when users feel an increased social presence in their nology leading to usage intentions (Rana, Dwivedi, Williams, & Weer­ interactions with smart technologies, they report higher levels of com­ akkody, 2016). Recent research on SVA suggests that fort, experience, and emotional satisfaction (Fernandes & Oliveira, anthropomorphism is significantly related to effort expectancy (Gursoy 2021). SVA offer an interactive medium where users can have two-way et al., 2019). Moreover, users look for functional and emotional communications. Hence, SVA’s social presence offers functional benefits compatibility with SVA (Karimova & Goby, 2020). Therefore, users as a mechanism to beat users’ loneliness (Savage, 2020) or to have fun anthropomorphize SVA as personal assistant or secretary to accomplish (Foehr & Germelmann, 2020). Hence, we hypothesize that: routine tasks (e.g., scheduling meetings and setting up reminders). Thus, we hypothesize that: H7. Social presence positively influences users’ utilitarian attitudes. H5b. Anthropomorphism positively influences users’ utilitarian 3.8. Moderating role of prestige attitudes. Prior research confirms the critical role of social influence in tech­ 3.6. Visual appeal nology adoption. The widely used UTAUT model included social influ­ ence as one of the four core constructs that can motivate individuals to Visual appeal echoes the beautiful design or physical attractiveness adopt and use technology (Venkatesh et al., 2003). Social influence re­ of a product or interface. The importance of visual appeal or esthetics is flects the perceptions of others about technology, gadgets, and devices. well established in IS (information systems) literature (Phan, 2019). For Individuals feel that what they wear or use reflects their social image as example, a visually appealing website interface offers higher experien­ conspicuous consumption. People believe that using emerging technol­ tial value to users leading to higher consumer engagement and purchase ogies like smartwatches, smart glasses, or playing augmented intentions (Mathwick et al., 2001). From a multisensory perspective, reality-based games can enhance their social image (Rauschnabel, esthetics manipulate the visual sense of users. Mostly, users’ first Rossmann, & tom Dieck, 2017). SVA offer symbolic benefits of a positive interaction with a product happens via eyes, and hence the brand logo image and work as status symbols (McLean & Osei-Frimpong, 2019). and packaging become a vital factor in influencing consumers’ percep­ Many people buy luxury products or high-end technology products (e.g., tions and purchase behavior (Krishna & Schwarz, 2014). While most iPhones) as these products fulfill their hedonic needs by signaling their research links esthetics to hedonic motivations of consumers, we find a high prestige. Hence, prestige affects users’ hedonic attitudes. handful of research exploring how visual appeal influences consumers’ perceptions of utility or functional performance of products (Mishra The uses and gratification theory proposes that individuals actively et al., 2021). For example, the concept of functional design stresses the seek media to satisfy their specific needs. Users are more likely to use importance of design that enhances the product’s function (e.g., SVA as a status symbol due to various gratifications obtained from its screw-driver). use. For example, such individuals may give more importance to SVA features such as playfulness, companionship, or an alternative to the Artificial intelligence and smart voice recognition technologies are mundane world to justify their use. Users who use SVA for fun will not physical products (Verma et al., 2021). Hence whether they need a develop a favorable hedonic attitude toward SVA believing that SVA physical body or not is an interesting question in itself. The theory of enhances their prestige. Users will spend more time with SVA to seek a embodied cognition suggests that cognitive processing depends on the higher status among their network. They may present it as a showpiece physical aspects of the objects, such as symmetry, shape, and colors to others to enhance their status. Thus, we hypothesize: (Krishna & Schwarz, 2014). Users consider the working of an iPhone ‘smooth’, which resembles the ‘smooth’ design of the phone itself. Most H8a. Prestige positively moderates the relationship between playful­ product designs (e.g., cars) have curves instead of corners to convey a ness and hedonic attitude. smooth ride. Similarly, users perceive a heavy and enormous-sized product as more robust and durable. A very famous example is Recent research on emergent technologies offering multisensory Titanic, where a fourth dummy smoke funnel was added to enhance experiences to users clearly outlines the importance of capabilities of symmetrical esthetics. Therefore, in this research, we focus on the devices to take users to an imaginary fantasy world (Mishra et al., 2021; functional value of visual appeal of SVA. Users may perceive that a Tamilmani et al., 2019). A common perception about prestige is that pleasant and innovative design delivers functional benefits efficiently. In products signal social prestige based on premium prices or scarcity. However, psychological gratifications like escapism are new parameters 5

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx that influence prestige (Holmqvist, Ruiz, & Pen˜aloza, 2020). Therefore, measures and have similar predictive accuracy. Moreover, single-item people who perceive SVA as a prestigious symbol may use them to measures have been used in similar research when the construct attri­ escape from everyday life. For example, they can experience a fantasy butes are concrete and unambiguous (e.g., Barnes, 2021). Since, usage is world where owning SVA elevates them to higher ranks of society a construct that can be easily and uniformly understood by respondents, (McLean & Osei-Frimpong, 2019). Whereas people who do not relate use of single-item for measurement is appropriate (Bergkvist & Rossiter, SVA to social status may use SVA primarily for utilitarian and realistic 2007). A 5-point Likert scale was used following the recent SVA research tasks. Hence, we hypothesize that: (e.g., Balakrishnan & Dwivedi, 2021a, 2021b; Gursoy et al., 2019). H8b. Prestige positively moderates the relationship between escapism 4.3. Common method variance (CMV) and hedonic attitude. Common method variance (CMV) is a source of bias in a survey People try to acquire and own products that are scarce and novel to method. CMV may happen because each respondent provides responses enhance their prestige. For example, people possess a rare piece of art or for both predictor and dependent variables (Podsakoff, MacKenzie, Lee, antiqueue that is considered prestigious. People treat their luxury cars as & Podsakoff, 2003). We applied recommended procedural steps during their companions or friends and assign names to them while proudly the survey design and administration process to handle the issue of CMV. discussing with their friends. Similarly, the ownership of SVA creates an As a preemptive approach, participants were assured anonymity and endowment effect, where people value SVA more on the dimensions of confidentiality. As a post hoc approach, marker-variable technique was competence and warmth (Hu et al., 2021). People treat SVA as a pres­ performed for CMV validity analysis. The results indicated that the tigious friend who helps them in their needs and accompanies them to difference between the original and CMV-adjusted correlations were beat their loneliness (Savage, 2020). Thus, people who believe that very small (≤0.07) for all the relevant constructs (Lindell & Whitney, owning SVA is prestigious are more likely to humanize SVA. Thus, we 2001). Hence, CMV does not seriously distort the results and predictions propose the following hypothesis: in this study. H8c. Prestige positively moderates the relationship between anthro­ 5. Results pomorphism and hedonic attitude. Descriptive analysis of demographic data is given in Table 1. The 4. Methodology final sample had 56.4% male, mostly between 26 and 35 years of age (38.1%), and working (63.3%). Most of the participants were from the 4.1. Sample and data collection upper middle class income group having a monthly income between INR 50,000–0.1 million (1 USD = 75 INR approx.) and were most likely to An online questionnaire-based survey was used to collect the data for afford SVA. Thus, the sample profile is appropriate to collect data for the this research. The survey approach was used because it allows gener­ study specific to the context of SVA adoption. alizability of outcomes, replicability of findings, and simultaneous evaluation of multiple factors (Bawack, Wamba, & Carillo, 2021). The The data was analyzed using structural equation modeling (SEM) survey method is a well-established and widely used method in the IS based on partial least squares (PLS) approach with SmartPLS 3.3 soft­ positivist research domain enabling researchers to reliably assess their ware (Ringle, Wende, & Becker, 2015). We selected the PLS-SEM predictive theories and research models (Straub, Boudreau, & Gefen, approach for the following reasons. This approach is widely used in 2004). Online surveys have been used in recent research related to recent research (e.g., Dwivedi, Hughes, et al., 2021; Hu et al., 2021) and AI-based artifacts and online impulsive shopping (e.g., Hu et al., 2021; it is based on component-based structural equation modeling (Hair, Wu et al., 2020). Ringle, & Sarstedt, 2011). Moreover, PLS is recommended for prediction-based models that focus on identifying the key predictor or The questionnaire had links to two online YouTube videos about SVA driver constructs (Hair et al., 2011), which aligns with the research to provide more information on SVA to participants. The survey link was objectives of this study. distributed to a diverse population including executives pursuing an executive MBA on an online learning platform, researchers, and aca­ 5.1. Measurement model demicians. Recent research in the domain of AI-based voice assistants has included student samples (e.g., Balakrishnan & Dwivedi, 2021b; We computed Cronbach’s alpha and composite reliability (CR) to Gursoy et al., 2019). The sample included executives who significantly assess the reliability of the research model. The values for both are differ from the typical university student sample used in research. higher than the recommended values of 0.7 (see Table 2). All the item Participation in the survey was voluntary and a total of 428 responses loadings were significant and more than 0.7. The AVE (average variance were received. Out of these, 68 responses were removed due to less than 10% completion of survey, resulting in the final sample size of 360. 4.2. Measures Table 1 Sample characteristics with sample size = 360. All measures in this study were adopted from the existing literature with minor modifications to reflect the SVA context (see Appendix A for Category Sub Category Frequency Percent detailed measurement items). Items were modified and verified with ten % research experts in the similar domain. Hedonic and utilitarian attitudes Gender Male 203 were measured using bipolar items. All the other constructs were Female 156 56.4 measured by multiple items (except usage) using a 5-point Likert scale Age (years) Prefer not to say 43.3 with suitable ranges (e.g., “strongly disagree” to “strongly agree”, 18–25 1 “never” to “always”, and “unlikely” to “likely”). One of the research Employment status 26–35 85 0.3 objectives of this study was to understand SVA usage, which was Monthly household income 36–45 137 23.6 measured using frequency with a single-item statement based on > 45 128 38.1 McLean and Osei-Frimpong (2019). Single items may have certain lim­ (INR) Not working 10 35.6 itations, however, for SEM context, Bergkvist and Rossiter (2007) sug­ Working 132 gest that single-item measures are equally valid as multiple-item Less than 25,000 228 2.7 25,000–50,000 46 36.7 50,000–0.1 million 109 63.3 More than 0.1 million 133 12.7 72 30.3 37.0 20.0 6

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx Table 2 Reliability and validity indices. Constructs and indicators Mean SD Factor Loadings Cronbach’s Alpha CR AVE Anthropomorphism 3.48 0.894 0.901 0.938 0.834 Anthro1 3.53 0.91 0.896 0.934 0.826 Anthro2 3.55 0.93 0.927 0.889 0.931 0.818 Anthro3 0.67 0.918 0.906 0.941 0.842 Escapism 2.98 0.888 0.93 0.816 Esc1 2.93 0.939 0.889 0.931 0.818 Esc2 2.80 0.82 0.88 0.926 0.807 Esc3 0.87 0.935 0.888 0.93 0.837 Social presence 3.28 0.83 0.849 Pres1 3.38 Pres2 3.21 0.901 Pres3 0.93 Playfulness 3.68 0.92 0.904 Play1 3.72 0.88 0.909 Play2 3.21 Play3 0.888 Visual appeal 3.71 0.97 Vis1 3.64 0.77 0.950 Vis2 3.64 0.98 0.914 Vis3 Hedonic attitude 3.90 0.921 Hedo1 4.01 0.88 Hedo2 3.86 0.99 0.926 Hedo3 0.87 0.863 Utilitarian attitude 3.96 Uti1 4.08 0.889 Uti2 4.04 0.93 Uti3 0.96 0.918 Word of mouth 3.85 0.99 0.906 WOM1 3.76 WOM2 3.63 0.921 WOM3 0.96 0.94 0.902 0.87 0.871 0.901 1.01 0.98 0.934 0.94 0.910 extracted) values were more than the suggested value of 0.50 indicating acceptable fit. Next, we examined adjusted R2, which shows the variance convergent validity (Hair, Risher, Sarstedt, & Ringle, 2019). Discrimi­ explained by the model that defines the quality of the overall model in nant validity was verified using two methods. First, the square root of PLS-SEM (Hair et al., 2019). Henseler and Sarstedt (2013) consider R2 each construct’s AVE was higher than its correlation with another values of 0.67, 0.33, and 0.19 as substantial, moderate, and weak, construct (see Table 3, Fornell & Larcker, 1981). Second, we checked for respectively. In our model, the R2 values for various constructs were as HTMT (Heterotrait-Monotrait Ratio of Correlations) values for estab­ follows: Hedonic attitude (0.43, moderate), utilitarian attitude (0.69, lishing discriminant validity. The HTMT values for all the constructs substantial), usage (0.53, moderate), and WOM (0.37, moderate). Next, were less than the recommended value of 0.85 (Henseler & Sarstedt, we used the blindfolding process (with omission distance D set to 7) and 2013, Table 4). the PLS Predict (by setting the number of folds k = 10) to assess the predictive relevance of the model (Hair et al., 2019). The resulting Q2 5.2. Structural model values (0.44) were larger than zero, suggesting the good predictive ac­ curacy of the model. We controlled for demographic variables in our We evaluated the structural model using the Bias-corrected and model. We did not find any significant effect of any demographic vari­ Accelerated (BCa) bootstrap method with 5000 bootsample re-sampling able in the model. approach. The variance inflation factors (VIF) values were less than 5 indicating no multicollinearity concern. We evaluated the goodness of 5.3. Path coefficient estimation fit of model through the score of the standardized root mean square residual (SRMR). The SRMR value for the estimated model was 0.055, The SEM results are shown in Table 5 and Fig. 2. The results show which was below the threshold of 0.08 (Henseler & Sarstedt, 2013). The standardized path coefficients (β values), t values, and p values. Hedonic value of Normed Fit Index (NFI) was 0.92, which was above 0.90 showed attitude has a positive and significant impact on SVA usage (β = 0.183, Table 3 Discriminant validity (Fornell & Larcker, 1981). AN ES HA PL SP US UA VA WOM Anthropomorphism (AN) 0.913 0.909 0.904 0.917 0.904 0.851 0.898 0.904 0.915 Escape (ES) 0.446 0.315 0.459 0.551 0.407 0.462 0.397 0.606 Hedonic Attitude (HA) 0.338 0.404 0.345 0.58 0.387 0.528 0.416 Playfulness (PL) 0.614 0.532 0.398 0.411 0.504 0.638 Social presence (SP) 0.682 0.331 0.787 0.628 0.487 Usage (US) 0.45 0.271 0.429 0.604 Utilitarian attitude (UA) 0.398 0.438 0.371 Visual appeal (VA) 0.573 0.341 Word of mouth (WOM) 0.538 Note: The numbers in the diagonal are the square root of the variance shared between the constructs and their measures. Off-diagonal elements are correlations among constructs. 7

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx SP US UA VA WOM Table 4 Discriminant validity (HTMT values). AN ES HA PL Anthropomorphism (AN) 0.492 0.343 0.511 0.614 0.548 0.624 0.444 0.684 Escape (ES) 0.374 0.44 0.381 0.764 0.433 0.697 0.465 Hedonic Attitude (HA) 0.679 0.6 0.529 0.458 0.573 0.845 Playfulness (PL) 0.764 0.438 0.821 0.702 0.546 Social presence (SP) 0.598 0.301 0.477 0.668 Usage (US) 0.447 0.489 0.41 Utilitarian attitude (UA) 0.64 0.373 Visual appeal (VA) 0.596 Word of mouth (WOM) Table 5 Hypothesis Path t p p < 0.001), supporting hypotheses H2a and H2b. SEM results. coefficient β Statistics Values Playfulness (β = 0.377, t = 5.89, p < 0.001) and escapism H1a 0.183 2.11 0.027 Hedonic attitude → Usage H1b 0.154 2.37 0.018 (β = 0.144, t = 2.89, p = 0.005) show a positive and significant impact Hedonic attitude → WOM H2a 0.391 4.49 <.001 on hedonic attitude, supporting H3 and H4. The results reveal that Utilitarian attitude → Usage H2b 0.325 3.93 <.001 anthropomorphism does not impact hedonic attitude (β = 0.041, Utilitarian attitude → WOM H3 0.377 5.89 <.001 t = 0.63, p = 0.53), but it has a significant influence on utilitarian atti­ Playfulness → Hedonic tude (β = 0.166, t = 1.98, p < 0.05). Hence, H5a is not supported, but H4 0.144 2.89 0.005 H5b is supported. Visual appeal (β = 0.222, t = 3.06, p = 0.002) and attitude social presence (β = 0.171, t = 2.25, p < 0.05) show a positive and Escapism → Hedonic H5a 0.041 0.63 0.53 significant relationship with utilitarian attitude. Thus, H6 and H7 are supported. attitude H5b 0.166 1.98 0.045 Anthropomorphism → 5.4. Multi-group PLS analysis for moderation H6 0.222 3.06 0.002 Hedonic attitude The moderating effects of prestige were examined using the multi- Anthropomorphism → H7 0.171 2.25 0.024 group analysis (PLS-MGA) in SmartPLS 3.3 (see Table 6). This method is a non-parametric significance test for the difference of group-specific Utilitarian attitude results based on PLS-SEM bootstrapping method (Hair et al., 2011, Visual appeal → Utilitarian 2019). The method tests whether the path coefficients significantly differ between the two or more groups (moderators). The results show attitude prestige moderates the two relationships: Playfulness → Hedonic atti­ Social presence → tude (βdiff = − 0.323, p = 0.048) and Escapism → Hedonic attitude (βdiff = 0.407, p = 0.031), whereas there is no significant difference in path Utilitarian attitude coefficients for relationship Anthropomorphism → Hedonic attitude t = 2.11, p < 0.05) and users’ WOM recommendations (β = 0.154, t = 2.37, p < 0.05) of SVA, supporting H1a and H1b. Similarly, utili­ tarian attitude appears as a strong determinant of SVA usage (β = 0.391, t = 4.49, p < .001) and WOM recommendations (β = 0.325, t = 3.93, Fig. 2. Results of structural model test. 8

A. Mishra et al. Hypothesis Difference in path p value (for International Journal of Information Management xxx (xxxx) xxx H8a coefficients (high vs. difference) Table 6 low prestige) why people are more likely to recommend SVA to others. These results PLS-MGA results. 0.048 complement the vast amount of prior research on WOM, which recom­ -0.323 0.031 mends that product quality and novel experiences are critical anteced­ Relationship 0.978 ents to WOM behavior (Berger, 2014; Mishra & Satish, 2016). H8b 0.407 Playfulness → Hedonic An examination of antecedents to hedonic attitude and utilitarian attitude H8c -0.005 attitudes indicates that playfulness and escapism strongly influence hedonic attitude, whereas the effect of anthropomorphism on hedonic Escapism → Hedonic attitude is non-significant. Human-SVA interactions are a playful ex­ attitude change of information in absorbing activities leading to intrinsic enjoyment (Mathwick & Rigdon, 2004). Users recognize playfulness as a Anthropomorphism → reflection of emotional attachment with SVA. The affective responses to Hedonic attitude SVA lead to warmth perceptions, which positively influence users’ usage intentions (Hu et al., 2021). The results extend the earlier findings on (βdiff = − 0.005, p = 0.978) for high and low prestige groups. Thus, the playfulness as an intrinsic utility for hedonic motives (Venkatesh et al., findings support H8a and H8b, but not H8c. 2012). Similarly, in the context of virtual reality store shopping, the experience of playfulness enhances consumers’ hedonic gratifications 6. Discussion (Kang, Shin, & Ponto, 2020). Moreover, the results endorse the similar positive effect of escapism on hedonic values found in recent tourism SVA are the next step of advancement in human-technology in­ research that focuses on experiential consumption (Ponsignon, Lunardo, teractions, which significantly impact the consumer journey and offer a & Michrafy, 2020). highly interactive and playful experience to users. Marketers and re­ searchers have recognized the immense potential of SVA that can sub­ Research offers evidence that characters, objects, or product pack­ stantially influence people’s productivity and consumer behavior. aging, which mimic humanness appeal to consumers’ hedonic values However, the adoption of SVA has not reached its full potential, which and perceptions (Epley et al., 2007). SVA deliver human and social cues can be attributed to users’ attitudes toward SVA. The first objective of while interacting with humans leading to higher engagement (Moriuchi, our research was to examine the effect of hedonic and utilitarian atti­ 2021). We did not find a meaningful relation between anthropomor­ tudes on SVA usage and WOM behavior. The second objective was to phism and hedonic attitude. A plausible explanation is that users may identify the antecedents to two types of attitudes. To address these ob­ perceive SVA as humans, but that perception is limited to the functional jectives, this study investigated the experiential aspect of using SVA aspect and may not lead to a strong emotional bonding. Research in IS from a users’ perspective, where attitude played a critical role in shaping area on SVA has mostly explored the effect of anthropomorphism using the behavioral intentions. task-fit perspective on constructs like performance expectancy and effort expectancy (e.g., Gursoy et al., 2019). Our results support prior research According to the statistical analysis presented in the previous sec­ and establish that anthropomorphism features of SVA appeal more to tion, the proposed research model shows an acceptable level of predic­ consumers’ utilitarian expectations rather than hedonic motives. tive power. The model meets the recommended criteria for reliability and validity indices and explains the significant amount of variance in Furthermore, visual appeal and social presence significantly influ­ all endogenous constructs: hedonic attitude (43%), utilitarian attitude ence utilitarian attitude. Most SVA available in markets comes in basic (69%), usage (53%), and WOM (37%). The obtained values of R2 are spherical or round shape in a single color (white, black, or blue). The within the highly acceptable level for research dealing with behavioral idea behind simple designs could be to allow users to put SVA nearby on predictions, especially in the domain of technology adoption (e.g., tables or bed rests. Moreover, SVA devices are compact and balanced so Alalwan et al., 2017; Hu et al., 2021). However, our results show a much that they do not fall easily. This explains why users relate visual appeal higher value of R2 for usage (53%) in comparison to R2 value (29.7%) of SVA to convenience and practical benefits. Similarly, SVA give an obtained by Hu et al. (2021) who studied the role of autonomy on SVA impression of having an ‘entity’ or a person in the vicinity, which can adoption. respond to users. Thus, SVA act as a solution to users’ loneliness ful­ filling users’ need for a human companion. Hedonic and utilitarian attitudes show a positive impact on SVA usage. However, utilitarian attitude has a stronger impact (vs. hedonic Lastly, SVA deliver symbolic benefits to owners as many users attitude) on SVA usage and users’ WOM recommendations. Thus, users perceive SVA prestigious. Individuals who perceive SVA as a symbol of perceive SVA more of a functional device leading to a favorable utili­ high prestige (vs. low prestige) are less likely to relate playfulness to tarian attitude. McLean and Osei-Frimpong (2019) examined the influ­ hedonic gratifications. In contrast, people attributing prestige to SVA are ence of hedonic and utilitarian benefits on SVA usage. They found that more likely to use it as an escape route from their mundane life. Thus, hedonic benefits were not a significant factor for determining SVA prestige moderates the influence of escapism and playfulness on hedonic usage. Our research examined hedonic attitudes instead of hedonic attitude in opposite directions. The results propose the boundary con­ benefits. In contrast to the findings of McLean and Osei-Frimpong dition of the degree of importance attached to the prestige associated (2019), the results indicate that hedonic attitude is a significant pre­ with SVA ownership, which affects users’ psychological gratifications dictor of SVA usage. We believe that our results are more in line with the (playfulness and escapism) and hedonic attitudes. influence of behavioral sciences research in IS domain, which outlines the importance of attitude in shaping human behavior (Ajzen, 1991; 6.1. Theoretical contributions Davis, 1989; Lee et al., 2020). Our findings suggest that users perceive SVA as a functional device that provides fun and entertainment. Thus, This paper makes several contributions to research on SVA usage and users’ hedonic motives shape hedonic attitudes that positively affect WOM behavior. First, it answers recent research calls on the adoption SVA usage (e.g., Tamilmani et al., 2019). and usage of SVA (e.g., Dwivedi, Ismagilova, et al., 2021; Dwivedi, Hugles, et al., 2021). The present study uses a cross-discipline approach The findings reveal that a favorable attitude toward SVA leads to a to extend the boundaries of SVA adoption by integrating the critical higher likelihood of WOM recommendations. From a task-fit perspec­ aspect of individual psychological experiences with design-centric tive, SVA can complete routine tasks with ease reducing users’ cognitive functional evaluations. We integrate elements from sociology and psy­ load. Borrowing the terminology from product performance literature, chology research to arrive at antecedents to attitudinal parameters. The we can say that SVA complete tasks efficiently and consistently. Thus, research model advances the current state of research on the adoption of SVA can be trusted and classified as high-quality products. In addition, emerging technologies. A majority of the extant literature on technology users experience joy and fun while interacting with SVA. This explains 9

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx adoption is based on the original or extended versions of TAM and adoption. This research is an attempt to fill this particular research gap. UTAUT frameworks (Tamilmani et al., 2019). However, recent literature A positive WOM is crucial for product success and drives viral diffusion reviews on emerging technologies clearly outline the need to integrate of advertising and marketing communications. Similarly, we expect theories and models from multiple disciplines to advance the knowledge favorable WOM should help in technology diffusion. We find that both in technology adoption in the context of emerging technologies (e.g., hedonic and utilitarian attitudes are critical for spreading WOM. Dwivedi et al., 2021). We also notice an increase in the usage of hedonic Moreover, hedonic satisfaction and functional performance of SVA are motivation models in research to predict the usage or adoption of crucial in influencing users’ attitudes. Thus, we contribute to the WOM emerging technologies like virtual reality stores (Kang et al., 2020) and literature by suggesting extrinsic gratifications as antecedents to WOM mobile payments (Kar, 2020). Our research examines psychological behavior in addition to the satisfactory performance of SVA. factors and design specific factors that influence users’ hedonic and utilitarian attitudes lading to SVA adoption and WOM recommenda­ 6.2. Implications for practice tions. The results cement the importance of psychological gratifications derived from using SVA. Our findings have several practical implications for marketers and other stakeholders working toward increasing the adoption of SVA and Second, the study focuses on users’ gratifications derived from SVA, generating positive WOM recommendations. The recent COVID-19 which is different from the typical organization-based technology pandemic has drastically changed the global business landscape. Most adoption research. This study complements and extends the work of of the physical retail businesses were forced to adopt digital platforms Tamilmani et al. (2019) about the role of hedonic motivations and (Swoboda, 2020). The pandemic brought substantial changes in con­ derived values from the tasks at hand. Hedonic motivations significantly sumer behavior, where consumer preferences for online shopping influence the outcomes that have hedonic values rather than utilitarian significantly increased. Many consumers experienced their first moment values. For example, people are more likely to use technology like mo­ of online shopping because they were scared of getting an infection due bile banking due to associated hedonic motives such as fun and enjoy­ to human touch at crowded places and physical retail stores. SVA should ment of using technology (Alalwan et al., 2017). The findings confirm be able to play a critical role in such uncertainties as they can alleviate that gratifications based outcomes derived from technology usage in­ users’ concerns about social distancing and touchless experiences. fluence utilitarian and hedonic attitudes that further influence SVA usage. SVA offer an element of playfulness to users who use SVA as a Industry data and research reports show that most users use smart gateway to escapism. We believe that users consider SVA as a com­ voice interaction technologies via their smartphones instead of smart panion with whom they can share experiences or feelings anytime. speakers or SVA (Foehr & Germelmann, 2020). Many users use voice However, SVA differ from a human companion in a sense that SVA abide interaction features on their phones to search information or to get di­ by what the owner says without its own thinking. Thus, interactions rections. Similarly, many users turn to SVA to perform a basic search, with SVA are without any conflict, enjoyable, and humorous. Interest­ play a song, or read news. However, SVA are used in a more convenient ingly, users perceive physical presence of SVA with anthropomorphism setting compared to a smartphone. For example, SVA are placed in a features as utilitarian benefits. The plausible explanation is in line with home at a designated place where users can easily access them using McLean and Osei-Frimpong (2019) findings, which suggest that the voice commands. Hence, marketers who have applied strategies to in­ companionship aspect and the virtual presence of a humanlike device crease the adoption of smartphones should alter their strategies for SVA. fulfill the functional benefits of reducing loneliness. Marketers should focus on making users’ attitudes more favorable to­ ward SVA. They must stress on hedonic benefits of SVA along with Third, the study reaffirms that attitude is a critical antecedent to functional appeal in their communications to differentiate form smart­ technology acceptance and usage (Dwivedi et al., 2019). People use SVA phones. Marketers can educate users that SVA can serve as a trustworthy for its functional as well as hedonic appeal. Hedonic motives are the and entertaining companion that can help users in dealing with loneli­ novel factors that may encourage users to try SVA and further usage ness and make them happy. SVA can be of extreme help in accom­ depends on the functional value. Thus, in the context of emerging plishing regular tasks that people tend to forget like scheduling technologies, which offer vivid, immersive, and highly interactive ex­ appointments, alarms, and even switching off electric lights. periences to users, hedonic motivations are at par with the functional benefits. Thus, a significant contribution of this research is toward the Social presence and playfulness highlight that SVA can tackle the hedonic motivation research that focuses on the extrinsic motivations to issue of loneliness and act as a substitute for humans at home or any use technology. SVA have unique characteristics of being a personal place. Hence, we see two types of opportunities here. First, the technical device that is mostly used in a private (home) environment. Hence, aspect where developers should make SVA more human in terms of technologies that resemble similar usage environments or objectives voice and AI capabilities to understand and interpret instructions. Sec­ should be evaluated on hedonic gratifications. ond, marketers should highlight the humanlike presence of SVA in their marketing communications. Marketers must understand and appreciate Fourth, a key finding of this research is the moderating role of that SVA fulfill functional benefits and provide psychological gratifica­ prestige. Prior research suggests that people are more likely to adopt a tions to users. Marketers can emphasize how users (especially kids and technology under social influence, which means technology is perceived teens) can have fun while interacting with SVA. Marketers can build prestigious and gets appreciation from others (Gursoy et al., 2019). specific content like trivia and quizzes related to education and knowl­ Emergent technologies like SVA or smart wearables signal a favorable edge that can stress learning with fun to encourage SVA adoption in social status in countries like India. The findings suggest that prestige families with kids. Similarly, the humanlike appearance should be part affects individuals’ hedonic attitudes. Generally, people use specific of the overall marketing communication strategy and product devel­ products to signal or enhance their prestige (conspicuous consumption). opment. We expect firms to improve the quality of voice accent and However, the findings bring a critical difference between playfulness support different languages so that people can freely talk to SVA in their and escapism due to the prestige associated with SVA, which contradicts native language. Such expansion will definitely increase SVA adoption. the established concept of conspicuous consumption similar to using augmented reality as an extension or replacement of physical products Esthetic or visual appeal is another crucial factor that affects users’ (Rauschnabel, 2021). attitudes. Since SVA are close to typical household items like white goods, they need better esthetics (e.g., more colors and shapes) to Lastly, a vast amount of research exists on the importance of WOM enhance their visual appeal and blend with surroundings. Amazon recommendations in the success of any product or service (Berger & continues to tweak the design of Alexa in new models and recently Milkman, 2012; Mishra & Satish, 2016). While an extensive amount of introduced the feature of a digital watch embedded in device appear­ WOM research exists in the marketing domain dealing with products ance. Marketers may learn a few lessons from blockbuster Hollywood and services, we notice scant research in the context of technology 10

