Identifying relevant segments of AI applications adopters : Expanding the UTAUT2’s variables

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY-NC-ND 4.0 https://creativecommons.org/licenses/by-nc-nd/4.0/ Identifying relevant segments of AI applications adopters : Expanding the UTAUT2’s variables © 2020 Elsevier Accepted version (Final draft) Cabrera-Sánchez, Juan-Pedro; Villarejo-Ramos, Ángel F.; Liébana-Cabanillas, Francisco; Shaikh, Aijaz A. Cabrera-Sánchez, J.-P., Villarejo-Ramos, Á. F., Liébana-Cabanillas, F., & Shaikh, A. A. (2021). Identifying relevant segments of AI applications adopters : Expanding the UTAUT2’s variables. Telematics and Informatics, 58, Article 101529. https://doi.org/10.1016/j.tele.2020.101529 2021 Identifying relevant segments of AI applications adopters – Expanding the UTAUT2’s variables Juan-Pedro Cabrera-Sánchez PhD Student [email protected] Business Administration and Marketing University of Sevilla, Spain Ángel F. Villarejo-Ramos Associate Professor [email protected] Business Administration and Marketing University of Sevilla, Spain Francisco Liébana-Cabanillas Professor [email protected] Marketing and Market Research Department University of Granada, Spain Aijaz A. Shaikh* Postdoctoral Researcher [email protected] Jyväskylä University School of Business and Economics, P.O.Box 35, FI-40014, University of Jyväskylä, Finland * Corresponding Author Identifying relevant segments of AI applications adopters – Expanding the UTAUT2’s variables Abstract Artificial intelligence (AI) is a future-defining technology, and AI applications are becoming mainstream in the developed world. Many consumers are adopting and using AI-based apps, devices, and services in their everyday lives. However, research examining consumer behavior in using AI apps is scant. We examine critical factors in AI app adoption by extending and validating a well-established unified theory of adoption and use of technology, UTAUT2. We also explore the possibility of unobserved heterogeneity in consumers’ behavior, including potentially relevant segments of AI app adopters. To augment the knowledge of end users’ engagement and relevant segments, we have added two new antecedent variables into UTAUT2: technology fear and consumer trust. Predictionorientated segmentation was used on 740 valid responses collected using a pre-tested survey instrument. The results show five segments with different behaviors that were influenced by the variables of the proposed model. Once known, the profiles were used to propose apps to AI developers to improve consumer engagement. The moderating effects of the added variables—technology fear and consumer trust—are also shown. Finally, we discuss the theoretical and managerial implications of our findings and propose priorities for future research. Keywords: Artificial intelligence; UTAUT2; segmentation; technology fear; consumer trust; heterogeneity 1. Introduction In recent years, advanced technologies, robotics, expert systems, and artificial intelligence (AI) applications and devices have become integral components of information technology (IT) policies and business strategies. Several developments in different AI-based gadgets have been introduced and deployed during the last decade, including voice recognition systems, virtual assistants, online recommendation systems, chat-bots, self-driving cars, and even search engines. These breakthroughs in AI technologies were, perhaps, due to the digital transformation initiatives undertaken in several countries, including members of the European Union. According to Gabriel and Goertzel (2019), by the end of 2020, the global AI market will exceed 45 billion USD. Considering this exponential growth in the demand for AI technologies, researchers (Sun and Medaglia, 2019) have compared the AI revolution to the industrial and digital revolutions. Reflecting on the capabilities of AI technologies, researchers (Kaplan and Haenlein, 2019; Duan et al., 2019) have argued that AI is a system or application that is capable of correctly interpreting both internal and external data, learning from such data, and using those learnings to achieve specific individual or organization-wide goals. AI enables greater personalization of information and services by considering a consumer’s needs and demands. This helps companies across various sub-sectors of the economy improve their decisionmaking, problem-solving (Androutsopoulou et al., 2019), customer service, personalization, conversion, and retention. Per prior findings (Waller and Fawcett, 2013), AI and similar technologies allow organizations to obtain and process a great quantity of valuable data in real time and use it to predict, describe, and even prescribe consumer and market behavior. This knowledge allows companies to become leaders and gain competitive advantages (Sivarajah et al., 2016), such as when Google recommends a restaurant nearby or when Booking.com shows consumers hotels in places they plan to visit. These companies are using AI applications to improve customers’ experiences and increase engagement. In addition to immense benefits, several challenges to the AI technologies, devices and services have been reported in the research. One of the major challenges, for example, include the ignorance, technology fear, and consumer distrust (Yaqoob et al., 2016). With regard to consumer preferences for convenient, innovative products, services, and devices, digital natives (i.e., consumers of the Gen Z and millennial generations) have distorted traditional business models, disrupted several business empires, and created a demand for more innovative, AI application-supported, and shared business models. It is widely believed that Big Data and AI are near ubiquitous (Kaplan and Haenlein, 2019) and they have received significant attention from tech-savvy consumers. Most previous empirical studies have focused on the technical side of Big Data, AI (Lecun et al., 2015; Triguero et al., 2015), app development, statistical algorithms, data mining cases, and analytics (Sivarajah et al., 2016). The focus of most of these studies is limited to the health and education sectors (Fan et al., 2018; Churamani et al., 2017), social networks (Liu, 2019), and the organizational perspective (Liu et al., 2020). We also found that a few studies have focused on the initial adoption of new and innovative technologies, systems, and applications, including AI, from the consumer perspective. Prior research (Schepman and Rodway, 2020) has argued that consumers’ general attitudes toward AI applications and systems likely play a major role in AI’s acceptance and prolonged use. Given the dearth of research examining consumer or end-user behaviors, attitudes, and beliefs toward the adoption and use of AI apps, the purpose of the present work is multi-fold. First, in the context of a developed country, we extend and validate one well-established unified theory of adoption and use of technology model, UTAUT2, that predicts an individual’s behavioral intentions and use of information technology, such as AI. Second, we determine the critical factors affecting the adoption and use of AI apps by adapting UTAUT2. Third, we add two new constructs to the UTAUT2, technology fear and consumer trust, as these could influence the individual’s behavioral intention to use AI apps. We added these two variables to improve the UTAUT2 and analyzed differences in consumers’ behaviors by describing a few segments that explain intention to behave a certain way regarding AI apps. These newly variables will allow us to make recommendations on how to design and promote AI apps. We also propose a segmentation that defines the unobserved heterogeneity of these consumers or end users. There are diverse papers on segmentation concerning technology adoption in different contexts, such as mobile banking (Shaikh and Karjaluoto, 2015), mobile financial services apps (Karjaluoto et al., 2019), information and communication technology adoption (FuentesBlasco et al., 2017), mobile phone use (Rondan-Cataluña et al., 2010), mobile