The Adoption and Use of Smart Assistants in Residential Homes: The Matching Hypothesis

Abstract

An increasing number of residential homes are equipped with smart assistants such as Cortana, Alexa, and Siri. Adoption rates and the frequency of the usage of smart assistants vary across users and residential homes. Building on the theory of uses and gratifications (UGT) and the unified theory of acceptance and use of technology 2 (UTAUT2), the objective of this paper was to examine whether the intended use of a digital assistant would moderate the effects of performance expectancy and hedonic motivation on its adoption. Two experiments (N = 345 and N = 351) tested the hypothesis that, for utilitarian purposes, devices with high performance appraisal are preferred, whereas for entertainment purposes, devices with high hedonic appraisal are preferred. The experiments manipulated the performance expectancy and hedonic motivation towards several digital assistants by varying how the assistants were introduced. Participants were asked which assistant they would choose for a variety of utilitarian and entertainment purposes. As expected, the experiments supported the proposed matching hypothesis, revealing that the devices that were high in performance appraisal were preferred for utilitarian tasks, whereas the devices high in hedonic appraisal were preferred for entertainment needs. These results suggest that a device’s introduction can change people’s perceptions of the device and subsequently their decision to use it.

1. Introduction

The technology-heavy societies of the 21st Century offer consumers an abundance of choices of new technologies. Annual patent applications for new technologies exceed three million every year, with over a million of those applications being granted [1]. Over 330,000 of the patents granted in 2020 were related to communication or computer technologies. The constant influx of new technology comes with continuous challenging decisions for consumers having to decide which technologies to use and which to ignore.

Voice-controlled digital assistants such as Alexa [2], Siri [3], and Google Assistant [4] are becoming more and more a part of residents’ daily lives. Voicebot.ai reported in 2021 that nearly half of the U.S. adult population has used voice assistants while driving [5]. Consumers use smart speakers for many different activities, including reading the news and weather, playing music and games, setting a timer, communication with others, and sports [6]. From a sustainability perspective, voice assistants provide an important tool to help consumers with energy consumption decisions by allowing them to control the heating and cooling of their homes, which has the potential to decrease energy usage by changing the temperature settings to more efficient schedules [7,8], digitally empowering citizens to have greater control [9]. Programmable thermostats can similarly alter temperature settings on an efficient schedule, but many consumers do not utilize the programmability of such thermostats [10], due to a lack of understanding about the potential for saving energy [11] and fear of misusing or damaging the device [12]. It can also be difficult for people with disabilities to manually set thermostats, while voice-controlled thermostats through digital assistants such as Alexa can increase accessibility [13]. Kim et al. [7] conducted a longitudinal study with residents in a multi-unit residential community, in which they introduced an energy-saving program using a modeling approach for personalized eco-feedback. Personalized feedback was provided through visual (wall-mounted tablet) and voice (Alexa) user interfaces.

Kim et al. [7] observed an increase in the indoor temperature during the cooling season, suggesting that people were spending less energy cooling their homes when it was warm outside, compared to a baseline level of cooling. This speaks to the potential of the energy-saving program to increase sustainable behavior. The authors recorded not only residents’ thermostat choices, but also their interactions and experiences with Alexa. This data revealed that there was variation (in particular, at the beginning of the intervention) in the adoption and use of the provided devices. Residents differed not only in the frequency in which they used the devices, but also in what they used them for. While some residents used the devices predominantly for controlling energy consumption, others relegated the same devices as a kids’ toy, with many occupants using the devices for several purposes [7]. This variation raises the question of whether the perceptions that a person has about smart assistants influence adoption, dependent on how a person would plan to use such a device. Specifically, we sought to shed some light on this variation in the usage of smart assistants by exploring if it makes a difference whether the introduction of a digital assistant stresses the benefits of using the assistant when getting a task done (increasing performance expectancy) or the fun and pleasure that is derived from using it (increasing hedonic motivation). We reasoned that the way in which a voice-controlled digital assistant is introduced may affect residents’ perceptions of how useful or entertaining the assistant is, which may subsequently affect what devices they prefer when choosing a device for different purposes. Understanding when and why residents adopt and use provided smart devices is important as the technology can only support sustainable behavior through feedback to the extent that it is adopted and used by consumers.

In the remainder, research on performance expectancy and hedonic motivation from the UTAUT2 model is reviewed first. Second, drawing on uses and gratifications theory (UGT), the hypothesis is introduced that the role performance expectancy and hedonic motivation have in technology adoption depends on the intended use of a technology. In the third step, two experimental studies are described that tested the proposed matching hypothesis. The implications of the studies for the adoption of smart assistants in residential homes are discussed.

