the Adoption of Virtual Voice Assistants as
the New Interface for Communication
between Organisations and Consumers
Linda Zieverink 10213376 Master’s Thesis
Graduate School of Communication
Master’s programme: Corporate Communication dr. G.L.A. (Toni) van der Meer
Abstract
A virtual voice assistant (VVA) is a type of virtual service agent that can execute actions in real-time by means of voice commands. The VVA is increasingly seen as a business opportunity for communicating with customers and strengthening the relationship. Using VVAs from a commercial perspective is also called conversational commerce. When the interaction between organizations and consumers takes place via a VVA, the structure of communicating changes structurally. As the screen interface has now been replaced by speech commands, commerce is possibly also changing and will respond to this evolution.
The technology behind the VVA is still in its infancy, but it is likely that it will soon become part of the consumer's expectations and omni-channel experience. As the VVA will increasingly facilitate the communication between companies and customers, it is important to investigate how consumers view these developments. However, little research has been done into the acceptance of VVAs and the developments surrounding conversational commerce. That is why this research investigated how open consumers really stand for the use of virtual voice assistants and conversational commerce.
The research model consists of a combination of the Technology Acceptance Model, supplemented with compatibility from the Innovation Diffusion Theory, a privacy risk factor and attitude towards advertising. Since the developments take place at a global level and differences between countries can be identified, it has been examined whether the acceptance of VVAs differs per country. This was done by sending out questionnaire to nine countries (N = 306). The results showed that only perceived usefulness and attitude towards mobile advertising were significant predictors on consumers attitude and behaviour in this context. In addition, the relationship between the two dependent variables was not moderated by country of residence. In view of that the latter may well be the case in a different research structure, these differences must be further investigated in the future.
1. Introduction
The year 2016 can be described as the year of the chatbots (Dale, 2016). A chatbot is a type of virtual service agent that can enhance customer relationships and provide for expectations by automating real-time human-computer interaction (Chung, Ko, Joung & Kim, 2018; Lowry, Romano, Jenkins & Guthrie, 2009). Critics defined chatbot-technology as having the potential to disrupt and eventually eliminate the need for websites and applications (Marchick, 2017). Chatbots already arose some time ago, but have only recently shifted to the mobile messenger domain. As a result, trends in the field of corporate communication and marketing are starting to focus more and more on the use of these conversational systems (Tuzovic & Paluch, 2018). The popularity of chatbots among consumers keeps on growing, and in the meantime, the technology continues to evolve. The chatbot from 2016 has now grown into a voice-enabled platform, or virtual voice assistant (VVA), where interaction takes place via voice commands instead of a screen interface. In addition, virtual voice assistants are becoming a business opportunity for a range of purposes, such as interacting with people or businesses for inquiries, purchases, leveraging customer service, generating sales or extending their brand experience (Shebat, 2016; Tell, 2017; Koksal, 2018). These commercial developments related to voice technology are also called conversational commerce (Messina, 2016). The goal of deploying virtual voice assistants for commercial purposes is to create a service that: a) clearly identifies needs and b) is successful in rectifying those needs. Large organizations respond to this evolution by developing voice assistants on a large scale, such as Apple's Siri, Google's Now, and Microsoft's Cortana (Messina, 2016).
However, with this new voice interface, it now seems that this will structurally change the way in which organisations communicate with consumers. Spoken conversations are hard to automate and differ structurally from screen interaction. As companies are expected to fully control the omni-channel experience to meet the needs of the consumer, it is likely that voice technology will soon fall within the expectation pattern of this consumer experience (Schreuder, Schreuder & Van Wijk, 2017). Nevertheless, conversations with a VVA occur in an open domain and can go into any direction.
Consumers already have high expectations of VVA that cannot be fulfilled yet (Heo & Lee, 2018). In addition, in general little is known about the consumer acceptance of this innovation and additional developments. How open are consumers really for the use of virtual voice assistants? And would the openness differ per country as adoption and sales figures are relatively higher in some countries (Exalto, De Jong, De Koning and Ravensteijn, 2018). Through a combination of the Technology Acceptance Model (Davis et al., 1989) and the
Innovation Diffusion Theory (Rogers, 1983), this research tries to gain insight into the current attitude of consumers towards virtual voice assistants and thus the possibilities for conversational commerce. In doing this, this research hopes to contribute to the understanding of this new communication technology and the structural changes in the way consumers are now starting to communicate with organisations.
To find an answer to this main question, the study is divided into four sections. First, the theoretical framework delves deeper into what virtual voice assistants are, the developments around them and their possible potential and barriers. Subsequently, hypotheses were developed on the basis of the TAM and IDT in order to find an answer to the main question: to what extent are consumers open to the use of virtual voice assistants and the associated growing degree of conversational commerce? Consumer data is collected in nine countries and is then analysed to gain more insight into consumer acceptance, its predictability power, and the disparities found between countries. The research concludes with a discussion and conclusion that discusses further implications and future perspectives.
2. Theoretical Background 2.1.Virtual Voice Assistants
The developments surrounding messenger chatbots are far from fully developed, but in the meantime, we are already heading towards the next phase (Reddy & Mahender, 2013). Nowadays, technological developments offer the opportunity to further develop chatbot’s automated way of communicating and even simulate various human-computer interaction scenarios on a large scale, such as answering questions, negotiating, tutoring, but also for e-commerce (Zang, Zhu, Wang & Liu, 2017). Hence, the conversational agent (bot) is slowly developing into a conversational voice agent, where communication is no longer occurring via a screen but via a virtual voice interface. A voice interface, enabled platform, or voice-agent, is a service that processes spoken input and provides audio output by making use of natural language processing (Danielsen, 2000). These voice-enabled platforms often take on the role of virtual personal assistant or partner (Bhattacharyya, 2016; Zang, Zhu, Wang & Liu, 2017). This specific form of a voice-enabled platform, the virtual voice assistant (VVA), will be the focus for this research from a communications perspective.
2.2. Conversational Commerce
The use of conversational robots for commercial purposes is also named “conversational commerce”. The term conversational commerce is mainly about “delivering convenience,
personalization, and decision support while people are on the go” (Messina, 2016). In addition, the commercial promise is to be “a service with which you can talk to clarify your needs or questions, with a result that is completed” (Van Manen, 2016). The term also refers to the crossover between messaging apps and online shopping which has resulted into a new trend of business (Piyush, Choudhury & Kumar, 2016). The inventor of the term envisioned an impactful future for this concept and VVA’s. Messina (2016) predicted that conversational agents would significantly change the way we communicate and interact with each other digitally. The ultimate goal of conversational agents is “to replace the most common interfaces we use on computers and in connected devices” (Newman, 2016). The adoption of this new phase is expected to go faster than with desktop apps (Heo & Lee, 2017). Organisations can utilize this development in order to strengthen the relationship with their consumers as virtual voice assistants aim to achieve the desired goal as quickly and efficiently as possible.
The reason that the developments are currently gaining momentum is probably because the current level of knowledge in the field of technology and artificial intelligence now allows and supports it (Dale, 2016). In addition, according to Dale (2016), people are now “entirely comfortable communicating via brief, textual interactions, and quite unfazed by carrying on several asynchronous conversations at the same time” (p. 815).