TV versus mobile news apps (Verdegem and De Marez, 2011), and even video games (Ramírez-Correa et al., 2018). However, none were found on a posteriori segmentation related to AI app adoption using latent class segmentation. Finally, we provide further insight into the role that demographic factors, such as age and income, could play in AI app adoption intention and usage. The ramifications of this study are clear for AI app developers, business executives, and policymakers. For example, it would be meaningful for the practice to know which among the variety of UTAUT2 constructs largely affect consumer intention, use, and success of AI apps. Also, how technology fear and consumer trust moderate the relationship between behavioral intention and use behavior toward the AI apps? Especially as they relate to consumer fear about AI technology and the lack of consumer trust in AI apps, the contributions of this study have important public policy ramifications for consumers’ personal data safety and security, particularly given that almost all AI apps communicate with the cloud. The remainder of this of this paper is as follows: section 2 provides the theoretical background and the state of AI in Spain, the context of the study; section 3 explains the research model and the hypotheses development; section 4 discusses the research methodology; section 5 reports the findings; and section 6 presents the main theoretical and practical conclusions as well as the study’s limitations. 2. Theoretical background 2.1 Artificial intelligence apps Mobile application software options for smartphones (apps) have grown enormously since the inception of app markets (Liao et al., 2018). Per Statista’s latest report (2018) on the use of mobile apps, more than 45% of participants remembered having between 1 and 15 applications on their smartphone. There are currently many apps available to carry out any type of activity, from solutions for increasing work productivity to help with learning a new language or leading a healthier life. However, the needs of users are changing, and applications for mobile devices are changing (Yen et al., 2019) with more AI options now embedded in various applications. As predicted by Gartner (2019), the exploration and implementation of AI-based applications and systems will rapidly become quite evident, with companies and consumers soon witnessing the presence of AI features in various applications on smart devices. Currently, the AI apps include voice assistive apps (e.g., Siri, Alexa), which are used in a variety of devices; facial recognition apps (e.g., AppLock, FaceApp), which are commonly used for security purpose such as unlocking phones or to recognize faces in a photo library, online recommendation apps (playlist generators for video and music services (e.g., Netflix, YouTube, and Spotify), which use recommendation algorithms based on permissions and functionalities; and geolocation apps (e.g., Google Maps, Bizzy), which are used to assist people with map navigation and nearby recommendations (Hoy, 2018; Peng et al., 2018; Oikonomidis and Fouskas, 2019). Different from expert systems, AI developments have been divided into three major eras (See Figure 1). The first era is Artificial Narrow Intelligence (ANI), which is highly predictive, reactive in nature, and based on predefined rules. ANI, which is sometimes referred to as “weak AI” or “special-purpose AI,” includes devices and solutions that perform specific tasks (e.g., smartphones are recognizing faces and other biometric features, weather forecasting, etc.). The second era is Artificial General Intelligence, which can sense and solve problems in tasks for which it was never designed (Kaplan and Haenlein, 2019). The third era is Artificial Super Intelligence (ASI), which comes quite close to the true definition and meaning of AI. ASIs are capable of innovative scientific creativity, social skills, and general wisdom (Kaplan and Haenlein, 2019), and they could make humans redundant. [Insert Figure 1 about here] 2.2 The state of AI in Spain Many countries are becoming aware of the significance and transformative power of AI for their economies, societies, public services, and labor markets. Consequently, they have increasingly recognized the need for comprehensive national AI strategies. In Spain, this approach is still far from reality. For example, in March 2019, Spain’s Ministry of Science and Innovation and universities published one report, titled “RDI Strategy in Artificial Intelligence,” setting the priorities for the AI field. These priorities include developing a framework for research, development, and innovation (RDI) in AI; identifying key research and innovation priority areas in AI; facilitating the transfer of knowledge and its return to society; and fostering the development of education and competences in the field of AI. A recent Organization for Economic Co-operation and Development report indicates that private equity investment in AI-focused startups in Spain from 2011 to mid-2018 represents 3% of the total amount invested in start-ups based in the EU, far behind France (13%), Germany (14%), and the United Kingdom (55%). According to a study carried out by the consulting firm Roland Berger, “Joining the Dots: A Map of Europe’s AI Ecosystem,” the four most important countries in AI in Europe are the United Kingdom, France, Germany, and Spain. These data show that there is a wide margin for improvement with a better system of cooperation between agents and that the technological investment made so far in Spain is insufficient. Ultimately, without solving these two aspects, Spain will not have an environment that favors AI technologies. In the private sector, AI activity is growing rapidly both through startups and in large companies and multinationals, with initiatives focused on the creation of R&D centers in AI technologies, so it is essential to encourage the analysis, study, and modeling of the factors that analyze the intention to use these tools. 2.3 The UTAUT2 and its constructs Although the UTAUT model adequately explains companies’ adoption of technology, it had to be revised and expanded to explain consumers’ adoption of technology, giving rise to the UTAUT2 (Venkatesh et al., 2012). Anecdotal evidence has suggested that the motivations for introducing the UTAUT2 included increased momentum in examining the consumer context due to growing interest in adopting and using new systems and technologies and to account for the latest technological developments. Venkatesh et al. (2012) included hedonic motivation, price-value, and habit in the UTAUT2, which was created to not only analyze IT adoption but also to predict future Use behaviour (Ramírez-Correa et al., 2019). The model comprises seven independent factors that influence the dependent factor behavioral intention, including performance expectancy, effort expectancy, social influence, facilitating condition, hedonic motivation, price-value, and habit. 