1.1. Performance Expectancy and Hedonic Motivation

UTAUT2 [14] aims to explain the adoption of technology through a series of predictors such as facilitating conditions, performance expectancy, hedonic motivation, or habit. This model was developed as an extension of UTAUT [15], adapting the earlier model to a consumer context. Both models have received substantial attention in the literature. Two constructs of the UTAUT2 model are particularly important in our context: performance expectancy and hedonic motivation. Performance expectancy refers to the extent to which a technology is perceived to benefit a consumer in performing various activities, while hedonic motivation is the fun or pleasure that is derived from using a technology [14]. Deviating from the tradition of previous research that measured users’ attitudes towards a technology often with a single measure (e.g., “I like this technology”; see [16]), the two constructs of performance expectancy and hedonic motivation include cognitive, as well as affective dimensions.

Both constructs, performance expectancy and hedonic motivation, have been identified to provide significant predictors of technology adoption in various studies [17,18,19]. However, several studies using these two constructs reported only one of the two being a significant predictor of technology adoption [20,21]. The predictive power of the two constructs is inconclusive as it greatly varies across studies.

1.2. Intended Use as a Moderator

We aimed to test as an explanation for these inconclusive findings if a person’s intended use of a digital assistant—the way in which they plan to use the device—moderates the effects of performance expectancy and hedonic motivation on the adoption of the device.

This possible explanation is primarily rooted in uses and gratifications theory (UGT) [22]. Katz et al. argued that people vary in their reasons to use media and technology because they vary in the needs they have to gratify. The theory predicts that when people come to media, they are motivated to do so by their needs. Recent examples of research on UGT have shown that people are motivated to use streaming services to gain emotional gratification through information and entertainment [23], that people use food delivery apps for the sake of ease [24], and that people use Facebook for affection, attention seeking, disclosure, information sharing, habit, and social influence [25]. These examples highlight two key ideas from the theory. First, people come to media and technology with needs, such as informational and emotional needs. Second, different media have differing levels of capability to gratify those needs. Together, these two components suggest that consumers will use media that they believe will gratify the needs that they have. We suggest from this theoretical background that if perceptions of digital assistants, such as Alexa, were manipulated by an introductory message, this could alter the perceived ability of such a device to gratify certain needs and, thereby, alter consumer choices of devices, depending on what needs they are trying to gratify.

To return to performance expectancy and hedonic motivation, UGT would suggest that people’s choices of technologies that are oriented toward utility or toward entertainment would depend on what needs they are coming to the technology with. Some technologies are clearly geared toward either utility (such as information technologies or a word processor) or entertainment (such as a gaming console or a gaming app). Based on UGT, performance expectancy would be expected to be a better predictor than hedonic motivation of the adoption of utility devices, such as spreadsheet applications [26]; conversely, hedonic motivation would be expected to be a better predictor than the performance expectancy of the choice of devices for entertainment such as gaming consoles or similarly fun-oriented technologies, such as entertainment websites [27]. Digital assistants are part of a third class of technologies that may be used for both utilitarian and entertainment purposes. Alexa, for example, has been shown to be used in a wide variety of ways that include both utilitarian and hedonic uses: check weather, find facts, listen to news, control other devices, set reminders/calendars, play music, set a timer, tell a joke, play a game, and check the time [28].

For smart assistants in residential homes, performance expectancy and hedonic motivation are both potentially relevant adoption factors. We sought to find out whether perceptions of performance expectancy or hedonic motivation would lead to different adoption choices if a person has tasks in mind (e.g., controlling a thermostat) or has entertainment purposes (e.g., playing a game) in mind when choosing a device. Indirect support for this hypothesis comes from a study by McLean and Wilson [29] on consumers’ online shopping experiences. The authors examined differences between shoppers who use the site for purposeful shopping and those who are simply browsing for fun. McLean and Wilson observed that the effects of interactivity, vividness, and novelty on perceived usefulness and enjoyment, respectively, were moderated by the intended use. Although the technology itself did not change, user experience did change based on the purpose for which consumers used the shopping platform. This study illustrates that the purpose for which technology is used can affect consumer perceptions of technology.

Additional preliminary, indirect evidence for the claim that the intended technology use moderates the effects of performance expectancy and hedonic motivation on adoption comes from studies that looked at technologies that can only be used for fun or work, but not both. Van der Heijden [27] reported a study in which the usage of a pleasure-oriented website was more strongly predicted by anticipated enjoyment than by anticipated usefulness. Davis et al. [30] tested the same idea by studying the use of video games, observing that hedonic motivation significantly predicted usage, but not utilitarian motivation. Additionally, and most similar to what we propose, Köse et al. [31] proposed that perceptions of MyDriveAssist, a gamified driving assistant app, as either utilitarian or hedonic would differentially predict continuance or discontinuance of using the app, dependent on whether users thought of the app as a tool or a toy. They found some support for their claims, reporting several significant interactions in their structural equation model that suggested this type of relationship. While Köse et al. [31] examined these relationships in a correlational design for current users, we explored this with an experimental design from the perspective of prospective consumers. Building on the provided rationale, we set out to propose and test the matching hypothesis. We anticipated that the adoption of technology would not simply be dependent on whether it was perceived as having high performance expectancy or high hedonic motivation. Rather, because certain technologies may gratify different needs better than others, we anticipated that those perceived to be high in performance expectancy would be preferred when adopting technology for task purposes, while those perceived to be high in hedonic motivation would be preferred when adopting technology for entertainment purposes:

H1. 