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx movies such as Star wars or Wall-E, which have turned the robots and 7. Conclusion droids into friends, companions, or saviors. We believe that marketers have yet to realize the immense potential of design for SVA. Therefore, Our study highlights the importance of hedonic and utilitarian atti­ they should continue experiments with out-of-the-box innovative tudes that lead to SVA usage and WOM recommendations. This study designs. proposed and tested a research model based on flow theory and the theory of anthropomorphism. While flow theory focuses on the We find that prestige associated with SVA influences users’ hedonic immersive and absorbing experiences of using SVA, anthropomorphism attitudes. In fact, it has opposite effects on how playfulness and escapism reflects the humanness of SVA. We identified five antecedents (play­ affect users’ hedonic attitudes. This finding calls for a distinct segmen­ fulness, escapism, anthropomorphism, visual appeal, and social pres­ tation strategy by marketers based on users’ valuation of prestige ence) to hedonic and utilitarian attitudes based on previous studies in attained from using SVA. Till now, marketers have rarely used the technology adoption and multisensory research. The moderating role of prestige factor to influence SVA adoption. We believe that marketers prestige (social status) was examined for hedonic attitude. The findings should try to highlight this aspect. They can create ads and post on social revealed that playfulness and escapism significantly influence hedonic media to target specific user-segment similar to users of premium luxury attitude, whereas anthropomorphism, visual appeal, and social presence goods. Marketers may think about launching different versions of SVA affect utilitarian attitude. Though both attitudes had a significant effect that are superior in design and get a price premium. on SVA usage and WOM recommendations, the effect of utilitarian attitude was relatively stronger than the hedonic attitude. Furthermore, Lastly, our results notice that favorable hedonic and utilitarian atti­ the influence of escapism and playfulness on hedonic attitude is tudes can lead to WOM recommendations, which further can influence moderated by prestige associated with SVA. These results provide SVA adoption. So, we suggest that marketers should encourage user- actionable insights to the various stakeholders to increase the SVA generated content (UGC) on how they use SVA and what they feel adoption by focusing on functional and psychological benefits offered by about SVA. SVA may politely (or humorously) nudge and encourage SVA. This research adds a new dimension of favorable WOM that can be users to post/share their experiences with SVA on online platforms. highly effective in accelerating SVA adoption. Thus, SVA can be promoters of themselves. This may help marketers to get insights directly from users, which should pave the way for im­ Author statement provements in future SVA models. All authors have participated equally and substantially in the 6.3. Limitations and future research direction concept, design, acquisition of data analysis, drafting, writing, and revision of the manuscript. As with any research study, our study has certain limitations. Therefore the findings should be interpreted according to the context. The authors certify that the material or similar material has not been First, the study uses convenience sampling for online survey. Due to the and will not be submitted to or published in any other publication before COVID19 situation and subsequent lockdown, only online data collec­ its appearance in the International Journal of Information Management. tion was feasible. Hence, only the respondents who had access to the Internet participated in the study. Thus, this may affect the generaliz­ Funding ability of the findings. Moreover, the majority of the sample is relatively young (less than 40 years). Though research shows that young adults are This research did not receive any specific grant from funding more likely to adopt new technologies, the elderly population may also agencies in the public, commercial, or not-for-profit sectors. use SVA for unique benefits. Therefore, further research can extend the sample to a broader age group and non-Internet users. Second, the study Appendix A. Measures uses survey method that suffers from certain biases. We have taken recommended measures like the assurance of confidentiality and ano­ Playfulness (Müller-Stewens, Schlager, Ha¨ubl, & Herrmann, 2017). nymity to respondents. We also tested for common method bias to verify that it does not confound our findings. However, we suggest that further 1. Talking to voice assistant is fun. research may employ experimental design or longitudinal research to 2. Interacting with voice assistant is enjoyable. corroborate our findings. Third, we believe that SVA have a novelty and 3. I feel happy when interacting with voice assistant. enjoyment factor, which may be prominent in initial usage but may wear out in due course of time. Since attitude formation is a function of time, Escapism (Mathwick et al., 2001). forthcoming research should extend our model and examine the differ­ ences in attitudinal parameters across a longer period of SVA usage. 1. Interacting with voice assistant takes me to another world. Fourth, we did not specifically look at the language aspect of voice as­ 2. Interacting with voice assistant makes me feel like I am in another sistants. In a diverse country like India, users use a variety of native languages. SVA like Alexa and Google home have started supporting world. various languages, but they are still far from being perfect. Thus, SVA 3. I get so involved when I interact with voice assistant that I forget may not be able to interpret voice commands at times leading to dissatisfaction. We think this could be another exciting aspect where everything else. upcoming research can explore users’ satisfaction when interacting with SVA using native languages or different accents. SVA (Alexa and Google Anthropomorphism (Guido & Peluso, 2015). home) interact with users in a female voice, which could influence users’ attitudes and acceptance in patriarchal cultures. Thus, researchers can 1. Voice assistant acts like a person. investigate the impact of gender SVA voice in a cross-cultural setting. 2. Voice assistant talks like a human. Our research model uses hedonic and utilitarian attitudes to predict 3. Voice assistant interacts like a person. adoption and WOM behavior, which can be extended based on the complexity of tasks. SVA may be used by an individual or by a group of Visual Appeal (Mathwick et al., 2001). people in a family. Moreover, the location of SVA (e.g., common hall or bedroom) may also influence its usage. Thus, we recommend future 1. Design of voice assistant is attractive. researchers to explore the specific usage situation to get deeper insights 2. Voice assistant is visually appealing. into SVA usage and attitude formation toward SVA. 3. I like how the voice assistant looks. 11

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx Social Presence (Roy et al., 2019). Bergkvist, L., & Rossiter, J. R. (2007). The predictive validity of multiple-item versus single-item measures of the same constructs. Journal of Marketing Research, 44(2), 1. Interacting with voice assistant makes me feel comfortable, as if I am 175–184. with a friend. Chatterjee, S., & Kar, A. K. (2020). Why do small and medium enterprises use social 2. There is a sense of human contact when I interact with voice media marketing and what is the impact: Empirical insights from India. International assistant. Journal of Information Management, 53, Article 102103. https://doi.org/10.1016/j. ijinfomgt.2020.102103. 3. Interacting with voice assistant gives me some sense of social life. Chatterjee, S., Kar, A. K., & Dwivedi, Y. K. (2021). Intention to use IoT by aged Indian Hedonic Attitude (Spangenberg, Voss, & Crowley, 1997). consumers. Journal of Computer Information Systems. , Article 1873080. https://doi. Interacting with voice assistant is. org/10.1080/08874417.2021.1873080. 1. Dull – exciting Conner, K., & Bradford, A. (2020). Need a good laugh? 55 funniest things to ask Alexa to 2. Not fun – Fun raise your spirits. Retrieved from 〈https://www.cnet.com/how-to/need-a-good- 3. Unenjoyable – enjoyable laugh-55-funniest-things-to-ask-alexa-to-raise-your-spirits/〉. Utilitarian attitude (Spangenberg et al., 1997). Dabholkar, P. A., van Dolen, W. M., & de Ruyter, K. (2009). A dual-sequence framework Interacting with voice assistant is. for B2C relationship formation: Moderating effects of employee communication style in online group chat. Psychology & Marketing, 26(2), 145–174. 1. Useless – useful 2. Unnecessary – Necessary Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of 3. Unproductive – productive information technology. MIS Quarterly, 13(3), 319–340. Prestige (Gursoy et al., 2019). Dwivedi, Y. K., Hughes, L., Ismagilova, E., Aarts, G., Coombs, C., Crick, T., & Williams, M. D. (2021a). Artificial Intelligence (AI): Multidisciplinary perspectives 1. Having a voice assistance improves my social status. on emerging challenges, opportunities, and agenda for research, practice and policy. 2. Having a voice assistance establishes my position in society. International Journal of Information Management, 57, Article 101994. 3. Having a voice assistance is prestigious for me. Dwivedi, Y. K., Ismagilova, E., Rana, N. P., & Raman, R. (2020). Security, privacy and Word-of-mouth behavior (Mishra et al., 2018). risks within smart cities: Literature review and development of a smart city interaction framework. Information Systems Frontiers: A Journal of Research and 1. I will share positive word of mouth about voice assistant. Innovation, 1–22. https://doi.org/10.1007/s10796-021-10106-y. 2. I will recommend voice assistant to my friends. 3. I will encourage my friends to use or buy voice assistant. Dwivedi, Y. K., Ismagilova, E., Hughes, D. L., Carlson, J., Filieri, R., Jacobson, J., & Wang, Y. (2021c). Setting the future of digital and social media marketing research: Usage (McLean & Osei-Frimpong, 2019). Perspectives and research propositions. International Journal of Information How frequently do you use voice assistant? Management, 59, Article 102168. References Dwivedi, Y. K., Rana, N. P., Jeyaraj, A., Clement, M., & Williams, M. D. (2019). Re- examining the unified theory of acceptance and use of technology (UTAUT): Agarwal, R., & Karahanna, E. (2000). Time flies when you’re having fun: Cognitive Towards a revised theoretical model. Information Systems Frontiers, 21(3), 719–734. absorption and beliefs about information technology usage. MIS Quarterly, 24(4), 665–694. Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. Ahn, T., Ryu, S., & Han, I. (2007). The impact of web quality and playfulness on user acceptance of online retailing. Information & Management, 44(3), 263–275. Fernandes, T., & Oliveira, E. (2021). Understanding consumers’ acceptance of automated technologies in service encounters: Drivers of digital voice assistants adoption. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Journal of Business Research, 122, 180–191. Decision Processes, 50(2), 179–211. Foehr, J., & Germelmann, C. C. (2020). Alexa, can I trust you? Exploring consumer paths Alalwan, A. A., Dwivedi, Y. K., & Rana, N. P. (2017). Factors influencing adoption of to trust in smart voice-interaction technologies. Journal of the Association for mobile banking by Jordanian bank customers: Extending UTAUT2 with trust. Consumer Research, 5(2), 181–205. International Journal of Information Management, 37(3), 99–110. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with Bag, S., Pretorius, J. H. C., Gupta, S., & Dwivedi, Y. K. (2021). Role of institutional unobservable variables and measurement error. Journal of Marketing Research, 18(1), pressures and resources in the adoption of big data analytics powered artificial 39–50. intelligence, sustainable manufacturing practices and circular economy capabilities. Technological Forecasting and Social Change, 163, Article 120420. https://doi.org/ Foroudi, P., Gupta, S., Sivarajah, U., & Broderick, A. (2018). Investigating the effects of 10.1016/j.techfore.2020. smart technology on customer dynamics and customer experience. Computers in Human Behavior, 80, 271–282. Balakrishnan, J., & Dwivedi, Y. K. (2021aaa). Role of cognitive absorption in building user trust and experience. Psychology & Marketing, 38(4), 643–668. Google Insights (2017). Voice-Activated Speakers: People’s Lives Are Changing. Retrieved from 〈https://www.thinkwithgoogle.com/data-collections/voice-assistan Balakrishnan, J., & Dwivedi, Y. K. (2021bbb). Conversational commerce: Entering the ce-emerging-technologies/〉. next stage of AI-powered digital assistants. Annals of Operations Research, 1–35. https://doi.org/10.1007/s10479-021-04049-5. Grover, P., Kar, A. K., & Dwivedi, Y. K. (2020). Understanding artificial intelligence adoption in operations management: Insights from the review of academic literature Barnes, S. J. (2021). Understanding the overvaluation of facial trustworthiness in Airbnb and social media discussions. Annals of Operations Research, 1–37. https://doi.org/ host images. International Journal of Information Management, 56, Article 102265. 10.1007/s10479-020-03683-9. Bawack, R. E., Wamba, S. F., & Carillo, K. D. A. (2021). Exploring the role of personality, Guido, G., & Peluso, A. M. (2015). Brand anthropomorphism: Conceptualization, trust, and privacy in customer experience performance during voice shopping: measurement, and impact on brand personality and loyalty. Journal of Brand Evidence from SEM and fuzzy set qualitative comparative analysis. International Management, 22(1), 1–19. Journal of Information Management, 58, Article 102309. https://doi.org/10.1016/j. ijinfomgt.2021.102309. Gursoy, D., Chi, O. H., Lu, L., & Nunkoo, R. (2019). Consumers acceptance of artificially intelligent (AI) device use in service delivery. International Journal of Information Belk, R. (2017). The soul and the machine: Human like machines and machine like Management, 49, 157–169. humans. In A. Gneezy, V. Griskevicius, & P. Williams (Eds.), NA—advances in consumer research (Vol. 45, pp. 164–169). Association for Consumer Research. Hair, J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19(2), 139–152. Berger, J. (2014). Word of mouth and interpersonal communication: A review and directions for future research. Journal of Consumer Psychology, 24(4), 586–607. Hair, J. F., Risher, J. J., Sarstedt, M., & Ringle, C. M. (2019). When to use and how to report the results of PLS-SEM. European Business Review, 31(1), 2–24. Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal of Marketing Research, 49(2), 192–205. Han, S., Min, J., & Lee, H. (2016). Building relationships within corporate SNS accounts through social presence formation. International Journal of Information Management, 36(6), 945–962. Henseler, J., & Sarstedt, M. (2013). Goodness-of-fit indices for partial least squares path modeling. Computational Statistics, 28(2), 565–580. Hernandez-Ortega, B., & Ferreira, I. (2021). How smart experiences build service loyalty: The importance of consumer love for smart voice assistants. Psychology and Marketing, 38(7), 1122–1139. Hoffman, D. L., & Novak, T. P. (2018). Consumer and object experience in the internet of things: An assemblage theory approach. Journal of Consumer Research, 44(6), 1178–1204. Holmqvist, J., Ruiz, C. D., & Pen˜aloza, L. (2020). Moments of luxury: Hedonic escapism as a luxury experience. Journal of Business Research, 116, 503–513. Hsu, C. L., & Lu, H. P. (2004). Why do people play on-line games? An extended TAM with social influences and flow experience. Information & Management, 41(7), 853–868. Hu, Q., Lu, Y., Pan, Z., Gong, Y., & Yang, Z. (2021). Can AI artifacts influence human cognition? The effects of artificial autonomy in intelligent personal assistants. International Journal of Information Management, 56, Article 102250. https://doi.org/ 10.1016/j.ijinfomgt.2020.102250. Hughes, L., Dwivedi, Y. K., Misra, S. K., Rana, N. P., Raghavan, V., & Akella, V. (2019). Blockchain research, practice and policy: Applications, benefits, limitations, 12

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx emerging research themes and research agenda. International Journal of Information Novak, T. P., & Hoffman, D. L. (2019). Relationship journeys in the internet of things: a Management, 49, 114–129. framework for understanding interactions between consumers and smart objects. Hur, J. D., Koo, M., & Hofmann, W. (2015). When temptations come alive: How Journal of the Academy of Marketing Science, 47(2), 216–237. anthropomorphism undermines self-control. Journal of Consumer Research, 42(2), 340–358. Park, K., Kwak, C., Lee, J., & Ahn, J. H. (2018). The effect of platform characteristics on Ismagilova, E., Rana, N. P., Slade, E. L., & Dwivedi, Y. K. (2020). A meta-analysis of the the adoption of smart speakers: Empirical evidence in South Korea. Telematics and factors affecting eWOM providing behaviour. European Journal of Marketing, 55(4), Informatics, 35(8), 2118–2132. 1067–1102. Jones, S., Cronin, J., & Piacentini, M. G. (2018). Mapping the extended frontiers of Phan, T. (2019). Amazon Echo and the aesthetics of whiteness. Catalyst: Feminism, escapism: Binge-watching and hyperdiegetic exploration. Journal of Marketing Theory, Technoscience, 5(1), 1–38. Management, 34(5–6), 497–508. Kang, H. J., Shin, J. H., & Ponto, K. (2020). How 3D virtual reality stores can shape Pillai, R., Sivathanu, B., & Dwivedi, Y. K. (2020). Shopping intention at AI-powered consumer purchase decisions: the roles of informativeness and playfulness. Journal of automated retail stores (AIPARS). Journal of Retailing and Consumer Services, 57, Interactive Marketing, 49, 70–85. Article 102207. https://doi.org/10.1016/j.jretconser.2020.102207. Kar, A. K. (2020). What affects usage satisfaction in mobile payments? Modelling user generated content to develop the “digital service usage satisfaction model”. Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method Information Systems frontiers: a Journal of Research and Innovation, 1–21. https://doi. biases in behavioral research: A critical review of the literature and recommended org/10.1007/s10796-020-10045-0. remedies. Journal of Applied Psychology, 88(5), 879–903. Karimova, G. Z., & Goby, V. P. (2020). The adaptation of anthropomorphism and archetypes for marketing artificial intelligence. Journal of Consumer Marketing, 38(2), Ponsignon, F., Lunardo, R., & Michrafy, M. (2020). Why are international visitors more 229–238. satisfied with the tourism experience? The role of hedonic value, escapism, and Kim, Y., & Sundar, S. S. (2012). Anthropomorphism of computers: Is it mindful or psychic distance. Journal of Travel Research. , Article 004728752096117. https://doi. mindless? Computers in Human Behavior, 28(1), 241–250. org/10.1177/0047287520961175. Krishna, A., & Schwarz, N. (2014). Sensory marketing, embodiment, and grounded cognition: A review and introduction. Journal of Consumer Psychology, 24(2), Poushneh, A. (2021). Humanizing voice assistant: The impact of voice assistant 159–168. personality on consumers’ attitudes and behaviors. Journal of Retailing and Consumer Kuo, A., Lutz, R. J., & Hiler, J. L. (2016). Brave new World of Warcraft: A conceptual Services, 58, Article 102283. https://doi.org/10.1016/j.jretconser.2020.102283. framework for active escapism. Journal of Consumer Marketing, 33(7), 498–506. Kushwaha, A. K., & Kar, A. K. (2020). Language model-driven chatbot for business to PWC (2019). Prepare for the voice revolution. Retrieved from 〈https://www.pwc.com/ address marketing and selection of products. In S. K. Sharma, Y. K. Dwivedi, us/en/services/consulting/library/consumer-intelligence-series/voice-assistants. B. Metri, & N. P. Rana (Eds.), Re-imagining Diffusion and Adoption of Information html〉. Technology and Systems: A Continuing Conversation. TDIT 2020. IFIP Advances in Information and Communication Technology, 617 pp. 16–28). Cham: Springer. https:// Rana, N. P., Dwivedi, Y. K., Williams, M. D., & Weerakkody, V. (2016). Adoption of doi.org/10.1007/978-3-030-64849-7_3. online public grievance redressal system in India: Toward developing a unified view. Kushwaha, A. K., Kar, A. K., & Dwivedi, Y. K. (2021). Applications of big data in Computers in Human Behavior, 59, 265–282. emerging management disciplines: A literature review using text mining. International Journal of Information Management Data Insights, 1(2), Article 100017. Rauschnabel, P. A. (2021). Augmented reality is eating the real-world! The substitution Lee, K., Lee, K. Y., & Sheehan, L. (2020). Hey Alexa! A magic spell of social glue?: Sharing of physical products by holograms. International Journal of Information Management, a smart voice assistant speaker and its impact on users’ perception of group 57, Article 102279. https://doi.org/10.1016/j.ijinfomgt.2020.102279. harmony. Information Systems Frontiers, 22(3), 563–583. Liao, Y., Vitak, J., Kumar, P., Zimmer, M., & Kritikos, K. (2019). Understanding the role Rauschnabel, P. A., Rossmann, A., & tom Dieck, M. C. (2017). An adoption framework for of privacy and trust in intelligent personal assistant adoption. International conference mobile augmented reality games: The case of Pok´emon Go. Computers in Human on information (pp. 102–113). Cham: Springer. Behavior, 76, 276–286. Lindell, M. K., & Whitney, D. J. (2001). Accounting for common method variance in cross-sectional research designs. Journal of Applied Psychology, 86(1), 114–121. Ringle, C. M., Wende, S., & Becker, J. M. (2015). SmartPLS 3. Boenningstedt: SmartPLS Lopatovska, I., Rink, K., Knight, I., Raines, K., Cosenza, K., Williams, H., … Martinez, A. GmbH. 〈http://www.smartpls.com〉. (2019). Talk to me: Exploring user interactions with the Amazon Alexa. Journal of Librarianship and Information Science, 51(4), 984–997. Roy, S. K., Singh, G., Hope, M., Nguyen, B., & Harrigan, P. (2019). The rise of smart Lovett, L. (2020). Tech’s role in helping seniors overcome loneliness amid COVID-19 consumers: role of smart servicescape and smart consumer experience co-creation. pandemic. Retrieved from 〈https://www.mobihealthnews.com/news/techs-role-he Journal of Marketing Management, 35(15–16), 1480–1513. lping-seniors-overcome-loneliness-amid-covid-19-pandemic〉. Mathwick, C., & Rigdon, E. (2004). Play, flow, and the online search experience. Journal Savage, M. (2020). Voice technologies using AI are being used to help combat loneliness of Consumer Research, 31(2), 324–332. in countries including Sweden and the UK. Should they be used more widely as Mathwick, C., Malhotra, N., & Rigdon, E. (2001). Experiential value: conceptualization, coronavirus spreads? Retrieved from 〈https://www.bbc.com/worklife/article measurement and application in the catalog and Internet shopping environment☆. /20200325-can-voice-technologies-using-ai-fight-elderly-loneliness〉. Journal of Retailing, 77(1), 39–56. McLean, G., & Osei-Frimpong, K. (2019). Hey Alexa… examine the variables influencing Spangenberg, E. R., Voss, K. E., & Crowley, A. E. (1997). Measuring the hedonic and the use of artificial intelligent in-home voice assistants. Computers in Human utilitarian dimensions of attitude: a generally applicable scale. In M. Brucks, & Behavior, 99, 28–37. D. J. MacInnis (Eds.), Advances in consumer research (Vol. 24, pp. 235–241). Provo, Mishra, A., & Satish, S. M. (2016). eWOM: Extant research review and future research UT: Association for Consumer Research. avenues. Vikalpa, 41(3), 222–233. Mishra, A., Maheswarappa, S. S., & Colby, C. L. (2018). Technology readiness of Straub, D., Boudreau, M. C., & Gefen, D. (2004). Validation guidelines for IS positivist teenagers: A consumer socialization perspective. Journal of Services Marketing, 32(5), research. Communications of the Association for Information systems, 13(1), 380–427. 592–604. Mishra, A., & Shukla, A. (2020). Psychological determinants of consumer’s usage, Swoboda, C. (2020). COVID-19 Is Making Alexa And Siri A Hands-Free Necessity. satisfaction, and word-of-mouth recommendations toward smart voice assistants. In Retrieved from 〈https://www.forbes.com/sites/chuckswoboda/2020/04/06/covid S. K. Sharma, Y. K. Dwivedi, B. Metri, & N. P. Rana (Eds.), Re-imagining diffusion and -19-is-making-alexa-and-siri-a-hands-free-necessity/〉. adoption of information technology and systems: a continuing conversation. TDIT 2020. IFIP Advances in Information and Communication Technology, 618 pp. 274–283). Tamilmani, K., Rana, N. P., Nunkoo, R., Raghavan, V., & Dwivedi, Y. K. (2020). Indian Cham: Springer. https://doi.org/10.1007/978-3-030-64849-7_24. travellers’ adoption of airbnb platform. Information Systems Frontiers. https://doi. Mishra, A., Shukla, A., Rana, N. P., & Dwivedi, Y. K. (2021). From “touch” to a org/10.1007/s10796-020-10060-1. “multisensory” experience: The impact of technology interface and product type on consumer responses. Psychology & Marketing, 38(3), 385–396. Tamilmani, K., Rana, N. P., Prakasam, N., & Dwivedi, Y. K. (2019). The battle of brain vs. Moriuchi, E. (2019). Okay, Google!: An empirical study on voice assistants on consumer heart: A literature review and meta-analysis of “hedonic motivation” use in UTAUT2. engagement and loyalty. Psychology & Marketing, 36(5), 489–501. International Journal of Information Management, 46, 222–235. Moriuchi, E. (2021). An empirical study on anthropomorphism and engagement with disembodied AIs and consumers’ re-use behavior. Psychology & Marketing, 38(1), Venkatesh, V., & Bala, H. (2008). Technology acceptance model 3 and a research agenda 21–42. on interventions. Decision Sciences, 39(2), 273–315. Müller-Stewens, J., Schlager, T., Ha¨ubl, G., & Herrmann, A. (2017). Gamified information presentation and consumer adoption of product innovations. Journal of Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of Marketing, 81(2), 8–24. information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. Nakamura, J., & Csikszentmihalyi, M. (2009). Flow theory and research. In C. R. Snyder, & S. J. Lopez (Eds.), Oxford handbook of positive psychology (pp. 195–206). Oxford, Venkatesh, V., Thong, J. Y., & Xu, X. (2012). Consumer acceptance and use of UK: Oxford University Press. information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly, 36(1), 157–178. Verma, S., Sharma, R., Deb, S., & Maitra, D. (2021). Artificial intelligence in marketing: Systematic review and future research direction. International Journal of Information Management Data Insights, 1(1), Article 100002. Vilpponen, A., Winter, S., & Sundqvist, S. (2006). Electronic word-of-mouth in online environments: Exploring referral networks structure and adoption behavior. Journal of Interactive Advertising, 6(2), 8–77. Vimalkumar, M., Sharma, S. K., Singh, J. B., & Dwivedi, Y. K. (2021). Okay google, what about my privacy?’: User’s privacy perceptions and acceptance of voice based digital assistants. Computers in Human Behavior, 120, Article 106763. https://doi.org/ 10.1016/j.chb.2021.106763. Voss, K. E., Spangenberg, E. R., & Grohmann, B. (2003). Measuring the hedonic and utilitarian dimensions of consumer attitude. Journal of Marketing Research, 40(3), 310–320. Walther, J. B. (1992). Interpersonal effects in computer-mediated interaction a relational perspective. Communication Research, 19(1), 52–90. Webster, J., & Martocchio, J. J. (1992). Microcomputer playfulness: Development of a measure with workplace implications. MIS Quarterly, 16(2), 201–226. Williams, M. D., Rana, N. P., & Dwivedi, Y. K. (2015). The unified theory of acceptance and use of technology (UTAUT): A literature review. Journal of Enterprise Information Management, 28(3), 443–488. Wu, L., Chiu, M. L., & Chen, K. W. (2020). Defining the determinants of online impulse buying through a shopping process of integrating perceived risk, expectation- 13

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx confirmation model, and flow theory issues. International Journal of Information Management, 52, Article 102099. 14

Computers in Human Behavior 124 (2021) 106914 Contents lists available at ScienceDirect Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh Exploring older adults’ perception and use of smart speaker-based voice assistants: A longitudinal study Sunyoung Kim *, Abhishek Choudhury School of Communication and Information, Rutgers University, New Jersey, USA ARTICLE INFO ABSTRACT Keywords: Thanks to their conversational capabilities, smart speaker-based voice assistants are gaining attention for their Older adults potential to support the aging population, though the empirical evidence is still scarce. This paper aims to obtain Voice assistant empirical evidence on older adults’ experiences with a voice assistant. We especially focused on how their Smart speaker perception and use change over time as they progress from novice to more experienced users through a longi­ Longitudinal study tudinal field deployment study. We deployed Google Home devices in the homes of twelve older adults aged 65 and above and studied their use for sixteen weeks. Results show that the benefits our participants perceived have incrementally changed from enjoying simplicity and convenience of operation in the early phase of the study to not worrying about making mistakes and building digital companionship as they got used to using it. Results also show that participants confronted several challenges that evolved from the unfamiliarity with a voice assistant in their first interactions to coping with the functional errors due to limited speech technology as they got used to using it. Based on the results, we discuss design implications that could foster better user experiences with a voice assistant among older adults. 1. Introduction that older adults generally have positive perceptions when introduced to a smart speaker (Blair & Abdullah, 2019), preferring voice-based user With the advancement of artificial intelligence and speech technol­ interfaces over traditional interaction modalities such as clicking or ogy, smart speaker-based voice assistants are becoming increasingly typing (Kowalski et al., 2019; Wulf et al., 2014). available in the market. It is estimated that in 2019, over 98 million units of smart speakers have been sold worldwide, and it is predicted to Despite the rapid growth of voice technology and its hyped antici­ reach up to 409.4 million units in 2025 (Vailshery, 2019). Nearly a pation for older adults, smart speakers’ actual adoption and use among quarter of households own a smart speaker in the US, and more than half older adults are very low. Older adults are slower to adopt new tech­ of them use two or more smart speakers (Richter, 2020). The smart nologies than younger adults (Vaportzis, Giatsi Clausn, & Gow, 2017), speaker is a type of speaker (e.g., Amazon Echo, Google Home) with an and a smart speaker is not an exception. The statistics show that younger integrated virtual assistant (e.g., Amazon Alexa, Google assistant) that Americans in the 18–29 age group are 75% more likely to own a smart responds to voice commands. It assists users in their daily lives, such as speaker than those over 60 in 2019 (Kinsella, 2019). Considering the playing music, checking weather forecasts, setting alarms and re­ potential of voice assistants for older adults and their low adoption rate, minders, controlling applicable smart home devices, and answering it is crucial to investigate the use of voice assistants for older adults. general questions. Because speech is one of the most natural ways of human communication, using speech to interact with devices can lower To date, researchers have extensively investigated various aspects of the barriers of technology use for those who are less familiar or have using voice assistants, including the use by specific user groups, (e.g., manual-dexterity and vision-related issues with typing- and people with disabilities (Abdolrahmani, Kuber, & Branham, 2018; screen-based interfaces. Therefore, this technology has gained particular Pradhan et al., 2018), children (Druga et al., 2017; Garg & Sengupta, attention as beneficial for older adults (Blair & Abdullah, 2019; Portet, 2020), low-income populations (Robinson et al., 2018), privacy con­ Vacher, Golanski, Roux, & Meillon, 2013). Recent studies have shown cerns (Lau et al., 2018; Liao et al., 2019), conversational aspects (Purington et al., 2017; Vtyurina & Fourney, 2018), and personification (Lopatovska & Williams, 2018; Pradhan et al., 2019). Recently, * Corresponding author. E-mail addresses: [email protected] (S. Kim), [email protected] (A. Choudhury). https://doi.org/10.1016/j.chb.2021.106914 Received 15 January 2021; Received in revised form 2 June 2021; Accepted 9 June 2021 Available online 14 June 2021 0747-5632/© 2021 Elsevier Ltd. All rights reserved.

