학술

Behavioral intention of gen Z towards society 6.0 in Nepal: An extension of UTAUT using PLS-SEM and ANN

PLOS ONE
CC BY
이 매체는 공공·자유 라이선스로 본문을 직접 표시합니다.

Figures
Abstract
Society is inherently dynamic and has continuously evolved from Society 1.0 to Society 6.0. As an emerging paradigm, Society 6.0 extends the technology-oriented vision of Society 5.0 by integrating technological advancement with ethical governance, ecological sustainability, and inclusive well-being. Nepal, one of the fastest urbanizing countries in South Asia, faces increasing pressure to align with such evolving societal models. In this context, this study examines the behavioral intention of Gen Z toward Society 6.0 in Nepal. Grounded in a post-positivist philosophy, the study adopts a cross-sectional research design. Primary data were collected from 1,039 Gen Z respondents across all seven provinces using a stratified random sampling technique. The study employs a hybrid analytical approach combining PLS-SEM and ANN, using SmartPLS 4.1 and SPSS 29.0. The Unified Theory of Acceptance and Use of Technology (UTAUT) model was extended by incorporating societal trust as an additional construct, thereby offering a novel contribution that captures the role of trust as a critical determinant in the adoption of Society 6.0. The PLS-SEM results reveal that performance expectancy (β = 0.345, p < 0.001), societal trust (β = 0.239, p < 0.001), social influence (β = 0.176, p < 0.001), and facilitating conditions (β = 0.160, p < 0.001) significantly influence behavioral intention, while effort expectancy shows no significant direct effect. Mediation analysis indicates that performance expectancy mediates the relationship between effort expectancy and behavioral intention, and societal trust mediates the relationship between social influence and behavioral intention. The model explains 61.6% of the variance in behavioral intention (R2 = 0.616). The ANN findings further confirm performance expectancy as the most influential predictor, followed by societal trust, facilitating conditions, and social influence. The study offers practical guidance for policymakers to strengthen trust-driven governance and for educators to redesign digital curricula, alongside valuable insights for technology developers and civil society organizations advancing Society 6.0 in Nepal and the world.
Citation: Siddique MR, Devkota N, Lawaju P, Lowar DB, Paudel UR, Chapagain R, et al. (2026) Behavioral intention of gen Z towards society 6.0 in Nepal: An extension of UTAUT using PLS-SEM and ANN. PLoS One 21(11): e0362300. https://doi.org/10.1371/journal.pone.0362300
Editor: Mohit Kumar, King Mongkut's University of Technology North Bangkok, THAILAND
Received: May 6, 2026; Accepted: October 21, 2026; Published: November 20, 2026
Copyright: © 2026 Siddique et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data required to replicate the findings of this study are provided in the Supporting Information files accompanying this article.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared no competing interests exists.
1. Introduction
The inherent nature of society is characterized by dynamism, which drives its evolution [1]. Accordingly, its development can be understood through successive stages, spanning from Society 1.0 to the emerging paradigm of Society 6.0 [2–,3]. Society 1.0 was the earliest version of human development, emerging before the 10th century in the pre-industrial era [4]. As this society relied on primitive means of livelihood, it is commonly known as a “Hunting and Gathering Society”, led by people referred to as “nomads” [5]. They depended on naturally available resources rather than organized production, resulting in a non-productive economy, and utilized simple tools made from stone or bone [6]. The second version of society is known as Society 2.0, which evolved between the 10th and 17th centuries [4,7]. According to Huang et al. [8], this stage introduced agriculture and the domestication of animals, leading to a transition from nomadic to sedentary lifestyles. Therefore, it is commonly referred to as the “Agrarian Society” [3,4]. This period marked the shift to a productive economy, accompanied by early mechanization and the continuation of manual craftsmanship [8,9].
Society 3.0 is known as the “Industrial Society,” as it emerged following the First Industrial Revolution in Britain between the 17th and mid-19th centuries [4,7,10]. It witnessed a significant transition from an agricultural economy to an industrial one, driven by the introduction of factories and machines [11,12]. In a similar context, Society 4.0 is known as the “Information/Smart Society,” representing the fourth stage of societal evolution [10,13]. It emerged between the 1950s and the 2015s and is primarily founded on the concept that “knowledge is power,” gaining momentum after World War II as globalization accelerated [3,4]. As per Turečková et al. [14], the rise of the internet and Information Communication Technologies (ICT) enabled global connectivity, online communication, and information sharing in Society 4.0. Therefore, this era shifted the focus from industrial economy to digital economy, paving the way for smart ecosystems [9,14].
