The Use of Artificial Intelligence: Exploring Using Motivations, Involvement, and Satisfaction with the Case of Alexa

Abstract

Whether it is asking Alexa to set a reminder or having Google Assistant place a call, AI-powered assistants are becoming an increasingly seamless part of our daily lives. This study aims to address what predicts the users’ satisfaction with Alexa by analyzing the using motives, cognitive involvement, and emotional involvement of its consumers. The variables include using motives, attention, elaboration, emotional involvement, and usage satisfaction. Alexa users (N = 299) completed a brief online survey, including Scales of Using Motives for Media, the Perceived Attention Scale, the Elaboration Scale, the Mood Adjective Check List Scale, and Television Viewing Satisfaction Scale. Participants who were at least eighteen years of age and owned and used Alexa were included in this study. An exploratory factor analysis revealed four distinct types of motivation for using Alexa: Companionship, Entertainment–Information, Work-Efficiency, and Pastime. The results from hierarchical regressions showed that Alexa usage satisfaction was predicted by Entertainment–Information and feeling positive emotions while using.

1. Introduction

Nowadays, artificial intelligence (AI) is a buzzword and an emerging technology hotly discussed in society, and cognitive assistants (Amazon’s Alexa, Google Assistant) might be the first generation of AI products that consumers can have. With the increasingly widespread growth of AI technology, its influence on the individual and even the whole of society is gradually becoming the focus of research within academia. Specifically, from the perspective of consumers, what predicts people’s satisfaction with AI products has become an issue worth studying.

Alexa (Amazon Echo), like iPhone, iPad, or iPod touch, is an intelligent personal assistant that helps you multitask and get things done. It conforms to three characteristics of the research object: First of all, it is a popular product. According to Consumer Intelligence Research Partners, by the end of September 2017, Amazon’s Echo devices (Alexa) were outselling their closest competitor, Google Home, by a ratio of three to one, with estimated sales reaching 20 million units compared to Google Home’s 7 million.1 Second, Alexa is a multifunctional device capable of performing a wide range of tasks and adapting to multi-user interactions (Purington et al., 2017). Third, Alexa is a highly engaging and intelligent product, encouraging users to personify it and incorporate it into their daily social lives (Purington et al., 2017). This paper aims to help both communication technology engineers and scholars better understand the potential demands of consumers, both technologically and psychologically, as well as to refine the existing relevant AI products and services.

This paper examines the case of Alexa through the lens of Uses and Gratifications Theory (UGT), using an online cross-sectional survey to explore the underlying factors—such as usage motives and cognitive and emotional involvement—that influence consumer satisfaction with AI products.

2. Literature Review

2.1. Demographics and Mobile Device Usage

Demographic factors have long been a fundamental consideration in the study of media consumption and usage, spanning both traditional media (Burgoon & Burgoon, 1980; Schoenbach et al., 1999; Stevenson, 1977; Westley & Severin, 1964) and newer forms, such as the Internet (Althaus & Tewksbury, 2000; D. S. Chung, 2008; Chyi et al., 2010; Kaye & Johnson, 2002; Larose & Eastin, 2004; Riffe et al., 2008) and mobile devices (Chan-Olmsted et al., 2013; Westlund, 2008).

Many studies on artificial intelligence focus on the speech, discourse, or rhetoric function of this technology (e.g., Hill et al., 2015; Porter, 2017; Prakken, 2011; Walton, 2015, 2016). Some studies pay attention to the application of artificial intelligence to different contexts. For example, several scholars studied AI technology’s automated surveillance in subway systems (e.g., Geoghegan, 2008; Mcclain, 2018). C. Edwards et al. (2018) examined the role of AI technology as machine agents within the context of education communication and instruction. Woods (2018) illuminated how feminine persona in artificially intelligent virtual assistants (AI VAs) such as Siri and Alexa contribute to surveillance capitalism in the platform economy. Mansell (2017) connected complex digital systems to enduring and multifaceted inequalities and warned that innovations in areas such as artificial intelligence, algorithmic computation, and machine learning could undermine human authority over the future direction of digital technologies. Galloway and Swiatek (2018) discussed the diversity and extent of AI’s uses in public relations practice. Neff and Nagy (2016) used the case of Tay, an experimental AI chatbot, to revisit theories of agency. Their study demonstrated how varying qualities of agency, expectations for technologies, and capacities for affordance emerged through interactions between people and artificial intelligence. Emerging problems with AI technology have been raised. Murdock (2018) highlighted the significant ethical concerns surrounding the social and environmental costs of AI products, arguing that the growing importance of digital technologies in shaping advanced capitalism, along with the swift expansion of artificial intelligence, brings the material aspects of media to the forefront of analysis. Interdisciplinary collaborative research was needed because AI technology involves knowledge from computer science, psychology, communication, and many other disciplines. Mansell (2017) advocated for critical interdisciplinary engagement to encourage digital economy policymakers to explore alternative innovation pathways and adopt more proactive policies for shaping a better future.