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 researchers have recognized the potentials that a smart speaker can offer in everyday contexts through a quantitative analysis of voice command in the aging society (Portet, Vacher, Golanski, Roux, & Meillon, 2013) data (Sciuto et al., 2018). As a result, researchers have identified music, and investigated older adults’ experiences with this technology. For search, and Internet of Things (IoT) as the most frequently used features instance, Pradhan et al. found reliability concerns as one of the key of voice-based commands (Ammari et al., 2019), popular types of challenges that prevent older adults with low technology use from commands in different times of the day (e.g., entertainment and home adopting this technology (Pradhan et al., 2020), and Trajova and automation commands peak in the evening while the weather and time Martin-Hammond found that lack of perceived utility contributes to request peak in the early morning hours) (Bentley et al., 2018), and increasing attrition rates among experienced older users (Trajkova & voice assistants integrating into people’s everyday routines by embed­ Martin-Hammond, 2020). Yet, missing is a holistic perspective on how ding in a range of conversational activities in the home (e.g., family older adults progress from novice to more experienced users and what dinners) (Porcheron, Fischer, Reeves, & Sharples, 2018). Other re­ contributes to or prevents their sustained use of a voice assistant. This searchers have found that individuals are motivated to use a voice as­ paper contributes to the growing body of literature by investigating how sistant by utilitarian, symbolic, and social benefits (McLean & novice older adults’ perception and use of a voice assistant change over Osei-Frimpong, 2019). However, its attrition rate is high due to unmet time as they become more experienced through a longitudinal deploy­ expectations and a lack of perceived utility (Cho et al., 2019). ment study. Specifically, we aim to answer the following research questions: Privacy was another major research topic due to a voice assistant’s privacy-intrusive potential to continuously listen to voices in intimate • What do older adults use a voice assistant for? spaces such as the home (Clark et al., 2019; Cowan et al., 2017; Myers • What different benefits do they perceive as they use the device over et al., 2018). The conversational nature and the “always-on” listening feature of a smart speaker have also drawn researchers’ attention to time? social aspects of using voice assistants, such as personification or • What different challenges do they face when using the device over anthropomorphism (Pfeuffer, Alexander, Gimpel, & Hinz, 2019). Re­ searchers have revealed that some users perceived the device as a time, and how do they progressively respond to or cope with those companion or a friend and exhibited personifying behaviors toward challenges? voice assistants by using human-like terms such as “she” or “her” when referring to a voice assistant (Purington et al., 2017). At the same time, To answer our research questions, we deployed a smart speaker in others have argued that such interaction patterns are “mindless” social the homes of twelve older adults aged 65 and above and studied its use responses that may not relate to the actual perception of personification for sixteen weeks. We conducted in-person interviews by visiting the (Lopatovska & Williams, 2018). Relatedly, researchers have investi­ homes every other week for the deployment duration and collected gated factors that constitute an effective conversation with voice assis­ usage logs from the system’s history repository for data collection. tants, though user experiences with voice assistants thus far remained disappointing due to the constrained and predefined turn-taking struc­ Our findings revealed the frequently used features, perceived and ture of question-answer rather than a realistic dialog of conversation experienced benefits, and difficulties with coping strategies that older (Cowan et al., 2017; Murad & Munteanu, 2019). adults might have when interacting with a voice assistant. We found that participants recognized, appreciated, and enjoyed several benefits that a Lastly, research has explored the potential utility of voice assistants voice assistant offers throughout the study period. However, they did as assistive technology for people with disabilities (Abdolrahmani, not acknowledge all the benefits in their first interactions. Instead, the Kuber, & Branham, 2018), specific health concerns (Maharjan et al., perceived benefits have incrementally changed from enjoying simplicity 2019), and the aging population. Recently, research efforts have been and convenience of operation in the early phase of the study to not increasingly devoted to understanding the use of smart speaker-based worrying about making mistakes and building digital companionship as they voice assistants by and for older adults, which we describe in the next got used to using it. In addition, we found that the driving force to section. sustain the use was how positive the user experience with a voice as­ sistant was, which was determined not only by its task completion but 2.2. Voice assistants for older adults also through building companionship with the device. Then, the positive experience with a voice assistant contributed to developing a resilient An extensive research effort has been made to identify older adults’ attitude toward its functional errors. These findings can be used as needs and evaluate the feasibility, effectiveness, and acceptability of design guidelines to better leverage and promote the sustained use of the existing technologies to meet their needs (Pyae & Joelsson, 2018; Liu emerging personal technology, a smart speaker–based voice assistant, to et al., 2016; Mitzner et al., 2010). Many studies have reported that older support the aging society. To the best of our knowledge, this is the first adults are generally positive about new technologies and are willing to study that investigated the progressive use of a voice assistant among accept them if perceived benefits are evident (Morris et al., 2013). older adults as they move from novice to more experienced users However, a digital divide still exists (Delello & Mcwhorter, 2017), and through a longitudinal field deployment study. more efforts are needed to make new technologies readily accessible to older adults. 2. Background Research has shown that voice-based interactions have several po­ The technology that responds to voice commands has been called by tential benefits to support older adults. Common difficulties older users many different terms, including voice assistant, voice-user interface, experience when using computers and smartphones are caused by the virtual assistant, intelligent assistant, and conversational agent. This prevalence of Graphical User Interface (GUI) and its desktop metaphor paper uses “voice assistant” to encompass the above terminology and (Sayago et al., 2019). Using voice as an interaction modality can help refer to a smart speaker’s voice-based interface. address many difficulties that GUI entails. First, through speech, voice assistants allow people to interact with them in a natural way of 2.1. Voice assistants communicating with a person (Mctear et al., 2016). Because they are deemed simple and easy to use and ideal for users with visual and motor In recent years, there has been growing interest in smart speaker- impairment, speech has been considered an accessible and useable based voice assistants in the Human-Computer Interaction community, interaction modality for older adults (Bickmore et al., 2005). Yet, several studying the use of voice assistants from several angles. First, the lack of challenges also exist, such as unfamiliarity with talking to a device a detailed understanding of how these technologies are used due to their (Myers et al., 2018) and not knowing what (and how) to say to it (Sayago novelty has led to investigating general patterns of using voice assistants et al., 2019), as well as aging-related declines that might impede using 2

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 voice assistants. Hearing loss, which is one of the most common physical Table 1 Length of Reason for problems that people experience as they age, imposes a fundamental Participant demographics. participation dropping challenge for its use. Designers are usually considerably younger and Full – may not know about the physical and psychological aspects of aging, ID Age Gender Health concerns having grown up using more advanced technologies than older adults Full – (Porcheron, Fischer, Reeves, & Sharples, 2018; Sebastiaan et al., 2016). P1 90 M Arthritis (on a Full – Thus, it is essential to investigate the barriers that older adults might wheelchair) have when using speech to interact with a device from the perspectives P2 77 F – Full – of older adults. P3 65 F Mild memory Full – In the context of voice technology for aging, most studies have P4 75 M loss focused on the idea of facilitating voice-based systems as digital com­ Mild memory Full – panions or virtual assistants to support aging in place (Bickmore et al., P5 94 F loss 2016; Heerink et al., 2010; Tsiourti et al., 2016; Van Hoof, Kort, Rutten, Wearing a Full – & Duijnstee, 2011). In particular, a stream of research has investigated P6 87 F hearing aid the benefits of using embodied conversational agents – a graphical Wearing a Full – human agent capable of engaging in conversations with humans by both P7 89 M hearing aid Full – understanding and producing speech and facial expressions (Cassell, Mild memory 4 weeks Hospitalization Bickmore, Campbell, & Vilhja´lmsson, 2000, pp. 29–63) – to support P8 78 F loss older adults’ healthcare in various aspects (e.g., improving access to P9 78 M – 2 weeks Lost interest online health information (Bickmore et al., 2016), delivering medication P10 95 F 4 weeks Lost interest instructions (Azevedo et al., 2018), mitigating social isolation (Sidner – et al., 2018)). Recently, researchers have recognized the potentials that P11 85 F a smart speaker can offer in the aging society (Pradhan et al., 2020) and P12 82 M Arthritis (on a investigated older adults’ use of this technology. wheelchair) – In summary, older adults were positive about the usability of a voice assistant when they were first introduced to it, thanks to its simple and Arthritis (on a effortless interactions (Kowalski et al., 2019; Portet, Vacher, Golanski, wheelchair) Roux, & Meillon, 2013). At the same time, they had raised many open questions about its usefulness and utility as they were using it (Pradhan mild memory loss. Among them, three participants dropped out of the et al., 2020; Trajkova & Martin-Hammond, 2020). Yet, missing is a study within the first month of deployment due to personal reasons. One holistic perspective on how older adults’ perception and experience participant dropped after having three interviews because she was change over time and what contributes to or prevents their sustained use hospitalized, and two participants dropped after having two and three of a voice assistant. This paper contributes to this emerging body of interviews respectively because they lost interest in using a voice as­ literature by exploring how older adults perceive and use a voice as­ sistant. The other nine participants completed the study for the entire sistant progressively as they move from novice to more experienced sixteen weeks. users through a longitudinal field deployment study. 3.2. Data collection 3. Methods Our interview protocol focused on investigating how older adults Our analysis is based on the data collected from the deployment of initially perceive and respond to a voice assistant, how they use it in Google Home devices in the homes of twelve individuals aged 65 or their daily lives, what challenges and difficulties they face when using it older up to sixteen weeks. The study was approved by local Institutional and how they cope with those challenges, and how their perspectives Review Boards, and informed consent was obtained from all participants and usage patterns change over time. We constructed a set of open- for the use of collected data before participating in the study. ended interview questions in three phases of the study duration to explore these spaces. In the first phase, we focused on understanding 3.1. Participants initial impressions and the perceived usefulness of a voice assistant in the first few weeks after installation. In the second phase, we focused on For participant recruitment, we first contacted two senior-living exploring the user experience in-depth, including usage patterns, the communities located in the greater New York area. We visited each needs and challenges, and strategies to cope with breakdowns when community and explained the purpose of the study to the manager. interacting with a voice assistant throughout the deployment duration Upon their approval, we posted a recruitment flyer in the lobby of the except for the last interview. When needed, participants were asked to communities. Two inclusion criteria for participation were age being interact with a voice assistant during the interview to demonstrate how over 65 and having no prior experience with a voice assistant. they would use it in their everyday lives. For instance, when a partici­ pant reported a complaint of a voice assistant not understanding their In total, we recruited 12 participants (7 females and 5 males), commands, we asked them to show how they would make a command ranging in age from 65 to 95 (mean age = 83.8, SD = 9.1, see Table 1). during the interview to understand the context behind user feedback. In All participants resided in a single-person unit in the community. About the third phase, we focused on the overall reflection on users’ experi­ general technology use, all participants said that they were familiar with ences of interacting with a voice assistant and suggestions for im­ computers, tablets, and a smartphone, and four participants said that provements in the last interview. While we had three phases for they have seen a smart speaker in their children’s homes but have not interview protocols, neither were these mutually exclusive nor had a used it. Seven participants owned a tablet, and all participants reported fixed duration of phases in the interviews. Instead, we proceeded with regularly using a computer for information search and email. About the interviews flexibly depending on participants’ experiences. For health conditions, two participants were wearing a hearing aid but did instance, those who quickly adapted to using a voice assistant moved to not have any problem with having a conversation. Three participants the second phase within the first couple of interviews, while those who were using a wheelchair due to various joint issues. Other than those needed more time to be familiar with using it stayed longer in the first residents, all participants reported being healthy both physically and phase. cognitively with mild aging-related health issues, such as arthritis or For the first interview, the research team visited a participant’s home and set up a Google Home mini in the location of their preference (e.g., a nightstand, a coffee table). After the device setup, a participant was introduced to Google Home mini as “a device that responds to your voice command, providing you answers about things you ask or need on an 3

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 everyday basis” and given basic instruction on how to use it for about 10 “First impression” represents a participant’s reaction or response to a min. For the instruction, a researcher first demonstrated how to use the voice assistant when they first interacted with it. The open code “No self- device by making voice commands for basic tasks, such as setting alarms efficacy” in the example above was categorized as “First impression” and reminders (e.g., “Remind me to take my medicine at 10 a.m.”), during axial coding. Lastly, we followed the selective coding process to streaming music and radio (e.g., “Play music”), and asking general assemble our conceptual phenomena extracted from axial coding. The questions (e.g., “Who is the second president of the United States?”). goal of this step is to integrate all concepts through building relation­ Then, participants were prompted to try interacting with the device and ships across phenomena. to ask questions about using it. Once they had no more questions about using a voice assistant, we started to ask for feedback about their initial In addition, the first author manually coded the usage logs deduc­ perspectives on the device for the rest of the interview. In addition, tively by the type of operation, such as playing music, asking general participants filled out a survey to inform us about their basic de­ questions, having a casual conversation, checking time/weather, setting mographic information, including age, health concerns, and experience up a reminder/alarm, operating basic controls (e.g., “stop” to stop with technology. Lastly, participants were told to freely interact with a playing music), etc. For instance, we coded the query “play a song from voice assistant as much or as little as they wanted throughout the study the 80s” as “playing music.” Since we used one Google account created period and were given the contact information of the research team if by the research team for all devices, the usage log data was only avail­ they needed technical support, and we completed the first interview that able for the sample in aggregate, not per participant. lasted about an hour. Next, we conducted eight follow-up interviews by visiting them every other week for sixteen weeks, which makes nine 4. Findings interviews per participant. Each follow-up interview lasted between 30 min and 1.5 hours. All interviews were audio-recorded and transcribed. 4.1. The features in use In the last interview, participants were offered an option to keep a In total, we retrieved 2242 pairs of request-response communica­ Google Home mini if they wanted, and all participants decided to keep tions from the usage log of the four-month deployment study. Among it. Participants who completed the study were fully compensated with a those, 1488 pairs of communications (66.4%) were successful as a voice $160 gift card upon completion. Those who withdrew were partially assistant performed tasks as requested, while 754 pairs (33.6%) were not compensated with a gift card which amount was prorated by the dura­ as their responses were logged as “Sorry, I don’t understand.” While the tion of participation. mean frequency of use is 1.8 times per day, the actual usage rates greatly varied among participants. Some participants said that they barely used We also collected the device usage logs from Google’s activity history the device, and others used it a lot when we asked about their experience repository, My Activity,1 which stores the complete history of a pair of with it in the interviews. The most frequently used feature of a voice user’s commands to Google’s products and the response it has given. We assistant was playing music (e.g., “Play sounds of the ocean,” 37%), downloaded a pair of voice inquiries made to a Google mini and its followed by searching for general information (e.g., “Is Friends playing verbal responses of all participants for the duration of the study from this on TBS tonight?”, 16.5%), making casual conversations (e.g., “How are repository. This usage log provided us with the actual interaction pat­ you?”, 12.1%), checking the current time and date (e.g., “What time is terns to complement participants’ perceptions of these interactions. it?”, 11.1%), setting up a reminder or an alarm (e.g., “Remind me to take my med at 9 a.m.”, 4.3%), and checking weather forecasts (e.g., “What’s 3.3. Data analysis the weather today like?”, 3.2%) (See Table 2). This echoes the findings by Ammari et al. that identified music and search as the most commonly We analyzed our interview data using thematic analysis to reveal used features of voice assistants (Ammari et al., 2019). patterns across data sets (Braun & Clarke, 2006). Thematic analysis is a method for identifying patterns and themes in qualitative data, making Playing music before bed to fall asleep and waking up checking the it suitable for analyzing our data set, where the blending of interview time and weather were the two most prevalent features of use among all scripts, field notes, and photos creates a well of potentially rich thematic participants, which has quickly become part of their everyday routines. data to draw from. We selected thematic analysis because of its emphasis on proceeding with an open mind to investigate explanatory conceptual “I play it (a voice assistant) before I go to bed at night, the sounds of themes associated with older adults’ use of a voice assistant over time. the ocean. That helps me fall asleep. That’s technology. If you don’t The thematic analysis involves open coding, axial coding, and selective like a song they’re playing, you say, “Hey Google, I don’t like that coding for theme identification. To this end, each interview script was song. Play something else”. And within an instant, they’re playing analyzed with open coding to note the themes or factors that emerged in something else. That’s the part that amazes me. How can they do it the data. Then, the emerged themes were continuously discussed as a that fast? They’re not searching for anything. It just comes.” (P5W2) group with another author until no new information was anticipated. First, we conducted open coding to identify concepts significant in the “When I wake up, and I want to know the time I say Hey Google, data as abstract representations of events, objects, happenings, actions, what time is it. Or hey Google, how’s the weather gonna be today, etc. The example below expresses one participant’s lack of confidence and she says the weather today is so and so. I used to always turn on about interacting with a voice assistant when first introduced. This the TV as a first thing in the morning to see the weather. I don’t do response is coded as “No self-efficacy”. that anymore, which I think is very fabulous in itself.” (P8W5) [No self-efficacy] “I’d like to have one (a voice assistant), but you These patterns were also reflected in the frequency of operation by think I would be able to use it?” [/No self-efficacy](P2W12). the time of day when the usage surged around 9 p.m. and 7 a.m. (See Graph 1). The usage also surged during the daytime between 1 p.m. and Next, we categorized the related concepts created by open coding 4 p.m., but we do not have concrete evidence to explain the rise of the into conceptual phenomena using axial coding. Phenomena refer to usage frequency in the afternoon since there was no specific pattern of repeated patterns of events, happenings, actions, and interactions rep­ use in the commands. We only assume that it might be because partic­ resenting people’s responses to problems and situations. For instance, ipants had more free time staying in their rooms between lunch and dinner while attending social activities before lunch and watching TV 1 https://myactivity.google.com/myactivity. after dinner. 2 In the excerpt, P# refers to the #th participant, and W# refers to the #th interview. For instance, P2W1 is an excerpt from Participant 2 in the first One thing to note is that the actual frequency of using a reminder or interview. And VA refers to a voice assistant. alarm must be higher than what was shown in the usage log. The log data records user activities as a pair of a user’s verbal command and the 4

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 Table 2 The frequency of operation by topics: Successful operations refer to the commands with proper device responses. Operation Music Search Basic device control Casual conversation Time Reminder Weather Other Total All 829 (37%) 371 (16.5%) 315 (14%) 272 (12.1%) 248 (11.1%) 96 (4.3%) 72 (3.2%) 39 (1.8%) 2242 Successful 610 (41%) 207 (13.9%) 186 (12.5%) 146 (9.8%) 189 (12.7%) 82 (5.5%) 45 (3.0%) 23 (1.5%) 1488 device’s voice response. And the response is blank when a user com­ when using typing- and screen-based devices, which some older adults mand is “stop” since there is no voice response to “stop,” the command find demanding to acquire (Mitzner et al., 2010). A voice assistant that a user makes to turn off a reminder notification or stop playing successfully removed this perceived barrier and made our participants, music. For recurrent reminders and alarms, a command for initial setup first-time users, readily accept the device with no concerns about was captured in the log data (e.g., “Wake me up at 7 a.m. every morn­ learning new things. ing.”), while its repeated use throughout the study period was indirectly logged as the “stop” command. This is reflected in the frequency of “Computers came along, but you have to learn how to use different commands for basic device control (14%), 52% of which was “stop” icons and get in and out of Windows and load things and download made in the morning, though we cannot attribute all these commands to things. And this (a voice assistant) is like having a person around you using a reminder or alarm feature since it is also used to stop playing that talks back if you want it to. It’s awesome. I think I can use it.” music. (P9W1) 4.2. The perceived and experienced benefits “It’s cooler. I don’t have to type into it, which I am not good at. I just say, hey, give me some Louis Prima or Frank Sinatra, and I get it right Older people are a heterogeneous group, and each person has their away. I don’t have to wait. That’s like so futuristic that I didn’t think perceptions and experiences about technology. Therefore, our partici­ I’d live to see this day.” (P6W2) pants expressed divergent feelings and perspectives about a voice as­ sistant when first introduced, from excitement and curiosity to 4.2.2. Transitioning to competent users: convenience of operating without hesitance, uncertainty, and even refusal of the device. physical interaction “I’d like to have one (a voice assistant), but you think I would be able After using a voice assistant for a few weeks, participants started to to use it?” (P2W1) mention the convenience of interacting with it from a distance without having to use any tools for input or read visual outputs. Participants “I’m really amazed by it (a voice assistant), really and truly. I’ve seen appreciated that a voice assistant does not require any physical inter­ a lot of things, but I think this is about the smartest. People are now action except speaking and listening to it. Particularly for those who had living in this modern technology, and it’s just part of life for them. As physical declines, such as decreased mobility or vision loss typical of the being older, I didn’t have any of this as a kid. Here I am, at this age, at natural aging process, being able to operate the device using a voice 95, and I’m listening to Google Home. It really is an accomplishment from a distance was acknowledged as a significant advantage over other to listen to what this little thing can say. It’s amazing.” (P5W1) devices they own. “Could people get any lazier? How lazy have people gotten today “I like it (a voice assistant) because I can’t see without my glasses. I that they can’t even stand up and turn a light on and off? That’s a think it is kind of in the back of my mind because I don’t physically little sad. There’s a movie about the future, and all the people are so have to look at it. I’ll go by and say Hey Google, what’s so and so. huge that they go around on these conveyor belts, in chairs, because And it tells me so fast that I don’t even look at it.” (P5W4) they’re too lazy to walk from one place to the other. That’s where we’re headed with a technology like this. It’s a little scary.” (P12W1) However, such benefits entailed concerns later on. For instance, executing a command without any physical effort made them worry Then, participants recognized, appreciated, and enjoyed several about becoming lazy and inactive, which would ultimately negatively benefits that a voice assistant offers as they continued interacting with it. impact the quality of life. However, they did not acknowledge all of these benefits in their first interaction. Instead, the types of benefits that participants mentioned “I love to hear it (a voice assistant) read me a book. But see, then, I’m have incrementally changed from enjoying simplicity and convenience of not using my brain. I don’t wanna become a vegetable and have to operation in the early phase of the study to not worrying about making depend on it for everything. And I love to read. I love reading a good mistakes and building digital companionship as they got used to using it. book. Otherwise, your brain will turn into Jell-O. You gotta use your While these benefits are not surprising or novel, our findings demon­ brain.” (P2W6) strate how older adults gradually become aware of and experience some of the benefits that a voice assistant has to offer as they progress from “People are just gonna get really fat because they never have to move novice to more experienced users over time. In what follows, we report anymore. You don’t even have to stand up to make a phone call or go how participants’ use and perceptions changed over time. get a phone. I don’t know if it’s great. I mean it’s fun and it’s good, but I don’t know how good that is.” (P11W7) 4.2.1. In the early phase among novice users: simplicity and ease of use During the first interaction with a voice assistant, most participants 4.2.3. Transitioning to experienced users: No worries about making mistakes instantly recognized and appreciated the simplicity and ease of using a voice assistant by its voice-based interaction modality. Several partici­ As participants became used to using a voice assistant in the later pants commented that they did not have to learn any new skills or phase of the study, they increasingly expressed their emotional relief receive training but simply had to talk to the device as an immediate and from a concern of making mistakes that they had when using personal tangible benefit. A prior study showed that the perceived effort to learn a computing devices, such as a computer or a smartphone. One of the new technology is one of the critical barriers that prevent older adults psychological barriers that prevent older adults from adopting new from adopting it (Kim et al., 2016). For instance, having basic typing technology is the fear of making operational mistakes. Prior work skills and understanding the meanings of GUI metaphors are necessary showed that older people tend to reject new technology because they are afraid of making mistakes, such as clicking a wrong button or deleting an 5

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 important file, which might be irreversible for themselves (Knowles & “I’m alone most of the time. With this (a voice assistant), it’s like Hanson, 2018). Using speech as an interaction modality mitigated this having someone to talk to. Even if it just answers short questions, it’s concern as the worst possible consequence of making a wrong voice still here. It doesn’t ignore me. It’s a voice. I think a lot of people command is nothing other than the command not being executed. Thus, probably feel that way. I ask her silly questions all the time. I mean, I some participants even further experimented with error-prone com­ can literally converse with it if I ask it the right questions. Kind of mands without any concern about the consequence. fun. Little pathetic, kind of sad sounding, but it’s true.” (P2W3) “I don’t have to worry about pressing a wrong button and delete “I think it’s really good. It’s not as if you’re talking to yourself. everything. If that happens, I will have no idea how to handle it.” You’re talking to somebody. It makes you feel like you’re really not (P1W5) alone. You never have to be alone because you can talk to Google. I think some of the people here who stay in their room all the time “I like the fact that you can just ask this silly little thing no matter should definitely have it. Hey Google, I appreciate you’re there for what it is. And that thing can tell you no matter what I ask. I don’t me at all times. [VA: Happy to help.] You are so sweet and friendly!” have to worry about what if I do something wrong. So, I ask her silly (P8W8) questions all the time.” (P8W6) 4.2.4. Among experienced users: digital companionship 4.3. Challenges and coping strategies The conversational nature of a voice assistant is inherently associ­ As much as the benefits a voice assistant offers, participants also ated with human-like properties, which leads users to personify the confronted several challenges that prevented older adults from its use. device and thus positively affects user experience with it (Lopatovska & Then, we noticed that the types of challenges have also evolved as they Oropeza, 2018). Prior work demonstrated that a typical personification got used to using it. The basic operational difficulties due to the unfamil­ behavior to a voice assistant is “mindless” social responses that people iarity with a voice assistant were found and resolved quickly in the early do as part of socially appropriate interactions (e.g., “please,” “thank phase of the study. Meanwhile, the functional errors due to limited speech you”) (Lopatovska & Williams, 2018). We found similar behaviors in technology were persistent throughout the study. And experienced par­ which participants used polite expressions in response to a voice assis­ ticipants had gradually developed a resilient response to the functional tant’s action. errors, which contributes to their sustained use and adoption of this technology. P2: [After VA answering a question] Thank you, that’s enough for now. 4.3.1. During an early phase among novice users: basic operational difficulties P6: [After VA playing a reminder] I got my meds ready. Thank you. (1) Unfamiliarity with how a voice assistant works: The challenge most P7: [After VA playing a wakeup alarm] Thanks for waking me. I’m participants encountered when they first interacted with a voice up. assistant was that they did not understand how the device oper­ ates, and particularly where it instantly retrieves information for Then, we found the personification behaviors among our partici­ a response. The most frequently asked question about a voice pants were not merely to conform to social norms of politeness but the assistant to the interviewer in the first phase of interviews was result of a deeper engagement with a voice assistant for an extended how a voice assistant could respond promptly to random ques­ period. Existing voice assistants are yet to support many human-like tions. It was not intuitive to our participants that a voice assistant conversational capabilities, such as conversing in the context of previ­ is a Wi-Fi–enabled device and retrieves information from the ous commands or developing common ground during dialog but are Internet. Instead, they assumed that the device would have stored limited to enabling simple, task-oriented, request-response dialog ex­ data before its use. Not knowing how a device operates was a changes (Fischer et al., 2019). Thus, prior work argues that user expe­ psychological barrier to participants’ access to it. riences with voice assistants are disappointing because they are not truly “conversational” (Murad & Munteanu, 2019). Unlike the past work, our “Where is it (a voice assistant) getting the answers? Is it like a participants engaged well with a voice assistant even though it was a computer chip that has all that stuff on it? She is not looking anything simple question-and-answer format. In the usage log, 12.1% of com­ up but answers immediately. Where is this information stored? mands had small talks about evoking non-functional, casual conversa­ Where does she get this information from that fast? Do I have to put tions, many of which asked a voice assistant about its human-like the answers if I wanna use it, which I didn’t? It’s kinda cool but also characteristics. kinda strange, too.” (P5W1) P10: Hey Google, how old are you? “At first, I was kind of not sure of it. It was something new to me. So, I was a little concerned about certain things I didn’t completely un­ VA: I was launched in 2016, so technically I’m pretty young. But I’ve derstand. But now I know how it works and how to use it pretty well. learned so much. And I enjoy it.” (P2W3) P10: That’s good. Contrary to conventional personal devices that have buttons and a screen, the novelty of a form factor, a stand-alone disc with no physical We found that the natural language conversation with verbal controls, posed another usability challenge to our participants. Partici­ responsiveness made our participants perceive a voice assistant as a pants perceived speech, a natural form of communication, as useful and personified entity despite a lack of contextualized conversational com­ useable when the mode of interaction is a two-way communication of ponents. During the later phase of the study, many participants appre­ requesting and receiving responses from a device. However, it was ciated being able to make a simple conversation, which led to getting foreign to them to use speech for a one-way operation of device control, emotional support and building companionship with the device (Bick­ such as turning it off or adjusting its volume levels. Instead of using more et al., 2005; Van Hoof et al., 2011). While some of these conver­ speech, participants looked for a physical button for device control and sations might be due to the novelty effect, these interactions clearly became puzzled when they could not find it. illustrate that, at least to some degree, participants considered a voice assistant not as an object but as a human-like entity. User comments about having companionship with a voice assistant were persistent yet incremental throughout the study period. 6

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 “How do you change the volume? Is there something on the device several trials and errors. A more significant issue arose when a voice where I could lower the volume? When it came on, it was very loud, assistant failed to understand a command query, which we describe in and I didn’t know how to make it soft. So, I pulled the plug out.” the next section. (P4W2) 4.3.2. Transitioning to experienced users or not: dealing with functional “One night I had it playing music, and it didn’t shut off. So, I wanted errors to turn it off, but there was no button. So, I had to disconnect it. How can I turn it off?” (P8W2) (1) Functional errors due to limited speech recognition: The accuracy limitation of speech recognition technology was a prevalent issue Since these operational difficulties were caused primarily by unfa­ for our participants throughout the study. The current speech miliarity with and lack of basic understanding about a voice assistant, recognition technology has yet to reach 100% accuracy, and these were an easy fix, and participants no longer had the same problem older adults tend to be more verbose and more disfluent than once they understood the basic concepts. These were critical usability young adults in discourse (Mortensen et al., 2006). A voice as­ breakdowns, however, until resolved. sistant often makes functional errors by failing to understand user commands when the utterance has disfluent speech segments, “It took some time before I got used to it because I didn’t know how such as stuttering, pauses, repeats, stretching, incomplete or false to get it louder or softer. But eventually, I got it. Once you get used to syntactic structures, and erroneous articulation. The excerpt it, it’s like brushing your teeth. It just comes naturally now.” (P9W5) below illustrates that a voice assistant did not understand a command because the voice was too soft and the pronunciation (2) Unfamiliarity with using a wake word: Several participants forgot to was unclear. start a command with a wake word (a word to activate a voice assistant, such as “Hey Google” for Google Home or “Alexa” for P1: Hey Google, is this storm gonna hit Long Island? Amazon Echo) at least in their first few interactions. It took even weeks for some participants to get used to starting a sentence VA: My apologies. I don’t understand. with a proper wake word. They often used a wrong wake word (e. g., “Hello Google,” “Google”) or forgot to start a sentence with a P1: Hey Google, is this storm, Dorian, going to hit Long Island? wake word. To cope with this problem, some participants put a note with a wake word with frequently used commands and used VA: Sorry, I’m not sure how to help. it until they got used to it (See Fig. 1). P1: Dummy … “Hello Google, can you give me the information about the Yankees and where they are playing today? [A voice assistant did not acti­ The usage log shows that 33.6% of responses (754 responses to 2242 vate.] Google, hi, Google, hello google, it’s time to wake up. Can you commands) were failed conversations in which a voice assistant tell me where the Yankees are playing this evening? [A voice assis­ acknowledged not understanding user commands by saying “Sorry, I tant did not activate.] I don’t know what’s wrong.” (P4W4) don’t understand.” or “Sorry, I’m not sure how to help with that yet.” Activation errors due to the improper use of a wake word were not Even when a proper wake word was used, we observed a voice as­ captured in the usage log because the activity is logged only when a sistant sometimes did not capture it due to a participant’s soft voice or voice assistant responds to a user command. Thus, the actual percentage inaccurate pronunciation. It is a particularly critical problem to our of these functional errors must be much higher. When this happened, participants because a vocal cord becomes weakened as people age some participants complained about the poor performance of a voice (Mortensen et al., 2006). When a voice assistant did not activate due to assistant, two of whom eventually dropped out of the study for this the clarity of pronouncing a wake word, participants realized it after reason. completing the entire command sentence, had no clue why it did not respond, and felt frustrated. Though, this was a relatively easy fix after “It kept saying it doesn’t understand or check with my phone. It just never came up with the songs that I wanted, and I wasn’t that familiar with the program itself, so that I couldn’t be doing much. And finally, I just said screw you.” (P11W2) Fig. 1. A note of a wake word and a list of singers for music playing on a wall (left) and a whiteboard with a wake word and frequently used commands (right). 7