2.3. Corporate Communication via Voice
The current technological developments have already affected the way in which organizations engage with their customers, as companies have become more approachable (Schreuder, Schreuder & Van Wijk, 2017). This has ensured that the power balance has shifted from an organizational-centric focus to a customer-led focus. The modern customer has become accustomed to automatic messages and ever-faster answers. This has changed the way organisations are communicating with their customers and consequently, the demand for automated reply to customers is exerting pressure on human resources (Schreuder et al., 2017). The customer experience and relationship are becoming more holistic and complex by nature, hence organizations are now expected to properly align all products, channels, customer segments and moments of contact (Verhoef, Lemon, Parasuraman, Roggeveen, Tsiros & Schlesinger, 2009). Consequently, managing the omni-channel experience of customers is becoming increasingly complex (Schreuder, Schreuder & Van Wijk, 2017). These developments offer new abilities and scale to organizations, given that customer experience now includes channels and interfaces that previously could not be considered as
options to interact with customers. Especially the voice interface will ensure for structurally different communication with customers. In not too long, it is likely that people will expect answers to their questions and problems via a virtual voice assistant. This will require companies to invest in voice technology for both service and commercial purposes. Voice will become part of the omni-channel experience, which creates opportunities for reach, sales and demand via voice.
2.4. The Paradox
At the same time, a paradox arises. Many organizations are convinced that they can rely on technology for creating the ultimate customer experience (Lee & Choi, 2017), while also talking about how the hearts and minds of the entire organization should ensure a warm human response (Lee, Oh & Choi, 2017). Previously, the gap between cold programmatic and warm empathic customer management was very large. However, with current developments in the field of artificial intelligence (AI) and natural language programming (NLP), it is becoming increasingly possible to simulate this warm empathic feeling through sentiment analysis and deep neural network learning (Schreuder, Schreuder & Van Wijk, 2017).
Nevertheless, the concept of conversational commerce is still in its infancy as it often involves simple human-computer interactions (Exalto et al., 2018). It is still, if ever, impossible to simulate the human aspect of communication via these devices. Modelling interaction between computers and humans via natural language is very complex. Long conversations are difficult to automate, especially in an open domain where the conversation can go in any direction (Britz, 2016). This interaction has therefore been an issue that has been studied extensively for years and remains a challenge (Zadrozny, Budzikowska, Chai, Kambhatla, Levesque and Nicolov, 2000; Hill, Randolph Ford, & Farreras, 2015). Currently, voice-agents can only process simple tasks and commands and they are more likely to take the form of a tool than that they are artificially intelligent (Hollister, 2016).
2.5. Consumer Adoption
Furthermore, what also plays a significant role in the successful implementation of innovations such as voice agents, is consumer adoption. Research indicates that consumer resistance is a strong predictor of the market failure of innovations (Ram & Sheth, 1989). Given virtual voice assistants currently are an inevitable and growing development (Tuzovic & Paluch, 2018), it is necessary to explore consumer attitude, resistance and possible reasons for adopting virtual voice assistants as the next interface for communication between
consumers and organisations. Hence, in order to gain more insight into consumer acceptance, and to reduce the gap between research on voice and voice in practice, this research will focus on the consumer acceptance of voice-enabled platforms. With this, this research hopes to contribute to getting somewhat closer to solving the paradox where there is a need for warm human contact with an organisation as we become more dependent on technology to facilitate in this need.
In their research, Davis, Bagozzi and Warshaw (1989) stress that technology cannot improve organizational performance if the technology is not used. For this reason, it is important to understand better why people would accept or reject a technology in order to better predict, explain and increase user acceptance. For instance, both consumers and organisations currently have high expectations of virtual voice assistants, Heo and Lee (2018) state: “Currently, there is a mix of an excessive expectation of the public and companies who do not understand the technology status of chatbots as well as the hyped marketing of chatbot-related companies” (p. 41). These high expectations are accompanied by a certain attitude towards virtual voice assistants that will be further investigated in this study. Still, little research has been done on the current acceptance of the phenomenon and it remains unclear what impact conversational commerce and virtual voice assistants will have on the current form of communication between companies and their consumers (Mool, 2017).
For this reason, the aim of this study will be to investigate how open consumers really are for the use of virtual voice assistants. What attitude do customers have towards virtual voice assistants and what is their possible behavioural intention? The results will be used to discover and possibly predict the intention of using virtual voice assistants as the new interface of (conversational) commerce. The main research question of this research will be as follows: to what extent are consumers open to the use of virtual voice assistants and the associated growing degree of conversational commerce?
2.6. Technology Acceptance Model
To find an answer to the main research question, this study will make use of the Technology Acceptance Model (TAM), originally created by Davis et al. (1989). The TAM can be used to model user acceptance of information systems. It provides insight into “the impact of external variables on internal beliefs, attitudes, and intentions” (Davis et al., 1989, p.985) on the (potential) use of an information system, such as a virtual voice assistant. In addition, the model offers the possibility to measure the impact of the information system as means of communication that can be deployed for and by companies.
The original TAM, as shown in Figure 1, is based on the finding that specific behaviour is determined by behavioural intent (BI). This behavioural intent is determined by attitude (A) towards the specific concept. The TAM is used as an indication for determining whether there is a causal relationship between perceived usefulness (PU), perceived ease of use (PEOU), a user attitude (A), intentions (BI) and the actual adoption of the information system (Davis et al., 1989).
Figure 1. The Original Technology Acceptance Model by Davis et al. (1989)
Davis et al. (1989) defined the experience of perceived usefulness as “the prospective user's subjective probability that using a specific application system will increase his or her job performance within an organizational context” (p.985). The variable perceived ease of use is defined in the model as “the degree to which the prospective user expects the target system to be free of effort” (p.985). The model indicates that PU and PEOU are influenced by external variables. The PU and PEOU are then decisive for the acceptable attitude towards the information system (A). The actual use of the system depends on specific behaviour that is triggered by BI. In addition, BI of the user depends on PU and A. There is also a direct effect of PEOU on PU.
Instead of looking at user satisfaction, studies show that the TAM is suitable for predicting actual use by users (Davis et al., 1989). Since it is fairly recent that virtual voice assistants are used on a large scale, and that the general attitude and acceptance of the technology has not been studied much, the TAM is a suitable model for analysing how open consumers at this stage are for using virtual voice assistants. By examining virtual voice assistants through the TAM, it is possible to investigate how consumers are currently viewing the technology and it will possible clarify which elements (variables) are most important for determining their attitude towards the virtual voice assistant.
The study of Davis et al. (1989) shows that mainly PU and PEOU are good predictors for the use of information systems. These variables will therefore also possibly be of influence on the acceptance of virtual voice assistants. Hence, the first two hypotheses evolved around those two predictors:
H1. There is a positive relation between perceived usefulness (PU) and the consumers’
attitude (A) towards virtual voice assistants.
H2. There is a positive relation between perceived ease of use (PEOU) and the
consumers’ attitude (A) towards virtual voice assistants.