As for earlier studies that were most related to this research, Purington et al. (2017) analyzed user reviews of the Echo posted on Amazon.com to examine the extent of device personification, sociability in interactions, factors associated with personification, and their impact on user satisfaction. Similarly, Ho et al. (2018) conducted an experiment to investigate the downstream effects of emotional versus factual disclosures in conversations with either a chatbot or a human. Their findings revealed that emotional disclosure had similar effects, regardless of whether participants believed they were interacting with a chatbot or a person. In all, few studies lay emphasis on the feelings and experiences of AI users.

2.2. Theoretical Framework

Traditionally, U&G research posits that media users are active agents who select media content, giving them gratifications. Hence, using motives predict activity (Levy & Windahl, 1984; Perse, 1990; Rubin & Perse, 1987). This active participation is especially relevant in the context of AI technologies, which rely on user input and feedback for functionality.

The Uses and Gratifications (U&G) theory offers a valuable framework to analyze individual motivations and experiences with AI-powered devices across different contexts (C. Xie et al., 2024). For example, Pitardi and Marriott (2021) investigated how utilitarian, technological, social, and hedonic gratifications influence users’ trust in Alexa, an AI voice assistant. Also, Shao and Kwon (2021) found that users’ satisfaction was positively influenced by functional utility and dynamic control, whereas companionship and entertainment did not have a significant impact. Similarly, M. Chung et al. (2020) analyzed customer satisfaction with online shopping e-service agents through the lens of utilitarian, technological, and hedonic gratifications. In the retail sector, Rese et al. (2020) studied the impact of a text-based chatbot, Emma, by applying the same gratification dimensions to assess its role in customer communication.

Using Motives. Rubin (1984) identified two general types of media: ritualistic and instrumental. These categories can also be applied to consumers’ use of AI technologies (Xu et al., 2022). Ritualistic use refers to habitual engagement—individuals interact with AI-enabled mobile devices out of habit, often to pass time, seek companionship, relax, become aroused, or escape from reality. In contrast, instrumental use involves more selective and purposeful interaction with specific AI features or content. Research suggests that instrumental use represents a more active and engaged form of media consumption (Perse, 1990; Rubin, 1984; Rubin & Perse, 1987). Contributing to prior U&G research on the distinction between the instrumental and ritual use of media, Xu et al. (2022) revealed that what has been traditionally perceived as instrumental use of media could also predict users’ affinity with the medium per se. Although there are many studies on using existing classifications of using motives from the lens of U&G theory to examine user satisfaction with AI communication tools, a few explored and further identified using motives related to AI products.

Involvement. Media user activity is inherently multidimensional (Levy & Windahl, 1984). Individuals engage with media to varying degrees across several dimensions, such as selectivity, utility, and involvement (Blumler, 1979; J. Kim & Rubin, 1997; Rubin & Perse, 1987). More specifically, media users exhibit different levels of selectivity, utility, and involvement (Levy & Windahl, 1984). These different forms of activity contribute to distinct media-related outcomes (J. Kim & Rubin, 1997). Involvement plays a crucial role in media user activity and is central to media effects research. For instance, Gomez et al. (2023) found that when users were actively involved in generating AI recommendations during tasks characterized by significant knowledge asymmetry, they were more likely to accept the AI’s suggestions and view both the AI agent and their collaborative interaction more favorably. Also, Doh et al. (2021) demonstrated that both user involvement and the perceived severity of a situation significantly influenced the acceptance of decisions made by AI judge and AI jury. Involvement reflects the extent of personal engagement with media content (Perse, 1990). Involvement is defined as both “the degree to which an audience member perceives a connection between oneself and mass media content; and the degree to which the individual interacts psychologically with a medium or its messages” (Levy & Windahl, 1985, p. 112). In order to be suitable for this study on AI, the latter definition of involvement is adopted.