2.3.1 Behavioral intention Prior research has examined consumers’ behavioral intentions in the context of online banking technology adoption (Guriting and Ndubisi, 2006; Luarn and Lin, 2005). These studies have reported a significant relationship between perceived usefulness/perceived ease of use and behavioral intention. Similarly, perceived playfulness in internet usage (Moon and Kim, 2001) and flow experience in online games (Hsu and Lu, 2004) correlate with behavioral intention. Baumann et al. (2007) took a different approach and segregated behavioral intention between short-term and long-term consumers. Here, a short-term behavioral intention allows a consumer to remain with a specific service provider for one year or less. However, in a long-term behavioral intention, this duration extends for five years or more. 2.3.2 Performance expectancy and effort expectancy The two key constructs of the UTAUT and its variant the UTAUT2 are performance expectancy and effort expectancy. Influenced by the technology acceptance model (TAM) that was introduced by Davis (1989), Venkatesh et al. (2003) included performance expectancy in place of “perceived usefulness” and effort expectancy in place of “perceived ease of use.” In addition, per Venkatesh et al. (2003) and Muhammad et al. (2018), performance expectancy comprises “perceived benefit” and “relative advantage,” while effort expectancy is considered similar to “perceived ease of use,” “complexity,” and “ease of use” (Magsamen-Conrad et al., 2015; Shaikh, 2016; Venkatesh et al., 2003) in influencing user behavioral intentions. In justifying the inclusion of performance expectancy in examining AI adoption, the authors understand that, until AI applications are fully developed and deployed, it will be difficult to predict their performance or usefulness. Moreover, it is widely believed that AI-based applications have better performance compared to other online applications. Previous studies (Ramírez-Correa et al., 2019; Venkatesh et al., 2003) have defined effort expectancy as the degree of ease that is associated with an individual’s use of a technology or an information system. Therefore, the perceived ease of use of an information system or technology assumes that such ease is more likely to induce behavioral intention and that effort expectancy has a significant influence on behavioral intention (Casey and WilsonEvered, 2012). For example, in the context of mobile banking, Alalwan et al. (2017) found that EE significantly and positively influenced behavioral intention. 2.3.3 Social influence With modern society always connected via various technologies, including social media, the interactions both between and among individuals and communities have increased tremendously and subsequently created a social atmosphere that has increased customer awareness and influenced the choices, intentions, and attitudes of individuals. This social atmosphere consists of reference groups, family and friends, colleagues, etc. According to Venkatesh et al. (2003), social influence is the extent to which a person perceives that others believe that it is important for him/her to apply a new information system or technology. Prior research has examined social influence in a variety of contexts, and its relationship with behavioral intention has been reported as both significant and insignificant. For example, in the context of mobile banking services, the relationship between social influence and behavioral intention was reported as insignificant (Alalwan et al., 2017; Merhi et al., 2019). 2.3.4 Hedonic motivations Several features are usually embedded into information systems and applications to increase consumer engagement, adoption, and usage, and they are broadly categorized under two major domains. The first domain includes productivity-oriented features that provide utilitarian or instrumental value to consumers and users. These productivity-oriented features, as explained by Van der Heijden (2004), promote extrinsic motivation, which demands rewards or external benefits. Popular examples of constructors that drive extrinsic motivation among users include perceived usefulness (or performance expectancy). The second category includes pleasure-oriented features that provide hedonic or self-fulfilling value to consumers and users of systems and applications. Per Van der Heijden (2004), hedonic features are strongly connected to family, home, and leisure activities and thereby target pleasure, happiness, and the fun aspect of using information systems or applications. The underlying purpose of embedding these hedonic features is to encourage the sustained or continuous usage of the application or system. Hedonic or self-fulfilling value promotes intrinsic motivation. Popular examples of the constructs that drive intrinsic motivation include hedonic expectancy, perceived enjoyment, flow experience, and perceived playfulness. Per Tamilmani et al. (2019, p. 223), hedonic motivation is the “fun or pleasure derived from using a technology or system, and it is an important determinant of a consumer’s technology acceptance and sustained use.” Prior research has found significant correlations between hedonic value and the Use behaviours of the mobile internet (Park, 2006) and mobile social network games (Paavilainen et al., 2012). Enjoyment has also been found to have a key role in virtual reality tourism (Tussyadiah et al., 2018). 2.3.5 Price-value Price-value is an individual’s cognitive trade-off between the perceived benefits of the applications and the monetary cost that is connected with using them (Ramírez-Correa et al., 2019). The benefits that drive price-value include convenience, quality, comfort, and ubiquity. The sacrifice is the monetary cost as well as the usage fees (if any) and the perceived risk of or actual privacy loss that is associated with using an application or system (Shaw and Sergueeva, 2019). Notably, some authors (e.g., Herrero and San Martín, 2019; Shaw and Sergueeva, 2019) have substituted price-value with perceived value (context: mobile commerce), privacy concerns (context: social network sites), etc. 2.3.6 Habit Habit is another factor that is considered significant to behavioral intention. Research has found that habitual behavior largely plays a role in the context of extended, sustained, or continuous usage of an information system. In the context of information system usage, Limayem et al. (2007) defined habit as the extent to which consumers tend to perform behaviors automatically and repeatedly because of learning. Although sufficient evidence (e.g., Ramírez-Correa et al., 2019) is available to suggest that habit is correlated with behavioral intention, a key question remains: What factors develop habit or the habitual behavior of a consumer? Per Limayem et al. (2007), the answer is consumer satisfaction, frequency of past behavior, and comprehensiveness of usage. 2.3.7 Facilitating conditions Shaw and Sergueeva (2019) defined facilitating conditions as the conceptualized knowledge, resources, and opportunities that are required to perform a specific behavior. Certain conditions as well as resources are required to facilitate the adoption and usage of an information system to complete a task. Developed countries have stable facilitating conditions, such as advanced telecommunication infrastructures, broadband and stable internet connectivity, and advanced education systems. These conditions allow greater adoption and usage of information systems and facilitating conditions; therefore, they are considered a significant predictor of the behavioral intention to use an information system. Prior research has examined facilitating conditions in a variety of contexts. For example, facilitating conditions can positively predict tablet use intentions (Magsamen-Conrad et al., 2015) and directly influence mobile banking adoption (Oliveira et al., 2014). 3. Research model and hypothesis development As a synthesis of previous models such as TAM (Davis, 1985), TPB (Ajzen, 1991) or TRA (Fishbein & Ajzen, 1975), we consider the UTAUT model (Venkatesh et al., 2003) one of the most suitable for measuring the acceptance level and use of tools that have been developed through AI. However, while the UTAUT focuses on companies, the UTAUT2 (Venkatesh et al., 2012) is orientated toward explaining the acceptance and use of technologies by end users. Therefore, this model will help us understand how end users adopt and use AI apps in a developed country. In addition to traditional UTAUT 2 constructs, in this research, we have used two new latent variables, technology fear and consumer trust, to assist in finding new clusters based on end-consumer behavior. This conceptual model is depicted in Figure 2. [Insert Figure 2 about here] 3.1 Performance expectancy on behavioral intention Performance expectancy is one of the most influential constructs regarding behavioral intention. In addition to Venkatesh et al. (2003), several other authors have also established a relationship between performance expectancy behavioral intention. For example, in the mobile banking context, Merhi et al. (2019) found that performance expectancy is an influential predictor of a consumer’s behavioral intention, while Chauhan et al. (2018) found similar results regarding voting machines,and Mosunmola et al. (2019) studied this phenomenon in mobile learning. However, this effect was insignificant in the use of online games on mobile devices (Ramírez-Correa et al., 2019). Therefore, we have hypothesized the following: H1: Performance expectancy is positively related to behavioral intention. 