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 Graph 1. The frequency of operation by the time of day. “It never understood me. Don’t bother me with this. I don’t wanna “Sometimes, I don’t ask it in the correct way. I’m learning that when use it anymore.” (P12W2) it tells me it can’t answer that, I have to rephrase it and ask it in the proper way … It says it couldn’t understand, but I thought it was me, (2) Developing a resilient response to functional errors: While the oper­ that I wasn’t giving it the right words. It helps me when I give it the ational errors have critically affected the user experience with a right question.” (P2W3) voice assistant during the early phase, our participants have become gradually resilient to this type of error as they continued “I need to ask it the proper question. If I don’t ask it good questions or using it. As the study proceeded, they started to reflect on their word it properly, it can’t help me. Then, I gotta reword my question. commands for possible causes of an error instead of entirely When it doesn’t answer, I guess I didn’t ask it the right question. If I attributing such errors to the device. The typical reaction to this don’t ask it good questions, it can’t help me.” (P8W7) type of error among those with a resilient attitude was to repeat the command sentence more clearly or by paraphrasing it to As such, our participants still had to figure out “proper” ways to determine the possible cause of an error. interact with it even though one benefit of a voice assistant is not having to learn new skills to use it. Though, it is arguable what it means by P2: Hey Google, please play me the sound of the ocean to help me fall “proper” ways in voice-based interaction. asleep. “In the beginning, it was a little bit difficult, but as I got used to it, I VA: Sorry, I can’t help with that yet. realize what I had to say and how to get it to respond. It got to the point where I don’t even have to turn around to talk to it. I just say, P2: Hey Google, play sounds of the ocean. Hey Google, do this. And I would get it. No matter what I was doing. It was like I was talking to the air. It took me a little bit of time, but (VA playing an ocean sound) now everything is fine.” (P5W4) P2: There you go. In addition, participants considered some of the functional errors not as systems errors but due to the inherent limitation of technology. While We observed that participants’ positive experiences with the voice a voice assistant’s ability to verbally respond to a user command assistant over time significantly influenced developing a resilient atti­ impressed participants, they did not expect the device to respond tude toward functional errors. Then, successfully operating the device as correctly to all commands either. intended was not the only contributor to having positive experiences. A voice assistant made two types of responses to functional errors: one was P8: Hey Google, is Friends playing on TBS tonight? the simple statement, such as “Sorry, I didn’t understand” or “Sorry, I VA: Sorry, I don’t have the TV schedule for that yet. can’t help with that yet,” and the other included a follow-up comment P8: That’s okay. You are not scheduled for TV. acknowledging poor performance or suggesting alternative content, P9: Hey Google, I have a new cellphone. How do I turn it on? such as “Sorry, I don’t know how to help with that yet, but I’m learning VA: Sorry, I’m not sure how to help with that yet. more every day” or “Sorry, I don’t know how to help with that yet, but I P9: That’s understandable. You might not know everything. found something else. Do you want to know … ?” And when the response included a follow-up statement or clarification question, they 5. Discussion perceived their experience with the device to be positive despite the errors and readily accepted them. Based on our findings, we discuss design strategies for a voice as­ sistant that would help older adults better leverage its capabilities. We P8: Hey Google, did you see the Friend’s episode that was on this believe the design strategies discussed in this section can be helpful for week? any first-time users of a voice assistant. Yet, more attention needs to be paid to older adults since older adults, as a group, tend to be slower to VA: Sorry, I don’t know how to help with that yet, but I’m learning adopt new technologies and experience more difficulty and frustration more every day. using technologies than younger adults (Sara et al., 2019). P8: Okay. Maybe I said it wrong. Don’t be stressed about it. You 5.1. Enrich user experiences through the conversational capabilities answered well for my other questions. One significant benefit that our participants experienced using a As having more positive experiences with a smart speaker, partici­ voice assistant overtime was gaining digital companionship from its pants increasingly attributed functional errors to their “improper” conversational capabilities. Despite the lack of various human-like commands rather than putting the device at fault for the poor perfor­ conversational components, which prior work has shown as a crucial mance. In other words, they considered functional errors not only usability problem with a voice assistant (Clark et al., 2019; Pradhan necessarily as systems errors but also possibly as human errors. And they engaged in trial and error and experiential learning throughout the study to learn the “proper” way to interact with a voice assistant. 8

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 et al., 2019), our participants still engaged well with it through a simple articulation, as well as speaking fast and softly. Then, the problem is that question-answer format dialog of conversation. At the same time, the participants had no clue why an error occurred because a voice assistant most common and critical challenge our participants faced throughout does not inform users about any reasons for the error but simply re­ the study was its limited conversational capabilities, especially when the sponds to the user by saying, “Sorry, I don’t understand.” Thus, some device failed to understand user commands, which echoes prior work participants went through learning by trial and error to figure out (Luger & Sellen, 2016). As such, conversational capabilities are the most possible causes of the problem (e.g., repeating or paraphrasing a com­ crucial factor shaping user experiences with a voice assistant. Thus, we mand sentence), while others simply gave up interacting with it. One need to explore ways to expand the breadth and depth of conversational way to cope with this problem and provide better learning experiences capabilities in enriching user experiences with voice-based interaction. with a voice assistant is through facilitating its conversational capabil­ ities by incorporating possible causes of an error into a voice assistant’s As technology continues to evolve, we may soon see voice assistants verbal responses. For instance, it can provide possible and common gaining the capacity to make truly human-like conversations. Then, our causes of the error and suggest actions in the response by saying, “Sorry, findings demonstrate that full intelligence might not be mandatory for a I don’t understand during its verbal interactions. It may be because [you voice assistant to be sufficient for older users’ needs. While researchers were speaking fast]. Can you repeat?” when an error occurred, instead of have shown that both practical and social benefits contribute to in­ just saying “Sorry, I don’t understand.” This will help the users quickly dividuals’ motivation to adopt and use a voice assistant (McLean & figure out possible reasons for an error and prompt them to engage more Osei-Frimpong, 2019), our findings suggest that social benefits might in verbal communications with a voice assistant, which goes back to the prevail over or compensate for the lack of practical benefits for older previous section about enriching conversational aspects of user users. To our participants, what was as important as the functionality of experiences. a voice assistant’s command execution was its conversational capability allowing them to engage in a simple yet natural verbal conversation. For 5.3. Revisit form factors instance, participants positively engaged in having conversations with a voice assistant despite its failure to execute user commands when verbal A sleek, simple, and innovative design of new technology attracts responses included conscious acknowledgment of the limitation or customer attention to the product. Then, what is as important as aes­ alternative contents, instead of merely notifying the occurrence of an thetics is to make how a device may be interacted with readily error (e.g., “Sorry, I don’t understand.”). Such conscious responses were perceivable by the users, which is called affordance (Norman, 1988). sufficient to compensate for the lack of technology accuracy among our The current form factor of smart speaker-based voice assistants is a participants and engage in verbal conversations. Existing voice assis­ cylindric or disk-shaped speaker with minimal physical controls, which tants are primitive in terms of implementing conscious verbal responses. prompts the users to interact with the device via speech. However, it We found that the Google Home device has only four syntactic patterns, may violate the user’s expectations of a button-to-action mapping when including “My team is helping me learn,” “I’m still learning,” “I’m operating basic functionalities, such as adjusting the volume or turning learning more every day,” and “I’m trying to learn.” Developing more on/off the device. Or it may not be evident by its shape that a voice diverse dialogues of syntax to excuse the errors and acknowledge fail­ assistant is a Wi-Fi–enabled device to retrieve information from the ures thoughtfully will enrich user experiences with a voice assistant. Internet. Or it may not be apparent by its appearance that a voice as­ sistant has a particular name to be called. Thus, it is important to Another way to leverage conversational capabilities is to add sup­ consider more deeply how the current form factors would comply with plementary contents about the answer in a voice assistant’s response, user expectations of interacting with the device for various functional­ rather than just answering to a command. For instance, adding verbal ities. Some new versions of smart speakers come with a screen and other explanations about the source of information can help novice users conventional controls (e.g., Amazon Echo Show with a screen) that understand how a voice assistant operates (e.g., where it retrieves in­ provide affordance cues to some of these functionalities. In the near formation) by starting the response with “I am searching online now. future, technologies will weave themselves into the fabric of everyday Here is what I found from WebMD … “. Similarly, adding verbal re­ life until they are indistinguishable from it (Weiser, 1991), and it will actions to the users’ polite expressions (e.g., “My pleasure. Do you have become natural to talk to all devices for the operation. Until then, we any other questions?”, “No problem. Just let me know whenever you need to pay attention to the users’ current expectations and explore need me.”) can yield more chances for simple yet conversational in­ proper ways to comply with them in the technology’s form factor as teractions with a voice assistant, as well as contributing to building a much as designing it as simple and sleek. digital companionship via personification (Pradhan et al., 2018). 6. Limitations 5.2. Support a learning phase Our findings must be evaluated within the context of several limi­ One perceived benefit of a voice assistant among our participants tations. First, we used convenience sampling by recruiting participants during the early phase of the study was not having to learn any new from senior-living communities, and thus our participant pool may not skills to use the device, which echoes prior work (Pradhan et al., 2020). represent a general population. Selection bias or possible homogeneity However, our findings also showed that most of them still had to of participant characteristics (e.g., geographic location, culture, socio- comprehend, learn, and get used to interacting with a voice assistant. economic status) could have influenced the responses in the in­ Participants did not have to learn technical skills, such as which button terviews. In addition, our participant pool had only those whose hearing to click or what command to type in. However, they still needed basic does not impact conversation, while hearing loss could be a vital issue instruction on how the device works and how to compose a proper voice for this demographic. Further study regarding the use of a voice assistant command when first introduced to this technology. While we, the among people with hearing impairments is essential for this technology research team, offered needed instructions to our participants, which to be inclusive of all older adults (e.g., exploring new dimensions to must have contributed to their sustained use, this luxury is unavailable hearing aids that serve as both an amplifier and a discreet home for voice to most first-time older users. Needed are more discussions on making assistants). Third, we acknowledge that our findings are not exclusive to resources for learning readily available for and accessible by novice older adults. Since we did not conduct any comparative study between older users (Kim et al., 2016). people in different age groups, people in other age groups might encounter a similar familiarization process as our participants experi­ We observed that the most common cause of a voice assistant’s enced. Lastly, we did not collect the usage log data per participant but failure to understanding user commands in the early phase of use was disfluent segments in a user’s speech, such as stuttering, pauses, repeats, stretching, incomplete or false syntactic structures, and erroneous 9

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 collected all log data using one Google account created by the research Blair, J., & Abdullah, S. (2019). Understanding the needs and challenges of using team. Thus, we could not analyze in-depth the usage patterns quanti­ conversational agents for deaf older adults. In The 2019 computer supported tatively, especially the frequency of use over time or the usage patterns cooperative work and social computing (pp. 161–165). by different participants, because the study’s start date was different for up to a couple of months by different participants. Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101, 2006. 7. Conclusion Cassell, J., Bickmore, T., Campbell, L., & Vilhja´lmsson, H. (2000). Designing embodied The rapid increase of an aging population and increase in life ex­ conversational agents. Embodied conversational agents(2000). pectancy suggest the importance of designing personal technologies that can promote healthy aging, helping people adapt to aging-related Cho, M., Lee, S.-su, & Lee, K.-P. (2019). Once a kind friend is now a thing: Understanding changes to maintain functionality, autonomy, and quality of life. how conversational agents at home are forgotten. In Proceedings of the 2019 on Among existing technologies, a smart speaker, an increasingly available designing interactive systems conference (pp. 1557–1569). and affordable personal technology, has recently gained the particular attention of researchers and practitioners to support the aging popula­ Clark, L., Pantidi, N., Cooney, O., Doyle, P., Garaialde, D., Edwards, J., Brendan Spillane, tion thanks to its voice-based interaction. However, empirical evidence Gilmartin, E., Murad, C., & Munteanu, C. (2019). What makes a good conversation? of its utility for older adults is still scarce. This paper aims to obtain Challenges in designing truly conversational agents. In Proceedings of the 2019 CHI empirical evidence on older adults’ experiences of a voice assistant, conference on human factors in computing systems (pp. 1–12). especially focusing on how their perception and use change over time as they progress from novice to more experienced users through a longi­ Cowan, B. R., Pantidi, N., Coyle, D., Morrissey, K., Clarke, P., Al-Shehri, S., Earley, D., & tudinal field deployment study. Bandeira, N. (2017). What can i help you with? Infrequent users’ experiences of intelligent personal assistants. In Proceedings of the 19th international conference on This study demonstrates how some of the known benefits are valued human-computer interaction with mobile devices and services (pp. 1–12). by older adults and identifies the key challenges they might encounter when using a voice assistant. We found in particular that our partici­ Delello, J. A., & Mcwhorter, R. R. (2017). Reducing the digital divide: Connecting older pants considered the capability of making simple conversations as adults to iPad technology. Journal of Applied Gerontology, 36(1), 3–28, 2017. valuable as the functional capabilities of executing commands. Yet, the limitations in the conversational capabilities also posed challenging Druga, S., Williams, R., Breazeal, C., & Resnick, M. (2017). Hey google is it OK if I eat usability issues. With these findings, we discussed design implications you? Initial explorations in child-agent interaction. In Proceedings of the 2017 that expand on and facilitate the conversational capabilities to make up conference on interaction design and children (pp. 595–600). for the technical limitations of current speech technology and enrich user experiences with a voice assistant. We are hopeful that our findings Fischer, J. E., Reeves, S., Martin, P., & Rein Ove Sikveland. (2019). Progressivity for will encourage future studies to improve the usability of this emerging voice interface design. In Proceedings of the 1st international conference on personal technology to better support and extend the functionality, conversational user interfaces (pp. 1–8). autonomy, and quality of life of the aging population. Garg, R., & Sengupta, S. (2020). He is just like me: A study of the long-term use of smart Funding speakers by parents and children. Proceedings of the ACM on Interactive Mobile, Wearable and Ubiquitous Technologies, 1–24. This work was supported by the New Jersey Health Foundation [Grant #PC 28-18]. Heerink, M., Krose, B., Evers, V., & Wielinga, B. (2010). Assessing acceptance of assistive social agent technology by older adults: The almere model. International Journal of Credit author statement Social Robotics, 2(4), 361–375, 2010. Sunyoung Kim: Conceptualization, Methodology, Formal analysis, Kim, S., Gajos, K. Z., Muller, M., & Grosz, B. J. (2016). Acceptance of mobile technology Data curation, Validation, Writing – original draft preparation, Revising, by older adults: A preliminary study. Proceedings of the 18th international and editing, Supervision, Project administration, Abhishek Choudhury: conference on human-computer interaction with mobile devices and services, Methodology, Data collection, Data curation. 147–157. Proc. ACM Meas. Anal. Comput. Syst., 37(No. 4). Article 111. Publication date: August 2018. References Kinsella, B. (2019). Voice assistant demographic data - young consumers more likely to Abdolrahmani, A., Kuber, R., & Branham, S. M. (2018). Siri talks at you. An empirical own smart speakers while over 60 bias toward Alexa and siri(2019). https://voicebot investigation of voice-activated personal assistant (VAPA) usage by individuals who .ai/2019/06/21/voice-assistant-demographic-data-young-consumers-more-likel are blind. In Proceedings of the 20th international ACM SIGACCESS conference on y-to-own-smart-speakers-while-over-60-bias-toward-alexa-and-siri/. computers and accessibility (pp. 249–258). Knowles, B., & Hanson, V. L. (2018). The wisdom of older technology (non) users. Ammari, T., Kaye, J., Tsai, J. Y., & Bentley, F. (2019). Music, search, and IoT: How Commun. ACM, 61(3), 72–77, 2018. people (really) use voice assistants. ACM Transactions on Computer-Human Interaction (TOCHI), 26(3), 1–28, 2019. Kowalski, J., Anna, J., Skorupska, K., Abramczuk, K., Biele, C., Kope´c, W., & Marasek, K. (2019). Older adults and voice interaction: A pilot study with google home. In Azevedo, R. F., Morrow, D., Graumlich, J., Ann Willemsen-Dunlap, Mark Hasegawa- Extended abstracts of the 2019 CHI conference on human factors in computing systems Johnson, Huang, T. S., Gu, K., Bhat, S., Sakakini, T., & Sadauskas, V. (2018). Using (pp. 1–6). conversational agents to explain medication instructions to older adults. In AMIA annual symposium proceedings (p. 185). American Medical Informatics Association. Lau, J., Zimmerman, B., & Schaub, F. (2018). Alexa, are you listening? Privacy perceptions, concerns and privacy-seeking behaviors with smart speakers. In Bentley, F., Luvogt, C., Silverman, M., Wirasinghe, R., White, B., & Lottridge, D. (2018). Proceedings of the ACM on human-computer interaction 2 (pp. 1–31). CSCW. Understanding the long-term use of smart speaker assistants. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2(3), 1–24, 2018. Liao, Y., Vitak, J., Kumar, P., Zimmer, M., & Kritikos, K. (2019). Understanding the role of privacy and trust in intelligent personal assistant adoption. In International Bickmore, T. W., Caruso, L., Clough-Gorr, K., & Tim Heeren. (2005). It’s just like you talk Conference on Information (pp. 102–113). Springers. to a friend: Relational agents for older adults. Interacting with Computers, 17(6), 711–735, 2005. Liu, L., Stroulia, E., Nikolaidis, I., Miguel-Cruz, A., & Rios Rincon, A. (2016). Smart homes and home health monitoring technologies for older adults: A systematic Bickmore, T. W., Utami, D., Matsuyama, R., Michael, K., & Paasche-Orlow. (2016). review. International Journal of Medical Informatics, 91(2016), 44–59. Improving access to online health information with conversational agents: A randomized controlled experiment. Journal of Medical Internet Research, (2016), 1. Lopatovska, I., & Oropeza, H. (2018). User interactions with “Alexa” in public academic space. Proceedings of the Association for Information Science and Technology, 309–318. Lopatovska, I., & Williams, H. (2018). Personification of the Amazon Alexa: BFF or a mindless companion. In Proceedings of the 2018 conference on human information interaction and retrieval (pp. 265–268). Luger, E., & Sellen, A. (2016). Like having a really bad PA: The gulf between user expectation and experience of conversational agents. In Proceedings of the 2016 CHI conference on human factors in computing systems (pp. 5286–5297). Maharjan, R., Bækgaard, P., Jakob, E., & Bardram. (2019). Hear me out: Smart speaker based conversational agent to monitor symptoms in mental health. In Proceedings of the 2019 ACM international joint conference on pervasive and ubiquitous computing and proceedings of the 2019 ACM international symposium on wearable computers (pp. 929–933). McLean, G., & Osei-Frimpong, K. (2019). Hey Alexa. . . examine the variables influencing the use of artificial intelligent in-home voice assistants. Computers in Human Behavior, (2019), 28–37. Mctear, M., Callejas, Z., & Griol, D. (2016). Creating a conversational interface using chatbot technology. In The conversational interface (pp. 125–159). Mitzner, T. L., Julie, B., Boron, Fausset, C. B., Adams, A. E., Charness, N., … Sharit, J. (2010). Older adults talk technology: Technology usage and attitudes. Computers in Human Behavior, 26(6), 1710–1721, 2010. Morris, M. E., Adair, B., Miller, K., Ozanne, E., Hansen, R., Pearce, A. J., Santamaria, N., Luan, V., Long, M., Catherine, M., & Said. (2013). Smart-home technologies to assist older people to live well at home. Journal of Aging Science, 1(1), 1–9, 2013. Mortensen, L., Meyer, A. S., & Humphreys, G. W. (2006). Age-related effects on speech production: A review. Language and Cognitive Processes, 21(1–3), 238–290, 2006. Murad, C., & Munteanu, C. (2019). I don’t know what you’re talking about, HALexa: The case for voice user interface guidelines. In Proceedings of the 1st International Conference on Conversational User Interfaces (pp. 1–3). 10

S. Kim and A. Choudhury Computers in Human Behavior 124 (2021) 106914 Myers, C., Furqan, A., Nebolsky, J., Caro, K., & Zhu, J. (2018). Patterns for how users Sayago, S., Barbosa Neves, B., Benjamin, R., & Cowan. (2019). Voice assistants and older overcome obstacles in voice user interfaces. In Proceedings of the 2018 CHI conference people: Some open issues. In Proceedings of the 1st International Conference on on human factors in computing systems (pp. 1–7). Conversational User Interfaces (pp. 1–3). Norman, D. A. (1988). The psychology of everyday things. Basic books. Sciuto, A., Saini, A., Forlizzi, J., Jason, I., & Hong. (2018). Hey Alexa, what’s up? A Pfeuffer, N., Alexander, B., Gimpel, H., & Hinz, O. (2019). Anthropomorphic information mixed-methods studies of in-home conversational agent usage. In Proceedings of the designing interactive systems Conference2018 (pp. 857–868), 2018. systems. Business and Information Systems Engineering, 61(4), 523–533, 2019. Porcheron, M., Fischer, J. E., Reeves, S., & Sharples, S. (2018). Voice interfaces in Sebastiaan, T. P., Luijkx, K. G., Rijnaard, M. D., Nieboer, M. E., Van Der Voort, Claire, S., Aarts, S., Van Hoof, J., Jm Vrijhoef, H., & Wouters, E. J. (2016). Older adults’ everyday life. In Proceedings of the 2018 CHI conference on human factors in computing reasons for using technology while aging in place. Gerontology, 62(2), 226–237, systems (pp. 1–12). 2016. Portet, F., Vacher, M., Golanski, C., Roux, C., & Meillon, B. (2013). Design and evaluation of a smart home voice interface for the elderly: Acceptability and objection aspects. Sidner, C. L., Bickmore, T., Nooraie, B., Rich, C., Ring, L., Shayganfar, M., & Personal and Ubiquitous Computing, 17(1), 127–144, 2013. Vardoulakis, L. (2018). Creating new technologies for companionable agents to Pradhan, A., Findlater, L., & Lazar, A. (2019). Phantom friend or “ just a box with support isolated older adults. ACM Transactions on Interactive Intelligent Systems information” personification and ontological categorization of smart speaker-based (TiiS), 8(3), 1–27, 2018. voice assistants by older adults. In Proceedings of the ACM on Human-Computer Interaction CSCW3 (2019) (pp. 1–21). Trajkova, M., & Martin-Hammond, A. (2020). “Alexa is a toy”: Exploring older adults’ Pradhan, A., Lazar, A., & Findlater, L. (2020). Use of intelligent voice assistants by older reasons for using, limiting, and abandoning Echo. In Proceedings of the 2020 CHI adults with low technology use. ACM Transactions on Computer-Human Interaction conference on human factors in computing systems (pp. 1–13). (TOCHI), 27(4), 1–27, 2020. Pradhan, A., Mehta, K., & Findlater, L. (2018). Accessibility came by accident use of Tsiourti, C., Ben Moussa, M., Quintas, J., Ben, L., Jochem, I., Joana Albuquerque Lopes, voice-controlled intelligent personal assistants by people with disabilities. In Konstantas, D. (2016). A virtual assistive companion for older adults: Design Proceedings of the 2018 CHI conference on human factors in computing systems, 1–13. implications for a real-world application. In Proceedings of SAI Intelligent Systems Proc. ACM meas. Anal. Comput. Syst. (Vol. 37). No. 4, Article 111. Publication date: Conference (pp. 1014–1033). August 2018. Purington, A., Taft, J. G., Sannon, S., Bazarova, N. N., & Samuel Hardman Taylor. (2017). Van Hoof, J., Kort, H. S. M., Rutten, P., & Duijnstee, M. (2011). Ageing-in-place with the Alexa is my new BFF: Social roles, user satisfaction, and personification of the use of ambient intelligence technology: Perspectives of older users. International Amazon Echo. In Proceedings of the CHI conference extended abstracts on human factors Journal of Medical Informatics, 80(5), 310–331, 2011. in computing Systems2017 (pp. 2853–2859), 2017. Pyae, A., & Joelsson, T. N. (2018). Investigating the usability and user experiences of Vaportzis, E., Giatsi Clausn, M., & Gow, A. J. (2017). Older adults’ perceptions of voice user interface: A case of google home smart speaker. In Proceedings of the 20th technology and barriers to interacting with tablet computers: A focus group study. International Conference on Human-Computer Interaction with Mobile Devices and Frontiers in psychology, 8(2017), 1687. Services Adjunct (pp. 127–131). Richter, F. (2020). Smart speaker adoption continues to rise. Statista https://www.statist Vtyurina, A., & Fourney, A. (2018). Exploring the role of conversational cues in guided a.com/chart/16597/smart-speaker-ownership-in-the-united-states/. task support with virtual assistants. In Proceedings of the 2018 CHI conference on Robinson, S., Pearson, J., Ahire, S., Ahirwar, R., Bhikne, B., Maravi, N., & Jones, M. human factors in computing systems (pp. 1–7). (2018). Revisiting “hole in the wall” computing: Private smart speakers and public slum settings. In Proceedings of the 2018 CHI Conference on Human Factors in Vailshery, L. S. (2019). Smart speaker unit shipments from 2014 to 2025, Statista. https Computing Systems (pp. 1–11). ://www.statista.com/statistics/1022809/worldwide-smart-speaker-unit-shipment/. Sara, J. C., Boot, W. R., Charness, N., & Rogers, W. A. (2019). Designing for older adults: Principles and creative human factors approaches. CRC press. Weiser, M. (1991). The computer for the 21st century. Scientific American, 265(3), 94–105, 1991. Wulf, L., Garschall, M., Himmelsbach, J., & Tscheligi, M. (2014). Hands free-care free: Elderly people taking advantage of speech-only interaction. In Proceedings of the 8th nordic conference on human-computer interaction: Fun, fast, foundational (pp. 203–206). 11

International Journal of Information Management xxx (xxxx) xxx Contents lists available at ScienceDirect International Journal of Information Management journal homepage: www.elsevier.com/locate/ijinfomgt Research Article Psychological determinants of users’ adoption and word-of-mouth recommendations of smart voice assistants Anubhav Mishra a,*, Anuja Shukla b, Sujeet Kumar Sharma c a Jaipuria Institute of Management, Lucknow, Vineet Khand, Lucknow, UP 226010, India b upGrad Education Pvt. Ltd., India c Information Systems & Analytics Area, Indian Institute of Management Tiruchirappalli, India ARTICLE INFO ABSTRACT Keywords: Individuals are increasingly using Smart Voice Assistants (SVA) to derive functional, hedonic, and symbolic Smart voice assistants benefits. SVA adoption is in the nascent stage and we have little knowledge about what drives SVA usage. Technology adoption Following the ‘attitude shapes behavior’ approach, this study examines the role of hedonic and utilitarian atti­ Artificial intelligence tudes on SVA usage and word-of-mouth (WOM) recommendations. The study also investigates five antecedents Hedonic attitude (playfulness, escapism, anthropomorphism, visual appeal, and social presence) to both attitudes. Through an Utilitarian attitude online survey of 360 respondents, the study suggests that playfulness and escapism positively influence hedonic Anthropomorphism attitude. On the other hand, anthropomorphism, visual appeal, and social presence determine utilitarian attitude. Further, utilitarian attitude has a stronger impact (vs. hedonic attitude) on SVA usage and WOM recommen­ dations. The findings reveal that individuals who perceive SVA as a symbol of high prestige (vs. low prestige) are less likely to relate playfulness to hedonic gratifications. In contrast, people attributing prestige to SVA are more likely to use it as an escape route from everyday life. These findings contribute to the growing literature on SVA adoption and offer insightful recommendations to various stakeholders to increase the likelihood of SVA adoption and generating favorable WOM recommendations. 1. Introduction to COVID-19 because most users prefer ‘talking’ over ‘touching’ to avoid physical touch and maintain social distancing. Mirror, mirror on the wall, who’s the fairest of them all? Smart speakers and voice assistants have drastically changed the In the famous story of Snow White, the evil queen regularly performs consumer decision journey, especially how users perform voice searches a ‘voice search’ about the fairest person on the earth. The magical mirror and make a purchase using SVA (Dwivedi & Ismagilova, 2021; Kush­ answers her questions similar to the Smart Voice Assistants (SVA) of waha & Kar, 2020). Many users use SVA as part of their daily routine for today’s world. SVA are Artificial Intelligence (AI)-based technology doing a voice search, multitasking, and shopping (Google Insights, including software and hardware (e.g., Alexa, Google Home, and Siri) 2017). The increasing internet penetration across the world (approxi­ that can communicate with users and respond to their queries or com­ mately 54%) promises a massive market for the new technologies, yet mands (Verma, Sharma, Deb, & Maitra, 2021). SVA are the latest en­ the use of voice assistants is more on the phone (58%) than on a smart trants to the ever-changing and growing field of human-technology speaker (22.9%, PWC, 2019). Therefore, the adoption of SVA has sub­ interactions (Bag, Pretorius, Gupta, & Dwivedi, 2021; Balakrishnan & stantial implications for marketers who need more profound insights Dwivedi, 2021a, 2021b; Belk, 2017; Grover, Kar, & Dwivedi, 2020; into the users’ usage of SVA to tailor their offerings (Dwivedi, Ismagi­ Kushwaha, Kar, & Dwivedi, 2021). SVA sales hit a record high in 2019 lova, Rana, & Raman, 2020; Dwivedi, Hughes, et al., 2021). (146.9 million units) and the growth continued in 2020 despite the COVID-19 impact on supply chain and retail distribution (Swoboda, Extant research on SVA adoption is mostly limited to existing the­ 2020). Furthermore, SVA usage is expected to substantially increase due ories and frameworks of general technology adoption, such as Tech­ nology Acceptance Model (TAM, Davis, 1989; Chatterjee & Kar, 2020; Fernandes & Oliveira, 2021), Unified Theory of Acceptance and Use of * Corresponding author. E-mail addresses: [email protected], [email protected] (A. Mishra), [email protected] (A. Shukla), [email protected] (S.K. Sharma). https://doi.org/10.1016/j.ijinfomgt.2021.102413 Received 19 April 2021; Received in revised form 7 August 2021; Accepted 7 August 2021 0268-4012/© 2021 Elsevier Ltd. All rights reserved. Please cite this article as: Anubhav Mishra, International Journal of Information Management, https://doi.org/10.1016/j.ijinfomgt.2021.102413