Society 5.0 is the successor to Society 4.0 and is defined as a “Super-Smart Society” [6–8]. It was first conceptualized by Japan in 2016 as part of its Fifth Science and Technology Basic Plan [3]. This era envisions a human-centered society in which economic growth and the resolution of social challenges coexist through the integration of cyberspace and physical space, enabled by Industry 4.0 technologies such as Artificial Intelligence (AI), Internet of Things (IoT), robotics, cloud computing, big data and analytics, additive manufacturing (3D printing), and cybersecurity [8,9,15]. However, critics argue that despite its promise, Society 5.0 has deepened social inequalities through the digital divide [13]. In Society 5.0, benefits remain concentrated among technologically empowered groups, while marginalized populations face exclusion due to limited access and capabilities [12,13]. Further, although digital innovations enhance efficiency, they also intensify competition and reduce opportunities for vulnerable groups [16,17]. Additionally, AI-driven automation is expected to displace millions of jobs globally, while also raising environmental concerns, as advanced technologies generate high emissions and resource consumption [18]. Consequently, this has paved the way for Society 6.0 that aims to balance technological progress with social and ecological well-being.
Society 6.0 represents the most recent and advanced stage of societal evolution, embodying a human-centered paradigm that redefines technological innovation as a means of promoting ethical governance, ecological sustainability, and inclusive well-being [19]. This era, conceptualized as a “Well-Being Society,” evaluates progress through the quality of human and planetary well-being rather than through technological efficiency or economic growth alone [2,3,10]. As a successor to Society 5.0, Society 6.0 is grounded in the value-oriented principles of Industry 5.0, formally proposed by the European Commission in 2021, which redirected industrial priorities from simple digitalization toward human physical and mental health and environmental safety [13,19]. Within this framework, technological innovation functions not as an end in itself but as an instrument for improving human life and fostering a socially responsible society [20,21]. For example, while Society 5.0 integrates cyberspace and physical space to address social issues and support economic development [22,23], Society 6.0 extends this vision by emphasizing AI and advanced technologies to enhance the real and virtual well-being of all stakeholders [2,3,24]. Society 6.0 is therefore distinguished from Society 5.0 not merely in degree but in foundational orientation: whereas Society 5.0 deployed Industry 4.0 technologies to optimize existing socio-economic systems, Society 6.0 repositions advance and next-generation technology as a subordinate instrument of human-defined values, ecological boundaries, and democratic governance [25,26].
The paradigm of Society 6.0 further resonates with alternative economic approaches, such as the Economy for the Common Good, which prioritize collective welfare, social justice, and sustainability over profit maximization [27]. This is a key point of departure from Society 5.0, which was framed around making economic development and the resolution of social issues compatible through cyber-physical integration, and so retains economic growth as a co-equal objective [26–28]. Society 6.0, by contrast, is proposed as a socially responsible society that extends social responsibility from the corporate level to society as a whole and seeks to move beyond market-driven models that benefit only a few, treating economic activity as a means to well-being rather than an end in itself [3,10–13]. In this context, Society 6.0 presents an integrative and value-driven framework in which technological, economic, and environmental dimensions are mutually reinforcing, thereby fostering a sustainable vision of future society globally [19,20,26,27].
Despite its promise, the adoption of such an advanced societal paradigm, where digital and AI-enabled technologies dominate almost sector, such as education [29], tourism [30,31], hospitality [32], agriculture [33], finance and banking [34,35], e-governance and smart cities [36–38], manufacturing and industry [21,39], accounting [40], and environmental health [41,42], remains a persistent challenge for developing nations. This is largely due to infrastructural gaps, limited digital literacy, weak institutional governance, and constrained public investment, which often confine broader policy visions to paper rather than translating them into practice [8]. These challenges are particularly pronounced in South Asia, where infrastructural constraints and socio-economic disparities intersect with the aspirations embedded in national digital transformation strategies [19,26]. In such contexts, low-income and middle-income economies frequently struggle to balance rapid technological adoption with equitable access, raising concerns that the well-being-oriented promise of Society 6.0 may remain concentrated in advanced and developed nations unless proactively addressed [41–44]. Nepal, therefore, emerges as a critical region of focus within South Asia.