When intended to be used for utilitarian purposes, the technology will be adopted that receives the highest performance expectancy rating.

H2. 

When intended to be used for entertainment purposes, the technology will be adopted that receives the highest hedonic motivation rating.

1.3. Overview of Current Research

The current project aimed to explain variation in the adoption of digital assistants by testing the matching hypothesis—the idea that the intended use of a digital assistant for utilitarian or entertainment purposes affects which digital assistant users prefer. Amazon’s Alexa is one such voice-controlled digital assistant, built into several Amazon products, such as the Amazon Echo and FireTV. Other similar assistants are Google’s Assistant and Apple’s Siri. Studies focusing on the adoption of digital assistants have identified several variables that are associated with the intended or actual usage of these recent technologies: relative advantage, complexity, emotions [32], performance, and price [33], as well as privacy concerns [34].

In the experiment that we conducted to test the proposed hypotheses, we introduced four hypothetically new digital assistants to participants in the form of smart speakers. For these four assistants, we described each of them in a paragraph that was designed to experimentally manipulate perceptions of performance expectancy and hedonic motivation. Specifically, two of the assistants were described in paragraphs designed to elicit high performance expectations, while the other two were designed to elicit hedonic motivation toward the devices. Following these manipulations, participants were asked to consider different situations in which to pick one of these digital assistants for each of eight different purposes that were classified as utilitarian or entertainment purposes. Beyond controlling one’s home (i.e., to control lights, security, and sound around the house), we included additional utilitarian and entertainment purposes (e.g., playing a game).

2. Materials and Methods

2.1. Sample

A total of 345 U.S.-based participants were recruited through Amazon’s MTurk to take a survey administered through Qualtrics in a convenience sample. Of the 345 participants, 17 responses were removed for failing to pass reverse-coding attention checks, leaving 328 usable responses. Of the respondents, 185 indicated that they were male, while 141 indicated that they were female and 2 chose not to indicate gender. Participants’ ages ranged from 22 to 70 years (M = 36.83, SD = 10.54). Most (80%) participants indicated that they had at least a Bachelor’s degree.

2.2. Procedure and Design

Participants were presented with a survey that introduced to them four supposedly new personal digital assistants via paragraph descriptions of each device. Each device was presented as a named device (e.g., Alice or Mason), and these names were systematically varied to prevent name bias from confounding the results. Two of the device descriptions were designed to trigger higher performance expectancy, and two were designed to trigger higher hedonic motivation. We refer to the high-performance-expectancy devices as performance devices and the high-hedonic-motivation devices as hedonic devices.

The performance device descriptions stressed relevant correlates of performance expectancy, namely perceived value [35] and job performance [36], along with language including cognitive attitude word pairs from Crites et al. [37]. Specifically, the first device in the performance condition was depicted as being simple, easy to use, and useful, while the second of these devices was described as high-tech, highly adaptable, and competent. Conversely, the first hedonic device was described as smooth and natural and with an engaging personality and the second hedonic device as intriguing, exciting, and entertaining [37,38,39]. Across all four of these descriptions, several key features were held constant. Each description specified the manufacturing location, availability in stores and online, WiFi and Bluetooth capability, the inclusion of a surge-protected power cord and setup instructions, and remote control through a companion app. In addition, each paragraph clearly described that each device was able to listen for its name, answer questions, remember things, and respond to tasks. In these ways, we ensured that there was a constant baseline expectation of device capability across all four devices, to counter the perception that one device would be unable to fulfill certain tasks.

Following the introduction of these devices, participants were asked to make a series of preference choices among the devices in a pairwise comparison, where each of the four devices was presented along with each of the other devices for a total of six comparisons. Within each of these choice sets, each participant chose between the two options eight times, once for each of eight different purposes for which the devices might be used. This study used a within-subject design where all participants read all four descriptions in a randomized order and made all choice comparisons in a randomized order. The eight purposes were comprised of four utilitarian purposes and four fun purposes, presented below. Thus, the independent variables were device type and purpose. As the dependent variable, we measured which device participants would choose for each of the tasks. In addition, participants’ experiences with digital assistants in general were measured.

2.3. Measurements

Digital assistant experience: Two items were adopted from the Pew Research Center [40], asking participants to indicate their experience level with digital assistants. A yes-or-no question asked whether they have used an assistant before, and the second item, “How often do you use digital assistants, such as Apple’s Siri, Amazon’s Alexa, or Google’s Assistant?” measured participants’ level of usage of the devices on a six-point scale from “Not at all” to “Several times a day”.

Performance expectancy: Performance expectancy was measured with items adapted from Venkatesh et al. [14]. These items were measured on a seven-point Likert scale from strongly agree (7) to strongly disagree (1), with neither agree nor disagree (4) in the middle. The three items were “I would find [device name] useful in my daily life”, “Using [device name] would help me accomplish things more quickly”, and “Using [device name] would increase my productivity”. These items obtained sufficient reliability for all four descriptions (Cronbach’s αs ≥ 0.88).