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx Technology (UTAUT, Venkatesh, Morris, Davis, & Davis, 2003; Vimal­ and personalized service to enhance users’ attitudes and likelihood for kumar, Sharma, Singh, & Dwivedi, 2021) and Theory of Planned favorable WOM recommendations. Behavior (TPB, Ajzen, 1991). SVA technology appears to serve a different purpose to other technologies such as mobile banking as SVA The remaining sections of this paper are structured as follows: The are used in more personal or private settings for immediate personal following Section 2 reviews existing research on the role of hedonic and gratifications (Lee, Lee, & Sheehan, 2020). Hence, future research on utilitarian attitudes in technology usage and WOM behavior, and ante­ SVA need to look beyond the concepts examined using the traditional cedents to both attitudes. Section 3 describes the conceptual model and adoption models. Recent research suggests that attitude is a critical presents hypotheses. Section 4 explains the research methodology, fol­ antecedent to technology usage and adoption behavior (Dwivedi, Rana, lowed by results in Section 5. Next Section 6 presents discussion, theo­ Jeyaraj, Clement, & Williams, 2019). Furthermore, identification of retical contributions, implications for practice, and limitations. Finally, factors that attribute to a favorable attitude toward technology is an the paper concludes with Section 7. underexplored research area in the context of SVA. There are specific calls for research to explore how brands can leverage emerging tech­ 2. Literature review nologies like SVA to influence consumer engagement behaviors (Dwi­ vedi & Ismagilova, 2021, p. 8). Similarly, McLean and Osei-Frimpong Rapid technological advancements have led to the development of (2019) call for future research on factors that influence SVA’s utilitarian many innovative products that fulfill the need of consumers of modern benefits. era. Every decade sees a radical change in how humans interact with technology. The journey of human-computer interactions encompasses SVA work as standalone devices delivering a spectrum of functional, the evolution from desktop to the Internet, touch-screens, SVA, and to hedonic, and symbolic benefits to users (Mishra & Shukla, 2020). SVA blockchain technology (Hughes et al., 2019; Lee et al., 2020). The topic technologies considerably differ from existing technologies as they offer of technology adoption has attracted many academicians and market almost humanlike interaction experiences (Fernandes & Oliveira, 2021). practitioners, who have used behavioral and cognitive theories to pro­ SVA have design concepts that predispose users to personify and interact pose various models and frameworks (e.g., Balakrishnan & Dwivedi, like conversing with a human being (Foehr & Germelmann, 2020). SVA 2021a; Hernandez-Ortega & Ferreira, 2021). It is important that the provide users an assurance of someone being present (social presence) in existing theoretical models should be updated to meet the challenges of their vicinity, who can empathize and follow their commands. More­ changing consumer behavior and preferences due to advances in tech­ over, users regard SVA as friends and communicate with SVA to deal nology (Dwivedi et al., 2019; Williams, Rana, & Dwivedi, 2015). with their loneliness or to have fun (Park, Kwak, Lee, & Ahn, 2018). Thus, users personify SVA and this behavior of assigning humanlike 2.1. Motivations and technology adoption traits to objects is known as anthropomorphism (Epley, Waytz, & Cacioppo, 2007). Users’ interactions with smart objects lead to devel­ Most of the research on technology adoption stems from behavioral oping humanlike relationships, which remains an unexplored research sciences and offers frameworks specific to the utilitarian perspective. area in the context of SVA adoption (Novak & Hoffman, 2019). Hence, For example, TAM builds on the Theory of Planned Behavior to inves­ this research specifically focuses on the anthropomorphism and psy­ tigate how perceived usefulness and ease-of-use determine attitude to­ chological perspective of gratifications derived from SVA devices. ward technology and subsequent behavioral intentions to use the technology (Davis, 1989). UTAUT is another widely used model to Users’ intentions to use a product, service, or technology greatly explain technology use and adoption (Venkatesh et al., 2003; Williams depend on the word-of-mouth (WOM) recommendations received from et al., 2015). The updated version UTAUT2 incorporates hedonic others (Mishra, Maheswarappa, & Colby, 2018). WOM plays a critical motivation to understand users’ motives to adopt a technology (Ven­ role toward the success of products and services via network effect. A katesh, Thong, & Xu, 2012). Similarly, TAM model has been updated to positive WOM motivates users to try and use new technology and can a third version (TAM3) that includes elements of experience, enjoyment, turn users into advocates (Tamilmani, Rana, Nunkoo, Raghavan, & and playfulness (Pillai, Sivathanu, & Dwivedi, 2020; Venkatesh & Bala, Dwivedi, 2020; Vilpponen, Winter, & Sundqvist, 2006). Individuals’ 2008). WOM behavior depends on their experiences and gratifications derived from the use of products or services (Berger, 2014). Prior research Extant research in technology adoption broadly uses two types of confirms the positive influence of hedonic motivations (e.g., joy and fun motivations - utilitarian-motivation and hedonic-motivation (Agarwal & of using technology) in adopting technology like mobile banking Karahanna, 2000; Lee et al., 2020). Utilitarian motives include the (Alalwan, Dwivedi, & Rana, 2017). However, the impact of hedonic functional aspects of technology such as performance, usefulness, and motives on WOM behavior, which can significantly influence further ease of use (e.g., TAM, Davis, 1989). In contrast, the second category technology adoption, is yet to be ascertained. Therefore, the main ob­ focuses on hedonic intentions such as enjoyment, immersion, and flow jectives of this research are: (e.g., UTAUT2, Tamilmani, Rana, Prakasam, & Dwivedi, 2019; Ven­ katesh et al., 2012). Moreover, SVA may lead to the state of cognitive • To examine the influence of hedonic and utilitarian attitudes on SVA absorption because they offer temporal escapism and heightened usage and WOM recommendations. enjoyment by addressing users’ curiosity (Liao, Vitak, Kumar, Zimmer, & Kritikos, 2019). A recent meta-analysis by Tamilmani et al. (2019) • To identify the effects of psychological antecedents to hedonic and reaffirms the crucial role of hedonic motivation in technology adoption. utilitarian attitudes. In fact, researchers have integrated the TAM model with flow theory to ascertain the role of immersion and interaction in human-technology The present study contributes to the extensive research on technol­ interactions such as online gaming (Hsu & Lu, 2004). Flow theory is a ogy adoption, especially in the area of emergent smart technologies. The psychological concept that explains individuals’ mental state when they findings indicate the importance of utilitarian and hedonic attitudes are completely immersed in an activity with extreme focus and enjoy­ toward users’ intentions to adopt SVA and share WOM recommenda­ ment (Nakamura & Csikszentmihalyi, 2009). For example, users can be tions. The results suggest that SVA provide many benefits to users such in a flow state while playing online games or online impulsive buying as playfulness, escapism, social presence, and, most importantly, as a (Wu, Chiu, & Chen, 2020). Therefore, flow state significantly influences humanlike companion. SVA usage depends on their esthetic appeal and users’ attitudes and behavioral intentions. figurative personification created by users. The findings provide actionable strategies to the various stakeholders who can tailor their offerings specific to users interacting with SVA. For example, manu­ facturers and marketers must emphasize esthetic, playful experience, 2

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx 2.2. SVA and multisensory experiences Germelmann, 2020). Hence, social influence can impact users’ SVA buying and usage intentions. One stream of research treats computers as SVA significantly differ from traditional technology like computers, social actors (CASA) to describe or explain the social interactions with laptop, or mobile banking in terms of usage situation, sensory experi­ computers (Kim & Sundar, 2012). In addition, people build relationships ences, and gratifications (Hoffman & Novak, 2018). Users generally do and emotional bonds with SVA during the regular interactions and not use SVA for increasing productivity in an organization, but they use discuss the experiences with others in the form of WOM behavior (Hur, SVA as a personal device in a relatively more private location (e.g., Koo, & Hofmann, 2015). Users show feelings of love, commitment, and home). SVA deliver a highly interactive experience where users can ask intimacy toward SVA as they do for fellow humans (Hernandez-Ortega SVA to do functional (e.g., setting an appointment or a reminder) or & Ferreira, 2021). One may easily notice how individuals proudly talk hedonic (e.g., telling a joke or playing a song) tasks (Foroudi, Gupta, about their emotional bonding with cars or watches. Moreover, SVA fill Sivarajah, & Broderick, 2018; Lopatovska et al., 2019). Furthermore, a conversational space and enable emotional venting, which are critical SVA provide a multisensory experience (esthetic and auditory) similar to antecedents to WOM behavior (Berger, 2014; Ismagilova, Rana, Slade, & emergent multisensory technologies like virtual reality and augmented Dwivedi, 2020). Therefore, we notice that users’ attitude toward SVA reality. Thus, we borrow from relevant research on multisensory tech­ and their usage depends on multiple factors. Hence, we consider the nologies to examine the importance of multisensory experiences (e.g., critical aspects of SVA usage like psychological gratifications to address playfulness and escapism) in SVA usage and WOM recommendations hedonic motivations, and anthropomorphism, to offer valuable insights (Mishra, Shukla, Rana, & Dwivedi, 2021). into users’ motivations to use SVA and WOM recommendations (Mishra & Shukla, 2020). 2.3. SVA and anthropomorphism 3. Conceptual model and research hypotheses Users consider SVA as humans like a friend or a companion (Poushneh, 2021). For ages, humans are anthropomorphizing objects A majority of human-technology interactions are gradually shifting and animals around them. Many examples exist in the mythological to SVA such as Alexa and Google Home. Many users use SVA for fun and literature, famous fables and stories (e.g., The Jungle Book and Alice in hedonic motives rather than only utilitarian benefits. Our study explores Wonderland), and movies (e.g., humanoids and R2D2 in Star Wars experiential and design factors that influence SVA usage and WOM movies). Similarly, brands utilize anthropomorphism to spread aware­ recommendations for SVA. To adequately address the gratifications ness and recognition (Guido & Peluso, 2015). Users can instantly derived from using SVA, the proposed model in our research (see Fig. 1) recognize and recall the associated brands, for example, Mickey Mouse builds upon two theories from the psychology area: flow theory and the (Disney), Cheetah (Cheeto), McDonald’s clown, and Michelin man. theory of anthropomorphism. The flow theory is used to capture the Therefore, we believe that the way users perceive SVA as humans will be immersive experiences of playfulness and escapism derived from using a critical factor for developing users’ attitudes and usage of SVA. SVA. The anthropomorphism concept explains users’ perceptions of SVA as humans and companions. We examine the role of various psycho­ 2.4. SVA and social influence logical antecedent variables that significantly influence users’ hedonic and utilitarian attitudes, which in turn affect SVA usage and WOM SVA deliver symbolic benefits of social prestige (McLean & recommendations. Osei-Frimpong, 2019). Many users consider SVA as a status symbol and use SVA to enhance their self-image among their social groups (Foehr & Fig. 1. Conceptual model. 3

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx 3.1. Hedonic attitude 3.3. Playfulness Users evaluate any product, service, or technology on two di­ Playfulness is defined as a degree of cognitive spontaneity in human- mensions: (1) hedonic or gratification value that is affective in nature, computer interactions (Webster & Martocchio, 1992). People interact and (2) utilitarian or instrumental value that is cognitive-intensive with technology to relish the experience. For example, users spend (Voss, Spangenberg, & Grohmann, 2003). Extant research suggests considerable time on mobile games to enjoy the thrills of racing, that attitude toward technology is a critical antecedent to technology adventurous journeys or missions, or building online empires. Users find usage and adoption behavior (Dwivedi et al., 2019). SVA offer users an element of playfulness in SVA interactions (Chatterjee, Kar, & Dwi­ enjoyment and entertainment by answering user’s questions in humor­ vedi, 2021; Lee et al., 2020). Most of the queries posted to SVA (e.g., ous ways. People find novelty in using personal assistants leading to Alexa) are functional in nature. Many users pose funny questions to SVA continuous usage (Gursoy, Chi, Lu, & Nunkoo, 2019). SVA provide an such as enquiring the marital status or asking to make a sandwich and element of playfulness and visually appealing esthetics, which satisfy get interesting humorous replies making the interaction amusing and the hedonic intrinsic experiential values (Mathwick, Malhotra, & Rig­ highly enjoyable. Playfulness stems from curiosity and enjoyment and don, 2001). A hedonic attitude increases the likelihood of trying new significantly impacts users’ attitude toward the use of technology (Ahn, technology such as eStores or mobile banking (Alalwan et al., 2017; Wu Ryu, & Han, 2007). Hence, we argue that users will actively use SVA for et al., 2020). Thus, we propose the following hypothesis: fun and enjoyment. Thus, we propose that: H1a. Hedonic attitude positively influences users’ SVA usage. H3. Playfulness positively influences users’ hedonic attitudes. When individuals derive psychological gratifications and satisfaction 3.4. Escapism from products and services, they are more likely to share their experi­ ences with others (Ismagilova et al., 2020). WOM behavior is a form of Escapism refers to people’s tendency to “get away from it all”. self-expression, where people share information to fulfill hedonic mo­ Escapism is defined as a state of psychological immersion in which a tives. Many people share their views related to experiential products like person is completely involved in the focal activity (Mathwick & Rigdon, much-awaited movies or highly rated restaurants immediately on online 2004). Individuals use escapism as a coping strategy to handle an un­ platforms (Mishra & Satish, 2016). Since SVA provide users an affective pleasant emotional state. Escapism offers individuals momentary satis­ experience, we expect users to display a similar WOM behavior. More­ faction and fulfillment of their needs. For example, people engage in over, SVA are considered as status symbols that enhance users’ recreational or leisure activities like shopping, consuming products (e. self-image among their social network. People display feelings of pride g., ice-cream), or binge-watching TV or web series as an escape route to by owning such advanced devices to influence others. Research suggests their routine life (Jones, Cronin, & Piacentini, 2018). that social influence and altruism are critical drivers for WOM activities (Berger, 2014). Thus, we propose the following hypothesis: Escapism as a concept has been widely used in research to explain users’ intensive engagement with virtual and highly interactive envi­ H1b. Hedonic attitude positively influences users’ WOM ronments like massively multiplayer online games (e.g., World of War­ recommendations. craft) and virtual worlds (e.g., Second Life, Kuo, Lutz, & Hiler, 2016). Multisensory technologies (e.g., augmented reality, virtual reality, and 3.2. Utilitarian attitude mixed reality) also offer an immersive experience leading to increased involvement with the environment. For example, the tourism industry Individuals use SVA to fulfill their functional needs, for example, uses virtual reality to connect with potential consumers and promote people regularly use Alexa for mundane tasks like setting up an alarm, activities like scuba diving in a highly immersive environment (Mishra scheduling meetings or appointments, and shopping. SVA reduce users’ et al., 2021). Such vivid experiences fulfill consumers’ desire for fantasy cognitive load by offering convenience. SVA are designed to learn and and novel experiences as evident from the growing usage of 3D visuals in adapt to the instrumental transactions and repeat them as desired by movies. In addition, many users take help of SVA to break the monot­ users at pre-configured intervals (McLean & Osei-Frimpong, 2019). onous routine by asking SVA to tell a joke or play entertaining music Many users use SVA extensively for seeking information on a variety of (Conner & Bradford, 2020). Thus, SVA perform the role of an avid topics like news, weather, products, sports scores, and cooking recipes. listener and entertainer who answers to the commands of its master. Recent research suggests that users will increasingly use SVA due to Hence, we propose the following hypothesis: convenience and competence offered by these personal digital assistants (Hu, Lu, Pan, Gong, & Yang, 2021). Thus, SVA usage becomes a habit for H4. Escapism positively influences users’ hedonic attitudes. many users to accomplish routine tasks. So, we hypothesize the following: 3.5. Anthropomorphism H2a. Utilitarian attitude positively influences users’ SVA usage. Anthropomorphism reflects human tendencies to label SVA as humans and to seek emotional support as a companion. Users consider Users with a favorable attitude toward SVA are more likely to use SVA as a child learning the tricks of communicating with the owner SVA to purchase and interact with customer support to resolve queries (Karimova & Goby, 2020). Recent research on personal assistants sug­ (Moriuchi, 2019). These consumers offer favorable recommendations of gests that anthropomorphism is a significant factor in human-SVA in­ firms that integrate SVA in their websites to enable direct communica­ teractions. For example, Hu et al. (2021) suggest that users evaluate SVA tion (Moriuchi, 2019). Moreover, consumers share novel, positive, and on two dimensions: warmth and competence. Warmth echoes the entertaining experiences online (Berger & Milkman, 2012). SVA in­ emotional benefits of SVA like benevolence, sociability, and caring. teractions deliver a highly interactive and entertaining experience to Users develop an emotional attachment with SVA due to persistent use. users. In a recent meta-analysis on eWOM behavior, Ismagilova et al. In fact, users enjoy communicating with SVA and humanize them as (2020) propose that altruism and homophily motivate people to share companions (Poushneh, 2021). WOM. We argue that users will share recommendations for SVA with others due to high levels of utilitarian value and convenience. Hence, we COVID-19 pandemic forced millions of people across the world to propose the following hypothesis: live in isolation under various restrictions on movement. Many people suffered a severe impact on mental well-being because of an unusually H2b. Utilitarian attitude positively influences users’ WOM long period of home confinement and lack of social gatherings and recommendations. celebrations (Lovett, 2020). In these tough times, SVA has helped people 4

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx to improve emotional and mental well-being and fight anxiety and the last decade, users have seen a drastic decrease in phones’ size with depression. People feel lonely when they do not have anyone to talk to, significant improvements in performance and features. Hence, SVA’s and in such cases, SVA played the role of users’ companion (or friend) to compact and symmetrical size can influence users’ perceptions of effi­ overcome people’s loneliness at home (Poushneh, 2021). In a ciency and utility of such devices. Thus, we hypothesize that: work-from-home culture, people start their home office by greeting SVA (Foehr & Germelmann, 2020). SVA’s critical role to fill the emotional H6. Visual appeal positively influences users’ utilitarian attitudes. void and loneliness of users is exemplified during the lockdown in 2020 when millions of people proposed to Alexa. Thus, SVA are emotionally 3.7. Social presence superior to other technologies in fulfilling hedonic needs and hence we posit the following hypothesis: Social presence reflects the sense or feeling of being with others while having a communication. In a technology-mediated interaction, H5a. Anthropomorphism positively influences users’ hedonic social presence is defined as “the degree to which users can feel others’ attitudes. presence in the result of interpersonal interactions during the commu­ nication process” (Walther, 1992, p. 54). Social presence encompasses According to Hu et al. (2021), competence (the second dimension to physical and psychological proximity to users (Roy, Singh, Hope, evaluate SVA) reflects the utility of SVA such as their capability to Nguyen, & Harrigan, 2019). Thus, the social presence of SVA can be complete the tasks efficiently as required by users. The UTAUT model explained as their ability to engage with users and to establish physical proposes two antecedents to users’ technology adoption - performance existence. Social presence is a critical factor that affects individuals’ expectancy and effort expectancy (Venkatesh et al., 2012). Performance affective and cognitive commitments toward their relationships with a expectancy of SVA measures relative performance compared to humans brand (Dabholkar, van Dolen, & de Ruyter, 2009). For example, when on parameters of accuracy, consistency, and errors. On the other hand, users interact with others on Twitter, a higher degree of social presence effort expectancy shows the degree of ease and learning curve required leads to better fulfillment of social connection needs (Han, Min, & Lee, to use SVA. Both antecedents influence users’ attitudes toward tech­ 2016). Likewise, when users feel an increased social presence in their nology leading to usage intentions (Rana, Dwivedi, Williams, & Weer­ interactions with smart technologies, they report higher levels of com­ akkody, 2016). Recent research on SVA suggests that fort, experience, and emotional satisfaction (Fernandes & Oliveira, anthropomorphism is significantly related to effort expectancy (Gursoy 2021). SVA offer an interactive medium where users can have two-way et al., 2019). Moreover, users look for functional and emotional communications. Hence, SVA’s social presence offers functional benefits compatibility with SVA (Karimova & Goby, 2020). Therefore, users as a mechanism to beat users’ loneliness (Savage, 2020) or to have fun anthropomorphize SVA as personal assistant or secretary to accomplish (Foehr & Germelmann, 2020). Hence, we hypothesize that: routine tasks (e.g., scheduling meetings and setting up reminders). Thus, we hypothesize that: H7. Social presence positively influences users’ utilitarian attitudes. H5b. Anthropomorphism positively influences users’ utilitarian 3.8. Moderating role of prestige attitudes. Prior research confirms the critical role of social influence in tech­ 3.6. Visual appeal nology adoption. The widely used UTAUT model included social influ­ ence as one of the four core constructs that can motivate individuals to Visual appeal echoes the beautiful design or physical attractiveness adopt and use technology (Venkatesh et al., 2003). Social influence re­ of a product or interface. The importance of visual appeal or esthetics is flects the perceptions of others about technology, gadgets, and devices. well established in IS (information systems) literature (Phan, 2019). For Individuals feel that what they wear or use reflects their social image as example, a visually appealing website interface offers higher experien­ conspicuous consumption. People believe that using emerging technol­ tial value to users leading to higher consumer engagement and purchase ogies like smartwatches, smart glasses, or playing augmented intentions (Mathwick et al., 2001). From a multisensory perspective, reality-based games can enhance their social image (Rauschnabel, esthetics manipulate the visual sense of users. Mostly, users’ first Rossmann, & tom Dieck, 2017). SVA offer symbolic benefits of a positive interaction with a product happens via eyes, and hence the brand logo image and work as status symbols (McLean & Osei-Frimpong, 2019). and packaging become a vital factor in influencing consumers’ percep­ Many people buy luxury products or high-end technology products (e.g., tions and purchase behavior (Krishna & Schwarz, 2014). While most iPhones) as these products fulfill their hedonic needs by signaling their research links esthetics to hedonic motivations of consumers, we find a high prestige. Hence, prestige affects users’ hedonic attitudes. handful of research exploring how visual appeal influences consumers’ perceptions of utility or functional performance of products (Mishra The uses and gratification theory proposes that individuals actively et al., 2021). For example, the concept of functional design stresses the seek media to satisfy their specific needs. Users are more likely to use importance of design that enhances the product’s function (e.g., SVA as a status symbol due to various gratifications obtained from its screw-driver). use. For example, such individuals may give more importance to SVA features such as playfulness, companionship, or an alternative to the Artificial intelligence and smart voice recognition technologies are mundane world to justify their use. Users who use SVA for fun will not physical products (Verma et al., 2021). Hence whether they need a develop a favorable hedonic attitude toward SVA believing that SVA physical body or not is an interesting question in itself. The theory of enhances their prestige. Users will spend more time with SVA to seek a embodied cognition suggests that cognitive processing depends on the higher status among their network. They may present it as a showpiece physical aspects of the objects, such as symmetry, shape, and colors to others to enhance their status. Thus, we hypothesize: (Krishna & Schwarz, 2014). Users consider the working of an iPhone ‘smooth’, which resembles the ‘smooth’ design of the phone itself. Most H8a. Prestige positively moderates the relationship between playful­ product designs (e.g., cars) have curves instead of corners to convey a ness and hedonic attitude. smooth ride. Similarly, users perceive a heavy and enormous-sized product as more robust and durable. A very famous example is Recent research on emergent technologies offering multisensory Titanic, where a fourth dummy smoke funnel was added to enhance experiences to users clearly outlines the importance of capabilities of symmetrical esthetics. Therefore, in this research, we focus on the devices to take users to an imaginary fantasy world (Mishra et al., 2021; functional value of visual appeal of SVA. Users may perceive that a Tamilmani et al., 2019). A common perception about prestige is that pleasant and innovative design delivers functional benefits efficiently. In products signal social prestige based on premium prices or scarcity. However, psychological gratifications like escapism are new parameters 5

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx that influence prestige (Holmqvist, Ruiz, & Pen˜aloza, 2020). Therefore, measures and have similar predictive accuracy. Moreover, single-item people who perceive SVA as a prestigious symbol may use them to measures have been used in similar research when the construct attri­ escape from everyday life. For example, they can experience a fantasy butes are concrete and unambiguous (e.g., Barnes, 2021). Since, usage is world where owning SVA elevates them to higher ranks of society a construct that can be easily and uniformly understood by respondents, (McLean & Osei-Frimpong, 2019). Whereas people who do not relate use of single-item for measurement is appropriate (Bergkvist & Rossiter, SVA to social status may use SVA primarily for utilitarian and realistic 2007). A 5-point Likert scale was used following the recent SVA research tasks. Hence, we hypothesize that: (e.g., Balakrishnan & Dwivedi, 2021a, 2021b; Gursoy et al., 2019). H8b. Prestige positively moderates the relationship between escapism 4.3. Common method variance (CMV) and hedonic attitude. Common method variance (CMV) is a source of bias in a survey People try to acquire and own products that are scarce and novel to method. CMV may happen because each respondent provides responses enhance their prestige. For example, people possess a rare piece of art or for both predictor and dependent variables (Podsakoff, MacKenzie, Lee, antiqueue that is considered prestigious. People treat their luxury cars as & Podsakoff, 2003). We applied recommended procedural steps during their companions or friends and assign names to them while proudly the survey design and administration process to handle the issue of CMV. discussing with their friends. Similarly, the ownership of SVA creates an As a preemptive approach, participants were assured anonymity and endowment effect, where people value SVA more on the dimensions of confidentiality. As a post hoc approach, marker-variable technique was competence and warmth (Hu et al., 2021). People treat SVA as a pres­ performed for CMV validity analysis. The results indicated that the tigious friend who helps them in their needs and accompanies them to difference between the original and CMV-adjusted correlations were beat their loneliness (Savage, 2020). Thus, people who believe that very small (≤0.07) for all the relevant constructs (Lindell & Whitney, owning SVA is prestigious are more likely to humanize SVA. Thus, we 2001). Hence, CMV does not seriously distort the results and predictions propose the following hypothesis: in this study. H8c. Prestige positively moderates the relationship between anthro­ 5. Results pomorphism and hedonic attitude. Descriptive analysis of demographic data is given in Table 1. The 4. Methodology final sample had 56.4% male, mostly between 26 and 35 years of age (38.1%), and working (63.3%). Most of the participants were from the 4.1. Sample and data collection upper middle class income group having a monthly income between INR 50,000–0.1 million (1 USD = 75 INR approx.) and were most likely to An online questionnaire-based survey was used to collect the data for afford SVA. Thus, the sample profile is appropriate to collect data for the this research. The survey approach was used because it allows gener­ study specific to the context of SVA adoption. alizability of outcomes, replicability of findings, and simultaneous evaluation of multiple factors (Bawack, Wamba, & Carillo, 2021). The The data was analyzed using structural equation modeling (SEM) survey method is a well-established and widely used method in the IS based on partial least squares (PLS) approach with SmartPLS 3.3 soft­ positivist research domain enabling researchers to reliably assess their ware (Ringle, Wende, & Becker, 2015). We selected the PLS-SEM predictive theories and research models (Straub, Boudreau, & Gefen, approach for the following reasons. This approach is widely used in 2004). Online surveys have been used in recent research related to recent research (e.g., Dwivedi, Hughes, et al., 2021; Hu et al., 2021) and AI-based artifacts and online impulsive shopping (e.g., Hu et al., 2021; it is based on component-based structural equation modeling (Hair, Wu et al., 2020). Ringle, & Sarstedt, 2011). Moreover, PLS is recommended for prediction-based models that focus on identifying the key predictor or The questionnaire had links to two online YouTube videos about SVA driver constructs (Hair et al., 2011), which aligns with the research to provide more information on SVA to participants. The survey link was objectives of this study. distributed to a diverse population including executives pursuing an executive MBA on an online learning platform, researchers, and aca­ 5.1. Measurement model demicians. Recent research in the domain of AI-based voice assistants has included student samples (e.g., Balakrishnan & Dwivedi, 2021b; We computed Cronbach’s alpha and composite reliability (CR) to Gursoy et al., 2019). The sample included executives who significantly assess the reliability of the research model. The values for both are differ from the typical university student sample used in research. higher than the recommended values of 0.7 (see Table 2). All the item Participation in the survey was voluntary and a total of 428 responses loadings were significant and more than 0.7. The AVE (average variance were received. Out of these, 68 responses were removed due to less than 10% completion of survey, resulting in the final sample size of 360. 4.2. Measures Table 1 Sample characteristics with sample size = 360. All measures in this study were adopted from the existing literature with minor modifications to reflect the SVA context (see Appendix A for Category Sub Category Frequency Percent detailed measurement items). Items were modified and verified with ten % research experts in the similar domain. Hedonic and utilitarian attitudes Gender Male 203 were measured using bipolar items. All the other constructs were Female 156 56.4 measured by multiple items (except usage) using a 5-point Likert scale Age (years) Prefer not to say 43.3 with suitable ranges (e.g., “strongly disagree” to “strongly agree”, 18–25 1 “never” to “always”, and “unlikely” to “likely”). One of the research Employment status 26–35 85 0.3 objectives of this study was to understand SVA usage, which was Monthly household income 36–45 137 23.6 measured using frequency with a single-item statement based on > 45 128 38.1 McLean and Osei-Frimpong (2019). Single items may have certain lim­ (INR) Not working 10 35.6 itations, however, for SEM context, Bergkvist and Rossiter (2007) sug­ Working 132 gest that single-item measures are equally valid as multiple-item Less than 25,000 228 2.7 25,000–50,000 46 36.7 50,000–0.1 million 109 63.3 More than 0.1 million 133 12.7 72 30.3 37.0 20.0 6

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx Table 2 Reliability and validity indices. Constructs and indicators Mean SD Factor Loadings Cronbach’s Alpha CR AVE Anthropomorphism 3.48 0.894 0.901 0.938 0.834 Anthro1 3.53 0.91 0.896 0.934 0.826 Anthro2 3.55 0.93 0.927 0.889 0.931 0.818 Anthro3 0.67 0.918 0.906 0.941 0.842 Escapism 2.98 0.888 0.93 0.816 Esc1 2.93 0.939 0.889 0.931 0.818 Esc2 2.80 0.82 0.88 0.926 0.807 Esc3 0.87 0.935 0.888 0.93 0.837 Social presence 3.28 0.83 0.849 Pres1 3.38 Pres2 3.21 0.901 Pres3 0.93 Playfulness 3.68 0.92 0.904 Play1 3.72 0.88 0.909 Play2 3.21 Play3 0.888 Visual appeal 3.71 0.97 Vis1 3.64 0.77 0.950 Vis2 3.64 0.98 0.914 Vis3 Hedonic attitude 3.90 0.921 Hedo1 4.01 0.88 Hedo2 3.86 0.99 0.926 Hedo3 0.87 0.863 Utilitarian attitude 3.96 Uti1 4.08 0.889 Uti2 4.04 0.93 Uti3 0.96 0.918 Word of mouth 3.85 0.99 0.906 WOM1 3.76 WOM2 3.63 0.921 WOM3 0.96 0.94 0.902 0.87 0.871 0.901 1.01 0.98 0.934 0.94 0.910 extracted) values were more than the suggested value of 0.50 indicating acceptable fit. Next, we examined adjusted R2, which shows the variance convergent validity (Hair, Risher, Sarstedt, & Ringle, 2019). Discrimi­ explained by the model that defines the quality of the overall model in nant validity was verified using two methods. First, the square root of PLS-SEM (Hair et al., 2019). Henseler and Sarstedt (2013) consider R2 each construct’s AVE was higher than its correlation with another values of 0.67, 0.33, and 0.19 as substantial, moderate, and weak, construct (see Table 3, Fornell & Larcker, 1981). Second, we checked for respectively. In our model, the R2 values for various constructs were as HTMT (Heterotrait-Monotrait Ratio of Correlations) values for estab­ follows: Hedonic attitude (0.43, moderate), utilitarian attitude (0.69, lishing discriminant validity. The HTMT values for all the constructs substantial), usage (0.53, moderate), and WOM (0.37, moderate). Next, were less than the recommended value of 0.85 (Henseler & Sarstedt, we used the blindfolding process (with omission distance D set to 7) and 2013, Table 4). the PLS Predict (by setting the number of folds k = 10) to assess the predictive relevance of the model (Hair et al., 2019). The resulting Q2 5.2. Structural model values (0.44) were larger than zero, suggesting the good predictive ac­ curacy of the model. We controlled for demographic variables in our We evaluated the structural model using the Bias-corrected and model. We did not find any significant effect of any demographic vari­ Accelerated (BCa) bootstrap method with 5000 bootsample re-sampling able in the model. approach. The variance inflation factors (VIF) values were less than 5 indicating no multicollinearity concern. We evaluated the goodness of 5.3. Path coefficient estimation fit of model through the score of the standardized root mean square residual (SRMR). The SRMR value for the estimated model was 0.055, The SEM results are shown in Table 5 and Fig. 2. The results show which was below the threshold of 0.08 (Henseler & Sarstedt, 2013). The standardized path coefficients (β values), t values, and p values. Hedonic value of Normed Fit Index (NFI) was 0.92, which was above 0.90 showed attitude has a positive and significant impact on SVA usage (β = 0.183, Table 3 Discriminant validity (Fornell & Larcker, 1981). AN ES HA PL SP US UA VA WOM Anthropomorphism (AN) 0.913 0.909 0.904 0.917 0.904 0.851 0.898 0.904 0.915 Escape (ES) 0.446 0.315 0.459 0.551 0.407 0.462 0.397 0.606 Hedonic Attitude (HA) 0.338 0.404 0.345 0.58 0.387 0.528 0.416 Playfulness (PL) 0.614 0.532 0.398 0.411 0.504 0.638 Social presence (SP) 0.682 0.331 0.787 0.628 0.487 Usage (US) 0.45 0.271 0.429 0.604 Utilitarian attitude (UA) 0.398 0.438 0.371 Visual appeal (VA) 0.573 0.341 Word of mouth (WOM) 0.538 Note: The numbers in the diagonal are the square root of the variance shared between the constructs and their measures. Off-diagonal elements are correlations among constructs. 7