Nepal, as one of the fastest urbanizing countries in South Asia with an annual urban population growth rate of 6%, faces increasing pressure to align with emerging societal models [45]. This shift is reflected in national policy initiatives, such as the introduction of the “Smart City” concept in the 2016/17 fiscal budget, which proposed the development of ten smart cities with an NPR 10 million for each, including Palungtar, Waling, Lumbini, Dadeldhura, and six unnamed cities [46,47]. The 15th Development Plan (2019/20–2023/24) defined smart cities as a pathway to sustainable economic and social prosperity through planned urbanization, a vision further continued in the 16th Development Plan (2024/25–2028/29), which aims to establish 13 smart cities by 2029 [48,49]. This direction is reinforced by the Digital Nepal Framework (2019), which declared the period 2024–2034 as the “IT Decade,” with an investment of NPR 7.5 billion allocated across 80 digital initiatives spanning eight sectors [50]. In a recent initiative, the National AI Policy, in 2025, aims to develop 5,000 skilled human resources within five years and position Nepal among the top 50 countries in the Global AI Readiness Index [51]. This reflects Nepal’s efforts to become a technologically advanced nation by strengthening its infrastructure to adapt to evolving societal models such as Society 6.0.
However, the transition toward Society 6.0 depends not only on technological advancement but also on broader societal acceptance. In this context, Generation Z or Gen Z, comprising 32% (9.2 million) of the total population of Nepal and being highly technology-oriented, plays a pivotal role in the adoption of such emerging societal concepts [52]. Despite this, their behavioral intention toward Society 6.0 remains largely unexplored, highlighting a significant research gap. While prior studies have extensively examined adoption behavior in the context of Industry 4.0, Society 5.0, Agriculture 5.0, smart cities, energy-saving and general digital technology adoption, limited attention has been given to Society 6.0, particularly among digital native Gen Z in developing countries [7–12,33,37,38,53,54]. Additionally, prior UTAUT-based studies have largely overlooked societal trust as a determinant of technology adoption, despite its growing relevance in contexts where governance transparency, data privacy, and ethical accountability critically shape public confidence in emerging technological systems [21,36,39,41,55–57]. Further, given that Society 6.0 is inherently a multi-dimensional concept, this study operationalizes Society 6.0 as behavioral intention of Gen Z to accept and engage with advanced, human-centric and sustainable technologies perceived to enhance human well-being and ecological sustainability through ethical and trustworthy governance [3,13,23,35,41]. Therefore, the present study extends the UTAUT framework by incorporating societal trust as a novel construct, thereby offering a more holistic understanding of the trust-based mechanisms driving adoption of Society 6.0 among Gen Z in the context of developing nations. Consequently, the primary aim of this study is to investigate the behavioral intention of Gen Z toward Society 6.0 in Nepal.
2. Research methodology
2.1. Theoretical framework and hypothesis
A theoretical framework provides a structured foundation of relevant theories that guide the development of arguments grounded in existing knowledge [58]. This study adopts the Unified Theory of Acceptance and Use of Technology (UTAUT), originally proposed by Viswanath Venkatesh, Michael G. Morris, Gordon B. Davis, and Fred D. Davis in 2003, as a comprehensive model for explaining user intentions and behaviors toward emerging technologies [59]. Aside from UTAUT, other models such as the Theory of Reasoned Action (TRA), the Theory of Planned Behavior (TPB), and the Technology Acceptance Model (TAM) were also reviewed. TRA and TPB explain behavior primarily through attitudes, subjective norms, and perceived behavioral control, while TAM narrows this further to perceived usefulness and ease of use, both frameworks being largely confined to individual-level cognitive and attitudinal predictors of discrete technology use [19,43,44,59].
Besides, UTAUT identifies four core determinants, performance expectancy (PE), effort expectancy (EE), social influence (SI), and facilitating conditions (FC), that collectively capture the multifaceted drivers of technology acceptance [39,55–57]. The selection of UTAUT over alternative behavioral models is theoretically justified by its integrative and holistic nature, as it synthesizes key constructs from prior theories, such as TRA, TBP and TAM, while simultaneously accounting for both individual and contextual dimensions [59]. Specifically, PE and EE reflect individual cognitive evaluations of usefulness and ease of use, whereas SI and FC incorporate the influence of social environments and enabling infrastructures [21,36,39,41]. Furthermore, the robustness of the UTAUT model has been verified across diverse technological and cultural contexts, including Industry 5.0 [21], AI-driven education [29], AI-enabled hospitality services [32], e-governance, smart cities [36–38], green innovation technologies [39], accounting information systems [40], and learning management systems [55,56]. Therefore, this dual-level explanatory capacity and robustness across different contexts make UTAUT particularly suitable for examining the preparedness intention of Gen Z to engage with the technology-driven ecosystem of Society 6.0 in Nepal, where technological competence is deeply intertwined with social and institutional contexts.