Hedonic motivation: Hedonic motivation was measured with items adapted from Venkatesh et al. [14]. These items were on a seven-point Likert scale varying from strongly agree (7) to strongly disagree (1), with neither agree nor disagree (4) in the middle. The three items were “Using [device name] would be fun”, “Using [device name] would be enjoyable”, and “Using [device name] would be very entertaining”. These items yielded high reliabilities for each of the four descriptions (Cronbach’s αs ≥ 0.89).

Adoption of the devices: To examine the choices that participants made in using the devices, eight unique general purposes for the devices were presented to the participants. There were two utilitarian purposes related to controlling one’s home: (1) to control lights, security, and sound around the house and (2) practical help around the house such as lists, timers, alarms, and reminders. In addition, we added two other utilitarian purposes, (3) community and world engagement for news, culture, local traffic, and weather and (4) communication with friends and family. There was an equal number of fun purposes: (1) a toy for kids to play with, (2) social entertainment for guests, friends, and family, (3) personal entertainment for streaming media such as music, movies, and TV shows, and (4) companionship. For each of these eight intended uses, two of the devices were presented, between which the participants chose with the item “If I were getting a digital assistant for [intended use] I would choose _______”. All possible pairs of the four devices were presented, for a total of six pair comparisons. The results are reported for four pair comparisons, as two of the comparisons compared the hedonic devices or performance devices with each other, respectively.

General use of digital assistants. To check whether participants perceived that digital assistants are generally used for either utilitarian or fun purposes, the following two items were created: “Digital assistants are intended for ______”, measured on a seven-point scale from 1 (having fun) to 7 (getting tasks done). The second item asked, “Which of the following statements best describes how you use (would use) digital assistants?”, where a person chose from three options: “I would use digital assistants for [fun things/getting tasks done/fun things and getting tasks done]”.

3. Results

The results were aggregated across devices and descriptions, such that the two descriptions intended to trigger high performance expectancy were combined and the two descriptions intended to trigger high hedonic motivation were combined.

3.1. Experience and Purpose of Assistants

There was 92% of the participants who indicated that they had used digital assistants in the past, with 72% indicating they use digital assistants at least once a week. Overall, participants said that digital assistants are intended more for getting tasks done (M = 5.19, SD = 1.35); this mean is higher than the midpoint (4) of the scale, t(321) = 15.83, p < 0.001.

3.2. Manipulation Check

As expected, participants’ reported performance expectancy of the performance devices (M = 5.36, SD = 1.14) was significantly higher than their performance expectancy of the hedonic devices (M = 5.07, SD = 1.25), t(327) = 6.21, p < 0.001. Likewise, participants’ reported hedonic motivation toward the hedonic devices was significantly higher (M = 5.41, SD = 1.16) than toward the performance devices (M = 5.08, SD = 1.14), t(327) = 6.91, p < 0.001). Thus, the manipulation of performance expectancy and hedonic motivation was successful.

3.3. Purposes and Choices

The hypotheses predicted that performance devices will be chosen at a higher frequency for utilitarian purposes and that hedonic devices will be chosen at a higher frequency for fun purposes. Each participant made four comparisons between performance and hedonic devices, for a total of 32 choices between device types (4 comparisons × 8 tasks). To test the main hypothesis, a χ² non-parametric test was run that examined the main effects of device type and interactions with the intended purpose by examining the frequency with which each device was chosen for each of the eight purposes. The main effects aggregated across all purposes, while interaction effects were aggregated within each type of purpose, whether utilitarian or entertainment. Overall, hedonic devices tended to be chosen more often than performance devices, χ²(1, N = 10,409) = 3.32, p = 0.068. As expected, there was a significant interaction indicating that performance devices were preferred for utilitarian purposes and hedonic devices were preferred for fun purposes, χ²(1, N = 10,409) = 393.29, p < 0.001 (see

Table 1. Preference for performance and hedonic devices, separate for task and entertainment purposes.

	Task Purposes	Entertainment Purposes
Performance devices	3061 (58.8%)	2050 (39.4%)
Hedonic devices	2143 (41.2%)	3155 (60.6%)
Main effect of devices	χ2(1, N = 10,409) = 3.32, p = 0.068
Interaction effect	χ2(1, N = 10,409) = 393.29, p < 0.001

Note: The table shows the frequencies by which participants chose each device type. The main effect compares across the first and second rows, while the interaction statistic indicates differences between device types while also accounting for purpose type.

).

Table 2. Individual purposes—Study 1.