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx SP US UA VA WOM Table 4 Discriminant validity (HTMT values). AN ES HA PL Anthropomorphism (AN) 0.492 0.343 0.511 0.614 0.548 0.624 0.444 0.684 Escape (ES) 0.374 0.44 0.381 0.764 0.433 0.697 0.465 Hedonic Attitude (HA) 0.679 0.6 0.529 0.458 0.573 0.845 Playfulness (PL) 0.764 0.438 0.821 0.702 0.546 Social presence (SP) 0.598 0.301 0.477 0.668 Usage (US) 0.447 0.489 0.41 Utilitarian attitude (UA) 0.64 0.373 Visual appeal (VA) 0.596 Word of mouth (WOM) Table 5 Hypothesis Path t p p < 0.001), supporting hypotheses H2a and H2b. SEM results. coefficient β Statistics Values Playfulness (β = 0.377, t = 5.89, p < 0.001) and escapism H1a 0.183 2.11 0.027 Hedonic attitude → Usage H1b 0.154 2.37 0.018 (β = 0.144, t = 2.89, p = 0.005) show a positive and significant impact Hedonic attitude → WOM H2a 0.391 4.49 <.001 on hedonic attitude, supporting H3 and H4. The results reveal that Utilitarian attitude → Usage H2b 0.325 3.93 <.001 anthropomorphism does not impact hedonic attitude (β = 0.041, Utilitarian attitude → WOM H3 0.377 5.89 <.001 t = 0.63, p = 0.53), but it has a significant influence on utilitarian atti­ Playfulness → Hedonic tude (β = 0.166, t = 1.98, p < 0.05). Hence, H5a is not supported, but H4 0.144 2.89 0.005 H5b is supported. Visual appeal (β = 0.222, t = 3.06, p = 0.002) and attitude social presence (β = 0.171, t = 2.25, p < 0.05) show a positive and Escapism → Hedonic H5a 0.041 0.63 0.53 significant relationship with utilitarian attitude. Thus, H6 and H7 are supported. attitude H5b 0.166 1.98 0.045 Anthropomorphism → 5.4. Multi-group PLS analysis for moderation H6 0.222 3.06 0.002 Hedonic attitude The moderating effects of prestige were examined using the multi- Anthropomorphism → H7 0.171 2.25 0.024 group analysis (PLS-MGA) in SmartPLS 3.3 (see Table 6). This method is a non-parametric significance test for the difference of group-specific Utilitarian attitude results based on PLS-SEM bootstrapping method (Hair et al., 2011, Visual appeal → Utilitarian 2019). The method tests whether the path coefficients significantly differ between the two or more groups (moderators). The results show attitude prestige moderates the two relationships: Playfulness → Hedonic atti­ Social presence → tude (βdiff = − 0.323, p = 0.048) and Escapism → Hedonic attitude (βdiff = 0.407, p = 0.031), whereas there is no significant difference in path Utilitarian attitude coefficients for relationship Anthropomorphism → Hedonic attitude t = 2.11, p < 0.05) and users’ WOM recommendations (β = 0.154, t = 2.37, p < 0.05) of SVA, supporting H1a and H1b. Similarly, utili­ tarian attitude appears as a strong determinant of SVA usage (β = 0.391, t = 4.49, p < .001) and WOM recommendations (β = 0.325, t = 3.93, Fig. 2. Results of structural model test. 8

A. Mishra et al. Hypothesis Difference in path p value (for International Journal of Information Management xxx (xxxx) xxx H8a coefficients (high vs. difference) Table 6 low prestige) why people are more likely to recommend SVA to others. These results PLS-MGA results. 0.048 complement the vast amount of prior research on WOM, which recom­ -0.323 0.031 mends that product quality and novel experiences are critical anteced­ Relationship 0.978 ents to WOM behavior (Berger, 2014; Mishra & Satish, 2016). H8b 0.407 Playfulness → Hedonic An examination of antecedents to hedonic attitude and utilitarian attitude H8c -0.005 attitudes indicates that playfulness and escapism strongly influence hedonic attitude, whereas the effect of anthropomorphism on hedonic Escapism → Hedonic attitude is non-significant. Human-SVA interactions are a playful ex­ attitude change of information in absorbing activities leading to intrinsic enjoyment (Mathwick & Rigdon, 2004). Users recognize playfulness as a Anthropomorphism → reflection of emotional attachment with SVA. The affective responses to Hedonic attitude SVA lead to warmth perceptions, which positively influence users’ usage intentions (Hu et al., 2021). The results extend the earlier findings on (βdiff = − 0.005, p = 0.978) for high and low prestige groups. Thus, the playfulness as an intrinsic utility for hedonic motives (Venkatesh et al., findings support H8a and H8b, but not H8c. 2012). Similarly, in the context of virtual reality store shopping, the experience of playfulness enhances consumers’ hedonic gratifications 6. Discussion (Kang, Shin, & Ponto, 2020). Moreover, the results endorse the similar positive effect of escapism on hedonic values found in recent tourism SVA are the next step of advancement in human-technology in­ research that focuses on experiential consumption (Ponsignon, Lunardo, teractions, which significantly impact the consumer journey and offer a & Michrafy, 2020). highly interactive and playful experience to users. Marketers and re­ searchers have recognized the immense potential of SVA that can sub­ Research offers evidence that characters, objects, or product pack­ stantially influence people’s productivity and consumer behavior. aging, which mimic humanness appeal to consumers’ hedonic values However, the adoption of SVA has not reached its full potential, which and perceptions (Epley et al., 2007). SVA deliver human and social cues can be attributed to users’ attitudes toward SVA. The first objective of while interacting with humans leading to higher engagement (Moriuchi, our research was to examine the effect of hedonic and utilitarian atti­ 2021). We did not find a meaningful relation between anthropomor­ tudes on SVA usage and WOM behavior. The second objective was to phism and hedonic attitude. A plausible explanation is that users may identify the antecedents to two types of attitudes. To address these ob­ perceive SVA as humans, but that perception is limited to the functional jectives, this study investigated the experiential aspect of using SVA aspect and may not lead to a strong emotional bonding. Research in IS from a users’ perspective, where attitude played a critical role in shaping area on SVA has mostly explored the effect of anthropomorphism using the behavioral intentions. task-fit perspective on constructs like performance expectancy and effort expectancy (e.g., Gursoy et al., 2019). Our results support prior research According to the statistical analysis presented in the previous sec­ and establish that anthropomorphism features of SVA appeal more to tion, the proposed research model shows an acceptable level of predic­ consumers’ utilitarian expectations rather than hedonic motives. tive power. The model meets the recommended criteria for reliability and validity indices and explains the significant amount of variance in Furthermore, visual appeal and social presence significantly influ­ all endogenous constructs: hedonic attitude (43%), utilitarian attitude ence utilitarian attitude. Most SVA available in markets comes in basic (69%), usage (53%), and WOM (37%). The obtained values of R2 are spherical or round shape in a single color (white, black, or blue). The within the highly acceptable level for research dealing with behavioral idea behind simple designs could be to allow users to put SVA nearby on predictions, especially in the domain of technology adoption (e.g., tables or bed rests. Moreover, SVA devices are compact and balanced so Alalwan et al., 2017; Hu et al., 2021). However, our results show a much that they do not fall easily. This explains why users relate visual appeal higher value of R2 for usage (53%) in comparison to R2 value (29.7%) of SVA to convenience and practical benefits. Similarly, SVA give an obtained by Hu et al. (2021) who studied the role of autonomy on SVA impression of having an ‘entity’ or a person in the vicinity, which can adoption. respond to users. Thus, SVA act as a solution to users’ loneliness ful­ filling users’ need for a human companion. Hedonic and utilitarian attitudes show a positive impact on SVA usage. However, utilitarian attitude has a stronger impact (vs. hedonic Lastly, SVA deliver symbolic benefits to owners as many users attitude) on SVA usage and users’ WOM recommendations. Thus, users perceive SVA prestigious. Individuals who perceive SVA as a symbol of perceive SVA more of a functional device leading to a favorable utili­ high prestige (vs. low prestige) are less likely to relate playfulness to tarian attitude. McLean and Osei-Frimpong (2019) examined the influ­ hedonic gratifications. In contrast, people attributing prestige to SVA are ence of hedonic and utilitarian benefits on SVA usage. They found that more likely to use it as an escape route from their mundane life. Thus, hedonic benefits were not a significant factor for determining SVA prestige moderates the influence of escapism and playfulness on hedonic usage. Our research examined hedonic attitudes instead of hedonic attitude in opposite directions. The results propose the boundary con­ benefits. In contrast to the findings of McLean and Osei-Frimpong dition of the degree of importance attached to the prestige associated (2019), the results indicate that hedonic attitude is a significant pre­ with SVA ownership, which affects users’ psychological gratifications dictor of SVA usage. We believe that our results are more in line with the (playfulness and escapism) and hedonic attitudes. influence of behavioral sciences research in IS domain, which outlines the importance of attitude in shaping human behavior (Ajzen, 1991; 6.1. Theoretical contributions Davis, 1989; Lee et al., 2020). Our findings suggest that users perceive SVA as a functional device that provides fun and entertainment. Thus, This paper makes several contributions to research on SVA usage and users’ hedonic motives shape hedonic attitudes that positively affect WOM behavior. First, it answers recent research calls on the adoption SVA usage (e.g., Tamilmani et al., 2019). and usage of SVA (e.g., Dwivedi, Ismagilova, et al., 2021; Dwivedi, Hugles, et al., 2021). The present study uses a cross-discipline approach The findings reveal that a favorable attitude toward SVA leads to a to extend the boundaries of SVA adoption by integrating the critical higher likelihood of WOM recommendations. From a task-fit perspec­ aspect of individual psychological experiences with design-centric tive, SVA can complete routine tasks with ease reducing users’ cognitive functional evaluations. We integrate elements from sociology and psy­ load. Borrowing the terminology from product performance literature, chology research to arrive at antecedents to attitudinal parameters. The we can say that SVA complete tasks efficiently and consistently. Thus, research model advances the current state of research on the adoption of SVA can be trusted and classified as high-quality products. In addition, emerging technologies. A majority of the extant literature on technology users experience joy and fun while interacting with SVA. This explains 9

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx adoption is based on the original or extended versions of TAM and adoption. This research is an attempt to fill this particular research gap. UTAUT frameworks (Tamilmani et al., 2019). However, recent literature A positive WOM is crucial for product success and drives viral diffusion reviews on emerging technologies clearly outline the need to integrate of advertising and marketing communications. Similarly, we expect theories and models from multiple disciplines to advance the knowledge favorable WOM should help in technology diffusion. We find that both in technology adoption in the context of emerging technologies (e.g., hedonic and utilitarian attitudes are critical for spreading WOM. Dwivedi et al., 2021). We also notice an increase in the usage of hedonic Moreover, hedonic satisfaction and functional performance of SVA are motivation models in research to predict the usage or adoption of crucial in influencing users’ attitudes. Thus, we contribute to the WOM emerging technologies like virtual reality stores (Kang et al., 2020) and literature by suggesting extrinsic gratifications as antecedents to WOM mobile payments (Kar, 2020). Our research examines psychological behavior in addition to the satisfactory performance of SVA. factors and design specific factors that influence users’ hedonic and utilitarian attitudes lading to SVA adoption and WOM recommenda­ 6.2. Implications for practice tions. The results cement the importance of psychological gratifications derived from using SVA. Our findings have several practical implications for marketers and other stakeholders working toward increasing the adoption of SVA and Second, the study focuses on users’ gratifications derived from SVA, generating positive WOM recommendations. The recent COVID-19 which is different from the typical organization-based technology pandemic has drastically changed the global business landscape. Most adoption research. This study complements and extends the work of of the physical retail businesses were forced to adopt digital platforms Tamilmani et al. (2019) about the role of hedonic motivations and (Swoboda, 2020). The pandemic brought substantial changes in con­ derived values from the tasks at hand. Hedonic motivations significantly sumer behavior, where consumer preferences for online shopping influence the outcomes that have hedonic values rather than utilitarian significantly increased. Many consumers experienced their first moment values. For example, people are more likely to use technology like mo­ of online shopping because they were scared of getting an infection due bile banking due to associated hedonic motives such as fun and enjoy­ to human touch at crowded places and physical retail stores. SVA should ment of using technology (Alalwan et al., 2017). The findings confirm be able to play a critical role in such uncertainties as they can alleviate that gratifications based outcomes derived from technology usage in­ users’ concerns about social distancing and touchless experiences. fluence utilitarian and hedonic attitudes that further influence SVA usage. SVA offer an element of playfulness to users who use SVA as a Industry data and research reports show that most users use smart gateway to escapism. We believe that users consider SVA as a com­ voice interaction technologies via their smartphones instead of smart panion with whom they can share experiences or feelings anytime. speakers or SVA (Foehr & Germelmann, 2020). Many users use voice However, SVA differ from a human companion in a sense that SVA abide interaction features on their phones to search information or to get di­ by what the owner says without its own thinking. Thus, interactions rections. Similarly, many users turn to SVA to perform a basic search, with SVA are without any conflict, enjoyable, and humorous. Interest­ play a song, or read news. However, SVA are used in a more convenient ingly, users perceive physical presence of SVA with anthropomorphism setting compared to a smartphone. For example, SVA are placed in a features as utilitarian benefits. The plausible explanation is in line with home at a designated place where users can easily access them using McLean and Osei-Frimpong (2019) findings, which suggest that the voice commands. Hence, marketers who have applied strategies to in­ companionship aspect and the virtual presence of a humanlike device crease the adoption of smartphones should alter their strategies for SVA. fulfill the functional benefits of reducing loneliness. Marketers should focus on making users’ attitudes more favorable to­ ward SVA. They must stress on hedonic benefits of SVA along with Third, the study reaffirms that attitude is a critical antecedent to functional appeal in their communications to differentiate form smart­ technology acceptance and usage (Dwivedi et al., 2019). People use SVA phones. Marketers can educate users that SVA can serve as a trustworthy for its functional as well as hedonic appeal. Hedonic motives are the and entertaining companion that can help users in dealing with loneli­ novel factors that may encourage users to try SVA and further usage ness and make them happy. SVA can be of extreme help in accom­ depends on the functional value. Thus, in the context of emerging plishing regular tasks that people tend to forget like scheduling technologies, which offer vivid, immersive, and highly interactive ex­ appointments, alarms, and even switching off electric lights. periences to users, hedonic motivations are at par with the functional benefits. Thus, a significant contribution of this research is toward the Social presence and playfulness highlight that SVA can tackle the hedonic motivation research that focuses on the extrinsic motivations to issue of loneliness and act as a substitute for humans at home or any use technology. SVA have unique characteristics of being a personal place. Hence, we see two types of opportunities here. First, the technical device that is mostly used in a private (home) environment. Hence, aspect where developers should make SVA more human in terms of technologies that resemble similar usage environments or objectives voice and AI capabilities to understand and interpret instructions. Sec­ should be evaluated on hedonic gratifications. ond, marketers should highlight the humanlike presence of SVA in their marketing communications. Marketers must understand and appreciate Fourth, a key finding of this research is the moderating role of that SVA fulfill functional benefits and provide psychological gratifica­ prestige. Prior research suggests that people are more likely to adopt a tions to users. Marketers can emphasize how users (especially kids and technology under social influence, which means technology is perceived teens) can have fun while interacting with SVA. Marketers can build prestigious and gets appreciation from others (Gursoy et al., 2019). specific content like trivia and quizzes related to education and knowl­ Emergent technologies like SVA or smart wearables signal a favorable edge that can stress learning with fun to encourage SVA adoption in social status in countries like India. The findings suggest that prestige families with kids. Similarly, the humanlike appearance should be part affects individuals’ hedonic attitudes. Generally, people use specific of the overall marketing communication strategy and product devel­ products to signal or enhance their prestige (conspicuous consumption). opment. We expect firms to improve the quality of voice accent and However, the findings bring a critical difference between playfulness support different languages so that people can freely talk to SVA in their and escapism due to the prestige associated with SVA, which contradicts native language. Such expansion will definitely increase SVA adoption. the established concept of conspicuous consumption similar to using augmented reality as an extension or replacement of physical products Esthetic or visual appeal is another crucial factor that affects users’ (Rauschnabel, 2021). attitudes. Since SVA are close to typical household items like white goods, they need better esthetics (e.g., more colors and shapes) to Lastly, a vast amount of research exists on the importance of WOM enhance their visual appeal and blend with surroundings. Amazon recommendations in the success of any product or service (Berger & continues to tweak the design of Alexa in new models and recently Milkman, 2012; Mishra & Satish, 2016). While an extensive amount of introduced the feature of a digital watch embedded in device appear­ WOM research exists in the marketing domain dealing with products ance. Marketers may learn a few lessons from blockbuster Hollywood and services, we notice scant research in the context of technology 10

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx movies such as Star wars or Wall-E, which have turned the robots and 7. Conclusion droids into friends, companions, or saviors. We believe that marketers have yet to realize the immense potential of design for SVA. Therefore, Our study highlights the importance of hedonic and utilitarian atti­ they should continue experiments with out-of-the-box innovative tudes that lead to SVA usage and WOM recommendations. This study designs. proposed and tested a research model based on flow theory and the theory of anthropomorphism. While flow theory focuses on the We find that prestige associated with SVA influences users’ hedonic immersive and absorbing experiences of using SVA, anthropomorphism attitudes. In fact, it has opposite effects on how playfulness and escapism reflects the humanness of SVA. We identified five antecedents (play­ affect users’ hedonic attitudes. This finding calls for a distinct segmen­ fulness, escapism, anthropomorphism, visual appeal, and social pres­ tation strategy by marketers based on users’ valuation of prestige ence) to hedonic and utilitarian attitudes based on previous studies in attained from using SVA. Till now, marketers have rarely used the technology adoption and multisensory research. The moderating role of prestige factor to influence SVA adoption. We believe that marketers prestige (social status) was examined for hedonic attitude. The findings should try to highlight this aspect. They can create ads and post on social revealed that playfulness and escapism significantly influence hedonic media to target specific user-segment similar to users of premium luxury attitude, whereas anthropomorphism, visual appeal, and social presence goods. Marketers may think about launching different versions of SVA affect utilitarian attitude. Though both attitudes had a significant effect that are superior in design and get a price premium. on SVA usage and WOM recommendations, the effect of utilitarian attitude was relatively stronger than the hedonic attitude. Furthermore, Lastly, our results notice that favorable hedonic and utilitarian atti­ the influence of escapism and playfulness on hedonic attitude is tudes can lead to WOM recommendations, which further can influence moderated by prestige associated with SVA. These results provide SVA adoption. So, we suggest that marketers should encourage user- actionable insights to the various stakeholders to increase the SVA generated content (UGC) on how they use SVA and what they feel adoption by focusing on functional and psychological benefits offered by about SVA. SVA may politely (or humorously) nudge and encourage SVA. This research adds a new dimension of favorable WOM that can be users to post/share their experiences with SVA on online platforms. highly effective in accelerating SVA adoption. Thus, SVA can be promoters of themselves. This may help marketers to get insights directly from users, which should pave the way for im­ Author statement provements in future SVA models. All authors have participated equally and substantially in the 6.3. Limitations and future research direction concept, design, acquisition of data analysis, drafting, writing, and revision of the manuscript. As with any research study, our study has certain limitations. Therefore the findings should be interpreted according to the context. The authors certify that the material or similar material has not been First, the study uses convenience sampling for online survey. Due to the and will not be submitted to or published in any other publication before COVID19 situation and subsequent lockdown, only online data collec­ its appearance in the International Journal of Information Management. tion was feasible. Hence, only the respondents who had access to the Internet participated in the study. Thus, this may affect the generaliz­ Funding ability of the findings. Moreover, the majority of the sample is relatively young (less than 40 years). Though research shows that young adults are This research did not receive any specific grant from funding more likely to adopt new technologies, the elderly population may also agencies in the public, commercial, or not-for-profit sectors. use SVA for unique benefits. Therefore, further research can extend the sample to a broader age group and non-Internet users. Second, the study Appendix A. Measures uses survey method that suffers from certain biases. We have taken recommended measures like the assurance of confidentiality and ano­ Playfulness (Müller-Stewens, Schlager, Ha¨ubl, & Herrmann, 2017). nymity to respondents. We also tested for common method bias to verify that it does not confound our findings. However, we suggest that further 1. Talking to voice assistant is fun. research may employ experimental design or longitudinal research to 2. Interacting with voice assistant is enjoyable. corroborate our findings. Third, we believe that SVA have a novelty and 3. I feel happy when interacting with voice assistant. enjoyment factor, which may be prominent in initial usage but may wear out in due course of time. Since attitude formation is a function of time, Escapism (Mathwick et al., 2001). forthcoming research should extend our model and examine the differ­ ences in attitudinal parameters across a longer period of SVA usage. 1. Interacting with voice assistant takes me to another world. Fourth, we did not specifically look at the language aspect of voice as­ 2. Interacting with voice assistant makes me feel like I am in another sistants. In a diverse country like India, users use a variety of native languages. SVA like Alexa and Google home have started supporting world. various languages, but they are still far from being perfect. Thus, SVA 3. I get so involved when I interact with voice assistant that I forget may not be able to interpret voice commands at times leading to dissatisfaction. We think this could be another exciting aspect where everything else. upcoming research can explore users’ satisfaction when interacting with SVA using native languages or different accents. SVA (Alexa and Google Anthropomorphism (Guido & Peluso, 2015). home) interact with users in a female voice, which could influence users’ attitudes and acceptance in patriarchal cultures. Thus, researchers can 1. Voice assistant acts like a person. investigate the impact of gender SVA voice in a cross-cultural setting. 2. Voice assistant talks like a human. Our research model uses hedonic and utilitarian attitudes to predict 3. Voice assistant interacts like a person. adoption and WOM behavior, which can be extended based on the complexity of tasks. SVA may be used by an individual or by a group of Visual Appeal (Mathwick et al., 2001). people in a family. Moreover, the location of SVA (e.g., common hall or bedroom) may also influence its usage. Thus, we recommend future 1. Design of voice assistant is attractive. researchers to explore the specific usage situation to get deeper insights 2. Voice assistant is visually appealing. into SVA usage and attitude formation toward SVA. 3. I like how the voice assistant looks. 11

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx Social Presence (Roy et al., 2019). Bergkvist, L., & Rossiter, J. R. (2007). The predictive validity of multiple-item versus single-item measures of the same constructs. Journal of Marketing Research, 44(2), 1. Interacting with voice assistant makes me feel comfortable, as if I am 175–184. with a friend. Chatterjee, S., & Kar, A. K. (2020). Why do small and medium enterprises use social 2. There is a sense of human contact when I interact with voice media marketing and what is the impact: Empirical insights from India. International assistant. Journal of Information Management, 53, Article 102103. https://doi.org/10.1016/j. ijinfomgt.2020.102103. 3. Interacting with voice assistant gives me some sense of social life. Chatterjee, S., Kar, A. K., & Dwivedi, Y. K. (2021). Intention to use IoT by aged Indian Hedonic Attitude (Spangenberg, Voss, & Crowley, 1997). consumers. Journal of Computer Information Systems. , Article 1873080. https://doi. Interacting with voice assistant is. org/10.1080/08874417.2021.1873080. 1. Dull – exciting Conner, K., & Bradford, A. (2020). Need a good laugh? 55 funniest things to ask Alexa to 2. Not fun – Fun raise your spirits. Retrieved from 〈https://www.cnet.com/how-to/need-a-good- 3. Unenjoyable – enjoyable laugh-55-funniest-things-to-ask-alexa-to-raise-your-spirits/〉. Utilitarian attitude (Spangenberg et al., 1997). Dabholkar, P. A., van Dolen, W. M., & de Ruyter, K. (2009). A dual-sequence framework Interacting with voice assistant is. for B2C relationship formation: Moderating effects of employee communication style in online group chat. Psychology & Marketing, 26(2), 145–174. 1. Useless – useful 2. Unnecessary – Necessary Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of 3. Unproductive – productive information technology. MIS Quarterly, 13(3), 319–340. Prestige (Gursoy et al., 2019). Dwivedi, Y. K., Hughes, L., Ismagilova, E., Aarts, G., Coombs, C., Crick, T., & Williams, M. D. (2021a). Artificial Intelligence (AI): Multidisciplinary perspectives 1. Having a voice assistance improves my social status. on emerging challenges, opportunities, and agenda for research, practice and policy. 2. Having a voice assistance establishes my position in society. International Journal of Information Management, 57, Article 101994. 3. Having a voice assistance is prestigious for me. Dwivedi, Y. K., Ismagilova, E., Rana, N. P., & Raman, R. (2020). Security, privacy and Word-of-mouth behavior (Mishra et al., 2018). risks within smart cities: Literature review and development of a smart city interaction framework. Information Systems Frontiers: A Journal of Research and 1. I will share positive word of mouth about voice assistant. Innovation, 1–22. https://doi.org/10.1007/s10796-021-10106-y. 2. I will recommend voice assistant to my friends. 3. I will encourage my friends to use or buy voice assistant. Dwivedi, Y. K., Ismagilova, E., Hughes, D. L., Carlson, J., Filieri, R., Jacobson, J., & Wang, Y. (2021c). Setting the future of digital and social media marketing research: Usage (McLean & Osei-Frimpong, 2019). Perspectives and research propositions. International Journal of Information How frequently do you use voice assistant? Management, 59, Article 102168. References Dwivedi, Y. K., Rana, N. P., Jeyaraj, A., Clement, M., & Williams, M. D. (2019). Re- examining the unified theory of acceptance and use of technology (UTAUT): Agarwal, R., & Karahanna, E. (2000). Time flies when you’re having fun: Cognitive Towards a revised theoretical model. Information Systems Frontiers, 21(3), 719–734. absorption and beliefs about information technology usage. MIS Quarterly, 24(4), 665–694. Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. Ahn, T., Ryu, S., & Han, I. (2007). The impact of web quality and playfulness on user acceptance of online retailing. Information & Management, 44(3), 263–275. Fernandes, T., & Oliveira, E. (2021). Understanding consumers’ acceptance of automated technologies in service encounters: Drivers of digital voice assistants adoption. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Journal of Business Research, 122, 180–191. Decision Processes, 50(2), 179–211. Foehr, J., & Germelmann, C. C. (2020). Alexa, can I trust you? Exploring consumer paths Alalwan, A. A., Dwivedi, Y. K., & Rana, N. P. (2017). Factors influencing adoption of to trust in smart voice-interaction technologies. Journal of the Association for mobile banking by Jordanian bank customers: Extending UTAUT2 with trust. Consumer Research, 5(2), 181–205. International Journal of Information Management, 37(3), 99–110. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with Bag, S., Pretorius, J. H. C., Gupta, S., & Dwivedi, Y. K. (2021). Role of institutional unobservable variables and measurement error. Journal of Marketing Research, 18(1), pressures and resources in the adoption of big data analytics powered artificial 39–50. intelligence, sustainable manufacturing practices and circular economy capabilities. Technological Forecasting and Social Change, 163, Article 120420. https://doi.org/ Foroudi, P., Gupta, S., Sivarajah, U., & Broderick, A. (2018). Investigating the effects of 10.1016/j.techfore.2020. smart technology on customer dynamics and customer experience. Computers in Human Behavior, 80, 271–282. Balakrishnan, J., & Dwivedi, Y. K. (2021aaa). Role of cognitive absorption in building user trust and experience. Psychology & Marketing, 38(4), 643–668. Google Insights (2017). Voice-Activated Speakers: People’s Lives Are Changing. Retrieved from 〈https://www.thinkwithgoogle.com/data-collections/voice-assistan Balakrishnan, J., & Dwivedi, Y. K. (2021bbb). Conversational commerce: Entering the ce-emerging-technologies/〉. next stage of AI-powered digital assistants. Annals of Operations Research, 1–35. https://doi.org/10.1007/s10479-021-04049-5. Grover, P., Kar, A. K., & Dwivedi, Y. K. (2020). Understanding artificial intelligence adoption in operations management: Insights from the review of academic literature Barnes, S. J. (2021). Understanding the overvaluation of facial trustworthiness in Airbnb and social media discussions. Annals of Operations Research, 1–37. https://doi.org/ host images. International Journal of Information Management, 56, Article 102265. 10.1007/s10479-020-03683-9. Bawack, R. E., Wamba, S. F., & Carillo, K. D. A. (2021). Exploring the role of personality, Guido, G., & Peluso, A. M. (2015). Brand anthropomorphism: Conceptualization, trust, and privacy in customer experience performance during voice shopping: measurement, and impact on brand personality and loyalty. Journal of Brand Evidence from SEM and fuzzy set qualitative comparative analysis. International Management, 22(1), 1–19. Journal of Information Management, 58, Article 102309. https://doi.org/10.1016/j. ijinfomgt.2021.102309. Gursoy, D., Chi, O. H., Lu, L., & Nunkoo, R. (2019). Consumers acceptance of artificially intelligent (AI) device use in service delivery. International Journal of Information Belk, R. (2017). The soul and the machine: Human like machines and machine like Management, 49, 157–169. humans. In A. Gneezy, V. Griskevicius, & P. Williams (Eds.), NA—advances in consumer research (Vol. 45, pp. 164–169). Association for Consumer Research. Hair, J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19(2), 139–152. Berger, J. (2014). Word of mouth and interpersonal communication: A review and directions for future research. Journal of Consumer Psychology, 24(4), 586–607. Hair, J. F., Risher, J. J., Sarstedt, M., & Ringle, C. M. (2019). When to use and how to report the results of PLS-SEM. European Business Review, 31(1), 2–24. Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal of Marketing Research, 49(2), 192–205. Han, S., Min, J., & Lee, H. (2016). Building relationships within corporate SNS accounts through social presence formation. International Journal of Information Management, 36(6), 945–962. Henseler, J., & Sarstedt, M. (2013). Goodness-of-fit indices for partial least squares path modeling. Computational Statistics, 28(2), 565–580. Hernandez-Ortega, B., & Ferreira, I. (2021). How smart experiences build service loyalty: The importance of consumer love for smart voice assistants. Psychology and Marketing, 38(7), 1122–1139. Hoffman, D. L., & Novak, T. P. (2018). Consumer and object experience in the internet of things: An assemblage theory approach. Journal of Consumer Research, 44(6), 1178–1204. Holmqvist, J., Ruiz, C. D., & Pen˜aloza, L. (2020). Moments of luxury: Hedonic escapism as a luxury experience. Journal of Business Research, 116, 503–513. Hsu, C. L., & Lu, H. P. (2004). Why do people play on-line games? An extended TAM with social influences and flow experience. Information & Management, 41(7), 853–868. Hu, Q., Lu, Y., Pan, Z., Gong, Y., & Yang, Z. (2021). Can AI artifacts influence human cognition? The effects of artificial autonomy in intelligent personal assistants. International Journal of Information Management, 56, Article 102250. https://doi.org/ 10.1016/j.ijinfomgt.2020.102250. Hughes, L., Dwivedi, Y. K., Misra, S. K., Rana, N. P., Raghavan, V., & Akella, V. (2019). Blockchain research, practice and policy: Applications, benefits, limitations, 12