Fig 1 presents the conceptual framework for assessing behavioral intention of Gen Z towards Society 6.0 based on UTAUT, adapted and modified from [21] and [36]. It shows that PE, EE, SI, and FC directly influence behavioral intention (BI) of Gen Z toward Society 6.0, with societal trust (ST) also having a direct effect. Additionally, PE mediates the relationship between EE and BI, while ST mediates the relationship between SI and BI.
Performance Expectancy (PE).
Performance Expectancy (PE) is the degree to which an individual believes that using a specific technology will help them attain gains in job performance [59]. Doven et al. [21] found that PE significantly affected office professionals’ intention to adopt robots in Industry 5.0 while also mediating the relationship between effort expectancy and intention. Similarly, Al-Haraizah et al. [57] demonstrated that PE positively shaped trust in smart city technologies, thereby reinforcing adoption intentions. Studies on sustainable innovations, such as Shahzad et al. [39], further confirm PE as a major predictor of behavioral intention. This is corroborated by Al-Mamary [29] and Azhar et al. [32], who found performance-related beliefs to be a key driver of AI adoption in education and hospitality contexts, respectively, reinforcing PE's cross-contextual relevance. Based on these findings, it is expected that Gen Z individuals will exhibit a stronger intention to engage with Society 6.0 technologies when they perceive them as enhancing future opportunities.
H1: Performance Expectancy positively influences behavioral intention of Gen Z to adopt Society 6.0.
Effort Expectancy (EE).
Effort Expectancy (EE) refers to the degree to which an individual perceives that using a technology will be free of effort [36,37]. Pujiati et al. [42] reported a strong effect of EE in shaping individuals’ engagement with AI technologies for enhancing environmental health. Similarly, Al-Haraizah et al. [57] demonstrated that EE significantly contributes to trust toward smart city innovations, which in turn strengthens adoption intentions. Similarly, Alquhaif and Al-Mamary [40] found that effort expectancy shaped behavioral intention. However, Doven et al. [21] found that while EE significantly influences PE, its direct effect on behavioral intention was insignificant. This inconsistency suggests EE's effect on behavioral intention may be context-dependent. Thus, it is reasonable to anticipate that Gen Z will exhibit higher readiness to adopt Society 6.0 technologies when they perceive them as requiring minimal effort.
H2: Effort Expectancy positively impacts Performance Expectancy in the context of Society 6.0.
H3: Effort Expectancy positively influences behavioral intention of Gen Z to adopt Society 6.0.
Facilitating Conditions (FC).
Facilitating Conditions (FC) refer to the availability of resources that enable effective technology adoption [21]. Al-Haraizah et al. [57] found that FC significantly enhance trust in smart city technologies, thereby increasing adoption willingness. Similarly, Popova and Zagulova [37] demonstrated that FC are positively associated with both behavioral intention and actual technology use. Mishra et al. [33] also reported a positive influence of FC on the intention to adopt Agriculture 5.0 technologies. This is also consistent with Al-Mamary [55], who found institutional and technical support significantly facilitated students’ continued use of learning management systems, reinforcing the role of FC across technology-adoption contexts. Consequently, when Gen Z perceives that sufficient resources, guidance, and institutional support are available, their intention to adopt Society 6.0 technologies is likely to increase.
H4: Facilitating Conditions positively influence behavioral intention of Gen Z to adopt Society 6.0.
Social Influence (SI).
Social Influence (SI) refers to the degree to which an individual perceives that important peers, groups, or societal expectations affect their decision to use a technology [41,42]. Doven et al. [21] demonstrated that SI significantly shapes behavioral intention toward robot use. In smart city contexts, Al-Haraizah et al. [57] found that SI positively affects trust in emerging technologies, thereby enhancing adoption willingness. This finding aligns with Al-Mamary and Alraja [44], who reported that subjective norms significantly influence digital entrepreneurship intentions. It suggests that social referents shape both behavioral intention and the trust evaluations that precede it [44]. Similarly, Hooda et al. [36] found that SI strengthens trust in e-government platforms. Given that Gen Z is highly responsive to peer norms, social trends, and digital communities, SI is expected to play a significant role in shaping their engagement with Society 6.0 technologies.
H5: Social Influence positively impacts the behavioral intention of Gen Z to adopt Society 6.0.
H6: Social Influence positively impacts societal trust in the context of Society 6.0.
Societal Trust (ST).