The TAM has been used, extended and criticized frequently in literature for quite some time. Studies indicate that the model can be extended with social influence and cognitive instrumental processes (Venkatesh, 2000), trust (Ha and Stoel, 2009), innovation (Zarmpou, Saprikis, Markos, & Vlachopoulou, 2012), or variables from the Innovation Diffusion Theory (IDT) (Rogers, 1983; Chen, Gillenson & Sherrell, 2002).
The TAM and the IDT can be considered as extremely similar in some constructs and complement each other (Wu & Wang, 2005). The first looks at the acceptance and the second at the adoption rate of innovations (Wu & Wang, 2005, Chen et al., 2002). The IDT is based on the variables relative advantage, compatibility, complexity, trialability, and observability that determine “the process in which an innovation is communicated through certain channels, over time, among the members of a social system.” (Rogers, 1983, p. 990). With the IDT, it is possible to study how these factors interact with each other in order to investigate user adoption and decision-making processes and to predict the implementation of innovations (Wu & Wang, 2005). A combined model with variables from the TAM and IDT is often used for research into e-shopping acceptance (Vijayasarathy, 2004; Ha and Stoel, 2008) and m-commerce service adoption (López-Nicolás, Molina-Castillo & Bouwman, 2008).
Previous research suggests that only relative advantage, compatibility, and complexity are consequent significant predictors of innovation adoption (Agarwal & Prasa, 1998). In reaction to this, Sonnenwald, Maglaughlin and Whitton (2003) stress that, in technology evaluation research, relative advantage is similar to perceived usefulness and complexity is comparable to perceived ease of use. Compatibility (C), the extent to which the technology fits the lifestyle, norms and values of the user, does not overlap with variables from the TAM. However, even though measuring construct compatibility leads to a faster adoption rate, it has not been studied often in combination with the TAM (Chen, Gillenson and Sherrell, 2002). The study of Vijayasarathy (2004) does look at C from IDT in combination with the variables PU, PEOU, BI and A. Their results show that this specific combination of the two models offers a good starting point for predicting and understanding technology acceptance.
Wu and Wang (2005) also used a combination of TAM and C for their research into user acceptance of the early phases of electronic mobile commerce. At the time, they recognized that “the rapid development of modern wireless communication technology, coupled with the increasingly high penetration rate of the Internet, is promoting mobile commerce (MC) as a significant application for both enterprises and consumers” (p. 719). To study the phenomenon, the TAM, supplemented with C, was used to understand which factors would influence users to use MC. It seems likely that with the current rise of the virtual voice assistants, conversational commerce is at the beginning of a similar development as electronic mobile commerce was at that time. Hence, compatibility was added to the research model in this study. The third hypothesis will therefore be:
H3. There is a positive relation between compatibility (C) and the consumers’ attitude
(A) towards virtual voice assistants.
Furthermore, studies that make use of the TAM for research into (mobile) commerce or mobile commercial information systems, often include a certain risk factor as external variable (Wu & Wang, 2005; Zarmpou et al., 2012; Van Eeuwen, 2017). These researchers found a significant relationship between perceived risk, or privacy concerns, and behavioural intent (BI). The privacy risk factor may also play a role in the adoption of virtual voice assistants, since a lot of valuable data is collected and processed by these virtual voice assistants. The creators and drivers behind conversational commerce have to be careful not to be perceived as too intrusive or irritating by the consumers (Sultan, Rohm, & Gao, 2009). This could ensure that the system would be seen as a threat to privacy. In this study, it was therefore expected that adding a privacy risk factor to the research model can help predict user acceptance of virtual voice assistants:
H4. There is a negative relation between internet privacy concern (ICP) and attitude
(A) towards virtual voice assistants.
In addition, users’ general attitude towards mobile advertising (ATMA) may also play a role in the acceptance of the use of conversational commerce via virtual voice assistants. Several studies show that the attitude towards mobile marketing (Karjaluoto & Alatalo, 2007; Watson et al., 2013) and advertising (Ling, Piew & Chai, 2010) is valuable to investigate in order to increase the effectiveness of a message. It can be important to take into account as a
commercial message of way of communicating can be seen as intrusive (Watson et al., 2013). With the rise of conversational commerce, as the successor of e-commerce and MC, it is important to take the attitude towards commerce into account, as this attitude will possibly extent towards conversational commerce, connected to virtual voice assistants. To investigate the influence of the general attitude towards mobile advertising on the potential attitude towards advertising via virtual voice assistants, the ATMA is included in this study:
H5. There is a positive relation between attitude towards mobile advertising (ATMA)
and attitude (A) towards virtual voice assistants.
In previous studies it became evident that attitude functions as a mediator between the external beliefs and behavioural intent. The previous hypotheses all focus on the different variables that influence the attitude towards using virtual voice assistants. As attitude (A) is expected to have a positive influence on behavioural intent (BI), the actual relationship between these two dependent variables needs to be investigated in order to properly draw conclusions on BI with regards to virtual voice assistants. Therefore, the following hypothesis was drawn up:
H6. There is a positive relation between attitude (A) towards using virtual voice
assistants and behavioural intent (BI).
Lastly, research shows that the results of the TAM differ per country or more specifically, per culture (Straub, Keil & Brenner, 1997). Given that the virtual voice assistant is currently developing at a global level, it is of value to investigate whether the TAM in this context also provides different outcomes per country. The differences are expected based on assumptions about current developments that vary per country. For example, China has been investing in voice since 2011 with WeChat, a free application that is currently used by nearly one billion people (Lien & Cao, 2014). WeChat is an integrated platform that generates both text and voice messages, where various services and products, such as taxis, food and cinema tickets can be ordered. Currently, there is no other country where a comparable app is used on this scale. Another example is a global study (“Talk to Me,” 2017), showing that 20% of all Americans who participated in the study intent to buy a virtual voice assistant in the near future. Only 6% of all Germans have the same intention. To investigate these differences further, it is expected that country of residence moderates the relationship between A and BI:
H7. The relationship between attitude (A) towards using virtual voice assistants and
behavioural intent (BI) is moderated by country of residence
In Figure 2, the proposed hypotheses are processed and brought together in a research model.
Figure 2. The Proposed Research Model
3. Methods 3.1. Research Design
To answer the main research question, a quantitative cross-sectional research approach was deployed for this study. The quantitative data was collected through an online questionnaire, designed in Upinion, written in English. Upinion is an application for real-time market research with which questions can be asked to an audience on a large scale. The advantage of this tool is that it becomes easier to reach a diverse, large sample of respondents in nine different countries. The disadvantage of using Upinion is that the respondents are used to short and easy questions and less to scientifically formulated questions. As a result, there is a risk that respondents are not motivated to start the questionnaire or, begin and then do not complete it. Because of the nature of the tool, it was decided to send out the questionnaire in three ‘batches’, spread over three days. Every batch contained one third of the questions. By doing this, the potential respondents received three notifications, i.e. triggers, to start or complete the questionnaire. To stimulate the response even more, the respondents had the possibility to win one of the three gift cards of €20.
3.2. Sample
The sample was generated by means of a non-probability purposive sampling technique. There was no other pre-selection for the respondents except for that they were able to answer the questionnaire in English. The sample consists of people from the Upinion database (N = 998). The people that are in this database have voluntarily clicked on an Upinion ad on Facebook stating that Upinion is looking for people who are interested in innovation and who are willing to, occasionally, answer questions about innovations in the market. Before the respondents answered the questionnaire, permission was asked again whether the data could be used for this academic study.