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx emerging research themes and research agenda. International Journal of Information Novak, T. P., & Hoffman, D. L. (2019). Relationship journeys in the internet of things: a Management, 49, 114–129. framework for understanding interactions between consumers and smart objects. Hur, J. D., Koo, M., & Hofmann, W. (2015). When temptations come alive: How Journal of the Academy of Marketing Science, 47(2), 216–237. anthropomorphism undermines self-control. Journal of Consumer Research, 42(2), 340–358. Park, K., Kwak, C., Lee, J., & Ahn, J. H. (2018). The effect of platform characteristics on Ismagilova, E., Rana, N. P., Slade, E. L., & Dwivedi, Y. K. (2020). A meta-analysis of the the adoption of smart speakers: Empirical evidence in South Korea. Telematics and factors affecting eWOM providing behaviour. European Journal of Marketing, 55(4), Informatics, 35(8), 2118–2132. 1067–1102. Jones, S., Cronin, J., & Piacentini, M. G. (2018). Mapping the extended frontiers of Phan, T. (2019). Amazon Echo and the aesthetics of whiteness. Catalyst: Feminism, escapism: Binge-watching and hyperdiegetic exploration. Journal of Marketing Theory, Technoscience, 5(1), 1–38. Management, 34(5–6), 497–508. Kang, H. J., Shin, J. H., & Ponto, K. (2020). How 3D virtual reality stores can shape Pillai, R., Sivathanu, B., & Dwivedi, Y. K. (2020). Shopping intention at AI-powered consumer purchase decisions: the roles of informativeness and playfulness. Journal of automated retail stores (AIPARS). Journal of Retailing and Consumer Services, 57, Interactive Marketing, 49, 70–85. Article 102207. https://doi.org/10.1016/j.jretconser.2020.102207. Kar, A. K. (2020). What affects usage satisfaction in mobile payments? Modelling user generated content to develop the “digital service usage satisfaction model”. Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method Information Systems frontiers: a Journal of Research and Innovation, 1–21. https://doi. biases in behavioral research: A critical review of the literature and recommended org/10.1007/s10796-020-10045-0. remedies. Journal of Applied Psychology, 88(5), 879–903. Karimova, G. Z., & Goby, V. P. (2020). The adaptation of anthropomorphism and archetypes for marketing artificial intelligence. Journal of Consumer Marketing, 38(2), Ponsignon, F., Lunardo, R., & Michrafy, M. (2020). Why are international visitors more 229–238. satisfied with the tourism experience? The role of hedonic value, escapism, and Kim, Y., & Sundar, S. S. (2012). Anthropomorphism of computers: Is it mindful or psychic distance. Journal of Travel Research. , Article 004728752096117. https://doi. mindless? Computers in Human Behavior, 28(1), 241–250. org/10.1177/0047287520961175. Krishna, A., & Schwarz, N. (2014). Sensory marketing, embodiment, and grounded cognition: A review and introduction. Journal of Consumer Psychology, 24(2), Poushneh, A. (2021). Humanizing voice assistant: The impact of voice assistant 159–168. personality on consumers’ attitudes and behaviors. Journal of Retailing and Consumer Kuo, A., Lutz, R. J., & Hiler, J. L. (2016). Brave new World of Warcraft: A conceptual Services, 58, Article 102283. https://doi.org/10.1016/j.jretconser.2020.102283. framework for active escapism. Journal of Consumer Marketing, 33(7), 498–506. Kushwaha, A. K., & Kar, A. K. (2020). Language model-driven chatbot for business to PWC (2019). Prepare for the voice revolution. Retrieved from 〈https://www.pwc.com/ address marketing and selection of products. In S. K. Sharma, Y. K. Dwivedi, us/en/services/consulting/library/consumer-intelligence-series/voice-assistants. B. Metri, & N. P. Rana (Eds.), Re-imagining Diffusion and Adoption of Information html〉. Technology and Systems: A Continuing Conversation. TDIT 2020. IFIP Advances in Information and Communication Technology, 617 pp. 16–28). Cham: Springer. https:// Rana, N. P., Dwivedi, Y. K., Williams, M. D., & Weerakkody, V. (2016). Adoption of doi.org/10.1007/978-3-030-64849-7_3. online public grievance redressal system in India: Toward developing a unified view. Kushwaha, A. K., Kar, A. K., & Dwivedi, Y. K. (2021). Applications of big data in Computers in Human Behavior, 59, 265–282. emerging management disciplines: A literature review using text mining. International Journal of Information Management Data Insights, 1(2), Article 100017. Rauschnabel, P. A. (2021). Augmented reality is eating the real-world! The substitution Lee, K., Lee, K. Y., & Sheehan, L. (2020). Hey Alexa! A magic spell of social glue?: Sharing of physical products by holograms. International Journal of Information Management, a smart voice assistant speaker and its impact on users’ perception of group 57, Article 102279. https://doi.org/10.1016/j.ijinfomgt.2020.102279. harmony. Information Systems Frontiers, 22(3), 563–583. Liao, Y., Vitak, J., Kumar, P., Zimmer, M., & Kritikos, K. (2019). Understanding the role Rauschnabel, P. A., Rossmann, A., & tom Dieck, M. C. (2017). An adoption framework for of privacy and trust in intelligent personal assistant adoption. International conference mobile augmented reality games: The case of Pok´emon Go. Computers in Human on information (pp. 102–113). Cham: Springer. Behavior, 76, 276–286. Lindell, M. K., & Whitney, D. J. (2001). Accounting for common method variance in cross-sectional research designs. Journal of Applied Psychology, 86(1), 114–121. Ringle, C. M., Wende, S., & Becker, J. M. (2015). SmartPLS 3. Boenningstedt: SmartPLS Lopatovska, I., Rink, K., Knight, I., Raines, K., Cosenza, K., Williams, H., … Martinez, A. GmbH. 〈http://www.smartpls.com〉. (2019). Talk to me: Exploring user interactions with the Amazon Alexa. Journal of Librarianship and Information Science, 51(4), 984–997. Roy, S. K., Singh, G., Hope, M., Nguyen, B., & Harrigan, P. (2019). The rise of smart Lovett, L. (2020). Tech’s role in helping seniors overcome loneliness amid COVID-19 consumers: role of smart servicescape and smart consumer experience co-creation. pandemic. Retrieved from 〈https://www.mobihealthnews.com/news/techs-role-he Journal of Marketing Management, 35(15–16), 1480–1513. lping-seniors-overcome-loneliness-amid-covid-19-pandemic〉. Mathwick, C., & Rigdon, E. (2004). Play, flow, and the online search experience. Journal Savage, M. (2020). Voice technologies using AI are being used to help combat loneliness of Consumer Research, 31(2), 324–332. in countries including Sweden and the UK. Should they be used more widely as Mathwick, C., Malhotra, N., & Rigdon, E. (2001). Experiential value: conceptualization, coronavirus spreads? Retrieved from 〈https://www.bbc.com/worklife/article measurement and application in the catalog and Internet shopping environment☆. /20200325-can-voice-technologies-using-ai-fight-elderly-loneliness〉. Journal of Retailing, 77(1), 39–56. McLean, G., & Osei-Frimpong, K. (2019). Hey Alexa… examine the variables influencing Spangenberg, E. R., Voss, K. E., & Crowley, A. E. (1997). Measuring the hedonic and the use of artificial intelligent in-home voice assistants. Computers in Human utilitarian dimensions of attitude: a generally applicable scale. In M. Brucks, & Behavior, 99, 28–37. D. J. MacInnis (Eds.), Advances in consumer research (Vol. 24, pp. 235–241). Provo, Mishra, A., & Satish, S. M. (2016). eWOM: Extant research review and future research UT: Association for Consumer Research. avenues. Vikalpa, 41(3), 222–233. Mishra, A., Maheswarappa, S. S., & Colby, C. L. (2018). Technology readiness of Straub, D., Boudreau, M. C., & Gefen, D. (2004). Validation guidelines for IS positivist teenagers: A consumer socialization perspective. Journal of Services Marketing, 32(5), research. Communications of the Association for Information systems, 13(1), 380–427. 592–604. Mishra, A., & Shukla, A. (2020). Psychological determinants of consumer’s usage, Swoboda, C. (2020). COVID-19 Is Making Alexa And Siri A Hands-Free Necessity. satisfaction, and word-of-mouth recommendations toward smart voice assistants. In Retrieved from 〈https://www.forbes.com/sites/chuckswoboda/2020/04/06/covid S. K. Sharma, Y. K. Dwivedi, B. Metri, & N. P. Rana (Eds.), Re-imagining diffusion and -19-is-making-alexa-and-siri-a-hands-free-necessity/〉. adoption of information technology and systems: a continuing conversation. TDIT 2020. IFIP Advances in Information and Communication Technology, 618 pp. 274–283). Tamilmani, K., Rana, N. P., Nunkoo, R., Raghavan, V., & Dwivedi, Y. K. (2020). Indian Cham: Springer. https://doi.org/10.1007/978-3-030-64849-7_24. travellers’ adoption of airbnb platform. Information Systems Frontiers. https://doi. Mishra, A., Shukla, A., Rana, N. P., & Dwivedi, Y. K. (2021). From “touch” to a org/10.1007/s10796-020-10060-1. “multisensory” experience: The impact of technology interface and product type on consumer responses. Psychology & Marketing, 38(3), 385–396. Tamilmani, K., Rana, N. P., Prakasam, N., & Dwivedi, Y. K. (2019). The battle of brain vs. Moriuchi, E. (2019). Okay, Google!: An empirical study on voice assistants on consumer heart: A literature review and meta-analysis of “hedonic motivation” use in UTAUT2. engagement and loyalty. Psychology & Marketing, 36(5), 489–501. International Journal of Information Management, 46, 222–235. Moriuchi, E. (2021). An empirical study on anthropomorphism and engagement with disembodied AIs and consumers’ re-use behavior. Psychology & Marketing, 38(1), Venkatesh, V., & Bala, H. (2008). Technology acceptance model 3 and a research agenda 21–42. on interventions. Decision Sciences, 39(2), 273–315. Müller-Stewens, J., Schlager, T., Ha¨ubl, G., & Herrmann, A. (2017). Gamified information presentation and consumer adoption of product innovations. Journal of Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of Marketing, 81(2), 8–24. information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. Nakamura, J., & Csikszentmihalyi, M. (2009). Flow theory and research. In C. R. Snyder, & S. J. Lopez (Eds.), Oxford handbook of positive psychology (pp. 195–206). Oxford, Venkatesh, V., Thong, J. Y., & Xu, X. (2012). Consumer acceptance and use of UK: Oxford University Press. information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly, 36(1), 157–178. Verma, S., Sharma, R., Deb, S., & Maitra, D. (2021). Artificial intelligence in marketing: Systematic review and future research direction. International Journal of Information Management Data Insights, 1(1), Article 100002. Vilpponen, A., Winter, S., & Sundqvist, S. (2006). Electronic word-of-mouth in online environments: Exploring referral networks structure and adoption behavior. Journal of Interactive Advertising, 6(2), 8–77. Vimalkumar, M., Sharma, S. K., Singh, J. B., & Dwivedi, Y. K. (2021). Okay google, what about my privacy?’: User’s privacy perceptions and acceptance of voice based digital assistants. Computers in Human Behavior, 120, Article 106763. https://doi.org/ 10.1016/j.chb.2021.106763. Voss, K. E., Spangenberg, E. R., & Grohmann, B. (2003). Measuring the hedonic and utilitarian dimensions of consumer attitude. Journal of Marketing Research, 40(3), 310–320. Walther, J. B. (1992). Interpersonal effects in computer-mediated interaction a relational perspective. Communication Research, 19(1), 52–90. Webster, J., & Martocchio, J. J. (1992). Microcomputer playfulness: Development of a measure with workplace implications. MIS Quarterly, 16(2), 201–226. Williams, M. D., Rana, N. P., & Dwivedi, Y. K. (2015). The unified theory of acceptance and use of technology (UTAUT): A literature review. Journal of Enterprise Information Management, 28(3), 443–488. Wu, L., Chiu, M. L., & Chen, K. W. (2020). Defining the determinants of online impulse buying through a shopping process of integrating perceived risk, expectation- 13

A. Mishra et al. International Journal of Information Management xxx (xxxx) xxx confirmation model, and flow theory issues. International Journal of Information Management, 52, Article 102099. 14

Telematics and Informatics 62 (2021) 101628 Contents lists available at ScienceDirect Telematics and Informatics journal homepage: www.elsevier.com/locate/tele “OK, Google, why do I use you?” Motivations, post-consumption evaluations, and perceptions of voice AI assistants Tae Rang Choi a,*, Minette E. Drumwright b a Department of Strategic Communication, Texas Christian University, 2805 S. University Drive, Fort Worth, TX, 76129, USA b Stan Richards School of Advertising & Public Relations, The University of Texas at Austin, 300 W Dean Keeton St, Austin, TX, 78712, USA ARTICLE INFO ABSTRACT Keywords: Voice-activated, artificial intelligence–based assistants (voice AI assistants) have become an in­ Artificial intelligence tegral part of everyday life because they can be easily activated to complete numerous tasks. Voice AI assistant However, little is known about what motivates people to use them or how these motivations Human− AI interaction influence users’ post-consumption evaluations and perceptions. This study develops measures to Motivation capture uses and gratifications related to voice AI assistants. It identifies five primary motivations Perception for use—social interaction, personal identity, conformity, life efficiency, and information. Results CASA show that the utilitarian motivations of life efficiency and information influence all post- consumption evaluations and behavioral intentions positively (i.e., users’ attitudes, satisfaction, and intentions to continue using voice AI assistants). However, social motivations are also important. Social interaction and conformity motivations also influence user satisfaction, and the conformity motivation shapes individuals’ intentions to continue using voice AI assistants. The findings further demonstrate that users’ motivations influence perceptions of their voice AI as­ sistants’ roles. In keeping with the “Computers Are Social Actors” (CASA) paradigm, users motivated by social interaction are likely to perceive a voice AI assistant as socially attractive and as a friend, whereas users motivated by life efficiency are likely to perceive it as an assistant. Users motivated by information are more likely to perceive a voice AI assistant as technology, while those motivated by social interaction are less likely to do so. The implications of these findings are discussed, and recommendations for future research are provided. 1. Introduction “OK, Google, …” “Alexa, …” These names are now familiar household words spoken in bedrooms, living rooms, and kitchens throughout the nation. They are “wake words” that users say to activate products powered by artificial intelligence (AI). These products, referred to as “voice AI assistants,” use automated speech recognition and natural language processing to perform various daily tasks for users. The number of consumer products equipped with AI has increased dramatically since Apple unveiled its AI- powered iPhone in 2011 (i.e., “Siri” in iPhone 4). It is predicted that 123 million people will use voice AI assistants by 2021 (Pet­ rock, 2019). Voice AI technology provides compelling benefits and has fundamentally reshaped many aspects of users’ lives in terms of communication, media use, entertainment, information searches, and purchases (Kinsella and Mutchler, 2019). The use of voice AI assistants is projected to continue expanding, especially in the branding and marketing realms (Kleinberg, 2018a). For instance, voice- * Corresponding author. E-mail addresses: [email protected] (T.R. Choi), [email protected] (M.E. Drumwright). https://doi.org/10.1016/j.tele.2021.101628 Received 18 November 2020; Received in revised form 7 April 2021; Accepted 13 April 2021 Available online 24 April 2021 0736-5853/© 2021 Elsevier Ltd. All rights reserved.

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 controlled shopping has a promising future: voice shoppers are anticipated to reach 38 million in the U.S. by 2021 (Koch, 2019), and by 2023, companies are expected to spend $7 billion annually to market their brands through smart technologies such as voice AI as­ sistants (MediaPost, 2019). The current study focuses on smart speakers, which are the most popular form of voice AI assistants (e.g., Google Assistant, Amazon Echo). Voice AI assistants offer users easy access to a range of real-life benefits through advanced voice technologies embedded in smart speakers. These benefits include practical tasks such as setting alarms and timers, checking the weather, looking up recipe ingredients, reminding users of their schedules or shopping lists, and playing music upon request. Voice AI assistants may also provide psycho­ logical and social benefits because they sound similar to humans and display an array of verbal abilities, such as offering greetings and even telling jokes. In fact, interactions between a user and a voice AI assistant can create the perception of a personal relationship. Studies have shown that users may consider a voice AI assistant a friend, a conversation partner, or a family member (Purington et al., 2017; Rhee and Choi, 2020; Wang et al., 2020; Zhao and Rau, 2020). Despite the growing popularity of voice AI assistants, little research has investigated interactions between users and these voice AI assistants. For instance, more research is needed to understand the social dimensions of interactions between users and voice AI as­ sistants (Acemoglu and Restrepo, 2018). Research is similarly lacking on the relationships between users’ motivations to use voice AI assistants and these individuals’ post-consumption evaluations and perceptions. Therefore, the present study takes a user-centered approach to explore people’s motivations for using a voice AI assistant and to examine the associations between users’ motivations and their post-consumption evaluations and perceptions of the roles of their voice AI assistants. 2. Theoretical background and research questions This research is conceptually and theoretically informed by earlier work on voice AI assistants, motivations, and the computers are social actors (CASA) research paradigm. 2.1. Voice AI assistants and multidimensional motivations The origins of voice AI assistants date back to early stages of the digital revolution. Drawing on Vannevar Bush’s work (1945), Alan Turing (1950) raised the provocative question “Can machines think?” and asserted that machines can be designed to be capable of accomplishing intelligent tasks generally completed by humans. Synthesizing these notions, McCarthy and colleagues (1955) coined the term and concept of “artificial intelligence” in reference to the science and engineering of creating intelligent machines and computer programs. More recently, AI has come to refer to programming computers that incorporate and simulate qualities of the human mind, such as the ability to understand language, acquire knowledge, memorize information, recognize images, and solve problems (Miller, 2019). Within social science, AI encompasses technology that enables digital computers or computer-controlled robots to perform tasks similar to intelligent beings (i.e., humans) (Miller, 2019). Even today’s nascent phase of AI technology (i.e., weak AI) has influenced society in broad ways such as improving energy performance, diagnosing human disease, caring for an aging population, and facilitating social interaction in children with autism (e.g., Marr, 2018). This technology has also been increasingly incorporated into individuals’ everyday lives (McLean and Osei-Frimpong, 2019). For example, text-based AI assistants (e.g., chatbots) can have conversations with consumers and provide customer service on websites or via messaging apps (e.g., Facebook Messenger). Leading technology companies (e.g., Amazon, Google) have introduced smart speakers with voice AI assistants to the consumer electric device market, and individuals have welcomed them into their homes, leading to skyrocketing sales (Kemp, 2019). However, much remains to be learned about how and why people use voice AI assistants (Kinsella and Mutchler, 2019). A long history of research in social psychology and communication has demonstrated the importance of understanding individuals’ motivations to gain insight into their attitudes and behaviors around media use. Katz et al. (1973) introduced the uses-and- gratifications approach to create a user-centered, theoretical framework delineating why and how individuals adopt and use media (Katz et al., 1973). This theory posits that users, who are active and goal-oriented agents, seek out and select media that gratify specific needs (Sundar and Limperos, 2013; Khan, 2017). “Uses” illuminate the motivations that prompt users to adopt a certain type of media, and “gratifications” capture the satisfaction gained from a particular type of media (Joinson, 2008). Sundar and Limperos (2013) called for a different approach to uses-and-gratifications research, termed “Uses and Grats 2.0.” They asserted that research on digital media had focused too much on gratifications and too little on the functions and capabilities of new media technologies. Specifically, they contended that the characteristics of technological innovation—the technical capabilities and features of new media—afford consumers new opportunities to fulfill their needs. As such, these attributes represent fundamental factors driving uses and gratifications and thus warrant closer scholarly attention. These technical features are conceptualized as “affordances.” For instance, “interactivity” is an affordance involving various types of new media that enable users to engage with content in a personal way; it provides users new possibilities for fulfilling their needs in ways that traditional media do not. As another example, smart speakers provide individuals the “affordance” of having voice-based interaction with AI (human − AI interaction) in ways that other computer-based media do not. Research has demonstrated that the distinct technical capabilities and features of new media (i.e., their affordances) are important aspects of media use and motivations, which have attitudinal and behavioral conse­ quences (Choi and Sung, 2018; Rauschnabel et al., 2017; Sundar and Limperos, 2013). However, Sundar and Limperos (2013) declared that too often researchers use measures of gratification designed for traditional media and/or conceptualize and operationalize gratifications too broadly. Recently, uses and gratifications theory has been applied in research on a variety of new media and digital communication technologies. It has been adopted to elucidate motivations for using the internet (Papacharissi and Rubin, 2000), social media (Buzeta 2

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 et al., 2020; Khan, 2017), mobile-based augmented reality (AR) games (Rauschnabel et al., 2017), mobile instant messaging (Kaur et al., 2020), chatbots (Brandtzaeg and Følstad, 2017), and smart devices (Cho and Lee, 2017; Park and Lee, 2012; Rauschnabel, 2018). Empirical studies have revealed that factors motivating people to use these new technologies include not only utilitarian benefits but also symbolic meanings and benefits related to personal identity and social integration. Particular motivations for using new tech­ nologies include information seeking and acquisition, efficiency, entertainment, self-expression, trendiness, communication, and social interaction (Buzeta et al., 2020; Cho and Lee, 2017; Park and Lee, 2012; Rauschnabel, 2018; Sundar and Limperos, 2013). Similarly, previous research on voice AI assistants has documented that individuals use these devices for purposes such as social interaction, information, utilitarian support, and symbolic expression (Lee et al., 2019; McLean and Osei-Frimpong, 2019; Moriuchi, 2019; Purington et al., 2017). However, little is known about users’ post-consumption attitudes and behavioral outcomes when interacting with voice AI assistants (e.g., users’ attitudes, satisfaction with, and intentions to continue using these devices). Scholars have demonstrated that understanding users’ motivations can clarify the drivers of post-use attitudinal and behavioral outcomes in various technological contexts (Cho and Lee, 2017; Kaur et al., 2020; Moriuchi, 2019; Papacharissi and Rubin, 2000). For instance, the self-presentation motivation has been shown to affect individuals’ intentions to continue playing mobile AR games (Rauschnabel et al., 2017); the motivation to have an emergency contact influenced continued use of smart devices among people with physical disabilities (Cho and Lee, 2017); and the affection motivation informed ongoing use of mobile instant messaging (Kaur et al., 2020). More research is needed to fully understand the multidimensional technological affordances offered by voice-controlled AI devices that influence users’ motivations along with their effects on users’ post-consumption evaluations and behavioral intentions. Studies involving uses and gratifications theory have encouraged researchers to explore media-specific motivations and related out­ comes rather than adopting established constructs (e.g., Papacharissi and Rubin, 2000; Sundar and Limperos, 2013). Therefore, the following research questions are proposed in this study: RQ1. What are users’ primary motivations for using a voice AI assistant? RQ2. What are the associations between users’ motivations and their attitudes, satisfaction, and intentions to continue using a voice AI assistant? 2.2. Computers are social actors (CASA) research paradigm The CASA research paradigm contends that individuals automatically and unconsciously perceive computers as if they are “social actors” and respond to them in ways similar to human-to-human interaction (Nass and Moon, 2000; Reeves and Nass, 1996). This paradigm was derived from the media equation theory (Reeves & Nass, 1996), which posits that “individuals’ interactions with computers, TV, and new media are fundamentally social and natural, just like interaction in real life” (p. 5). The CASA paradigm focuses on individuals’ perceptions of and responses to the computers they use. Numerous studies have indicated that individuals respond to computers based on social rules and that people have social expec­ tations of computers similar to those in human social relationships. For instance, scholars have found that people politely interact with computers as they would in human-to-human communication (Reeves and Nass, 1996) and perceive computers as cooperative teammates and credible information sources (Nass et al., 1996). Once a computer divulges information about itself, individuals tend to disclose more personal information (Moon, 2000). Furthermore, researchers have found that people ascribe human personalities to computers and evaluate robots with personalities similar to their own as socially attractive (Lee et al., 2006; Nass and Lee, 2001; Nass and Moon, 2000; Reeves and Nass, 1996). In line with the logic of the CASA research paradigm, scholars have asserted that minimal social cues, such as the human-like voice of a voice AI assistant, can enable human–technology interaction that approximates human social interaction. Today’s ever-evolving technologies have enabled voice AI assistants to use advanced natural language and have human-like voices (e.g., verbal cues in spoken communication), making these assistants likely to be perceived as social actors (Feine et al., 2019; Go and Sundar, 2019; Lee et al., 2019; Zhao and Rau, 2020). Some individuals even use expressions of politeness (e.g., “please,” “thank you”) when interacting with voice AI assistants and thus perceive themselves as having a conversation with another person (Kleinberg, 2018b). In addition, the usage environment can render interactions between users and voice AI assistants more engaging and more similar to social interaction. To illustrate, most individuals use smart speakers in their homes or other personal spaces (Kleinberg, 2018b). Also, occupying the same space in close quarters facilitates social interaction, which encourages users to interact with their voice AI assistants as peers or companions (Vinciarelli et al., 2009). Previous studies have revealed that the more individuals perceive a voice AI assistant as a person, the more satisfied they are with the device and the more likely they are to continue using it (Ki et al., 2020; Lee et al., 2019; Purington et al., 2017). Research has also suggested that voice AI assistants can be perceived as social agents that play different roles, such as that of a friend, assistant, servant, or expert (Kim et al., 2019; Schweitzer et al., 2019; Wang et al., 2020; Zhao and Rau, 2020). For instance, when users develop social relationships with voice AI assistants (Ki et al., 2020), people have been shown to perceive the devices as more socially attractive and to be more inclined to like their recommendations (Rhee and Choi, 2020; Sundar et al., 2017). To some extent, voice-controlled AI assistants are designed to mimic friends because they use everyday language to engage in seemingly friendly conversation. In addition, just as people address others using their names in social settings, users must speak to voice AI assistants using their unique names (e.g., Alexa), which serve as “wake words”; these greetings further lead users to perceive these devices as friends (Branham and Roy, 2019). Purington et al.’s (2017) analysis of Amazon reviews indicated that people who described their voice AI assistant by name (e.g., Alexa) and referred to it using personal pronouns, particularly gendered pronouns (i.e., “she”), 3

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 were more likely to use it for social communication and to perceive it as a conversation partner, friend, or family member. In contrast, people who described their voice AI assistants as an object (i.e., “it”) were more likely to consider it a technological device that provides functionality. As such, it is reasonable to argue that users perceive their voice AI assistants in a manner that aligns with their motivations for using them. For example, people who are motivated to use a voice AI assistant for utilitarian purposes (e.g., completing daily tasks) will presumably perceive it as an assistant or mere technological device. Conversely, users who use voice AI assistants for social conversations should perceive them as friends. Although studies have demonstrated that users have distinct perceptions of voice AI assistants, scarce research has investigated what drives users’ perceptions of these devices. Accordingly, the following research questions are proposed to examine the influences of individuals’ motivations for using voice AI assistants on users’ perceptions of the devices while interacting with them: RQ3. What are the associations between users’ motivations and their perceptions of the social attraction of a voice AI assistant? RQ4. What are the associations between users’ motivations and their perceptions of the role of a voice AI assistant as (a) a friend, (b) an assistant, and (c) technology? 3. Methods 3.1. Sample A survey method was employed to gain an understanding of individuals’ underlying motivations for using smart speakers with voice AI assistants with high adoption rates in the consumer market. Respondents were recruited from a large university in the southeastern United States. The use of smart speakers is mainstream among Gen Z and Millennials (YPulse Inc., 2020); thus, users between 17 and 38 years old constituted the target population for this research. Focusing on a population who generally used smart speakers was appropriate given that this study did not intend to explore motivations for first-time adoption but rather the associations among individuals’ motivations for using voice AI assistants and their post-consumption attitudes and behaviors related to continued use. The initial sample consisted of 268 respondents who identified themselves as current users of a smart speaker with a voice AI assistant during the screening procedure. Respondents who submitted surveys that were incomplete or that contained extreme or abnormally consistent rating patterns were eliminated; the final sample included 256 respondents. Among them, 37.5% were men and 61.7% were women; 0.8% preferred not to identify their gender. Respondents’ average age was 21.45 years. The sample was approximately 52.0% Caucasian, 17.2% Hispanic, 13.3% African American, 12.9% Asian/Asian American, and 4.6% “other.” 3.2. Procedure After respondents consented to participate, three steps were taken to ensure that respondents were eligible to take part in the study. First, those who had never owned a voice AI assistant were redirected to the end of the survey. Second, respondents who used a voice AI assistant were asked to identify the specific device they used most (e.g., Amazon Echo) in an open-ended question. They were then asked a second screening question to gauge the frequency of their use of a voice AI assistant; those who answered that they “almost never” used their voice AI assistant were redirected to the end of the survey. Finally, respondents were asked their age, and those younger than 17 or older than 38 (i.e., outside the target population) were redirected to the end of the survey. Respondents who met survey qualifications were next asked to answer a series of questions about the voice AI assistant they had specified. Respondents were first asked to rate a series of statements, each of which described a reason for using a voice AI assistant. Next, respondents were instructed to complete measures regarding their attitudes toward their voice AI assistants, their satisfaction with these devices, and their intentions to continue using them. They were then asked a set of questions about the social attraction of their voice AI assistant and how they would describe the assistant from one of the following options: friend, assistant, or technology. After completing the main measures, respondents were asked to indicate the number of devices they owned, the average duration of time per use of their voice AI assistant, when they used the device most, and its location. After answering multiple demographic items, re­ spondents were debriefed and thanked for their participation. 3.3. Measures As noted earlier, previous uses and gratifications research in the context of new media has been criticized because scholars often adopt measures of uses and gratifications designed for traditional media and/or conceptualize gratifications so broadly that they are not tailored to the unique affordances of new media (Sundar and Limperos, 2013). As such, this study involved extensive measure development specific to voice AI assistants. Measures were devised to (1) assess users’ motivations for using a voice AI assistant; (2) explore the relationships among users’ motivations and post-consumption responses (i.e., attitudes, satisfaction, and intentions to continue using a voice AI assistant); and (3) understand users’ perceptions of their voice AI assistants (i.e., social attraction and perceived role of a voice AI device as a friend, an assistant, or technology). 3.3.1. Motivations for using voice AI assistants The current study integrated quantitative and qualitative approaches to gain an understanding of consumers’ reasons for using a voice AI assistant. Prior to the main study, 107 current users of voice AI assistants were asked to write down the reasons they use their 4