Societal Trust (ST) refers to users’ confidence in the safety, fairness, and responsible governance of advanced technologies [36]. Trust reduces perceived risks and strengthens willingness to engage with emerging digital systems [60]. For example, Alkdour et al. [38] found that societal trust significantly enhances users’ readiness to adopt smart city innovations. Azhar et al. [32] similarly found trust central to consumer acceptance of AI-driven hospitality services. Beyond this direct effect, trust has also been found to exert an indirect influence on behavioral intention. For instance, Hooda et al. [36] found that trust mediates the relationship between social influence and behavioral intention. This is due to the fact that when peers, institutions, or media support a technology, people are more likely to trust it, which then increases their behavioral intention [38,60]. Consequently, when Gen Z perceives technologies within Society 6.0 as trustworthy, their intention to adopt these technologies is likely to increase.
H7: Societal Trust positively influence Gen Z’s behavioral intention to adopt Society 6.0.
Mediating Hypotheses.
PE and ST are proposed as mediators in this study. PE is proposed to mediate the relationship between EE and the BI of Gen Z to adopt Society 6.0. This is because when technologies are perceived as easy to use, they are also seen as more useful, which in turn enhances adoption intention [39,41,42]. This mechanism is supported by Doven et al. [21], who found that PE mediated the relationship between EE and office professionals’ behavioral intention to adopt robots in Industry 5.0. This reflects cognitive appraisal processes in technology acceptance, where ease of use functions as an antecedent belief shaping utility judgments rather than independently driving adoption [22,59]. Likewise, ST is hypothesized to mediate the relationship between SI and behavioral intention, as peer norms and social expectations can enhance perceptions of trustworthy technological practices, thereby encouraging adoption [38]. Evidence for this mediating effect is provided by Hooda et al. [36], who reported that social influence strengthens trust in e-government platforms. This aligns with the broader trust literature, where social referents shape behavior indirectly by first cultivating trust evaluations, which then function as the proximal mechanism driving behavioral intention [32,37,59]. Therefore, the mediating hypothesis are:
H8: Performance Expectancy mediates the relationship between Effort Expectancy and behavioral intention toward Society 6.0.
H9: Societal Trust mediates the relationship between Social Influence and behavioral intention toward Society 6.0.
2.2. Variable table
In this study, the technological orientation of Society 6.0 is framed around intelligent systems, referring to advanced and future-oriented AI-enabled digital systems that sense, learn from data, and support or automate human decisions and tasks. These technologies are conceptualized not merely as tools for efficiency but as mechanisms for advancing human well-being, ethical responsibility, and social value, which constitute key foundations of Society 6.0. Consistent with this perspective, Society 6.0 connects next-generation AI with public-interest technologies guided by societal and environmental priorities [13,20], while Industry 5.0 emphasizes human-centric, sustainable, and resilient technological development [2,3,19,20,26,27]. Accordingly, the measurement items incorporate ecological sustainability, privacy, fairness, transparency, ethical AI use, and responsible governance. PE, FC, and BI were further contextualized around emerging, AI-enabled, sustainable, and human-centric intelligent systems, while ST captures fairness, privacy protection, transparency, and responsible use of such advanced technologies. Thus, the measurement framework integrates technology acceptance with the broader societal values that characterize Society 6.0 rather than relying solely on generic technology acceptance indicators. Table 1 presents the constructs and corresponding measurement items used in the study.
2.3. Research design
Research design is the overall plan outlining how data are collected, measured, and analyzed to achieve research aims. Following the research onion by Saunders et al. [61], this study adopts a post-positivist philosophy with a deductive approach, using a survey strategy and cross-sectional time horizon. A mono-method quantitative design was employed, with primary data collected through a structured questionnaire and analyzed using both descriptive and inferential techniques.
2.4. Study area and population
For this study, Nepal is selected as the research area, which is administratively divided into seven provinces and 77 districts (Fig 2). The selection of Nepal is driven by its ongoing technological transition. Currently classified as a Least Developed Country, Nepal is in a transformative phase and is expected to graduate to a developing, middle-income nation by 2026 [63]. This shift is characterized by increasing digitalization and a gradual move toward automation, particularly in urban sectors such as governance, finance, education, and service delivery. National initiatives, including the Digital Nepal Framework (2019) and the AI Policy (2025), further reflect the country’s commitment to technological advancement. Therefore, Nepal provides a relevant context to examine adoption of Society 6.0. Similarly, the target population of this study comprises Gen Z individuals in Nepal, defined as those born between 1997 and 2012 [52]. The rationale for selecting Gen Z lies in their role as primary users and future drivers of emerging digital technologies, making them a relevant population for examining Society 6.0 [23,35,41].