3.2.1. Participants
The total amount of respondents that filled in the complete survey was N = 306, of which 125 were female (40.8%) and 181 were male (59.2%). The respondents came from nine different countries, including Austria (N = 38), Greece (N = 43), Italy (N = 37), Mexico (N = 40), Portugal (N = 28), Singapore (N = 23), Spain (N = 26), The Netherlands (N = 46), and the United Kingdom (N = 25). The average age of the respondents was 29 years (SD = 13.06). Only 13 respondents had never heard of VVAs and 68 had never used a VVA before. Of the people who had used a virtual voice assistant before, 143 people (46.7%) had a positive first impression. To gather more information about the context of the current behaviour of consumers, they were questioned about their knowledge of VVAs, mobile phone usage, and shopping behaviour (see Appendix A1, A2 and A3).
3.3. Materials
The questionnaire consists of two parts and can be found in Appendix B. The first part concerns demographic information and provides insight into gender, the age and mobile phone usage, online shopping behaviour and the country of residence. The respondents then received a brief explanation of what virtual voice assistants are with two example images of Siri and Alexa (see Appendix B). The second and largest part of the questionnaire consists of questions about the dependent and independent variables PU, PEOU, C, IPC, ATMA, A and BI, in that order. An overview of how the variables were operationalized can be found in Table 1. Every construct was measured with three to five items that were adopted from different relevant studies using the same variables. Answering occurred based on Likert's five-point scale to determine whether the respondent strongly disagrees (-2); disagrees (-1);
neither agrees nor disagrees (neutral) (0); agrees (1); or strongly agrees (2) with the questions presented.
Table 1. Variable information as used for this research
Variable (Modified) Definition Items References
Perceived Usefulness (PU)
The degree to which a person believes that using a virtual voice assistant would enhance his or her job performance
5 Chen, et al., 2002; Pikkarainen et al., 2004; Wu & Wang, 2005; Zarmpou et al. 2012 Perceived Ease of Use (PEOU)
The degree to which a person believes that a virtual voice assistant would be free from effort
4 Chen, et al., 2002; Pikkarainen et al., 2004; Wu & Wang, 2005; Zarmpou et al. 2012 Compatibility (C) The degree to which a virtual voice assistant
is perceived as consistent with existing values, past experiences, and needs of a potential adopter
3 Dinev and Hart, 2006; Van Eeuwen, 2017
Internet Privacy Concern (IPC)
Concerns opportunistic behaviour related to the personal information submitted over virtual voice assistants by the respondent in particular 3 Chen et al. (2002) Attitude Towards Mobile Advertisement (ATMA)
A consumer’s positive or negative response towards mobile advertisement send through a virtual voice assistant
5 Ling et al., 2010; Van Eeuwen, 2017
Attitude (A) An individual’s positive or negative feelings about using a virtual voice assistant
4 Zarmpou et al., 2012; Van Eeuwen, 2017 Behavioural Intent
(BI)
A person’s subjective probability that he or she will use a virtual voice assistant for commercial purposes
3 Zarmpou et al., 2012; Van Eeuwen, 2017
3.3.1. Descriptive Statistics
Attitude is one of the two dependent variables in this study. The results show that the data is normally distributed and most respondents have a neutral attitude towards virtual voice assistants (N = 306, M = .66, SD = 3.33). Scores lower than -1.5 are considered as strongly negative, scores between -1.5 and -0.5 as negative and scores between -0.5 and 0.5 as neutral. Scores between 0.5 and 1.5 as positive and above 1.5 as extremely positive. The frequencies
are displayed in Table 2Error! Reference source not found. and show that almost half of the respondents (40.8%) have a neutral attitude towards virtual voice assistants. The majority of the remaining respondents had a positive attitude (32.3%) and a slightly smaller share had a more negative attitude (26.8%).
Table 2. Frequency table Attitude towards VVA
Frequency Percent % Cumulative %
Strongly Negative 12 3.9 3.9
Negative 70 22.9 26.8
Neutral 125 40.8 67.6
Positive 91 29.7 97.4
Strongly Positive 8 2.6 100.0
Behavioural intent is the second dependent variable in the research model of this study (N = 306, M = .73, SD = 2.69). The data is normally distributed and the results are slightly skewed to the right. For BI, the respondents answers are spread and distributed, which indicates a diversity in the respondents’ intention of using virtual voice assistants. The results indicate that attitude towards virtual voice assistants is strongly related to behavioural intent r = .71, p <.001. The means differ slightly, with BI having a slightly higher average. This indicates that the two variables almost measure the same phenomenon. An overview of descriptive statistics of both the independent and dependent variables can be found in Table 3.
Table 3. Descriptive statistics independent and dependent variables
Variable N M SD Skewness Kurtosis
PU 306 -.65 4.28 -.00 -.23 PEOU 306 2.72 2.28 -.67 1.31 C 306 -.57 2.40 -.16 -.33 IPC 306 1.50 2.92 -.42 -.63 ATMA 306 -1.21 4.29 -.17 -.73 A 306 .66 3.33 -.31 -.35 BI 306 .73 2.69 -.30 -.30 3.3.2. Covariate Testing
By means of independent T-tests and ANOVA analysis, the data was tested for a difference in A and for BI based on gender, age, shopping behaviour and VVA knowledge to see if there were any covariates needed in the analyses. At first it seemed that men have a slightly more
positive attitude (N = 181, M = .91, SD = 3.39), compared to women (N = 125, M = .29, SD = 3.24), but the results of the independent t-test were again not significant (t = 1.61, df = 304, p = .71). The same analyses that were performed for A, were performed for BI. The difference between females (N = 125, M = .52, SD = 2.65) and males (N = 181, M = .87, SD = 2.72) was initially not significantly different when it came to their behavioural intent (t = 1.11, df = 304,
p = .27).
When looked at age, divided into five categories, ‘18-28 years old’, ‘29-38’, ‘39-48’, ‘49-58’, and ‘59 and older’, the age group of ‘39-48’ has the highest mean of 1.25 (N = 28), which indicates a positive to strongly positive behavioural intent. The age group 49-58 has the lowest mean (N = 9, M = .23). The majority of respondents are 28 years or younger and score an average of M = .74 (N = 188). But, even though Levene’s test for equality of variances was found to be violated for the present analysis, F(4,301) = .69, p = .60, which indicates that the variances in the sample are not equal, the differences were not significant (F(4,301) = .57, p = .69).
Lastly, the relationship between the shopping frequency and the general attitude towards advertisements and the attitude towards virtual voice assistants were examined. This shows that the ATMA is only positive (N = 6, M = 1.67, SD = 5.47) among people who shop online daily (see Appendix A3Error! Reference source not found.). The ATMA is negative for the other respondents. The attitude towards virtual voice assistants is also highest for the daily shoppers (N = 6, M = 1.67, SD = 4.84). The attitude towards virtual voice assistants is the lowest among people who never shop online. The daily shopper also scores highest on behavioural intent to use a virtual voice assistant (N = 6, M = 2.00, SD = 1.67). Nevertheless, no reliable conclusions can be derived from the shopping behaviour as they are not significant for ATMA, A and BI (p > .05).