Message content is perceived as a sign of cognitive involvement (Rubin & Perse, 1987). Cognitive involvement refers to paying attention to media content, including the mental effort devoted to engaging with the media and evaluating its messages during consumption (Perse, 1990). A deeper aspect of this involvement is elaboration—the process by which media users interpret, derive meaning from, and respond to messages (Eveland, 2001; Perse, 1990; Rubin & Perse, 1987). Both attention and elaboration reflect an instrumental use of media, where users actively engage with content to fulfill specific goals (Levy & Windahl, 1985; Rubin & Perse, 1987).

In addition to cognitive involvement, using AI also includes emotional involvement. Emotional involvement encompasses a spectrum of feelings, from satisfaction and happiness to frustration and anger (e.g., Nabi et al., 2006). Nabi et al. (2006) found that positive emotions during TV viewing increased enjoyment, while negative emotions diminished it.

Existing research on AI mainly treated involvement as a unified concept rather than a stratified one. Therefore, this study examined the role of specific cognitive and emotional involvements in AI usage satisfaction.

Interactive Experience and Satisfaction. Satisfaction refers to an affective response to media use, stemming from the extent to which users feel their motivations for engaging with the media have been fulfilled. This concept is significant because it is linked to personal fulfillment (e.g., Hecht, 1978), enjoyment, favorable evaluations of media content, and increased media consumption (e.g., Perse & Ferguson, 1993).

Relevant research has shown that satisfaction often stems from more instrumental and active media use (J. Kim & Rubin, 1997; Perse & Rubin, 1988). Similarly, satisfaction is the consequence of instrumentally and actively using AI products. User satisfaction serves as a key indicator of a technology’s success, as it often drives both initial adoption and continued use over time (Luo & Remus, 2014). The purpose of designing AI products is to meet the different needs of consumers through the machines’ interaction with them and further improve their life quality and effectiveness.

Drawing on U&G theory as a user-centered framework, this study focuses on diverse uses of artificial intelligence by consumers and their satisfaction with Alexa:

RQ: What predicts customers’ satisfaction with using Alexa?

3. Method

Amazon Mechanical Turk (MTurk) was used to distribute the survey, and Qualtrics Survey Software was used to collect the data. If the participants were not at least 18 years of age and owned or used Alexa, they were excluded from the study immediately. If the participants were eligible based on the requirements, they moved on to complete the survey.

Table 1. Descriptive statistics for demographic variables.

	N	Min	Max	Mean	S.E.	SD	Variance	Skewness
Age	299	18.00	61.00	29.44	0.398	6.881	47.35	1.682
Gender	299	0.00	1.00	0.39	0.028	0.489	0.24	0.445
Education	299	1.00	6.00	4.91	0.054	0.926	0.86	−1.654
Income	299	1.00	9.00	3.32	0.100	1.731	3.00	0.645

presents the descriptive statistics for demographic variables. Notably, the mean value of age is close to the minimum value, indicating that limited variation may affect the results.

4. Measurement

4.1. Exposure to Alexa

Respondents indicated how frequently they used Alexa on a scale from 1 (never) to 5 (always). The responses were then summed to generate a measure of exposure to Alexa.

4.2. Using Motives

Prior studies identified a number of key motivations that drive users’ media usage, such as Information–Learning, Exciting Entertainment, Habit–Pastime, Companionship, Relax–Escape, Social Interaction, and Work-Efficiency. In this study, all measurement items were adapted from existing U&G literature with minor modifications to suit the context of the current research (e.g., Godlewski & Perse, 2010; Rubin, 1983; Y. Kim et al., 2016). To measure these using motives with a total of 26 motivation items, respondents evaluated the statements based on their own reasons for using Alexa on a Likert-type scale from 1 to 5 (1 = not at all, 5 = exactly). These scales were chosen over others because of their flexibility in being applied to emerging media technologies and their well-documented multidimensional structure.