	Lights and Security	Practical Help	News and Traffic	Communication
Performance devices	823 (63.3%)	811 (62.3%)	772 (59.3%)	655 (50.3%)
Hedonic devices	477 (36.7%)	490 (37.7%)	530 (40.7%)	646 (49.7%)
	χ2(1, N = 1300) = 92.09, p < 0.001	χ2(1, N = 1301) = 79.20, p < 0.001	χ2(1, N = 1302) = 44.98, p < 0.001	χ2(1, N = 1301) = 0.062, p = 0.803
	Kids’ Toy	Social Entertainment	Personal Entertainment	Companion
Performance devices	461 (35.4%)	448 (34.4%)	599 (46.1%)	542 (41.6%)
Hedonic devices	840 (64.6%)	854 (65.6%)	701 (53.9%)	760 (58.4%)
	χ2(1, N = 1301) = 110.41, p < 0.001	χ2(1, N = 1302) = 126.6, p < 0.001	χ2(1, N = 1300) = 8.00, p = 0.005	χ2(1, N = 1302) = 36.50, p < 0.001

Note: This table shows the frequency by which participants chose each device type within each individual stated purpose.

contains the results for each of the eight individual purposes separately. For each of the described utilitarian purposes, the performance device was preferred, except for the purpose of communicating with family members and friends. All four of the entertainment purposes showed that the hedonic devices were chosen a significant majority of the time. Thus, the two main hypotheses were confirmed overall and also held for seven out of the eight single purposes that had been pre-classified as predominantly utilitarian or hedonic. This included the two utilitarian purposes related to controlling one’s home: for the purposes of controlling lights, security, and sound around the house and practical help around the house such as lists, timers, alarms, and reminders, 63% and 62%, respectively, preferred the performance device.

4. Discussion

This study shed light on purpose as a moderator for the adoption of smart digital assistants. The matching hypothesis predicted devices that receive high performance expectancy ratings would be chosen at a higher rate when the intended use was utilitarian, whereas devices that receive high hedonic motivation ratings would be adopted more when the intended use of the device is hedonic. This pattern is exactly what was observed.

Examining the four utilitarian purposes and the four hedonic purposes separately, there was a strong interaction supporting the hypotheses. Specifically, participants were asked which devices they would prefer if they had to control their home (control lights, security, and sound around the house) and needed practical help around their house (including timers, alarms, and reminders). As expected, for these utilitarian tasks, participants preferred in the majority of cases (in over 60%) digital assistants that were described as simple, easy to use, useful, high-tech, highly adaptable, and competent (performance devices) over assistants that were described as smooth, natural, intriguing, exciting, and entertaining (hedonic devices). Conversely, hedonic devices were preferred at a similar rate for entertainment purposes such as using the device as a toy for children or for social and personal entertainment and companionship.

One intended use, communication with friends and family, was included as a utilitarian purpose, but did not match the predicted pattern like the other purposes did. This was an unexpected result. In hindsight, we reasoned that participants may not have viewed this purpose as a utilitarian purpose. The second study aimed to replicate the findings of Study 1 and included a measure of how participants view each of the eight purposes and which purposes participants classify as entertainment or utilitarian.

5. Study 2

The first study was intended to test the proposed matching hypothesis that the way a digital assistant is used impacts the adoption of a digital assistant. The second study was designed to replicate the findings of the first study and to test whether the interaction effect found in the first study would also hold across different levels of effort expectancy.

Venkatesh et al. [14] used effort expectancy in UTAUT2 as a predictor of technology adoption along with performance expectancy and hedonic motivation. Effort expectancy refers to a person’s expectation for how difficult a technology will be for them to use. Created from previous work by Venkatesh et al. [15] for the UTAUT model, it was found to be associated with behavioral intention to adopt four different workplace technologies in four different work settings. More recent examples of studies supporting the association between technology adoption and effort expectancy include the adoption of e-Learning [41], social media apps [42], and social recommender systems [43]. Specifically, Gaiser and Utz [32] noted that complexity is a significant theme related to the adoption of digital assistants, complexity being a part of the inspiration for the term effort expectancy.

The UTAUT2 model suggests that, in addition to the effects of performance expectancy and hedonic motivation, perceptions about effort will also affect adoption intentions and use. We sought to explore if effort expectancy has an effect on adoption beyond the effects of performance expectancy and hedonic motivation. Based on previous research, we anticipated finding that devices that are perceived to require less effort to use would be chosen at a higher rate than those perceived to require more effort to use.

6. Materials and Methods

6.1. Sample

A total of 351 U.S.-based participants were recruited through the MTurk system to take a survey administered through Qualtrics as a convenience sample and gave informed consent. There were 10 participants removed as a result of attention checks, leaving 341 responses used in the data analysis. Of the participants, 205 were male, 135 participants female, and 1 chose not to answer. The participants’ reported ages ranged from 18 to 72 (M = 36.26, SD = 10.62). A majority (76%) of respondents indicated that they had at least a Bachelor’s degree.

6.2. Procedure and Design

Participants were presented with a survey that introduced them to the same four mock-up new digital assistants as in the first study, two designed to trigger high performance expectancy ratings and two designed to trigger high hedonic motivation ratings. As in Study 1, participants viewed each of the four digital assistants in a complete pair comparison for eight different purposes, which were either utilitarian or entertainment purposes.