3.4. Procedure
The first part of the questionnaire was analysed by means of descriptive statistics in SPSS. For the second part, a confirmatory factorial analysis was performed to validate the reliability of the constructs used in the research model. Hereafter, several single regressions were performed to test the first six hypotheses and a multiple regression was executed to measure the full research model.
3.4.1. Construct Measurement
Before the hypotheses were tested, a confirmatory factor analysis was performed to validate that the constructs used for this study actually measure what was intended. For this, the different items were first checked for reliability. The model consists of two dependent variables (A and BI), five independent variables (PU, PEOU, C, ICP, ATMA) and 27 items. The total set of combined items gives a reliability of α = .83. The individual items give a Crohnbach’s alpha score not lower than α = .82, which indicates a good reliability of the dataset (see Appendix A4).
Subsequently, a confirmatory factor analysis was performed with varimax rotation. KMO and Bartlett’s Test gives a significant value of .81, df = 190, p < .001, which indicates that the sample is adequate. The results show that of the 27 items included in the factor analysis, when the items from A and BI are included, six components are extracted that account for 66.7% of the variance. When looked at the rotated component matrix, it appears that as expected, the items of A are under the same component as the items from BI. When performing the same confirmatory analysis with only the dependent variables, including 20 items, five components are extracted that account for 68.5% of the variance. The other items fit the variables as intended, except for PEOU-2. The scale reliability analysis shows that when item PEOU-2 is removed, the reliability hardly increases (0.01). This is not enough reason to remove the item from the analysis, and since the item is believed to be more related to PEOU (.59) than to C (.45), the item will be assigned to the PEOU component. Hence, it was decided to include all items in the analyses for testing the hypotheses.
4. Results 4.1. Measurement Model
After the constructs were confirmed, the variables were tested to find the strength and direction of the relationship between the constructs and to determine whether the hypotheses are supported. By means of simple regressions, the hypothesis 1 through 6 were tested. Hypothesis 1, PU has a positive influence on A, hypothesis 3, C has a positive influence on A, hypothesis 5, ATMA has a positive influence on A, and hypothesis 6, A has a positive influence on BI, are significant and thus supported as is shown in Table 4. This implies that when these variables increase, the attitude increases as well. The results for H2, PEOU has a positive influence on A, and H4, IPC has a negative influence on A, are not significant. Therefore, H2 and H4 were rejected. In addition, the effect of IPC on A for H4 is positive instead of the expected negative. Furthermore, the analyses show that initially PU and C are
probably responsible for the largest part of the variance in A. This was confirmed by an additional bi-variate analysis that was performed on the data.
Table 4. Simple regressions to test the hypotheses
Hypothesis Path Direction B t-value p Rejected/Accepted
H1 PU A + .81 4.39 <.001 Accepted H2 PEOU A + .53 1.78 >.001 (.76) Rejected H3 C A + .78 4.03 <.001 Accepted H4 IPC A - .14 2.56 >.001 (.011) Rejected H5 ATMA A + .98 5.22 <.001 Accepted H6 A BI + .34 3.16 <.001 Accepted
Hereafter, a multiple linear regression was calculated stepwise to predict the attitude towards virtual voice assistants based on PU, PEOU, C, IPC and ATMA. A signification regression equation was found for two models. The first model contains variable ATMA (F(1,303) = 38.11, p < .001), with an R² of .11. The second model contains ATMA and PU (F(2,302) = 33.02, p <.001), with an R² of .18.
The respondents predicted Attitude is equal to 1.09 + .24 (ATMA) + .20 (PU), where the variables were coded between -2 and 2, ranging from strongly negative to strongly positive. An increase in both ATMA and PU provided for a more positive attitude towards virtual voice assistants as they turned out to be significant predictors of the level of A (p < .001). Variables PEOU, C and IPC were not significant predictors of the level of A in this setting.
4.2. Country Differences
To test the last hypothesis, tests were performed to measure if country of residence moderates the relationship between A and BI. As shown in Section 4.1, attitude was significantly related to behavioural intent. For hypothesis 7, it was expected that the country of residence moderates this relationship. The interaction effect of this moderation was measured with Hayes process macro in SPSS. The results re-confirm that attitude was significantly related to behavioural intent with a regression coefficient of .54 (p < .001), slightly lower than when measured with a single regression. However, when looked at the interaction effect, the country of residence does not significantly moderate the relationship with a coefficient of .01 (p = .56). In addition, the Johnson-Neyman technique showed was not able to detect a significant relationship between attitude and behavioural intent when moderated by country of
residence at a certain level of standard deviations. The regression results of testing the hypotheses, including H7, are displayed in Figure 3.
Figure 3. Results multiple regression calculated for the research model (** p<0.01)
To cross-check the results, the country of residence variable was coded into nine dummy variables to measure if certain countries would have a significant effect in the regression equation. The interaction between the separate countries and attitude was added to the regression between A and BI. The results show an R² of .53 with a Durbin-Watson value of 1.57, but with a significance of p = .65. When looked at the interaction coefficients, no country was found to have a significant interaction (p < 0.05). For this reason, the null hypothesis is accepted and it is assumed that the country of residence does not moderate the relationship between A and BI.
Lastly, additional tests were run to see if there were differences in attitude and behavioural intent when measured separately. When it comes to attitude towards virtual voice assistants in the different countries, it appears that there is a difference detected (F(8,297) = 2.47, p = .01. However, when looked at Levene’s test for homogeneity, it turns out that the variances in the dataset are possibly equal (p < .05). When looked at the results for differences per country when it comes to BI, the homogeneity of variances gives a non-significant result (p = .06), which indicates that there is a difference in BI between the countries included in the research (F(8,297) = 3.49, p = .001). An overview of the differences can be found in Appendix C. Austria has the lowest behavioural intent (N = 38, M = -.76, SD
5. Discussion and Conclusion 5.1. Theoretical Implications
This study focuses on mapping consumers’ current view on using virtual voice assistants as the new interface for communication between customers and organisations. To analyse the current perspective of consumers, the TAM was deployed in combination with construct compatibility from IDT, a privacy risk factor and the general attitude towards advertising. Results showed that not every hypothesis was supported. Not every variable from the research model was able to explain the variance in the level of attitude or behavioural intent towards VVA’s in this context. PEOU (H2) and IPC (H4) were considered as not influencing the attitude towards virtual voice assistants in this context. In addition, it was striking that IPC had a positive (non-significant) effect, instead of the expected negative. The reasons for this opposite direction are not clear, but several reasons might exist. The analysis of the mobile phone usage and online shopping behaviour show that many respondents are accustomed to using a mobile phone for several hours per day. In addition, the respondents shop online frequently using their mobile phone. This can be a sign that consumers are now more accustomed to technologies that are more complex and think that they understand it. An assumption which would be in line with the findings of Wu and Wang (2005) from their study on mobile commerce. As far as PEOU is concerned, it is possible that the respondents in this study are more technologically savvy than the average person in their country because of the way they were recruited for the survey. The people in the database ended up there because they were especially interested in innovations. This again, is an assumption and needs to be investigated further as PEOU is often significant in this research set-up in different studies (Wu & Wang, 2005; Van Eeuwen, 2017). Also unexpected was that country of residence (H7) did not have a moderating effect on the relationship between attitude and behavioural intent. When the two dependent variables were measured separately, the difference per country was significant.