The responses to the 26 motive statements underwent exploratory factor analysis with oblique rotation, acknowledging the interrelated nature of the motives for using Alexa. These variables are sufficiently correlated to perform EFA because Bartlett’s test of sphericity was significant, χ²(325, n = 299) = 4765.15, p < 0.001, and the Kaiser–Meyer–Olkin Measure of Sampling Adequacy (KMO) was 0.95 > 0.60. Four meaningful factors composed of 16 items were finally created based on the 0.50/0.30 decision rule and the content of the variables for each factor.

Table 2. Exploratory factor analysis of consumers’ Alexa use motives.

	Factor 1	Factor 2	Factor 3	Factor 4
Factor 1: Relax–Companionship
Because it relaxes me	0.835	0.071	0.029	−0.122
Because it makes me feel less lonely	0.727	−0.085	−0.076	0.084
Because it’s a pleasant rest	0.680	0.255	−0.017	−0.067
When there’s no one else to talk to or be with	0.639	−0.021	0.002	0.257
Because it allows me to unwind	0.569	0.089	−0.215	−0.007
So I won’t have to be alone	0.556	−0.005	−0.080	0.240
So I can forget about school, work, or other things	0.553	0.047	0.032	0.206
Factor 2: Entertainment–Information
Because it’s enjoyable	−0.013	0.746	0.059	0.016
Because it helps me learn about lots of information	0.002	0.656	−0.265	−0.029
Because it entertains me	0.118	0.627	0.090	0.046
Because by using it, I can know what I want to know	0.058	0.598	−0.203	−0.113
Factor 3: Work-Efficiency
To do something related to my job	−0.004	0.099	−0.742	0.153
To make it easier to do my work	0.133	0.038	−0.692	0.037
Factor 4: Habit–Pastime
Because I had nothing better to do	−0.041	0.002	−0.088	0.678
Because it passes the time away, particularly when I’m bored	0.132	0.109	0.021	0.589
Just because it was on	0.106	−0.088	−0.181	0.549

Note. Factor loadings at least 0.50 were bolded to indicate items that meaningfully load onto a given factor.

summarizes the exploratory factor analysis.

Factor 1, Relax–Companionship, included four Relax–Escape items and three Companionship items. Relaxation and companionship are typically ritualistic using motives (Rubin, 1984). The first factor accounts for a 45.70% variation among the original variables. Factor 2, Entertainment–Information, included two Exciting Entertainment items and two Information–Learning items. Entertainment and Information learning are typically instrumental using motives (Rubin & Perse, 1987; Y. Kim et al., 2016). The second factor accounts for a 7.68% variation among the original variables. Factor 3, Work-Efficiency, included two Work-Efficiency items. Work-Efficiency is an instrumental using motive (Y. Kim et al., 2016). The third factor accounts for a 4.50% variation among the original variables. Factor 4, Habit–Pastime, included three Habit–Pastime items. This factor is a motive driven by ritualistic use (Rubin, 1984). The fourth factor accounts for a 4.43% variation among the original variables.

The item scores were averaged to create motive scores. The most strongly endorsed motive was Entertainment–Information (M = 3.72, SD = 0.74; Cronbach’s α = 0.81), followed by Relax–Companionship (M = 3.56, SD = 0.91; Cronbach’s α = 0.91), Work-Efficiency (M = 3.55, SD = 1.00; Cronbach’s α = 0.82), and Habit–Pastime (M = 3.20, SD = 0.91; Cronbach’s α = 0.75). All of the four using motives were interrelated. All the motives examined were positively and significantly associated with exposure to Alexa.

Table 3. Pearson correlations between the variables.

	Relax Companionship	Entertainment Information	Work Efficiency	Habit Pastime	Attention	Elaboration	Positive Emotion	Negative Emotion	Satisfaction
Entertainment–Information	0.557 **
Work-Efficiency	0.597 **	0.506 **
Habit–Pastime	0.648 **	0.374 **	0.510 **
Attention	0.622 **	0.553 **	0.606 **	0.597 **
Elaboration	0.729 **	0.543 **	0.642 **	0.633 **	0.742 **
Positive Emotion	0.530 **	0.725 **	0.526 **	0.364 **	0.605 **	0.622 **
Negative Emotion	0.358 **	−0.014	0.272 **	0.450 **	0.406 **	0.394 **	0.059
Satisfaction	0.492 **	0.753 **	0.523 **	0.359 **	0.552 **	0.550 **	0.699 **	0.044
Exposure	0.481 **	0.470 **	0.486 **	0.425 **	0.533 **	0.453 **	0.339 **	0.299 **	0.446 **

Note. N = 299. ** Correlation is significant at the 0.01 level (2-tailed).

displays the Pearson correlations between the various variables in this study.