In addition to repeating the manipulation from Study 1, half of the participants received an effort manipulation, where some of the performance devices would take more effort to use, whereas for others, the hedonic devices would take more effort. Some descriptions included a very easy setup process that can be aided by downloading an app onto a smartphone, while other descriptions indicated that consumers would need to register their device, create an account over the phone through a customer service line, and take extended time to learn how to use the device. The descriptions from Study 1 were used as the low-effort devices, and four new descriptions were added as high-effort devices. The four new descriptions were identical to the original four descriptions from the first study, other than the manipulation of effort expectancy.

6.3. Measurements

In addition to several new items, measurements for experience with digital assistants, performance expectancy, hedonic motivation, and device adoption were repeated from the first study. See Appendix A for the full list of items.

Effort expectancy: Three items were adapted from Venkatesh et al. [14] dealing with effort expectancy. The items were “Learning how to use [device name] would be easy for me”, “My interaction with [device name] would be clear and understandable”, and “It would be easy for me to become skillful at using [device name]”. The items were measured on a seven-point Likert scale, from strongly disagree (1) to strongly agree (7). Dissimilar to some of the other items, lower effort expectations are a higher number on this scale. Thus, a higher score indicates lower perceived effort. These items obtained sufficient reliability for all four low-effort expectancy descriptions (Cronbach’s αs ≥ 0.78) and all four high-effort expectancy descriptions (Cronbach’s αs ≥ 0.83).

Perception of the eight purposes: To provide insight into how participants thought about each of the eight purposes provided, eight items were added that asked whether they thought the purposes were fun purposes or task purposes. The eight items were “Using a digital assistant as a kids’ toy is a (fun purpose/task purpose)”, “Using a digital assistant for social entertainment for guests, friends, and family is a ______”, “Using a digital assistant for personal entertainment for streaming media like music, movies, and TV shows is a ______”, “Using a digital assistant as a companion is a ______”, “Using a digital assistant as a way to control lights, security, and sound is a ______”, “Using a digital assistant to keep lists, timers, alarms, and reminders is a ______”, “Using a digital assistant to keep up with news, culture, local traffic, and weather is a ______”, and “Using a digital assistant to communicate with family and friends is a ______”.

7. Results

As in the first study, the names of the digital assistants were presented with different descriptions in random order. The results were aggregated with each device type and across the name variations, as in Study 1.

7.1. Experience and Purpose of Assistants

Among the participants, 90.6% indicated that they had used digital assistants in the past, with 71.8% indicating they use digital assistants at least once a week. As in Study 1, overall, participants said that digital assistants are intended more for getting tasks done than for having fun, with four being neutral on a seven-point scale (M = 5.01, SD = 1.35), t(340) = 13.78, p < 0.001.

7.2. Manipulation Check

Participants’ performance expectancy of the performance devices (M = 5.31, SD = 1.25) was significantly higher than their performance expectancy of the hedonic devices (M = 4.92, SD = 1.35), t(340) = 7.16, p < 0.001. Likewise, participants’ reported hedonic motivation toward the hedonic devices was significantly higher (M = 5.34, SD = 1.29) than toward the performance devices (M = 4.91, SD = 1.27), t(340) = 8.13, p < 0.001), as expected.

There was a marginally significant main effect for effort expectancy t(340) = 1.92, p = 0.056, the high-effort devices (M = 5.49, SD = 1.11) being rated as slightly more difficult to use than the low-effort devices (M = 5.58, SD = 1.01). However, there was also a main effect of device type on effort expectancy t(340) = 2.42, p = 0.016, the performance devices (M = 5.56, SD = 1.02) being rated as slightly easier to use than the hedonic devices (M = 5.46, SD = 1.04), which may be related to the language in the description of one performance device that describes it as relatively simple to use. Thus, the independent manipulation of effort expectancy was not successful. Of note, all of these effort ratings were above the midpoint of the seven-point Likert scale, indicating that none of the devices were perceived as being very difficult to use.

7.3. Purposes and Choices

The hypotheses predicted that, for the four utilitarian purposes, participants would choose the performance devices at a higher frequency, and for the four entertainment purposes, participants would choose the hedonic devices at a higher frequency. To test this main hypothesis, we ran a chi-squared test in the same way as in Study 1. Across all participants and all eight purposes, there was a main effect of a preference for the hedonic devices, χ²(1, N = 10,906) = 9.51, p = 0.002. However, in line with the matching hypothesis, there also was a significant interaction: for utilitarian purposes, performance devices were chosen more often than hedonic devices, whereas for entertainment purposes, hedonic devices were chosen more often than performance devices, χ²(1, N = 10,906) = 642.56, p < 0.001 (see

Table 3. Preference for performance and hedonic devices—Study 2.

	Task Purposes	Entertainment Purposes
Performance devices	3308 (62.5%)	2146 (38.2%)
Hedonic devices	1984 (37.5%)	3468 (61.8%)
Main effect of devices	χ2(1, N = 10,906) = 9.51, p = 0.002
Interaction effect	χ2(1, N = 10,906) = 642.56, p < 0.001

Note: The table shows the frequencies by which participants chose each device type. The main effect compares across the first and second rows, while the interaction statistic indicates differences between device types while also accounting for purpose type.