3.2 Effort expectancy on behavioral intention Like performance expectancy, the relationship between effort expectancy and behavioral intention is well established. Several studies (e.g., Al-Gahtani et al., 2007; Chauhan and Jaiswal, 2016; Kim et al., 2007; Lee and Song, 2013; Yu, 2012) have reinforced the sense and weight of the effect that effort expectancy has on behavioral intention. In the context of mobile technology, multiple studies have verified that behavioral intention is significantly and positively influenced by effort expectancy, including mobile banks (Alalwan et al., 2017), mobile cloud services (Park and Kim, 2014), mobile maps (Park and Ohm, 2014), mobile Short Message Services (Beza et al., 2018), and mobile learning (Ho et al., 2010). Thus, the following is hypothesized: H2: Effort expectancy is positively related to behavioral intention. present study. Two of the previously discussed variables were also examined: technology fear and consumer trust. We wanted to determine whether any of these variables were significantly different in one or more of the five segments. Therefore, we performed an analysis of variance (ANOVA) and found no significant differences in any sociodemographic variables of the UATUT2 model. In the case of behavioral intention, effort expectancy, hedonic motivation, price-value, and the newly included variables technology fear and consumer trust affected the different segments (see the results in Table 12). Only these variables (in bold) had significant differences between the segments. [Insert Table 12 about here] We also performed a Chi-square test to search for any relation between the categorical sociodemographic variables (gender, city of residence, kind of job, civil status, studies, and familial income) and the different segments, with no association found. Because all the tests were negative, we can conclude that none of the socio-demographic variables affects the segments that were obtained. 6. Discussion, implications, and limitations Using the UTAUT2 model, this article examines the factors that affect AI adoption among a heterogeneous group of people in Spain. This study offers important contributions in the context of AI application adoption, such as that behavioral intention yields the strongest effects on consumer use behavior and that performance expectancy and hedonic motivation have the greatest influence on behavioral intention. 6.1 Theoretical Implications Our findings provide some useful theoretical contributions. For example, our research shows that the revised and extended UTAUT2 model is consistent and that the behavioral intention toward AI apps is positively and significantly influenced by certain variables, notably performance expectancy and hedonic motivations. The perception of how AI applications are useful (performance expectancy) in achieving in day-to-day objectives is in line with the earlier findings reported by Lee and Song (2013) on e-governments and by Yu (2012) on the behavioral intention to use internet banking. The positive influence of hedonic motivation on using AI apps follows studies in the same context of online purchasing (Chen and Zhang, 2014) and in tourist recommendations for online reservations (Gupta and Dogra, 2017). Regarding the remaining variables of the original UTAUT2 model, we highlight a relevant influence (although at a lower level of demand) on the behavioral intention to use AI apps: The effort expectancy (the AI application’s ease of use) is in line with the findings of Cabrera-Sánchez and Villarejo-Ramos (2018) as are the social influence (what others consider appropriate to use), the perceived price-value (the value that end users associate with cost), and habit, as measured by the habitual use of these systems. The facilitating conditions (ease of access to the application) do not have a significant influence on behavioral intention, although they do on use behavior, for which habit also shows a significant favorable effect. One of the significant theoretical findings of our study is that the extension of the model proposed with two new variables (technology fear and consumer trust) has been significant improving the results of the original UTAUT2 model and supports the results as earlier reported by Wang & Jeong (2018), Wang et al. (2019), and Zhou (2012). The results of this research show that the new variables that were proposed to cluster end users had significance. These variables, although antecedents of use intention of use, indirectly affect use: they have a mediating effect on the relationship between behavioral intention and use behavior. In this sense, those users who have technology fear will reduce their use behavior due to negative perceptions of the technology itself. However, if AI apps users have more confidence, they will increase their use behavior, believing that these systems have numerous advantages. Therefore, technology fear and consumer trust are useful for explaining the unobserved heterogeneity of end users. Some literature exists about a posteriori segmentation with the UTAUT (Ramírez-Correa et al., 2018), but none exists for the UTAUT2. Even with this complex model (the UTAUT is more parsimonious than the UTAUT2), we were able to run a POS-PLS segmentation. After analyzing all data, we discovered great improvement in the explained variance in the endogenous variables. The original model had an explained variance of use behavior of 0.454, and the model with five segments improved this explained variance to 0.735. Same goes with the usage behavior: the original model had an explained variance of 0.178, while the model with five segments reached 0.602. See Table 9. From the ANOVA (Table 12), only the variables effort expectancy, hedonic motivation, price-value, technology fear, consumer trust, and behavioral intention showed significant differences for the five segments of end users. Table 13 briefly analyzes each segment, which are described below. The segment-1 (Players): This segment is the smallest (3.78%). The players have the largest value of hedonic motivation, the second largest value of trust, and the smallest value of technology fear, while the rest of the variables have small values. Therefore, we can conclude that those in this segment know the technology, trust it, and enjoy playing with it (although they do not have a significant intention to use it). The segment-2 (Home end users): This segment is the second largest (28.51%). They have small values for almost every variable, with the exceptions of hedonic motivation and behavioral intention. They think that these apps are not worth the price offered. Those in this segment are not afraid of this technology and even trust it; therefore, they have the highest behavioral intention to use it. They use it for some enjoyment, with a bit of trust and with no fear, but they are not going to pay for it. The segment-3 (Reluctant): This segment has a medium size (21.08%). They have the smallest behavioral intention of any segment, the largest technology fear, and almost no trust in this new technology. They also think that it is difficult to use and that there is no pleasure in doing it. They are not going to use this technology (at least not in the short term) because they do not trust it. The segment-4 (Professional end users): This segment is the largest (35.54%), and its users are ready to pay for these apps if they are perceived as worth the cost. They have the largest perceived price-value and a large behavioral intention. They also trust the technology, and they are not afraid of these apps. They enjoy them a bit less than