Construct compatibility (H3) did influence attitude as a single predictor, but not as part of the complete research model. This is not in line with expectations and previous comparable research where C is the most important predictor (Wu & Wang, 2005). It could be that past experience and user requirements are less relevant for the user as VVAs offer a new type of interface that requires habituation. However, this is not clear. Furthermore, PU (H1) and ATMA (H5) are the strongest predictors of attitude towards VVA’s in this context.
Whereas in other studies, the combination of the TAM and IDT, supplemented with external variables ATMA and IPC, explains a large percentage of the variance in A (Wu and
Wang, 2005; Van Eeuwen, 2017), the research model in this study explains only a small percentage of the total variation in attitude towards using virtual voice assistants. Since the TAM looks at intrapersonal differences between respondents and, according to some studies, is seen as outdated and incomplete (Zarmpou, Saprikis, Markos & Vlachopoulou, 2012), future research may look at a new set-up of the research model. This would include finding new variables that would consistently be significant predictors of the users’ acceptance and adoption of innovations.
5.2.Practical Implications
The results from this study may provide researchers and businesses with a better understanding of virtual voice assistants, consumers’ attitude towards these devices and the intention to eventually use a VVA for commercial purposes. Chatbots, the predecessor of the virtual voice assistant, are increasingly used for communication between companies and customers and help facilitate an omni-channel customer experience. This habituation to chatbots ensures that customers are increasingly accustomed to being in contact with companies that are approachable via an easy and fast channel. But, because it’s a structural change in the communication between consumers and companies, or even people and computers, it will probably take some time before this form of interaction becomes the norm on a large scale. Still, it remains important to continue studying virtual voice assistants and the acceptance by consumers during the unfolding of the technology in society. The technology continues to develop and keeps triggering various developments in society. In addition, VVA and conversational commerce are still fairly new and only in the beginning of the innovation cycle (Rogers, 1983).
5.2.1. Consumer Acceptance
The results of this study show that perceived usefulness and the attitude towards mobile advertising explain the attitude towards VVAs in this context. These findings indicate that providers of VVAs and managers must increase the perceived usefulness by closely listening to the needs and requirements of the consumer (Fenech, 1998). The expansion of conversational commerce may contribute to this. Where chatbots were previously mainly used to provide communicative service, VVAs offer a possibility to go one step further. By bringing together both commercial service, help in finding the right solutions, products and services, and traditional customer service in a VVA, the perceived usefulness can potentially
be increased. This is an opportunity for companies that, according to Messina (2016), will continue to be deployed in the future.
In this future scenario, it will be important to take internet privacy into account even though the variable has not proven to be a significant predictor of attitude in this study. It is important to consider privacy related risks as privacy risk is increasingly playing a role and will possibly affect people's behaviour and attitude. Virtual voice assistants always listen to conversations within reach from their default mode (Verhoeven, 2018). In addition, the devices identify who is in the house based on the sound of footsteps and choice of music. For the further development of conversational commerce, it is therefore important for companies not to be intrusive and to take into account the privacy wishes of consumers. Especially because the general attitude towards advertising in this research can be seen as a variable that strongly influences attitude and behavioural intent. When this attitude towards advertisements and conversational commerce becomes negative, for instance, when the consumers feel that their privacy is violated, the attitude towards virtual voice assistants can also become more negative.
Furthermore, even though it became clear that country of residence does not have a moderating influence on the relationship between A and BI, it can still be assumed that there are differences in how people view VVAs per country (Exalto, De Jong, De Koning and Ravensteijn, 2018). It is possible that the adoption, innovation speed, and the use of VVAs differs both now and in the future. The non-significant results do show that at present there is still little behavioural intent in certain countries, for example in Europe, when it comes to VVAs, certainly for commercial purposes.
Austria with its rather negative, and Mexico with the most positive attitude and behavioural intent, are the most deviant countries. It is not clear whether these differences can be explained by culture, prosperity or other developments. When looking at Hofstede's cultural dimensions (Hofstede, 1984), consisting of power distance, individualism, masculinity, uncertainty avoidance, long-term orientation and indulgence, there are certainly differences between the two countries. Hofstede has developed a tool (hofstede-insights.com) based on his theory to compare culture in countries. From this tool, it becomes clear that Austria has a very low power distance and Mexico a very high one. Power distance in a country indicates the degree to which members of a society expect that power is distributed unequally. An assumption may therefore be that Mexico is open to the use of VVAs because it can give the respondents a form of control and power. With a VVA, they can gain control over various actions at home and outside. Austria may not have a need for this. Austria is also
more oriented towards the long term variable and Mexico towards the short term. A high score on this variable stands for traditions and a low score for creativity and is less linked to the past. The VVA is a disruptive technology as it changes the way we communicate (Argenti, 2006), which is why Austria may have more difficulty with adoption because they adhere to traditions.
For future research, it would be interesting to investigate why these differences have manifested itself. In addition, large Asian countries, such as China and Japan, and the United States were not included in this study. As these three countries are already reasonably far with developments surrounding VVAs, compared to other countries (Lien & Cao, 2014), it may be an addition to include them in the analysis.
5.2.2. Innovation Diffusion
This study did not investigate where exactly the VVA is in the innovation cycle (Rogers, 1983), but for companies it may be interesting to use the TAM and IDT to determine this. When the location in the cycle is localized, companies can react and invest at the right time. According to the IDT, this would be the moment the 'chasm' was crossed (Moore, 1991), a point in the innovation diffusion cycle between the early adopters and early majority. At that moment, the innovation can be stopped or going from invention to penetration.
Currently, people and companies are not yet used to and specialized in the field of VVAs. The system has not yet been standardized and the technology behind it has not yet been fully developed (Exalto, De Jong, De Koning and Ravensteijn, 2018). It could therefore be assumed that the VVAs and conversational commerce are not yet beyond the chasm. But, as e-commerce once was at this point of the innovation cycle (Wu & Chang, 2005), it is possible that the VVA and conversational commerce are going to make the same developments.
In conclusion, it can be stated that consumers are fairly open to adopting virtual voice assistants. On average, the respondents have quite a positive attitude towards technology and with that a relatively positive behavioural intent. Here it is important to note that most respondents do not make use of a VVA for commercial purposes, but for personal activities, as was shown by the results. Despite the fact that the technology is in its infancy, and will develop at a rapid pace. The technology is far from being at a its peak and the gap between warm human response and cold programmatic answers is still intact. Whether it will ever be possible to simulate these human aspects, remains unknown. What is clear is that virtual voice assistants advance at the right time and that consumers are open to this new form of
communication. It is therefore increasingly interesting for both businesses and science to continue studying the phenomenon of virtual voice assistants.