). As expected and in line with Study 1, for the two utilitarian tasks of controlling one’s home and using digital assistants for personal tasks (see

Table 4. Individual purposes—Study 2.

	Lights and Security	Practical Help	News and Traffic	Communication
Performance devices	905 (66.3%)	909 (66.7%)	857 (62.8%)	637 (46.7%)
Hedonic devices	459 (33.6%)	454 (33.3%)	507 (37.2%)	726 (53.3%)
	χ2(1, N = 1364) = 145.83, p < 0.001	χ2(1, N = 1363) = 151.89, p < 0.001	χ2(1, N = 1364) = 89.81, p < 0.001	χ2(1, N = 1363) = 5.81, p = 0.016
	Kids’ Toy	Social Entertainment	Personal Entertainment	Companion
Performance devices	425 (31.2%)	445 (32.7%)	602 (44.1%)	512 (38.5%)
Hedonic devices	938 (68.8%)	917 (67.3%)	761 (55.9%)	852 (62.5%)
	χ2(1, N = 1363) = 193.08, p < 0.001	χ2(1, N = 1362) = 163.57, p < 0.001	χ2(1, N = 1363) = 18.55, p = 0.005	χ2(1, N = 1364) = 84.75, p < 0.001

Note: This table shows the frequency by which participants chose each device type within each individual stated purpose.

), participants preferred in the majority of cases (in over 66%) digital assistants that were described as performance devices (simple, easy to use, useful, high-tech, highly adaptable, and competent) over hedonic devices (smooth, natural, intriguing, exciting, and entertaining).

We also examined choice behaviors within the group of participants who received an effort manipulation and observed that the interaction held the same across the two conditions, one condition with low-effort performance devices and high-effort hedonic devices and the other with high-effort performance devices and low-effort hedonic devices, χ²(1, N = 5498) = 358.92, p < 0.001, with participants preferring the performance devices for utilitarian purposes and the hedonic devices for entertainment purposes. There was no significant main effect for effort, in that neither the high- nor low-effort devices were preferred over one another, χ²(1, N = 5498) = 0.841, p = 0.359.

To add insight into participants’ perceptions about the eight purposes given in the study, questions were added to provide data on whether they perceived them to be task purposes or fun purposes.

Table 5. Perceived purposes.

	Lights and Security	Practical Help	News and Traffic	Communication
Task Purpose	301 (88.3%)	286 (83.9%)	268 (78.6%)	154 (45.2%)
Fun Purpose	40 (11.7%)	55 (16.1%)	73 (21.4%)	187 (54.8%)
	χ2(1, N = 341) = 199.77, p < 0.001	χ2(1, N = 341) = 156.48, p < 0.001	χ2(1, N = 341) = 111.51, p < 0.001	χ2(1, N = 341) = 3.19, p = 0.074
	Kids’ Toy	Social Entertainment	Personal Entertainment	Companion
Task Purpose	38 (12.2%)	56 (16.5%)	102 (29.9%)	83 (24.3%)
Fun Purpose	303 (88.8%)	284 (83.5%)	239 (70.1%)	258 (75.7%)
	χ2(1, N = 341) = 205.94, p < 0.001	χ2(1, N = 340) = 152.89, p < 0.001	χ2(1, N = 341) = 55.04, p < 0.001	χ2(1, N = 341) = 89.81, p < 0.001

Note: This table shows participants’ labeling of each stated purpose as either a task purpose or a fun purpose.

lists the eight purposes and how they were perceived. The table indicates that controlling one’s home (control lights, security, and sound around the house) and utilizing digital assistants for practical help in one’s house (including timers, alarms, and reminders) were clearly perceived as being rather utilitarian than fun purposes. Of all participants, 88% and 84%, respectively, said that these tasks are related to utilitarian purpose.

The results suggested that participant perceptions of the purposes matched our pre-classified conceptualizations except for the purpose communication with friends and family, which tended to be rather perceived as a hedonic than a utilitarian purpose, χ²(1, N = 341) = 3.19, p = 0.074. This notable diversion from the intended utilitarian purpose provides an explanation for the finding that the interaction did not hold for pair comparisons involving communication with friends and family, indicating that the mismatch from our original hypothesis was due to different conceptualizations of that purpose as being a task or being fun.

8. Discussion

The central focus of this second study was to see whether the results of the first study could be successfully replicated while introducing and controlling for perceived effort expectancy. Using the same methods as in the first study, this second study yielded very similar findings. The two main hypotheses were supported in the same way as in the first study, with seven out of the eight purposes supporting the hypotheses. The same purpose that did not match the hypotheses in the first study, communication with friends and family, again did not match the hypothesis, as predicted.