those in Segment 2, but they differ in that they are willing to pay if the app is good. They also see this technology as being very easy to use, which makes them willing to use it. The segment-5 (Skeptical users): This segment is the second smallest (11.08%). They think that this new technology is very difficult to use (the biggest effort expectancy) and has no hedonic motivation, no price-value, no technology fear (the smallest one), and they fear this technology. Even with these characteristics, they have a greater behavioral intention than Segment 3, although it is not significant. They will use this new technology voluntarily in the short term, but they may use it in the future. [Insert Table 13 about here] 6.2 Managerial Implications AI apps, devices, wearables, and sensors are on our wrists, in our pockets, in our homes, in our cars, and in our workplaces, and they are already making a real difference to how we experience life and the world around us (Arm, 2020). This study used the revised UTAUT2 model to explore the acceptance and use of AI apps among a set of consumers and professionals using various AI apps in their daily routines. The UTAUT2 variables were combined with two new variables (technology fear and consumer trust) to create a consistent model for finding different segments of consumers or end users. With the profiles of these groups known, we can make recommendations to app developers, business executives, and policymakers. 6.2.1 For AI app developers For segment 1 (Players), developers may need to spend more resources to develop and deploy intuitive, simple, and enjoyable AI apps that consumers can access and use frequently. The developers should also make AI apps more useful for specific tasks, no matter the complexity of the programming, and communicate the usefulness or performance expectancy of AI apps to end users. For segment 2 (Home end users), app developers should communicate to the consumers and prospects that AI apps are valuable by explaining their benefits, including their 24/7/365 availability and noticeable time and money savings. Developers should present the AI apps’ capabilities for different tasks, increase consumers’ trust, and remove fears, such as the fear of privacy intrusion, as in the case of Alexa, a popular virtual/voice assistant developed by Amazon. After all, consumers are trading privacy for convenience. After the promulgation of consumer privacy regulations such as the General Data Protection Regulation (GDPR), consumer awareness of the amount of their personal data that AI apps and devices need to perform well has grown, and so has demand for security (Arm, 2020). For segment 3 (Reluctant), AI app developers should promote safety and trust in the acquisition and use of AI apps, devices, and services among various consumer segments and prospects interested in acquiring and using the AI apps, devices, and services. For segment 4 (Professional end users), not all AI apps, devices, and services are created equal. Various tendencies were found when choosing and using the AI apps to complete various personal and professional tasks. With this in mind, AI app developers should create apps that help users in their professional tasks and reflect benefits for their professional performance. Finally, for segment 5 (Skeptical users), app developers should promote safety and trust in the acquisition and use of AI apps and communicate the usefulness of these apps, including their ability to save time and money. Messages for this segment should be utilitarian. 6.2.2. For the business executives and policy makers Respondents placed greater emphasis on the usefulness of AI apps, the hedonic features, and habit. In addition, technology fear and consumer trust play significant roles in developing consumer behavioral intent to use AI apps. These findings provide significant business and marketing guidance to companies developing and deploying AI apps and other technologies. For example, the practice should take greater caution when offering AI apps, considering the growing privacy laws, such as GDPR. The storage and retrieval of private consumer data and how the personal consumer data is processed should be explicit and shared with the consumers. This would reduce the technology fear among consumers and increase their trust in AI apps. More than a third of respondents said they would switch to a competitor’s product should an AI device they use be hacked, and another third would consider stopping using that device category altogether. Clearly, the development of AI apps must be supported with an effective marketing communication strategy. Hedonic features in AI apps are significant. Consumers expect fun and entertainment when using AI-supported devices, apps, and other services. For example, issuing commands, asking questions, and getting a reply or the required information increase hedonic feelings and make a technology feel more intelligent. AI apps that do not create a hedonic experience will make the technology less attractive and ultimately fail. According to Arm (2020), consumers do not feel comfortable when AI apps are either too autonomous or too dictatorial. To make consumers happy, apps should be neither too independent nor too manipulative, but just right. The industry and consumers have been galvanized by the disruptive challenges created by COVID-19. Out of these crises, a new consumer segment called Generation N has emerged. According to Solis (2020), members of Generation N, or Novel, are tech-savvy digital-centric consumers who have emerged from the fear and anxiety created by the novel coronavirus. Companies and business and marketing executives must prioritize studying and understanding the behavior of Gen N, which is poised to increase exponentially, as the pandemic has accelerated digital behavior among those consumers and prospects who were previously either slow or unmotivated to adopt and use digital products and services. Moreover, the current pandemic has created complex challenges and uncertainties for businesses and government organizations, including regulators and policymakers. AI apps could play a greater role, and AI-based technology and business models could provide products and services in new contactless ways. Regulators and policymakers should also address their citizens’ growing concerns, fear, and lack of trust in AI apps. According to the IFC (2020), a sister organization of the World Bank, policymakers and regulators should take necessary steps to mitigate these concerns, promote responsible stewardship of AI apps, and develop good practices, especially with regard to data protection. 6.3 Limitations and future research directions Future research should address some limitations associated with the present study. For example, first, only including two variables when expanding the UTAUT2 model (or in this case, to help us segment the database) could have caused bias because the effect of other possible constructs, such as perceived risk, resistance to use, or the conditions of privacy, were not considered. Second, it seems necessary to explore new moderator variables other than those of the original UTAUT2 not only to expand the model but also to help segmentations, with the aim of evaluating possible new effects not previously contemplated. New moderators can enable us to establish differences in the behavior of consumers and set up possible new market segments. Third, although the observations were collected via online questionnaires, we could not avoid the biases of age (very young) and educational level (mostly university students). Fourth, the use of AI apps is evident in developed countries and slowly gaining popularity in emerging and developing countries as well. Future research may include empirical studies in emerging country contexts and cross-country assessments comparing developed countries with emerging countries. This would provide new insights into consumer behavioral intentions and usage behavior concerning AI apps, devices, and services. Fifth, COVID-19 and other health crises have disrupted