5.3. Limitations
The results of this study have a number of limitations that must be taken into account when conclusions are drawn. Firstly, the sample size of more than 300 respondents could be considered as a limitation. Given that the respondents come from nine countries, it concerns an average of 34 respondents per country. Ideally, the number of respondents per country is a higher in order to draw reliable statements from the data. Due to time and the scope of this research, it was not possible to recruit more respondents. For future research, it would be interesting to comparatively analyse the differences per country in greater depth. More detailed insights into these differences can be valuable for companies.
In addition, the tool that was used for data collection is may cause for a bias in the sample. The tool was only recently set-up, which means that the size of the database had not reached its maximum. The tool has the potential to grow in the future, which would probably also result in a growth in the percentage of people that complete the questionnaire.
Lastly, the variables explain 66.7% of the variance in the research model. The variables were carefully chosen in advance, but every construct turned out to be significant in this setting. In order to develop a better predictive model for adoption, the research model needs to be adapted with other constructs for further research. Additional research will have to be done to find other variables that can explain the variation, such as social influence and cognitive instrumental processes (Venkatesh, 2000), trust (Ha and Stoel, 2009), and innovation (Zarmpou, Saprikis, Markos, & Vlachopoulou, 2012).
References
Agarwal, R., & Prasa, J. A Conceptual and Operational Definition pf Personal Innovativeness in the Domain of Information Technology, Information Systems Research 9(2), 1998, 204–301.
Argenti, P. A. (2006). How Technology Has Influenced The Field Of Corporate
Communication. Journal of Business and Technical Communication, 20(3), 357-370. Britz, D. (2016). Deep Learning for Chatbots, Part 1, Introduction. WildML. Retrieved from
http://www.wildml.com/2016/04/deep-learning-forchatbots- part-1- introduction/. Chen, L., Gillenson, M. L., & Sherrell, D. L. (2002). Enticing Online Consumers: An
Extended Technology Acceptance Perspective. Information & management, 39(8), 705-719.
Chung, M., Ko, E., Joung, H., & Kim, S. J. (2018). Chatbot E-Service and Customer Satisfaction Regarding Luxury Brands. Journal of Business Research.
Dale, R. (2016). The Return of the Chatbots. Natural Language Engineering, 22(5), 811-817. Danielsen, P. J. (2000). The Promise of a Voice-Enabled Web. Computer, 33(8), 104-106. Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User Acceptance of Computer
Technology: A Comparison of Two Theoretical Models. Management Science, 35, 982–1003.
Dinev, T., & Hart, P. (2006). An Extended Privacy Calculus Model for E-Commerce Transactions. Information systems research, 17(1), 61-80.
Eeuwen, M. V. (2017). Mobile Conversational Commerce: Messenger Chatbots as the Next Interface between Businesses and Consumers. University of Twente.
Exalto, M., de Jong, M., de Koning, T., & Ravesteijn, A. G. P. (2018, October).
Conversational Commerce, the Conversation of Tomorrow. In ECMLG 2018 14th
European Conference on Management, Leadership and Governance (p. 76).
Academic Conferences and publishing limited.
Fenech, T. (1998). Using Perceived Ease of Use and Perceived Usefulness to Predict
Acceptance of the World Wide Web. Computer Networks and ISDN Systems, 30(1-7), 629-630.
Ha, S., & Stoel, L. (2009). Consumer E-Shopping Acceptance: Antecedents in a Technology Acceptance Model. Journal of Business Research, 62(5), 565-571.
Heo, M., & Lee, K. J. (2018). Chatbot as a New Business Communication Tool: The Case of Naver TalkTalk. Business Communication Research and Practice, 1(1), 41-45.
Hill, J., Ford, W. R., & Farreras, I. G. (2015). Real Conversations with Artificial Intelligence: A Comparison between Human–Human Online Conversations and Human–Chatbot Conversations. Computers in Human Behavior, 49, 245-250.
Hofstede, G. (1984). Cultural Dimensions in Management and Planning. Asia Pacific journal
of management, 1(2), 81-99.
Hollister, S. (2016, April 15). Why does Facebook want me to chat with dumb robots? CNET. Retrieved from http://www.cnet.com/news/chatbots-facebook-bots- aresuper-stupid-right-now/.
Karjaluoto, H., & Alatalo, T. (2007). Consumers' Attitudes Towards and Intention to Participate in Mobile Marketing. International Journal of Services Technology and
Management, 8(2-3), 155-173.
Koksal, Ilker. “Voice-First Devices Are The Next Big Thing - Here's Why.” Forbes.
Retrieved from: www.forbes.com/sites/ilkerkoksal/2018/02/01/voice-first-devices-are-the-next-big-thing-heres-why/#e59df0168739.
Lee, D., Oh, K. J., & Choi, H. J. (2017). The Chatbot Feels You - A Counseling Service Using Emotional Response Generation. Big Data and Smart Computing (BigComp),
2017 IEEE International Conference, 437-440.
Lee, S., & Choi, J. (2017). Enhancing User Experience with Conversational Agent for Movie Recommendation: Effects of Self-Disclosure and Reciprocity. International Journal of
Human-Computer Studies, 103, 95-105.
Lien, C. H., & Cao, Y. (2014). Examining WeChat Users’ Motivations, Trust, Attitudes, and Positive Word-of-Mouth: Evidence from China. Computers in Human Behavior, 41, 104-111.
Ling, K. C., Piew, T. H., & Chai, L. T. (2010). The Determinants of Consumers’ Attitude Towards Advertising. Canadian social science, 6(4), 114-126.
López-Nicolás, C., Molina-Castillo, F. J., & Bouwman, H. (2008). An Assessment of Advanced Mobile Services Acceptance: Contributions from TAM and Diffusion Theory Models. Information & Management, 45(6), 359- 364.
Lowry, P. B., Romano, N. C., Jenkins, J. L., & Guthrie, R. W. (2009). The CMC Interactivity Model: How Interactivity Enhances Communication Quality and Process Satisfaction in Lean-Media Groups. Journal of Management Information Systems, 26(1), 155-196. Marchick, Adam. “The 2017 Voice Report by Alpine (VoiceLabs).” Medium. Retrieved from:
https://medium.com/@marchick/the-2017-voice-report-by-alpine-fka-voicelabs-24c5075a070f.
Messina, C. (2016, January 19). 2016 Will Be the Year of Conversational Commerce.
Medium. Retrieved from:
https://medium.com/chris-messina/2016-will-be-the-yearof-conversational-commerce-1586e85e3991#.bsdskkyji.
Mool, T.J. (2017). What is Conversational Commerce? Native Msg. Retrieved from: https://blog.nativemsg.com/what-is-conversational-commerce-five-useful-examples-of-e-commerce-in-messaging-6265cc56dfa5
Moore, G. A. (1991). Crossing the Chasm: Marketing and Selling High-Tech Products to Mainstream Customers (Collins Business Essentials). HarperBusiness, New York. Newman, D. (2016, May 24). Chatbots and the Future of Conversation Based Interfaces.
Forbes. Retrieved from: http://www.forbes.com/sites/danielnewman/2016/05/24/
chatbots-and-the- future-of-conversation-basedinterfaces/#6c40cc51220d.