Map created by the authors using ArcGIS. Base map data for the provincial boundaries of Nepal and the South Asia inset were obtained from geoBoundaries (gbOpen release; CC BY 4.0) [62].
2.5. Sampling technique and sample size
this study employed a combination of stratified and random sampling technique. In the first stage, the Gen Z population aged 18–28 was stratified based on seven provinces of Nepal: Koshi, Madhesh, Bagmati, Gandaki, Lumbini, Karnali, and Sudurpaschim. In the second stage, one university was randomly selected from each province, except for Bagmati, where two universities were included because of its relatively larger population and concentration of higher education institutions. In the third stage, students from the selected universities were invited to participate voluntarily. As provincial Gen Z population estimates and university enrollment frames were unavailable, respondents within each institution were selected using probability-based random sampling. Most Gen Z individuals in Nepal are in their early twenties and pursuing university education, while those in their late twenties are comparatively fewer due to overseas employment [35,41]. Therefore, the sample primarily represents university-enrolled Gen Z individuals across all seven provinces. However, as many universities in Nepal are concentrated in urban areas, this sampling approach may underrepresent Gen Z populations from rural and geographically remote regions. Further, given that the size of Gen Z in Nepal is 9.2 million, as reported by Karki [52], the required sample size for this research was determined using the following formula as suggested by Ahmed [64].
In this formula,
and Zα/2 denotes the critical value for a two-tailed test (2.576 at 99% confidence), MOE is the margin of error (5%), P is the estimated proportion (0.5), and N is the population size. The sample size is calculated as follows:
When this sample size adjusted to finite population size, the adjusted sample size for the study is:
Further, the study accounts for a 5% non-response rate, acknowledging that some respondents may decline to participate. Thus, the actual sample size is calculated as: 663.53+ (663.53 × 0.05) = 663.53 + 33.18 = 696.7 ~ 700. The selection of 700 respondents is further justified by the study’s stratified sampling technique, with seven strata corresponding to the seven provinces of Nepal, each contributing 100 respondents. However, the researcher obtained 1,039 valid responses from all seven provinces of Nepal. Of the total responses, 22.81% were from Bagmati, 19.54% from Sudurpaschim, 13.47% from Karnali, 11.65%from Madhesh, 11.26% from Koshi, 10.88% from Lumbini, and 10.39% from Gandaki (Table 2). Further, a detailed province-wise breakdown of the sample collection is presented visually in Fig 2, which was prepared by authors using ArcGIS software.
2.6. Research instruments and data collection
2.7. Data analysis tools and technique
This study employed both descriptive and inferential techniques for data analysis. Descriptive statistics were used to summarize respondents’ socio-demographic characteristics using MS Excel. Data were downloaded from KoboToolbox in Excel format and summarized using frequencies and percentages. Inferential analysis was conducted using Partial Least Squares Structural Equation Modeling (PLS-SEM) in SmartPLS 4.1 and Artificial Neural Network (ANN) analysis in SPSS 29.0. PLS-SEM was conducted in three phases. First, common method bias and the measurement model were assessed for reliability and validity. Second, the structural model was evaluated using 10,000 bootstrap samples to test the proposed hypotheses and mediation effects. Finally, robustness tests for linearity, unobserved heterogeneity, and endogeneity were conducted to verify the stability of the findings.
Following PLS-SEM, latent variable scores were extracted from SmartPLS 4.1 and imported into SPSS 29.0 for ANN analysis. Because ANN does not establish causal relationships, it was used only as a complementary approach to assess the relative predictive importance of the constructs. Consistent with prior methodological guidance, only predictors that were significant in the PLS-SEM analysis were included [65,66]. Three ANN models were specified: Model 1, with FC, PE, SI, and ST predicting BI; Model 2, with EE predicting PE; and Model 3, with SI predicting ST. In each model, the outcome was treated as the dependent variable and the predictors as covariates, rescaled by standardization, while the dependent variable was rescaled by adjusted normalization (correction = 0.02).
All 1,039 valid responses were randomly partitioned into 70% training and 30% testing samples. Based on this partition, a custom ANN architecture with two hidden layers was specified, with the number of units in each layer determined automatically. To capture nonlinear relationships, the hyperbolic tangent function was applied to the hidden layers, while the sigmoid function was used in the output layer to constrain predictions between 0 and 1, consistent with the rescaled dependent variable [67]. The networks were then trained in batch mode using the scaled conjugate gradient algorithm (initial lambda = 0.0000005, initial sigma = 0.00005, interval center = 0, interval offset = ±0.5), with training terminated after one consecutive step without a decrease in error. All other training parameters were kept at SPSS default settings.