many industries and services. Future research should examine how AI could play a role in mitigating these challenges and bringing the world back to normal, especially when there is no pandemic playbook to follow. References Ajzen, I., 1991. The theory of planned behavior. Orgnizational Behavior and Human Decision Processes, 50, 179–211. Ajzen, I., 2002. Residual Effects of Past on Later Behavior. Personality and Social Psychology Review, 6 (2), 107–122. Alalwan, A. A., Dwivedi, Y. K., Rana, N. P., 2017. Factors influencing adoption of mobile banking by Jordanian bank customers: Extending UTAUT2 with trust. International Journal of Information Management, 37 (3), 99–110. Alalwan, A. A., Dwivedi, Y. K., Rana, N. P., and Algharabat, R., 2018. Examining factors influencing Jordanian customers’ intentions and adoption of internet banking: Extending UTAUT2 with risk. Journal of Retailing and Consumer Services, 40, 125–138. Al-Gahtani, S. S., Hubona, G. S., and Wang, J., 2007. Information technology (IT) in Saudi Arabia: Culture and the acceptance and use of IT. Information and Management, 44 (8), 681– 691. Ameen, N., Willis, R., Shah, M. H., 2018. An examination of the gender gap in smartphone adoption and use in Arab countries: A cross-national study. Computers in Human Behavior, 89, 148–162. Androutsopoulou, A., Karacapilidis, N., Loukis, E., Charalabidis, Y., 2019. Transforming the communication between citizens and government through AI-guided chatbots. Government Information Quarterly, 36 (2), 358–367. Arenas-Gaitán, J., Villarejo-Ramos, Á.F., Peral-Peral, B., 2019. A posteriori segmentation of elderly internet users: Appling PLS-POS. Marketing Intelligence and Planning. https://doi.org/10.1108/MIP-01-2019-0057 (In-press). Arm, 2020. AI today, AI tomorrow. Retrieved on September 15, 2020 from https://www.arm.com/resources/report/ai-today-ai-tomorrow. Baumann, C., Burton, S., Elliott, G., Kehr, H. M., 2007. Prediction of attitude and behavioural intentions in retail banking. International Journal of Bank Marketing, 25 (2), 102–116. Becker, J.-M., Rai, A., Ringle, C. M., Völckner, F., 2013. Discovering Unobserved Heterogeneity in Structural Equation Models to Avert Validity Threats. MIS Quarterly, 37 (3), 665–694. Beza, E., Reidsma, P., Poortvliet, P. M., Belay, M. M., Bijen, B. S., Kooistra, L., 2018. Exploring farmers’ intentions to adopt mobile Short Message Service (SMS) for citizen science in agriculture. Computers and Electronics in Agriculture, 151, 295–310. Bock, G.-W., Zmud, R. W., Kim, Y.-G., Lee, J.-N., 2005. Behavioral Intention Formation in Knowledge Sharing : Examining the Roles of Extrinsic Motivators. Social-Psychological Forces, and Organizational Climate, 29 (1), 87–111. Brown, S. Venkatesh, V., 2005. Model of Adoption of Technology in Households: A Baseline Model Test and Extension Incorporating Household Life Cycle. MIS Quarterly, 29 (3), 399–436. Cabrera-Sánchez, J.-P., Villarejo-Ramos, Á.F., 2018. Extendiendo el Modelo UTAUT para evaluar los factores que afectan la adopción del Big Data en empresas españolas. En J.A. Trespalacios, R. Vázquez, E. Estrada y C. González (Coord.) Nuevos horizontes del marketing y de la distribución comercial. Ed. Cátedra Fundación Ramón Areces de Distribución Comercial, (pp.181–200). Casey, T., Wilson-Evered, E., 2012. Predicting uptake of technology innovations in online family dispute resolution services: An application and extension of the UTAUT. Computers in Human Behavior, 28 (6), 2034–2045. Chauhan, S., Jaiswal, M., 2016. Determinants of acceptance of ERP software training in business schools: Empirical investigation using UTAUT model. International Journal of Management Education, 14 (3), 248–262. Chauhan, S., Jaiswal, M., Kar, A. K., 2018. The acceptance of electronic voting machines in India: a UTAUT approach. Electronic Government, an International Journal, 14 (3), 255–275. Chin, W. W., Dibbern, J., 2010. An Introduction to a Permutation Based Procedure for MultiGroup PLS Analysis: Results of Tests of Differences on Simulated Data and a Cross Cultural Analysis of the Sourcing of Information System Services Between Germany and the USA. In V. E. Vinzi, W. Chin, J. Henseler, H. Wang (eds.) Handbook of Partial Least Squares. Concepts, Methods and Applications. Springer, (pp.171–193). Churamani, N., Anton, P., Brügger, M., Fließwasser, E., Hummel, T., Mayer, J., … Wermter, S., 2017. The Impact of Personalisation on Human-Robot Interaction in Learning Scenarios. Proceedings of the Fifth International Conference on Human Agent Interaction (HAI), accepted. Davis, F., 1985. A Technology Acceptance Model for Empirically Testing New End-User Information Systems. Massachusetts Institute of Technology, (December 1985), 291. Davis, F., 1989. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13 (3), 319–340. Duan, Y., Edwards, J. S., Dwivedi, Y. K., 2019. Artificial intelligence for decision making in the era of Big Data–evolution, challenges and research agenda. International Journal of Information Management, 48, 63–71. Duyck, P., Pynoo, B., Devolder, P., Voet, T., Adang, L., Ovaere, D., Vercruysse, J., 2010. Monitoring the PACS implementation process in a large university hospital-discrepancies between radiologists and physicians. Journal of Digital Imaging, 23 (1), 73–80. 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): Towards a revised theoretical model. Information Systems Frontiers, 21 (3), 719–734. Falk, R. F., Miller, N. B., 1992. A Primer for Soft Modeling. University of Akron Press. Akron, Ohio (USA). Fan, W., Liu, J., Zhu, S., Pardalos, P. M., 2018. Investigating the impacting factors for the healthcare professionals to adopt artificial intelligence-based medical diagnosis support system (AIMDSS). Annals of Operations Research, 1–26. Fishbein, M., Ajzen, I., 1975. Belief, attitude, intention, and behavior : an introduction to theory and research. Philosophy & Rhetoric (Vol. 10). Reading, Mass.: Addison-Wesley. Fuentes-Blasco, M., Moliner-Velázquez, B., Gil-Saura, I., 2017. Exploring relationship variables and Information and Communication Technologies use in industrial segmentation. Management Decision, 55 (7), 1441–1459. Gabriel, A.M., Goertzel, B., 2019. Distributed, decentralized, and democratized artificial intelligence. Technological Forecasting and Social Change, 141, 354–358. García, A. S., Fernández-Sotos, P., Fernández-Caballero, A., Navarro, E., Latorre, J. M., Rodriguez-Jimenez, R., González, P., 2019. Acceptance and use of a multi-modal avatarbased tool for remediation of social cognition deficits. Journal of Ambient Intelligence and Humanized Computing, 1–12. Gartner, 2019. Top 10 Strategic Technology Trends for 2020. Retrieved on April 10, 2020 from https://www.gartner.com/smarterwithgartner/gartner-top-10-strategic-technologytrends-for-2020/ Gelbrich, K., Sattler, B., 2014. Anxiety, crowding, and time pressure in public self-service technology acceptance. Journal of Services Marketing, 28 (1), 82–94. Gharaibeh, M. K., Arshad, M. R. M., 2018. Determinants of intention to use mobile banking in the North of Jordan: extending UTAUT2 with mass media and trust. Journal of Engineering and Applied Sciences, 13 (8), 2023–2033. Guo, X., Sun, Y., Wang, N., Peng, Z., Yan, Z., 2013. The dark side of elderly acceptance of preventive mobile health services in China. Electronic Markets, 23 (1), 49–61. Gupta, A., Dogra, N., 2017. Tourist Adoption of Mapping Apps: A UTAUT2 Perspective of Smart Travellers. Tourism and Hospitality Management, 23 (2), 145–161. Guriting, P., Oly Ndubisi, N., 2006. Borneo online banking: evaluating customer perceptions and behavioural intention. Management Research News, 29 (1/2), 6–15. Hair, J. F., Sarstedt, M., Ringle, C. M., Mena, J. A., 2012. An assessment of the use of partial least squares structural equation modeling in marketing research. Journal of the Academy of Marketing Science, 40 (3), 414–433. Heinssen Jr., R. K., Glass, C. R., Knight, L. A., 1987. Assessing computer anxiety: Development and validation of the Computer Anxiety Rating Scale. Computers in Human Behavior, 3 (1), 49–59. Henseler, J., Ringle, C. M., Sarstedt, M., 2014. A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43 (1), 115–135. Herrero, Á, San Martín, H., 2017. Explaining the adoption of social networks sites for sharing user-generated content: A revision of the UTAUT2. Computers in Human Behavior, 71, 209– 217. Ho, C. T. B., Chou, Y. T., O'Neill, P., 2010. Technology adoption of mobile learning: A study of podcasting. International Journal of Mobile Communications, 8 (4), 468–485. Holzmann, P., Schwarz, E. J., Audretsch, D. B., 2018. Understanding the determinants of novel technology adoption among teachers : the case of 3D printing. The Journal of Technology Transfer, 1–17. Hoy, M. B., 2018. Alexa, Siri, Cortana, and more: an introduction to voice assistants. Medical Reference Services Quarterly, 37 (1), 81–88. 