4.3. Involvement

Involvement is measured in three different subscales. Those subscales include attention, elaboration, and emotional involvement. These scales were selected because they capture both cognitive and emotional involvement, which are central to understanding user engagement with interactive media. They were preferred over broader engagement scales, which often conflate emotional, cognitive, and behavioral aspects, making it difficult to isolate the specific role of cognitive and emotional efforts in satisfaction.

Attention. To assess attention to Alexa, participants indicated their level of agreement with seven statements using the Perceived Attention Scale (Cegala, 1981; Perse, 1998). Responses were measured on a 5-point scale, ranging from 1 (strongly disagree) to 5 (strongly agree), reflecting participants’ typical thoughts and feelings while using Alexa. Questions three and six needed reverse coding so as to uphold the validity of the scale. Previous studies have demonstrated that this scale is reliable: Cronbach’s α = 0.90 (Perse, 1998). The scale demonstrated acceptable reliability in this study (Cronbach’s α = 0.78).

Elaboration. Respondents indicated their level of agreement with each item on the Elaboration Scale (Perse, 1998), using a scale ranging from 1 (strongly disagree) to 5 (strongly agree). Previous studies demonstrate this scale is reliable: Cronbach’s α = 0.83 (Perse, 1998). The scale demonstrated acceptable reliability in this study (Cronbach’s α = 0.87).

Emotional Involvement. Emotional reactions to using Alexa were measured using an adapted version of the Mood Adjective Check List (Nowlis, 1965; Perse, 1998). Participants rated their agreement on a 5-point scale (1 = strongly disagree to 5 = strongly agree) with statements reflecting the intensity of 10 positive emotions (e.g., amused, content, happy) and 10 negative emotions (e.g., angry, bored, worried) experienced during their interaction with Alexa. The MACL scale upholds a high level of reliability in other studies. The positive emotion scale demonstrated high reliability (Cronbach’s α = 0.90), and the negative emotion scale was also reliable (Cronbach’s α = 0.97) in this study.

4.4. Satisfaction

On the Television Viewing Satisfaction Scale (Ferguson & Perse, 2004; Godlewski & Perse, 2010), ranging from 1 (not at all) to 5 (completely), respondents indicated how pleasing and satisfying they found their experience using Alexa. This scale was selected for its conceptual alignment with situational satisfaction in AI research and its prior use in studies involving utility-driven technology adoption. Previous studies have demonstrated that this scale is reliable: Cronbach’s α = 0.86 (Ferguson & Perse, 2004). The scale demonstrated acceptable reliability in this study (Cronbach’s α = 0.79).

5. Results

Following scale construction and reliability analyses, several steps were taken to address the research question. First, two-tailed correlations were examined among motives, involvement, and usage satisfaction. Next, hierarchical multiple regression was conducted to assess the multivariate relationships between satisfaction and the independent variables. Age, gender, education level, and income level were included in the regression models as control variables to account for potential demographic influences on the dependent variable.

Alexa Using Satisfaction

The research question examined how usage motives, as well as cognitive and emotional involvement, are associated with users’ satisfaction with Alexa. Two-tailed Pearson correlations revealed several significant relationships. First, satisfaction with Alexa usage was positively correlated with all identified usage motives. The correlation between Entertainment–Information and Alexa usage satisfaction was positively significant, r(297) = 0.75, p = 0.000. The correlation between Work-Efficiency and Alexa usage satisfaction was positively significant, r(297) = 0.52, p = 0.000. The correlation between Relax–Companionship and Alexa usage satisfaction was positively significant, r(297) = 0.49, p = 0.000. The correlation between Habit–Pastime and Alexa usage satisfaction was positively significant, r(297) = 0.36, p = 0.000.

Second, Alexa usage satisfaction is positively linked to cognitive and feeling positive emotions, but satisfaction is unrelated to feeling negative emotions. The correlation between Attention and Alexa usage satisfaction was positively significant, r(297) = 0.55, p = 0.000. The correlation between Elaboration and Alexa usage satisfaction was positively significant, r(297) = 0.55, p = 0.000. The correlation between Positive emotions and Alexa usage satisfaction was positively significant, r(297) = 0.70, p = 0.000. The correlation between Negative emotions and Alexa usage satisfaction was not significant, r(297) = 0.04, p = 0.453. See Table 3.