Piyush, N., Choudhury, T., & Kumar, P. (2016). Conversational Commerce a New Era of E- Business. In System Modeling & Advancement in Research Trends (SMART),
International Conference (pp. 322-327). IEEE.
Pikkarainen, T., Pikkarainen, K., Karjaluoto, H., & Pahnila, S. (2004). Consumer Acceptance of Online Banking: an Extension of the Technology Acceptance Model. Internet
research, 14(3), 224-235.
Ram, S., & Sheth, J. N. (1989). Consumer Resistance to Innovations: the Marketing Problem and Its Solutions. Journal of Consumer marketing, 6(2), 5-14.
Reddy, B. R., & Mahender, E. (2013). Speech to Text Conversion Using Android
Platform. International Journal of Engineering Research and Applications
(IJERA), 3(1), 253-258.
Rogers, E. M. (1983). Diffusion of lnnovations. New York: Free Press, 1, 14-15.
Schreuder, A., Schreuder, A., & van Wijk, J. (2017). Customer Centric Artificial Intelligence- Using Text and Sentiment Analysis & Deep Neural Network Learning to make Chatbots Reply in a more Customer Centric Fashion.
Shebat, A. (2016, April 29). Five Scenarios for How Humans and Bots Will Work Together.
VentureBeat. Retrieved from
http://venturebeat.com/2016/06/15/5-scenarios-for-howhumans-and-bots-will- work-together/.
Sonnenwald, D. H., Maglaughlin, K. L., & Whitton, M. C. (2001). Using Innovation Diffusion Theory to Guide Collaboration Technology Evaluation: Work In Progress. Straub, D., Keil, M., & Brenner, W. (1997). Testing the Technology Acceptance Model
Across Cultures: A three country study. Information & management, 33(1), 1-11. Sultan, F., Rohm, A. J., & Gao, T. T. (2009). Factors Influencing Consumer Acceptance of
Mobile Marketing: A Two-Country Study of Youth Markets. Journal of Interactive
Marketing, 23(4), 308-320.
Talk to Me. Increasing Demand and Deeper Interactions with Digital Voice Assistants Expected. (2017). Accenture, The Future of Work. Retrieved from: www.accenture. com/_acnmedia/PDF-65/Accenture-Voice-Assistant-Infographic.pdf.
Tell, J. (2017, August, 24). What Are Amazon Allexa Skills? GearBrain. Retrieved from: https://www.gearbrain.com/what-are-amazon-alexa-skills-2471456002.html. Tuzovic, S., & Paluch, S. (2018). Conversational Commerce – A New Era for Service
Business Development?. In Service Business Development (p. 81-100). Springer Gabler, Wiesbaden.
Van Manen, T. (2016, April 14). Bot or Not: Dit Is Waarom Facebook Inzet op Chatbots.
Marketingfacts. Retrieved from:
http://www.marketingfacts.nl/berichten/chatbots-facebook-inzet-chatbots-messenger.
Venkatesh, V. (2000). Determinants of Perceived Ease of Use: Integrating Control, Intrinsic Motivation, and Emotion into the Technology Acceptance Model. Information
Systems Research, 11(4), 342-365.
Verhoef, P. C., Lemon, K. N., Parasuraman, A., Roggeveen, A., Tsiros, M., & Schlesinger, L. A. (2009). Customer Experience Creation: Determinants, Dynamics and Management Strategies. Journal of retailing, 85(1), 31-41.
Verhoeven, A. (2018, March 28). Audiotrends: Smart Speakers Zijn de Toekomst.
Frankwatching. Retrieved from: https://www.frankwatching.com/archive/2018/03/28/
audiotrends-smart-speakers-zijn-de-toekomst/.
Vijayasarathy, L. R. (2004). Predicting Consumer Intentions to Use On-Line Shopping: the Case for an Augmented Technology Acceptance Model. Information & Management,
41(6).
Watson, McCarthy & Rowley. (2013). Consumer Attitudes Towards Mobile Marketing in the Smart Phone Era. International Journal of Information Management, 33(5).
Wu, J. H., & Wang, S. C. (2005). What Drives Mobile Commerce?: An Empirical Evaluation of the Revised Technology Acceptance Model. Information & management, 42(5), 719-729.
Zadrozny, W., Budzikowska, M., Chai, J., Kambhatla, N., Levesque, S., & Nicolov, N. (2000). Natural Language Dialogue for Personalized Interaction. Communications of
the ACM, 43(8), 116-120.
Acceptance of Mobile Services. Electronic Commerce Research, 12(2), 225-248. Zhang, W. N., Zhu, Q., Wang, Y., Zhao, Y., & Liu, T. (2017). Neural Personalized Response
Appendix
Appendix A: Tables and Figures, Methods Section Appendix A1
Appendix A1 contains information on the usage and first impression of virtual voice assistants.
Table 1. Experience with Virtual Voice Assistants
Frequency Percent (%) Cumulative (%) Ever heard of virtual voice
assistants?
No 13 4.2 4.2
Yes 293 95.8 100.0
Ever used virtual voice assistants?
No 68 22.2 22.2
Yes 238 77.8 100.0
First impression of virtual voice assistants Very negative 0 0.0 0.0 Negative 35 11.4 14.7 Neutral 24 7.8 24.8 Positive 143 46.7 84.9 Very Positive 36 11.8 100.0
Table 2. Activities carried out by the respondents with a VVA
Voice Assistant Activity Frequency
Ask a quick question 158
Search for information that you would normally type into a search engine 145
Play music 102
Check the weather/news 86
Set a timer or reminder 83
Check traffic/navigation 51
Send a text or email 37
Control other smart devices 28
Appendix A2
Appendix A2 contains information on the mobile phone usage of the respondents.
Table 3. Mobile phone usage
Frequency Percent (%) Cumulative (%)
Time spent per day on mobile phone on average 0-1 hour 17 5.6 5.6 1-2 hours 58 19.0 24.5 2-3 hours 79 25.8 50.3 3-4 hours 74 24.2 74.5 > 4 hours 78 25.5 100.0
Use of messaging apps per day
0 - 10 minutes 3 1.0 62.1
10 - 20 minutes 16 5.2 63.1
20 - 30 minutes 43 14.1 68.3
30 – 40 minutes 54 17.6 82.4
> 40 minutes 190 62.1 100.0
Table 4. Activities carried out on the mobile phones of the respondents
Activity on Phone Frequency
Social media 291
Messaging via app 284
Browsing the web 261
Calling 228
Making Photo's 192
Online Shopping 153
Appendix A3
Appendix A3 contains information on the online shopping behaviour of the respondents.
Table 5. Devices used for shopping online
Device for shopping Frequency
Computer or laptop 260
Mobile phone 235
Tablet 59
Virtual Voice Assistant 11
Table 6. Differences in ATMA, A and BI when looked at shopping behaviour
Shopping online ATMA A BI
N M SD M SD M SD Daily 6 1.67 5.47 1.67 4.84 2.00 1.67 Weekly 73 -.58 4.07 .51 3.73 .60 2.95 Monthly 168 -1.45 4.45 .57 3.01 .56 2.50 Annually 47 -1.26 3.77 1.17 3.67 1.53 2.85 Never 12 -2.92 4.12 .33 3.34 .00 2.99