Given that overfitting is a key concern in ANN analysis, the architecture was selected automatically, the testing sample was excluded from model tuning, and the procedure was repeated across 10 random partitions to enhance model generalizability. Predictive accuracy was subsequently evaluated using Root Mean Squared Error (RMSE) = √(SSE/N) (where SSE is the sum of squared errors and N is the number of cases in the sample) for both training and testing samples, expressed on the rescaled (0–1) scale, with comparable RMSE values indicating satisfactory generalization [66,68]. The SSE and N were obtained from SPSS29.0, while the RMSE was computed in MS Excel for both training and testing datasets. Finally, sensitivity analysis was conducted for Model 1 to determine the relative importance of each predictor, with normalized importance calculated by dividing each predictor's average importance by the highest observed importance value [69,70]. Since Models 2 and 3 each contain a single predictor, their importance is 100% by construction, and they were used only to assess predictive accuracy.
3. Results
3.1. Descriptive analysis
Socio-demographic Information.
The study collected data from 1,039 Gen Z respondents across Nepal (Table 3). Regarding sex distribution, 51.11% were male and 48.89% were female, indicating a nearly balanced sample. Participants’ ages ranged from 18 to 28 years, aligning with the typical Gen Z cohort born between 1997 and 2012 [23]. The largest shares were 34.65% each for the 18–20 and 21–23 groups, followed by 18.77% for ages 24–26 and 11.93% for 27–28. This distribution shows a strong concentration in the early twenties, representing the core segment of Gen Z [41]. Most respondents were single (89.32%), while 10.39% were married and 0.29% were divorced, consistent with Herawati et al. [71], who observed that Gen Z individuals are typically unmarried in their early years.
As illustrated in Fig 3, the provincial origin of the respondents is led by Bagmati (22.81%), followed by Sudurpaschim (19.54%), Karnali (14.47%), Madhesh (11.65%), Koshi (11.26%), Lumbini (10.88%), and Gandaki (10.39%). Currently, a majority reside in Bagmati (56.59%), with notable shares in Sudurpaschim (17.04%) and Karnali (12.51%), while other provinces contribute marginally. This concentration reflects strong pull factors of Kathmandu Valley, including education, digital infrastructure, and employment opportunities [35]. Additionally, the educational background indicate that most respondents have bachelor’s degrees (53.9%), followed by +2 (25.41%) and master’s (17.23%), indicating a highly educated Gen Z cohort [55].
Tribhuvan University dominates enrollment (30.9%), reflecting its national prominence as the largest university [72], followed by other major universities. In terms of discipline, management leads (54.57%), followed by science and technology (21.17%) and law (11.16%), highlighting a preference for career-oriented fields. Economically, most respondents are students (78.06%), with only 18.67% employed, indicating early career stages and limited financial independence [35,71]. Among the employed, income levels remain generally low, with most earning below NPR 20,000 and only a small fraction exceeding NPR 100,000, reflecting modest earning capacity among Nepalese Gen Z.
3.2. Inferential analysis
Structural Equation Modelling Analysis.
Common Method Bias and Measurement Model Assessment: Henseler et al. [73] note that SEM-based analysis may suffer from systematic measurement error when both independent and dependent variables are collected using the same method, potentially inflating relationships. Thus, testing for common method bias (CMB) is essential. In this study, CMB was assessed using the full collinearity test with Variance Inflation Factor (VIF), where values below 3.3 indicate no bias [74]. The highest VIF was 2.268 for Facilitating Conditions (FC) (Table 4), suggesting that CMB is not a concern.
The measurement model in SEM specifies the relationships between observed indicators and their underlying latent constructs, ensuring the reliability and validity of the measures [59,61]. In this study, it was evaluated using internal consistency reliability, convergent validity, and discriminant validity. Internal consistency reliability was assessed through Cronbach’s alpha (CA) and composite reliability (CR). According to Hair et al. [75], CA assesses inter-item consistency based on the average correlations among indicators, whereas CR incorporates indicator loadings, offering a more robust estimate of reliability in SEM. Hair et al. [76] suggest that CA values above 0.70 are acceptable, although values below 0.60 may be tolerated when CR exceeds 0.70. Similarly, CR values of 0.70 or higher indicate adequate reliability [73]. As shown in Table 4, all constructs met these thresholds, confirming adequate internal consistency reliability.