Hierarchical multiple regression was used to examine the multivariate relationships between the variables in this study, which was used to explore the research question.

Table 4. Hierarchical regression predicting Alexa using satisfaction.

Predictor Variables	Step	B	SE B	Final β	R2 Change	Adjusted ΔR2
Demographics	1				0.005	−0.008
Age		−0.003	0.004	−0.031		0
Gender		−0.023	0.059	−0.014
Education		0.000	0.031	−0.001
Income		−0.010	0.017	−0.022	-
Using Motives	2				0.59 **	0.59
Relax–Companionship		−0.053	0.052	−0.061
Entertainment–Information		0.470	0.062	0.447 **
Work-Efficiency		0.066	0.040	0.084
Habit–Pastime		−0.005	0.045	−0.006
Exposure	3	0.068	0.037	0.086	0.002	0.001
Involvement	4				0.04 **	0.04
Attention		0.056	0.068	0.051
Elaboration		0.064	0.054	0.079
Positive Emotion		0.286	0.065	0.261 **
Negative Emotion		−0.032	0.030	−0.049

Note. Step 1: F(4, 294) = 0.39, p = 0.813; Step 2: F(8, 290) = 53.37, p < 0.001; Step 3: F(9, 289) = 47.71, p < 0.001; Step 4: F(13, 285) = 38.88, p < 0.001. ** p < 0.001.

presents a summary of the regression results.

Multicollinearity is not problematic because, in model 4, all values of TOL are greater than 0.29, ranging from 0.29 to 0.93. According to Meyers et al. (2013), some scholars say TOL should be greater than 0.40; others say TOL values should be greater than 0.10.

The demographic variables (age, gender, education level, and income level) entered in the first step explained 0.5% of the variance in Alexa satisfaction: R²-change = 0.005, F-change(4, 294) = 0.39, p = 0.813. At this stage, age, gender, education level, and income level were not significant predictors. In the second step, the portion of the variation in Alexa satisfaction explained by Entertainment–Information, Work-Efficiency, Relax–Companionship, and Habit–Pastime was statistically significant: R²-change = 0.59, F-change(4, 290) = 105.79, p < 0.001. In the third step, exposure to Alexa contributed an additional 0.2% to the variance in Alexa satisfaction: R²-change = 0.002, F-change(1, 289) = 1.55, p = 0.214. At this stage, exposure was not a significant predictor. In the fourth step, the inclusion of cognitive and emotional involvement (such as attention, elaboration, positive emotion, and negative emotion) accounted for a significant proportion of variation in Alexa satisfaction: R²-change = 0.04, F-change(4, 285) = 8.25, p < 0.001.

In the final analysis, the proportion of variation in Alexa using satisfaction explained by the set of predictors was statistically significant: R² = 0.64, adjusted R² = 0.62; F(13, 285) = 38.88, p < 0.001. Both Entertainment–Information (β = 0.45, p < 0.001) and experiencing positive emotions while using Alexa (β = 0.26, p < 0.001) were positive predictors of Alexa satisfaction. However, there is no significant negative predictor of Alexa using satisfaction.

6. Discussion

As for using motives, previous research has indicated that program satisfaction is influenced by program expectations (e.g., Perse & Ferguson, 1993; Perse & Rubin, 1988). In this study, correlation results show that every using motive is significantly linked to all other using motives, which is similar to the findings of previous research on traditional media that using motives were interrelated (Godlewski & Perse, 2010). Moreover, correlation results also illustrate that usage satisfaction is positively related to all of the using motives; however, among these using motives, only Entertainment–Information (instrumental motive) is a significantly positive predictors of Alexa satisfaction. This finding aligns with previous research, which suggests that many users interact with Alexa primarily for entertainment or assistant functions (Purington et al., 2017; Orden-Mejía et al., 2023).