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 devices. An initial pool of 69 items was generated based on thematic analysis of textual data from the survey and a review of the literature (e.g., Cho & Lee, 2017; Hsu & Lin, 2016; Koo et al., 2015). Sixteen items were eliminated to avoid redundancy, and several items were modified slightly to ensure applicability to voice AI assistants. Ultimately, 53 motivation items were retained. To test the items’ face validity, 13 current users of voice AI assistants reviewed the initial 53 items. Users were asked to evaluate whether each item was relevant to the context of a voice AI assistant, and they were allowed to suggest additional motivations for using such devices. Eleven items were then removed due to irrelevancy or duplicate meanings. In the end, 42 unique statements illustrating reasons why people use a voice AI assistant in their everyday lives were retained. All items were scored on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree). A single index for each identified motivation was generated by averaging the item scores. See section 4.2 for details. 3.3.2. Attitudes toward voice AI assistants A measure of attitudes toward smart devices, one of the main dependent variables in this study, was adopted from research on the internet of things (Hsu and Lin, 2016) with slight modifications to suit the context of a voice AI assistant. The measure included three items (e.g., “Overall, my attitude toward using a voice AI assistant is favorable”) measured on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree; α = 0.89). Item scores were averaged to generate an index score. 3.3.3. Satisfaction with voice AI assistants The extent to which users were satisfied with their voice AI assistants was evaluated with four items, such as “Overall, I am satisfied with the voice AI assistant” (Lee et al., 2019). Relevant items were revised slightly to align with the context of the interaction between an individual and a voice AI assistant. Respondents were asked to indicate the extent to which they agreed or disagreed with each item on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree; α = 0.82). A single index was created by averaging the item scores. 3.3.4. Intentions to continue using voice AI assistants Users’ intentions to keep using a voice AI assistant was measured based on three items adapted from Hsu and Lin (2016), which were modified to suit the context of this research. A sample item was “I plan to keep using the voice AI assistant in the future.” Each item was rated on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree; α = 0.82). An index score was generated by averaging the item scores. 3.3.5. Social attraction of voice AI assistants Social attraction, also known as social appeal, concerns the extent to which individuals perceive a communication experience with others positively (Davis and Perkowitz, 1979). The social attraction of a voice AI assistant was evaluated using four items, such as “I Table 1 Sample Characteristics. Sample Profile Age, mean (min–max) 21.5 (17–38) years Number of devices owned 80.1 Gender 1 16.0 37.5 2 Male 61.7 3 2.0 Female 4 or more 1.9 Prefer not to say 0.8 Single-person household 84.8 Duration per use 50.8 Brand of devices Less than 10 min 35.9 Amazon (Alexa, Echo) 69.5 10 to less than 30 min 11.0 Google (Home) 24.6 30 to less than 60 min Apple (Homepod) 60 min or longer 2.3 Samsung (Bixby) 4.7 Usage period Microsoft (Cortana) 0.8 Less than 3 month 24.9 Usage frequency 0.4 3 to less than 6 months 21.9 Several times a day 6 months to less than 1 year 18.8 Daily or almost daily 33.0 1 year or longer 28.1 At least weekly 34.0 At least monthly 26.4 Bathroom 8.6 Time of use Patio 3.9 Morning (06:00 am–11:59 am) 6.6 Basement 2.7 Afternoon (12:00 pm–04:59 pm) Others 1.2 Evening (05:00 pm–07:59 pm) 26.6 Night (08:00 pm–05:59 am) 26.9 Device location* 37.1 Bedroom Living room 9.4 Kitchen Study room 60.1 Dining room 55.1 29.3 16.8 10.5 Note. Values are expressed in percentages. * indicates that respondents chose multiple locations. 5

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 think I could have a good time with the voice AI assistant” (Lee et al., 2006), and measured on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree; α = 0.89). Item scores were averaged to produce an index score. 3.3.6. Perceived roles of voice AI assistants To explore perceptions of the roles of their voice AI assistants, respondents were presented with three options—friend, assistant, and technology—and asked to evaluate how well each of the three terms described their voice AI assistants during users’ interactions with them. The response scales were anchored by 1 = least appropriate descriptor and 7 = most appropriate descriptor, and each of the three terms was used as a variable. The specific items composing the measures are in Appendix A. 4. Results 4.1. Sample characteristics Most respondents (80.1%) owned one voice AI assistant, and nearly 70% had owned the device for more than 3 months. Re­ spondents reported using their voice AI assistants frequently; almost 70% used them daily. Roughly 65% used their voice AI assistants in the afternoon or evening. The majority of respondents used their voice AI assistants most in their bedrooms (60.1%) and living rooms (55.1%). Almost 85% of respondents lived in single-person households. Detailed sample characteristics are listed in Table 1. 4.2. Motivations for using voice AI assistants RQ1 focused on identifying the primary reasons why consumers use voice-controlled AI devices. To examine this research question, principal component analysis (PCA) with varimax rotation was performed on respondents’ answers to the 42-item scale to identify the underlying structure of motivations to use a voice AI assistant. The PCA results were assessed based on eigenvalues (greater than 1.0), the amount of variance explained by each component, the loading score for each factor (≥|0.60|), and the relevance and meaning­ fulness of each dimension. After eliminating 11 items that had either loading scores of less than |0.60|or high loadings on more than one component, the PCA was rerun. Additionally, a parallel analysis was conducted to determine the number of components to be extracted in the analysis. The first five components from the PCA results were retained because their eigenvalues were greater than those derived from the random data (nDatasets = 1000, % = 95). These analyses resulted in a significant and interpretable five-factor Table 2 1 2 3 4 5 Motivations for using voice AI assistants. 0.88 0.15 0.05 0.07 0.19 Social Interaction (α = 0.90) 0.85 0.17 0.12 0.00 0.07 Because it is always there for me when I want to talk 0.83 0.21 0.07 0.02 0.16 Because I can have a conversation when I feel down 0.77 0.25 0.08 0.03 0.06 Because I can talk without having to consider other people 0.62 0.36 0.34 − 0.08 0.01 Because I can have a casual conversation 0.61 0.31 0.20 0.09 0.09 Because having one helps me fit into my social groups Because having one helps me maintain friendships 0.13 0.77 0.30 − 0.03 − 0.02 0.26 0.75 0.21 0.13 0.24 Personal Identity (α = 0.89) 0.33 0.74 0.26 0.03 0.00 Because I follow trends 0.35 0.72 0.14 0.07 0.16 Because using it seems sophisticated 0.41 0.71 0.19 0.15 Because using it makes me look cool − 0.04 Because I can express myself as an early adopter 0.13 0.28 0.81 0.05 Because it enhances my self-image 0.12 0.18 0.79 0.01 0.19 0.22 0.15 0.77 0.09 0.18 Conformity (α = 0.87) 0.08 0.29 0.76 0.10 0.09 Because my friends recommended it to me 0.15 Because most people around me get excited about using it − 0.03 0.01 0.09 0.09 Because other people told me it is useful 0.01 0.01 0.11 0.87 0.21 Because my family/close friends use it 0.05 0.07 0.03 0.87 0.23 0.08 0.00 0.07 0.79 0.27 Life Efficiency (α = 0.84) 0.63 Because it saves time and my effort − 0.07 0.12 0.02 0.77 Because it makes my life easier 0.25 0.09 0.08 0.28 0.75 Because it improves quality of my life 0.12 0.14 0.21 0.20 0.75 Because it is useful for multitasking 0.24 0.01 0.25 0.27 0.70 4.25 3.38 3.03 0.16 2.65 Information (α = 0.82) 14.71 13.18 2.84 11.50 Because I can stay informed (e.g., updates on weather, traffic, news, etc.) 18.48 33.19 46.37 12.33 70.20 Because I can find new information 18.48 58.70 Because I can get accurate information Because I can get information on things that interest me (e.g., sports scores, brands, sales, etc.) Eigen value % of Variance Cumulative % Note. Loadings that were 0.60 or larger are set in bold. 6

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 principal component solution; these five components explained 70.20% of the total variance. As shown in Table 2, the first component, “social interaction,” consisted of six items and accounted for 18.48% of the variance (α = 0.90). The second component, “personal identity,” included five items and explained 14.71% of the variance (α = 0.89). The third component, “conformity,” contained four items and accounted for 13.18% of the variance (α = 0.87). The fourth component, “life efficiency,” encompassed four items and explained 12.33% of the variance (α = 0.84). Lastly, the fifth component, “information,” contained four items accounting for 11.50% of the variance (α = 0.82). Given acceptable reliability, five motivational indices were created by averaging the corresponding items to be used as independent variables in subsequent analysis. Table 2 presents the full factor loadings, variance explained by each component, and items constituting each component. 4.3. Relationships between motivations and post-consumption evaluations For RQ2, the relationships between the five primary motivations for using voice-powered AI devices and individuals’ post- consumption evaluations were investigated. Partial correlations among the variables of interest were computed to control for the effects of gender, usage period, and usage frequency. As displayed in Table 3, all five identified motivations were positively correlated with users’ attitudes (ps < 0.01), satisfaction (ps < 0.001), and intentions to continue using their device (ps < 0.05). Next, hierarchical multiple regression analyses were conducted to investigate the relative effects of motivational factors on post- consumption response variables. In particular, three dependent variables were respectively regressed onto the average ratings of independent variables. Multicollinearity was diagnosed by calculating the variance inflation factor (VIF) in each regression model. The VIF values (VIF < 5), ranging from 1.06 to 2.12, demonstrated that no significant multicollinearity problem appeared in any of the regression models (Hair et al., 1998). As listed in Table 4, two sets of independent variables were entered into each of two blocks. In Block 1, the effects of gender, usage period, and usage frequency were controlled (Model 1), and the five identified motivations were added in Block 2 (Model 2). Overall, compared with Model 1, the variance in outcome variables was better explicated by Model 2, which contained the five motivations for using a voice AI assistant. The life efficiency (β = 0.48, p < .001) and information motivations (β = 0.17, p < .01) were significant predictors of individuals’ attitudes toward using a voice AI assistant. Social interaction (β = 0.18, p < .01), conformity (β = 0.14, p < .05), life efficiency (β = 0.42, p < .001), and information motivations (β = 0.24, p < .001) were determinants of user satisfaction. Finally, conformity (β = 0.15, p < .05), life efficiency (β = 0.41, p < .001), and information motivations (β = 0.12, p < .05) significantly predicted one’s likelihood of continuing to use a voice AI assistant in the future. The personal identity motivation was a nonsignificant predictor of all three post-consumption response variables. 4.4. Relationships between motivations and perceptions To explore RQ3 and RQ4, associations among the five motivations and users’ perceptions of voice AI assistants were tested. Table 5 indicates that the social attraction of a voice AI assistant was positively correlated with all five identified motivations (ps < 0.05). Perceived-as-friend was correlated with the social interaction motive (r = 0.40, p < .001) and the personal identity motive (r = 0.22, p < .01). Perceived-as-assistant was correlated with the motivations of conformity (r = 0.17, p < .01), life efficiency (r = 0.79, p < .001), and information (r = 0.37, p < .001). Perceived-as-technology was negatively correlated with the social interaction motive (r = − 0.33, p < .001) and personal identity motive (r = − 0.15, p < .05). Another set of hierarchical multiple regression analyses were carried out to investigate the relative effects of motivational factors on perceptual outcomes, including the social attraction and perceived roles of a voice AI assistant. First, the social attraction of a voice AI assistant and the three perceived roles of a device were respectively regressed onto average ratings of the five motivations while controlling for individuals’ gender, usage period, and usage frequency. Multicollinearity was tested, and the VIF values (VIF < 5; range: 1.06–2.12) revealed no multicollinearity concerns in any of the regression models (Hair et al., 1998). Table 6 shows that, overall, the variance in dependent variables was better explained by Model 2. Interestingly, the social inter­ action motive was the only significant predictor of the social attraction of a voice AI assistant (β = 0.67, p < .001) and of users’ perceptions of their voice AI assistant as a friend (β = 0.44, p < .001). The more individuals were motivated by social interaction to use their voice AI assistant, the more likely they were to think of the device as socially attractive and as a friend. Further, the life efficiency Table 3 Partial correlations and descriptive statistics. Variables 12 3 4 5 6 7 8 M SD 1. Social interaction – 0.62*** 0.39*** 0.09 0.34*** 0.17** 0.30*** 0.16* 3.39 1.53 0.56*** 0.33*** 0.19** 0.24*** 0.13* 3.46 1.45 2. Personal identity – 0.11 0.27*** 4.33 1.42 0.48*** 5.39 1.08 3. Conformity – 0.23*** 0.39*** 0.29*** 0.36*** 0.38*** 5.24 1.13 0.71*** 5.47 1.10 4. Life efficiency – 0.48*** 0.58*** 0.57*** 0.71*** 5.38 0.98 – 5.51 1.08 5. Information – 0.45*** 0.53*** 6. Attitude – 0.77*** 7. Satisfaction – 8. Intention Notes. Control variables: gender, usage period, usage frequency. M = mean; SD = standard deviation. *p < .05, **p < .01, ***p < .001. 7

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 Table 4 Hierarchical multiple regression analyses. Attitude Satisfaction Intention Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Control 0.01 − 0.01 0.04 0.03 0.04 0.01 Gender − 0.01 − 0.00 − 0.04 − 0.01 − 0.03 − 0.02 Usage period − 0.21** − 0.01 − 0.28*** − 0.06 − 0.30*** − 0.12* Usage frequency 0.04 0.03 0.07 0.18** 0.09 0.08 Motivation 0.03 0.01 0.06 − 0.07 0.08 − 0.08 Social interaction 3.84* 0.10 6.72*** 7.92*** Personal identity 0.48*** 0.14* 0.15* Conformity 0.17** 0.42*** 0.41*** Life efficiency 0.41 0.24*** 0.12* Information 0.39 0.49 0.35 21.32*** 0.47 0.33 R2 30.45*** 29.65*** 16.88*** Adjusted R2 0.37 40.26*** 20.42*** F 0.42 0.27 ΔF ΔR2 Notes. All values indicate standardized β value. *p < .05, **p < .01, ***p < .001. Table 5 Partial correlations and descriptive statistics. Variables 12 3 4 5 6 7 8 9 M SD 1. Social interaction - 0.62*** 0.39*** 0.09 0.34*** 0.71*** 0.40*** 0.07 − 0.33*** 3.39 1.53 2. Personal identity - 0.56*** 0.11 0.33*** 0.47*** 0.22** 0.10 − 0.15* 3.46 1.45 3. Conformity - 0.23*** 0.39*** 0.26*** 0.07 0.17** − 0.05 4.33 1.42 4. Life efficiency - 0.48*** 0.16* − 0.07 0.79*** 0.10 5.39 1.08 5. Information - 0.34*** 0.09 0.37*** 0.06 5.24 1.13 6. Social attraction - 0.47*** 0.14* − 0.33*** 3.58 1.60 7. Perceived-as-Friend - − 0.08 − 0.61*** 3.29 1.88 8. Perceived-as-Assistant - 0.11 5.23 1.46 9. Perceived-as-Technology - 5.38 1.63 Notes. Control variables: gender, usage period, usage frequency. M = mean; SD = standard deviation. *p < .05, **p < .01, ***p < .001. Table 6 Hierarchical multiple regression analyses. Social attraction Perceived-as-friend Perceived-as-assistant Perceived-as-technology Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Control − 0.07 − 0.03 0.08 0.11 0.08 0.06 0.02 0.00 Gender − 0.18* − 0.03 − 0.18* − 0.09 − 0.01 − 0.03 0.16* 0.07 Usage period − 0.08 0.06 − 0.11* − 0.09 − 0.34*** − 0.08* 0.02 0.01 Usage frequency 0.04 0.67*** 0.05 0.44*** 0.13 − 0.00 0.02 − 0.43*** Motivations 0.03 0.08 0.03 − 0.00 0.12 0.03 0.01 0.04 Social interaction 3.48* − 0.09 3.93* − 0.09 10.38** 2.20 0.02 Personal identity 0.08 − 0.10 − 0.02 0.06 Conformity 0.08 0.79*** 0.16* Life efficiency 0.53 0.03 0.16 Information 0.54 0.21 − 0.01 0.14 36.56*** 0.19 0.67 6.07*** R2 54.20*** 8.23*** 0.66 8.20*** Adjusted R2 0.50 10.37*** 0.14 F 0.17 59.53*** ΔF 79.35*** ΔR2 0.55 Notes. All values indicate standardized β value. *p < .05, **p < .01, ***p < .001. motivation was the only determinant of users’ perceptions of a voice AI assistant as an assistant (β = 0.79, p < .001). Finally, users motivated by social interaction were less likely to perceive a voice AI assistant as technology (β = − 0.43, p < .001), whereas those motivated by information purposes were more likely to think of the device as technology (β = 0.16, p < .05). See Table 7 for a summary of the findings. 8

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 5. Discussion and implications This research makes theoretical and empirical contributions by responding to two criticisms of previous uses and gratifications research on new technologies and digital media: (1) initial constructs and measures designed for traditional media were applied to new media and (2) limited attention had been given to the affordances created by technological innovation, which often alter uses and gratifications (Sundar and Limperos, 2013). The authors of the present research conducted extensive research to identify and measure constructs uniquely related to the uses and gratifications provided by voice AI assistants. Ultimately, five primary motivations for using voice AI assistants were identified: social interaction, personal identity, conformity, life efficiency, and information. As detailed below, these findings provide a more nuanced understanding of individuals’ motivations for using voice AI assistants. This study is the first to investigate the relationships between users’ motivations and their post-consumption attitudes, satisfaction, and intentions to continue using their voice AI assistants. Notably, the life efficiency and information motivations—both utilitarian- related motivations—were the only significant antecedents of all post-consumption response outcomes and behavioral intentions. These results highlight the primacy of utilitarian motivations in influencing post-consumption outcomes. However, it is important to note that satisfaction with a voice AI assistant was significantly predicted by all primary motivations except the personal identity motivation; this finding implies that users were generally satisfied with their voice AI assistants whether their motivations were social or utilitarian. The conformity, life efficiency, and information motivations significantly predicted users’ intentions to continue using their voice AI assistant. The fact that the conformity motivation, which captures social influence factors, significantly affected in­ dividuals’ intentions to continue using their voice AI assistant aligns well with several theories predicting people’s technology acceptance behavior (e.g., the technology acceptance model). Such theories assert that social groups’ opinions are critical in explaining users’ post-consumption evaluations and intentions (Rauschnabel and Ro, 2016). Another notable finding was that individuals’ motivations for using voice AI assistants played key roles in shaping users’ per­ ceptions of their voice AI assistant’s persona, character, or roles. In brief, only people who were motivated by social interaction were likely to perceive their voice AI assistants as a friend and socially attractive, whereas only those who were motivated by life efficiency were likely to deem their voice AI assistants an assistant. Only individuals who were motivated by information were likely to consider their voice AI assistant technology. Whether intentionally or unintentionally, users appeared to project their own meanings, which were based on their primary motivations, onto voice AI assistants. Doing so may have led users to feel that their devices added value to their lives. These findings echo those of earlier studies (e.g., Schweitzer et al., 2019; Natale, 2020; Ki et al., 2020; Zhao & Rau, 2020). This research further demonstrated the prevalence of the social interaction motivation. Specifically, users sought and valued some degree of casual social communication when interacting with their voice AI devices. Items assessing social interaction focused on users’ perceptions that voice AI assistants were always present, good listeners, and seldom enacted social judgment. The findings suggested that respondents conceived of their voice AI assistants as familiar and supportive conversation partners who were always there, could engage in conversation when users felt blue, and provided an atmosphere where users could speak freely without concern for others’ reactions or social impressions. Speaking is a more natural way of interacting with a device than typing, and voice AI assistants have a range of verbal abilities and natural language that enable them to engage with users through spoken dialogue that includes greetings and social talk (Feine et al., 2019). These features can create the impression of social presence, the sense that someone is there, and cause individuals to think of their voice AI assistants as social entities (Reeves and Nass, 1996). This explanation coincides with McLean and Osei-Frimpong’s (2019) discovery of social presence as a key variable motivating the use of in-home voice AI assistants and that social-benefit motivations were stronger among households with two or fewer people. Interestingly, most re­ spondents in the present study lived alone and placed their devices in personal locations (e.g., bedroom). These factors may have influenced respondents’ motivations to use their voice-activated AI devices as conversation partners or interlocutors. These findings also resonated with suggestions that voice AI assistants can be perceived as social agents (Feine et al., 2019; Go and Sundar, 2019; Ki et al., 2020) and fulfill users’ psychological needs for communication (Sundar et al., 2017). Table 7 Key Findings Summary of findings. • Five motivations: Social interaction, Personal identity, Research Questions Conformity, Life efficiency, and Information RQ1: What are users’ primary motivations for using a voice AI assistant? • Satisfaction determined by Social interaction, Conformity, Life RQ2: What are the associations between users’ motivations and their attitudes, efficiency, and Information motivations satisfaction, and intentions to continue using a voice AI assistant? • Intention to continue using a voice AI assistant determined by RQ3: What are the associations between users’ motivations and their perceptions of Conformity, Life efficiency, and Information motivations the social attraction of a voice AI assistant? • Social attraction of a voice AI assistant determined only by RQ4: What are the associations between users’ motivations and their perceptions of Social interaction motivation the role of a voice AI assistant as (a) a friend, (b) an assistant, and (c) technology? • Perceptions of voice AI assistant as “friend” determined only by Social interaction motivation • Perceptions of voice AI assistant as “assistant” determined only by Life efficiency motivation • Perceptions of voice AI assistant as “technology” determined only by Information motivation 9

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 The findings of this research add depth and precision to our understanding of utilitarian motivations for using voice AI assistants (e. g., McLean and Osei-Frimpong, 2019). Specifically, results unveiled the life efficiency motivation and differentiated it from another utilitarian motivation, information, which refers to the ability to access accurate and timely user-desired information. The information motivation indicated that voice AI assistants can serve as useful gadgets that provide information for everyday life (e.g., weather forecasts and traffic reports) along with information tailored to an individual’s personal interests (e.g., sports scores, entertainment options). In contrast, the life efficiency motivation captured the manner in which voice AI assistants helped users stay organized and made their lives more efficient and productive, thereby enhancing the overall quality of individuals’ lives. This motivation is directly tied to unique aspects of voice AI assistants such as their ability to make lists, set alarms and timers, make recommendations, an enable users to multitask. The hands-free activation of voice AI assistants also facilitates multitasking and likely contributes to the life effi­ ciency motivation. In addition, this study enriches our understanding of symbolic expression (e.g., Rauschnabel, 2018). Two motivations related to symbolic expression were identified: the personal identity motivation and the conformity motivation. The personal identity motivation focused on voice AI assistants as a means of expressing and enhancing a user’s self-image. For example, because voice AI assistants are used mostly at home, they can help some users convey their self-images as the devices are part of home furnishings (McLean and Osei- Frimpong, 2019). This finding is consistent with prior research demonstrating that individuals use newly developed technologies (e.g., wearable devices, smart glasses) to create personal social images (e.g., trendy, cool, tech-savvy, early adopter) and reflect their life­ styles, tastes, and personalities (Cho and Lee, 2017; Rauschnabel, 2018; Rauschnabel and Ro, 2016; Rettberg, 2014). The conformity motivation indicated that people use voice AI assistants to fit in with others. This outcome converged with other research indicating that social influence variables are pivotal in explaining technology adoption and use behavior (Rauschnabel and Ro, 2016). The recent proliferation of voice AI assistants among young adults suggests that using a voice AI assistant may serve as an especially effective means of conforming to one’s social group and sharing something in common. Our findings offer valuable managerial implications for the designers and marketers of voice AI assistants. This study’s results showed that individuals seek utilitarian and social functions when using voice AI assistants. Designers of these voice AI assistants should therefore develop more elaborate functions to fulfill individuals’ utilitarian, psychological, and social needs; doing so will cause users exhibiting both types of motivations to be more likely to continue using voice AI assistants. For instance, designers could add an option for users to select a role (e.g., utilitarian vs. social), and the voice AI assistant’s expressions and tone could vary accordingly to enhance the user experience (Brüggemeier et al., 2020; Rhee and Choi, 2020). In addition, this research provides useful implications for marketers and brand communication managers interested in tapping into this new marketing domain and employing voice AI assistants as part of their strategies. As voice AI assistants have become one of the main shopping channels, users are likely to be major contributors to the rising trend of voice-controlled shopping (Kinsella & Mutchler, 2019). Brands and companies could therefore leverage AI assistants as a venue for providing information about promotions, products and product categories, and other shopping opportunities. For example, the laundry detergent brand Tide created an Alexa skill called “Stain Remover” that informs users about how to remove more than 200 types of stains with Tide products. Nonprofit brands can also employ voice AI assistants: Boston Children’s Hospital launched an Amazon skill called “KidsMD” that enables parents to ask questions about their children’s symptoms and receive answers from experts at the hospital, which are more credible than what parents would otherwise obtain from a Google search (Kemp, 2019). As such, understanding why people use voice AI assistants can enable practitioners in diverse industries to enhance the effectiveness of branding and marketing activities by offering what consumers need and want through voice AI assistants. Device designers may also be able to enhance these devices’ persuasiveness by providing messages that sound as though they are being conveyed by a voice AI assistant in the social role that the consumer prefers (e.g., friend vs. assistant). 5.1. Limitations and future research This research makes noteworthy contributions that pave the way for a timely research agenda on voice AI assistants, but it is not without limitations. First, the study sample was limited to respondents between 17 and 37 years old. Given the potential for moti­ vational differences across age groups (Biele et al., 2019), future studies should employ samples that include young children and older adults to explore whether the findings of this research hold for different age brackets. The study methodology is also worthy of expansion. This research relied on data from self-report measures. Subsequent studies could integrate other methods to develop a broader understanding of user behavior. For example, scholars could collect qualitative data via a diary method or in-depth interviews to explore consumer behavior and more thoroughly assess activities, experiences, emotions, attitudes, motivations, and other contextual information that influence technology use in everyday life (Lopatovska et al., 2018). Further, future studies should include ethnographic approaches to capture consumers’ behavior with an eye toward how their motivations inform their communication choices. Areas for exploration include users’ frequency of device usage, usage length, and style of communication with voice AI assistants. AI is being used in a variety of professional and industrial settings, and voice AI assistants also have many potential applications in these contexts, which warrant investigation. For example, AI is being integrated into smart cities to make them run more effectively and efficiently (Yigitcanlar et al., 2020), and AI can enhance the lives of people with disabilities (Branham and Roy, 2019). AI clearly has utilitarian uses and gratifications in professional and industrial settings, but do AI in general and voice AI assistants in particular also give rise to social uses and gratifications in these settings? More needs to be known about both utilitarian and social uses, gratifications, and motivations in professional and industrial contexts. This research made certain assumptions that could be examined in future research. For example, it investigated the impacts of users’ motivations on their perceptions of voice AI assistants’ roles; however, one could argue that perceptions of the roles of voice AI 10

T.R. Choi and M.E. Drumwright Telematics and Informatics 62 (2021) 101628 assistants determine motivations. As such, research that considers the manner in which role perceptions influence motivations could provide additional insight. The use of voice AI assistants is experiencing exponential growth in the global voice AI market for consumers (Tankovska, 2020), and a promising avenue for future work involves assuming a cross-cultural perspective on users’ adoption of voice AI assistants. For example, this study used a sample of current voice AI assistant users in the United States, which is a prime example of an individualistic culture, and cross-cultural psychology has revealed that members of individualistic cultures tend to prioritize personal benefits, success, and self-enhancement and emphasize personal privacy (Hofstede, 1984). As such, respondents in this study may have exhibited positive impressions of voice AI assistants and been inspired to use them because these devices directly benefit various aspects of their daily lives. In contrast, members of collectivistic cultures value social relationships, in-group harmony, and an agreeable interpersonal atmosphere (Hofstede, 1984); therefore, group-oriented motivations may come into play. For example, a conformity motivation might lead to better post-consumption outcomes among individuals from collectivistic cultures. 6. Conclusion This study expanded the uses and gratifications framework by developing measures to explore multidimensional motivations for using voice AI assistants. The findings enhanced our understanding of individuals’ primary motivations for using voice AI assistants and contributed to a richer understanding of users’ post-consumption perceptions and behavioral intentions. The results shed addi­ tional light on human − voice AI interaction by demonstrating that individuals perceived a voice AI assistant’s social attraction and roles in a manner that corresponded with their motivations. As such, this research extended the body of empirical work on what drives users’ motivations and perceptions about technology. Furthermore, findings provided additional empirical support for the CASA research paradigm by confirming that users perceive voice AI assistants’ social attractiveness and roles differently, converging with the burgeoning literature on the various roles of AI (e.g., Ki et al., 2020; Zhao & Rau, 2020). Results regarding the social interaction motivation especially underscored the potential social value of voice AI assistants and implied that this dimension deserves closer attention. This study also pointed to pertinent managerial implications for voice AI designers and brands. However, much remains to be learned about the uses and gratifications of voice AI assistants, which have the potential to enhance users’ lives in compelling ways. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Appendix A Measurement items Attitude toward voice AI assistants (α ¼ 0.89) I like using a voice AI assistant I feel good about using a voice AI assistant Overall, my attitude toward using a voice AI assistant is favorable Satisfaction with voice AI assistants (α ¼ 0.82) Overall, I am satisfied with my voice AI assistant Overall, interacting with a voice AI assistant is emotionally satisfying and pleasant Overall, the functions provided by a voice AI assistant meet my needs I am satisfied with my decision to purchase a voice AI assistant device Intention to continue using voice AI assistants (α ¼ 0.82) I plan to keep using a voice AI assistant in the future I want to continue using a voice AI assistant I intend to recommend a voice AI assistant to my friends Social attraction of voice AI assistants (α ¼ 0.89) I think I could have a good time with my voice AI assistant I think my voice AI assistant could be a friend of mine I would enjoy a casual conversation with my voice AI assistant I would like to spend more time with my voice AI assistant References Acemoglu, D., Restrepo, P., 2018. Artificial intelligence, automation and work. Natl. Bur. Econ. Res, No, p. w24196. Biele, C., Jaskulska, A., Kopec, W., Kowalski, J., Skorupska, K., Zdrodowska, A., 2019. How might voice assistants raise our children?. In: International Conference on Intelligent Human Systems Integration. Springer, Cham, pp. 162–167. Brandtzaeg, P.B., Følstad, A., 2017. Why people use chatbots. In: Proceedings of the 4th International Conference on Internet Science. Springer, pp. 377–392. Branham, S.M., Roy, A.R.M., 2019. Reading between the guidelines: How commercial voice assistant guidelines hinder accessibility for blind users. 21st Int. ACM SIGACCESS Conf. Comput. Access. 446–458. 11