To understand why participants’ preferences for either performance or hedonic devices related to this purpose were not successfully predicted in the first study, the second study also examined participants’ perceived purpose of each of the eight tasks or usages. Although seven out of the eight usages were perceived as was anticipated, as utilitarian or entertainment purposes, respectively, the purpose communication with friends and family was split relatively evenly on whether people viewed it as a fun purpose or task purpose. This result adds credibility to the underlying hypothesis of the matching purpose to performance expectancy or hedonic motivation ratings. In the case where there was not a clear choice between the performance device and the hedonic device, we also observed that participants did not agree on whether that purpose was utilitarian or entertainment.

The impact of the interaction that was found held across different levels of effort expectancy. Even when the expectation for effort was higher for one type of device, whether performance devices or hedonic devices, the interaction effect was reliable. In this study, we did not find a main effect of effort expectancy; that is, participants did not select devices more often that needed less effort than devices that needed more effort. Because effort expectancy has been found to have an effect on a variety of technology adoption situations [41,42], higher adoption rates for lower-effort devices were expected in this study also. It was unexpected that the manipulation of effort expectancy did not substantially affect adoption. It is possible that the manipulation was not strong enough to make a difference, especially given only a marginally significant manipulation, and that participants judged the expected effort to use the digital assistant to be sufficiently low in both conditions. Moreover, we reasoned that the characteristic of the performance devices as being “simple to use” may have triggered the observed lower effort expectancy in performance devices than in hedonic devices.

9. General Discussion

Although some technologies are very distinctly and clearly utilitarian or hedonic in nature [27], other technologies may be able to fulfill both entertainment or utilitarian purposes, such as voice-controlled digital assistants such as Google Assistant and Amazon’s Alexa. Even among these dual-purpose devices, there may still be variation in people’s perceptions of the task and entertainment value of the technologies. The matching of people’s perceptions of these values to the device was shown in this study to depend on their own needs that they are seeking to gratify [22]. The hypotheses predicted that high performance expectancy ratings would predict the selection of the device for utilitarian motivations, whereas high hedonic motivation toward the device would predict the selection of the device for entertainment-oriented purposes. This study shed light on purpose as a moderator for the adoption of technology, specifically for digital assistants.

Examining the four utilitarian purposes and the four entertainment purposes separately, there was a strong interaction in the direction that the hypotheses predicted: for utilitarian purposes, performance devices were chosen more often than hedonic devices, whereas for entertainment purposes, hedonic devices were preferred over performance devices. As predicted, the choices participants made about which device they preferred depended on how the device would be used. This interaction effect held not only across all utilitarian and hedonic choices, but was also evident for seven of the eight purposes, as hypothesized. One purpose, communication with friends and family, was included as a utilitarian purpose, but did not match the predicted pattern like the other purposes did. This was an unexpected result. Through further examination, we found that participants did not generally view this purpose as a task-oriented purpose, which seems to explain this result.

The conducted study sought to build on the current technology adoption literature in two primary ways. First, we examined purpose as a factor moderating the effects of performance expectancy and hedonic motivation on adoption behavior. The study thus provides support for a conceptual hypothesis that may be applicable across a variety of different technologies and expands on the UTAUT2 model to add purpose as a useful moderator among adoption factors. Second, we aimed to contribute to our understanding of the potential adoption of a specific new technology, voice-controlled assistants.

The basic proposal of this paper is simple: people match their technology choices to the specific needs that they want to gratify. However, this important concept has largely been overlooked in the technology adoption literature. Our results imply that it may be misleading and incomplete to examine these two factors of adoption without consideration of how a consumer purposes to use the technology. This, we believe, is the central contribution of this paper, as the field has largely looked at performance expectancy and hedonic motivation without this consideration. Uses and gratifications theory [22] offers the foundation for how the diffusion of these technologies, which is a social and communicative process, may change from person to person based on their beliefs about what a technology is best used for.

Limitations and Future Directions

It would be interesting to explore in future studies if it is possible that, after beginning to use Alexa for entertainment, consumers would be open to Alexa initiating a conversation about energy consumption. Further research may also consider having messaging prompts that include both appeals, performance expectancy and hedonic motivation language, in comparison to only one. The presented studies did not include a condition in which participants saw both types of language in the description of a device. It would be telling to see whether both aspects can be equally emphasized in a single prompt and if it makes a difference or not whether characteristics appealing to performance expectancy or to hedonic motivation are emphasized first within the message.

10. Conclusions

The results of this study suggest that the way in which a device is introduced can lead to different perceptions of the device, which can subsequently alter decisions to begin using the device. We demonstrated that preference for performance or hedonic devices is dependent on the intended use of the device. The study thus provides support for a conceptual hypothesis of adoption in which initial consumer perceptions of a device may need to be matched with the purpose or need with which a user comes to the device. This is of significant importance in a context like Kim et al. [7] in which consumers had devices available to them, but not everybody used them to the same extent and for the same purpose. Because smart technology has significant implications for energy conservation and the efficient use of resources [9,44], understanding the adoption of digital assistants and how it depends on the intended use of the device is vital to the implementation of energy-saving programs and interventions that use smart technologies in residential homes.