The non-significant effects of other motivational dimensions, including Relax–Companionship, Work-Efficiency, and Habit–Pastime, suggest that user satisfaction with voice-based AI systems may depend more on instrumental gratification than on habitual or situational one. The findings are inconsistent with the studies by Yu et al. (2024), Arpaci et al. (2022), and Z. Xie et al. (2024) that habit influences usage behavior in an interactive relationship such as with AI, and the relationship exhibits cultural variations.

In relation to cognitive involvement (attention and elaboration), it was not a significant predictor of Alexa’s satisfaction, which was different from other online activities, like reality TV programs. It suggests that interactive communication did not have a main effect on cognitive involvement (Jiang et al., 2010). It contradicts many prior results that cognitive involvement is a good predictor of satisfaction (K. Kim, 2008). One possible explanation is that the nature of interaction with Alexa may be more task-oriented and less cognitively demanding compared to content-rich platforms that require sustained attention and critical engagement (Vtyurina & Fourney, 2018). Although interactivity has been shown to create cognitively involving experiences (Liu & Shrum, 2002), the type of interactivity offered by voice assistants may not be rich enough to elicit deep cognitive processing, especially in routine or command-based interactions.

The non-significant effects of both cognitive involvement and several usage motives suggest important boundary conditions for existing theories of media satisfaction. These results underscore the need for more nuanced theoretical models that account for the specific interaction modalities and usage contexts of AI-driven interfaces like Alexa. From a design perspective, the findings also highlight potential areas for enhancement—such as developing features that stimulate deeper cognitive engagement or broaden the range of satisfying experiences beyond entertainment and utility.

Concerning emotional involvement, the results from Pearson correlations and the regression indicated that satisfaction is irrelevant to feeling negative emotions, and feeling negative emotions was also not a predictor of Alexa usage satisfaction. It is understandable that feeling positive emotions while using was a positive predictors of Alexa satisfaction. A similar research finding on reality television programs is that more satisfied viewers are those who feel positive emotions while viewing (Godlewski & Perse, 2010). It is surprising, however, that experiencing negative emotions did not significantly predict lower satisfaction. Even though there are some negative feelings from consumers, negative feelings are much less than positive ones. However, it may be too early to say that feeling negative emotions is not a significant predictor of usage satisfaction in the whole population of Alexa users because of the current development stage of AI products (Fidler, 1997).

7. Limitations and Future Directions

This study, like all studies, has limitations. One limitation of this study is that different diversities of AI products were not considered because they were hard to control. Obviously, different kinds of AI products for different uses highlight different features, which might give rise to different pictures of using motives, cognitive and emotional involvement, and satisfaction. Future research should examine and compare how motives, involvement, and satisfaction differ across various AI product types, such as those based on assistance, gaming, romance, art creation, image recognition, and more. Also, further studies should investigate the intrinsic and extrinsic factors influencing user satisfaction with different AI products in the evolving and interactive media landscape.

Furthermore, the data may be relatively small and less typical. Although an online survey was used in order to collect responses from people of different ages and backgrounds, the majority of survey respondents were still young people, which will possibly bring about less accurate results. More observation or experiment participants should be recruited. Future studies could adopt more diversified sampling strategies—such as quota sampling or stratified random sampling—to ensure broader demographic representation and enhance external validity. The data for this study were collected when that use of AI products was gradually rising. The popularity of some types of AI products has even increased. According to Clark (2015), 2015 was a landmark year for artificial intelligence, with the number of software projects that use AI within Google increasing from a “sporadic usage” in 2012 to more than 2700 projects.

In addition, although Entertainment–Information and feeling positive emotions while using were identified as all positive predictors of Alexa satisfaction in this study, all of the scales were drawn from traditional media research and social media research, which may not best fit the research on AI products. Admittedly, AI products, traditional media, and social media have a lot in common, but there are still many detailed and nuanced differences. While these measures provided a useful starting point, they may not fully capture the nuanced dynamics unique to human–AI interactions. Future research could consider validating these scales specifically within the AI context or developing new instruments that more accurately reflect the affordances, interactivity, and perceived agency of AI systems. In order to self-construct more suitable scales to measure the relationship of Alexa satisfaction with using motives and cognitive and emotional involvement, a mixed-methods approach, incorporating both qualitative and quantitative research methods, should be employed. Also, it is possible that some ignored variables contribute to usage satisfaction, such as other using motives. The mixed research method will not only help us deeply explain the reasons behind the existing research findings but also help us discover many new variables and relations.