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. International Finance Corporation, 2020. Artificial intelligence in emerging markets. Retrieved on September 15 from https://www.ifc.org/wps/wcm/connect/95a40480-27b54b99-8b4c-7768ae6a53a2/AI-Report_Web.pdf?MOD=AJPERES&CVID=nhLrsRc Juaneda-Ayensa, E., Mosquera, A., Sierra Murillo, Y., 2016. Omnichannel customer behavior: key drivers of technology acceptance and use and their effects on purchase intention. Frontiers in Psychology, 7, 1–11. Kaplan, A., Haenlein, M., 2019. Siri, Siri, in my hand: Who’s the fairest in the land? On the interpretations, illustrations, and implications of artificial intelligence. Business Horizons, 62 (1), 15–25. Karjaluoto, H., Shaikh, A. A., Saarijärvi, H., Saraniemi, S., 2019. How perceived value drives the use of mobile financial services apps. International Journal of Information Management, 47, 252–261. Khasawneh, O. Y., 2018. Technophobia without boarders: The influence of technophobia and emotional intelligence on technology acceptance and the moderating influence of organizational climate. Computers in Human Behavior, 88, 210–218. Kim, D., Park, J., Morrison, A. M., Management, R., Sciences, C., Sciences, F., 2008. A Model of Traveller Acceptance of Mobile Technology, International Journal of Tourism Research, 10 (5), 393–407. Kim, H. W., Chan, H. C., Gupta, S., 2007. Value-based Adoption of Mobile Internet: An empirical investigation. Decision Support Systems, 43 (1), 111–126. Kock, N., 2015. Common Method Bias in PLS-SEM: A Full Collinearity Assesment Approach. International Journal of E-Collaboration, 11 (4), 1–10. Kock, N., Lynn, G. S., 2012. Lateral Collinearity and Misleading Results in Variance-Based SEM : An Illustration and Recommendations. Journal of the Association for Information Systems, 13 (7), 546–580. Kwon, O., Lee, N., Shin, B., 2014. Data quality management, data usage experience and acquisition intention of big data analytics. International Journal of Information Management, 34 (3), 387–394. LeCun, Y., Bengio, Y., Hinton, G., 2015. Deep learning. Nature, 521(7553), 436–444. Lee, J.-H., Song, C., 2013. Effects of trust and perceived risk on user acceptance of a new technology service. Social Behavior and Personality: An International Journal, 41 (3), 587– 598. Liao, Y. W., Liu, C. G., Lo, C. L., Su, Z. Y., Wang, C. S., 2018. The Intention to Use the APPs of Smartphone: An Integration of Individual Differences and Browsing Experiences Perspective. Journal of Internet Technology, 19 (2), 439–449. Limayem, M., Hirt, S. G., Cheung, C. M. K., 2007. How habit limits the predictive power of intention: the case of information systems continuance. MIS Quarterly, 31 (4), 705–737. Liu, X., 2019. A big data approach to examining social bots on Twitter. Journal of Services Marketing, 33 (4), 369–379. Liu, J., Chang, H., Forrest, J. Y. L., Yang, B., 2020. Influence of artificial intelligence on technological innovation: Evidence from the panel data of china's manufacturing sectors. Technological Forecasting and Social Change, 158, 1–11. Luarn, P., Lin, H. H., 2005. Toward an understanding of the behavioral intention to use mobile banking. Computers in Human Behavior, 21 (6), 873–891. Magsamen-Conrad, K., Upadhyaya, S., Joa, C. Y., Dowd, J., 2015. Bridging the divide: Using UTAUT to predict multigenerational tablet adoption practices. Computers in Human Behavior, 50, 186–196. Martínez-Córcoles, M., Teichmann, M., Murdvee, M., 2017. Assessing technophobia and technophilia: Development and validation of a questionnaire. Technology in Society, 51, 183–188. Martins, C., Oliveira, T., Popovič, A., 2014. Understanding the internet banking adoption: A unified theory of acceptance and use of technology and perceived risk application. International Journal of Information Management, 34 (1), 1–13. Merhi, M., Hone, K., Tarhini, A., 2019. A cross-cultural study of the intention to use mobile banking between Lebanese and British consumers: Extending UTAUT2 with security, privacy and trust. Technology in Society, 59, 1–12. Moon, J.-W., Kim, Y.-G., 2001. Extending the TAM for a World-Wide-Web context. Information & Management, 38 (4), 217–230. Moorman, C., Deshpande, R., Zaltman, G., 1993. Factors affecting trust in market research relationships. Journal of Marketing, 57 (1), 81–101. Morosan, C., DeFranco, A., 2016. It's about time: Revisiting UTAUT2 to examine consumers’ intentions to use NFC mobile payments in hotels. International Journal of Hospitality Management, 53, 17–29. Mosunmola, A., Mayowa, A., Okuboyejo, S., Adeniji, C., 2018. Adoption and use of mobile learning in higher education: the UTAUT model. In Proceedings of the 9th International Conference on E-Education, E-Business, E-Management and E-Learning (pp. 20–25). ACM. Muhammad, S. S., Dey, B. L., Weerakkody, V., 2018. Analysis of factors that influence customers’ willingness to leave big data digital footprints on social media: A systematic review of literature. Information Systems Frontiers, 20 (3), 559–576. Muñoz-Leiva, F., Climent-Climent, S., Liébana-Cabanillas, F., 2017. Determinants of intention to use the mobile banking apps: An extension of the classic TAM model. Spanish Journal of Marketing-ESIC, 21(1), 25–38. Niemelä‐Nyrhinen, J., 2007. Baby boom consumers and technology: shooting down stereotypes. Journal of Consumer Marketing, 24 (5), 305–312. Nunnally, J. C., 1978. Psychometric Theory (2nd ed.). McGrawHill College. Nysveen, H., 2005. Intentions to Use Mobile Services: Antecedents and Cross-Service Comparisons. Journal of the Academy of Marketing Science, 33 (3), 330–346. Oikonomidis, T., Fouskas, K., 2019. A Literature Review of Smartphone Geolocation Technologies. In Strategic Innovative Marketing and Tourism (pp. 1085–1092). Springer. Oliveira, T., Faria, M., Thomas, M. A., Popovič, A., 2014. Extending the understanding of mobile banking adoption: When UTAUT meets TTF and ITM. International Journal of Information Management, 34 (5), 689–703. Paavilainen J., Alha K., Korhonen H., 2012. Exploring Playability of Social Network Games. In: Nijholt A., Romão T., Reidsma D. (eds) Advances in Computer Entertainment. ACE 2012. Lecture Notes in Computer Science, vol 7624. Springer, Berlin, Heidelberg List of Tables Table 1: VIF from all variables to check CMB Variables variable_CMB Behavioral Intention (BI) 1,940 Effort Expectancy (EE) 1,907 Facilitating Conditions (FC) 1,906 Habit (HT) 1,372 Hedonic Motivation (HM) 1,887 Performance Expectancy (PE) 1,945 Price-value (PV) 1,394 Social Influence (SI) 1,483 Technology Fear (TF) 1,264 Consumer Trust TR) 1,326 Use Behavior UB) 1,254 Table 2: Composite reliability and convergent validity Cronbach’s alpha rho_A Composite reliability Average variance extracted (AVE) Behavioral Intention 0,899 0,900 0,937 0,832 Effort Expectancy 0,926 0,932 0,948 0,819 Facilitating Conditions 0,809 0,819 0,875 0,638 Habit 0,864 0,873 0,908 0,711 Hedonic Motivation 0,937 0,938 0,960 0,888 Performance Expectancy 0,856 0,858 0,903 0,699 Price-value 0,866 0,887 0,918 0,789 Social Influence 0,935 0,939 0,958 0,885 Technology Fear 0,845 0,963 0,891 0,672 Consumer Trust 0,725 0,744 0,842 0,641 Use Behavior 1,000 Table 3: Discriminant Validity (Ratio Heterotrait-Monotrait -HTMT) BI EE FC HT HM PE PV SI TF TR BI EE 0,436 FC 0,433 0,734 HT 0,469 0,366 0,394 HM 0,604 0,483 0,499 0,417 PE 0,649 0,390 0,343 0,457 0,607 PV 0,443 0,328 0,462 0,342 0,442 0,402 SI 0,441 0,276 0,220 0,346 0,424 0,574 0,324 TF 0,232 0,410 0,439 0,131 0,207 0,134 0,142 0,062 TR 0,478 0,263 0,373 0,315 0,423 0,462 0,446 0,326 0,199 Table 4: R2 of the model R2 Adjusted R2 Behavioral Intention 0,473 0,467 Use Behavior 0,182 0,179 Table 5: Structural Model Estimates (Path Coefficients) Original Sample P-values H1. Performance Expectancy → Behavioral Intention 0,261*** 0,000 H2. Effort Expectancy → Behavioral Intention 0,064 (ns) 0,095 H3. Social Influence → Behavioral Intention 0,081 * 0,020 H4. Hedonic Motivation → Behavioral Intention 0,211 *** 0,000 H5. Price Value → Behavioral Intention 0,085 * 0,021 H6a. Habit → Behavioral Intention 0,118 ** 0,001 H6b. Habit → Use Behavior 0,146 *** 0,000 H7a. Facilitating Conditions → Behavioral Intention 0,010 (ns) 0,778 H7b. Facilitating Conditions → Use Behavior 0,090 * 0,023 H8. Technology Fear → Behavioral Intention -0,084 * 0,010 H9 Consumer Trust → Behavioral Intention 0,110 ** 0,002 H10. Behavioural Intention → Usage Beahaviour 0,297 *** 0,000 ***p<0.001, **p<0.01, *p<0.05. (It based in 1-tail test and Bootstrap with 10.000 sub-samples). (ns) Non-significant. Table 6: Multigroup Analysis with gender Path Coefficients -dif. (men - women) New p-value (men vs women) Behavioral Intention → Use Behavior -0,045 0,573 Effort Expectancy → Behavioral Intention -0,110 0,158 Facilitating Conditions → Behavioral Intention -0,058 0,449 Facilitating Conditions → Usage Beahaviour 0,236 0,006 Habit → Behavioural Intention 0,087 0,205 Habit → Usage Beahaviour -0,069 0,413 Hedonic Motivation → Behavioral Intention -0,044 0,608 Performance Expectancy → Behavioural Intention 0,133 0,111 Price Value → Behavioural Intention -0,034 0,637 Social Influence → Behavioral Intention 0,023 0,743 Technology Fear → Behavioral Intention -0,100 0,101 Consumer Trust → Behavioral Intention -0,019 0,796 Table 7: Multigroup Analysis with Income Path Coefficients - dif. (<1800€- <1800 €) New p-value (<1800€ vs <1800 €) Behavioral Intention → Use Behavior -0,011 0,900 Effort Expectancy → Behavioral Intention 0,004 0,955 Facilitating Conditions → Behavioral Intention 0,092 0,252 Facilitating Conditions → Use Behavior -0,151 0,122 Habit → Behavioral Intention 0,000 0,991 Habit → Usage Behavior -0,028 0,765 Hedonic Motivation → Behavioral Intention -0,229 0,008 Performance Expectancy → Behavioral Intention 0,088 0,294 Price-value → Behavioral Intention -0,080 0,274 Social Influence → Behavioral Intention 0,113 0,125 Technology Fear → Behavioral Intention -0,046 0,483 Consumer Trust → Behavioral Intention 0,017 0,819 Table 8: Indirect effects of TF and Trust in Usage Path P-Values H11a Technology Fear → Behavioral Intention → Use Behavior -0,025 * 0,010 H11b Consumer Trust → Behavioral Intention → Use Behavior 0,033 ** 0,002 ***p<0.001, **p<0.01, *p<0.05. (It based in 1-tail test and Bootstrap with 10.000 sub-samples). Table 9: Average R2 R squared (by number of segments) Original sample (without segments) 2 segments 3 segments 4 segments 5 segments 6 or more Behavioural Intention 0.454 0.483 0.493 Error 0.735 Error Usage Behaviour 0.178 0.195 0.275 Error 0.602 Error Average R 2 0.316 0.339 0.384 Error 0.669 Error Table 10: Path coefficients and p-values (original model vs segmented model) Original Seg 1 (28) Seg 2 (211) Seg 3 (156) Seg 4 (263) Seg 5 (82) PE → BI 0.286 -0.849 0.002 0.404 0.000 0.450 0.000 0.274 0.000 -0.120 0.044 EE → BI 0.084 0.383 0.011 -0.212 0.000 -0.134 0.031 0.532 0.000 0.014 0.417 SI → BI 0.088 0.006 0.490 -0.171 0.000 0.163 0.009 0.257 0.000 0.341 0.000 HM → BI 0.230 0.753 0.000 0.431 0.000 -0.106 0.007 0.084 0.037 0.950 0.000 PV → BI 0.107 -0.147 0.185 0.062 0.073 0.554 0.000 -0.241 0.000 0.056 0.130 HT → BI 0.092 0.604 0.000 -0.044 0.165 0.053 0.205 0.184 0.000 -0.061 0.156 HT → UB 0.098 0.537 0.000 -0.145 0.001 0.468 0.000 -0.058 0.131 0.192 0.000 FC → BI 0.055 0.262 0.063 0.433 0.000 -0.189 0.002 -0.043 0.225 -0.319 0.002 FC → UB 0.103 0.106 0.047 -0.777 0.000 0.427 0.000 0.758 0.000 -0.233 0.001 Bold: significant path. Table 11: R2 of endogenous variables R squared original sample POS-Seg 1 POS-Seg 2 POS-Seg 3 POS-Seg 4 POS-Seg 5 Behavioral Intention 0.454 0.979 0.719 0.715 0.694 0.861 Use Behavior 0.178 0.979 0.632 0.673 0.408 0.880 Table 12: ANOVA and p-values of every variable and segment Significant differences between segments Not significant differences between segments Variable Segment N Mean Significance Variable Segment N Mean Significance Behavioral Intention 1 28 - 0.0412 0.009 Facilitating Conditions 1 28 0.1708 0.076 2 211 0.0949 2 211 0.0597 3 156 - 0.2402 3 156 - 0.0727 4 263 0.0914 4 263 0.0560 5 82 - 0.0661 5 82 - 0.2528 Total 740 0.0000 Total 740 0.0000 Effort Expectancy 1 28 0.0247 0.010 Habit 1 28 - 0.0427 0.102 2 211 0.0618 2 211 - 0.0008 3 156 - 0.1828 3 156 - 0.1793 4 263 0.1203 4 263 0.0726 5 82 - 0.2053 5 82 0.1244 Total 740 0.0000 Total 740 - 0.0001 Hedonic Motivation 1 28 0.3299 0.013 Performance Expectancy 1 28 - 0.0363 0.145 2 211 0.1007 2 211 0.0637 3 156 - 0.1813 3 156 - 0.1258 4 263 0.0400 4 263 0.0757 5 82 - 0.1554 5 82 - 0.1548 Total 740 0.0000 Total 740 0.0000 Price-value 1 28 0.0429 0.029 Social Influence 1 28 0.2306 0.178 2 211 - 0.0249 2 211 - 0.0258 3 156 - 0.1200 3 156 - 0.1189 4 263 0.1444 4 263 0.0887 5 82 - 0.1859 5 82 - 0.0700 Total 740 - 0.0001 Total 740 0.0000 Technology Fear 1 28 - 0.3162 0.027 Usage Behaviour 1 28 - 0.1058 0.201 2 211 - 0.0811 2 211 0.0232 3 156 0.2002 3 156 - 0.1553 4 263 - 0.0357 4 263 0.0468 5 82 0.0501 5 82 0.1220 Total 740 0.0000 Total 740 0.0000 Consumer Trust 1 28 0.1068 0.041 2 211 0.0099 3 156 - 0.1281 4 263 0.1195 5 82 - 0.2013 Total 740 0.0000 Table 13: Segments obtained in the POS-PLS latent class segmentation N 28 211 156 263 82 Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 BI - 0.0412 0.0949 - 0.2402 0.0914 - 0.0661 EE 0.0247 0.0618 - 0.1828 0.1203 -0.2053 HM 0.3299 0.1007 - 0.1813 0.0400 - 0.1554 PV 0.0429 -0.0249 -0.1200 0.1444 -0.1859 TF -0.3162 - 0.0811 0.2002 - 0.0357 0.0501 TR 0.1068 0.0099 -0.1281 0.1195 - 0.2013 - 37 - ANNEX I: Measurement Scales Effort expectancy EE1: I find it easy to learn to use AI tools. EE2: My interaction with AI tools is clear. EE3: I find it easy to use AI. EE4: I believe that learning to use an AI application is easy for me. Performance expectancy PE1: I believe that AI is useful for me in my day - to - day life. PE2: I believe that AI will help me achieve things that are important to me. increase. PE3: I believe that AI helps me carry out my tasks quickly. PE4: I believe that AI improves my performance. Social influence SI1: People who I care about think I should use AI applications. SI2: People who influence my behavior think that I should use AI. applications. SI3: People whose opinion I value believe that I should use AI apps. Hedonic motivations HM1: Using AI applications is fun. HM2: I enjoy using AI applications. HM3: Using AI applications is very entertaining. Price value PV1: AI applications are reasonably priced. PV2: AI applications are worth what they cost. PV3: At the current price, AI gives good value. Habit HT1: The use of AI has become a habit for me. HT2: I am an AI addict. HT3: I must use AI applications. HT4: Using AI has become something natural for me. Facilitating condition FC1: I have the necessary resources to use AI. FC2: I have the necessary knowledge to use AI applications. FC3: AI is compatible with other applications that I use. FC4: When I have trouble using AI applications, I can get help. Behavioral intention BI1: I intend to use AI applications soon. BI2: I will always try to use AI applications in my daily life. BI3: I plan to use AI applications frequently. Use behavior UB1 Maps: What is your current use of the maps and routes? UB2 Recommendations: What is your current use of the recommendations? systems UB3 Voice: What is your current use of voice recognition? Technology fear - 38 - Consumer trust