Convergent validity was assessed to determine whether indicators adequately represent their respective latent constructs, using factor loadings and average variance extracted (AVE). Factor loadings reflect the strength of the relationship between indicators and constructs, while AVE captures the proportion of variance explained [77]. According to Hair et al. [78], AVE values of 0.50 or higher signify adequate convergent validity. Similarly, factor loadings of 0.70 or above indicate strong indicator reliability, although values between 0.60 and 0.70 may be acceptable if AVE exceeds 0.50 [73,76]. However, items with loadings below 0.40 should be removed due to weak contribution [75]. In this study, all constructs achieved AVE values above 0.50, and most loadings exceeded 0.60 (Table 4), confirming convergent validity. Similarly, discriminant validity was evaluated to ensure that constructs are empirically distinct using the Heterotrait–Monotrait (HTMT) ratio and the Fornell–Larcker (F-L) criterion. The HTMT ratio assesses the degree of similarity between constructs, with values ≤ 0.85 indicating a strict threshold and ≤ 0.90 a more lenient criterion [33,34]. As reported in Table 5, all values were below 0.90, confirming adequate discriminant validity. The F-L criterion compares the square root of the AVE of each construct with its inter-construct correlations, requiring the former to be higher [73–75]. Table 6 shows that this condition was satisfied for all constructs, further supporting discriminant validity.
After confirming the reliability and validity of the instrument, model fit was assessed. As per Aalam et al. [34], model fitness denotes the extent to which an SEM adequately represents observed data by capturing relationships among constructs and their indicators. In PLS-SEM, the Standardized Root Mean Square Residual (SRMR) is a widely used goodness-of-fit measure that evaluates the discrepancy between observed and model-implied correlations [77]. As cited by Hair et al. [78], SRMR values below 0.08 are considered acceptable. In this study, the SRMR value of 0.061 falls within the acceptable range, indicating a satisfactory model fit.
Structural Model Assessment: Structural model assessment evaluates the relationships among latent constructs and tests the proposed hypotheses [75,76,78]. This assessment begins with examining the inner model for collinearity, followed by path coefficient analysis and hypothesis testing. In this study, all independent variables were assessed against the dependent variable using the VIF. The findings show that all the VIF values were below the recommended threshold of 3.3 [74], indicating that multi-collinearity does not threaten the model’s validity. Subsequently, path analysis was performed using 10,000 bootstrap samples in SmartPLS 4.1 (Fig 4).
The path analysis shows that all endogenous constructs, namely PE, EE, FC, SI and ST, have a positive impact on BI. Additionally, in this study, coefficient of determination (R2) indicates that the model explains 61.6% of the variance in BI through its independent variables, reflecting moderate explanatory power. This is based on the threshold values of R2, which are commonly interpreted as 0.75 for substantial, 0.50 for moderate, and 0.25 for weak explanatory power [77,78]. Similarly, EE explains 37.2% of PE, while SI accounts for 24.4% of ST, both demonstrating weak but acceptable explanatory power.
Hypothesis Testing: In this study, seven direct hypotheses were examined, of which six are statistically significant at the 5% level (Table 7). PE shows a strong positive effect on BI (β = 0.345, p = 0.000), supporting H1 and indicating that Gen Z is more likely to adopt Society 6.0 when it is perceived as useful. EE significantly influences PE (β = 0.610, p = 0.000), supporting H2, suggesting that ease of use enhances perceived usefulness. However, EE does not have a significant direct effect on BI (β = 0.038, p = 0.239), leading to the rejection of H3, as the confidence interval includes zero and the t-value is below 1.96. FC significantly affects BI (β = 0.160, p = 0.000), supporting H4 and highlighting the importance of infrastructure and resources in Nepal. Similarly, SI has a significant positive impact on BI (β = 0.176, p = 0.000), supporting H5, indicating the role of social and peer influence. SI also strongly affects ST (β = 0.494, p = 0.000), supporting H6, showing that social context contributes to trust formation. Furthermore, ST significantly influences BI (β = 0.239, p = 0.000), supporting H7, emphasizing the role of trust in technology adoption. Furthermore, mediation analysis reveals that PE significantly mediates the relationship between EE and BI (β = 0.210, p = 0.000), supporting H8, while ST significantly mediates the relationship between SI and BI (β = 0.118, p = 0.000), supporting H9. These results highlight the mediating roles of PE and ST in explaining how EE and SI influence Gen Z’s behavioral intention to adopt Society 6.0 in Nepal.
Similarly, additional analyses, including effect size (f2) and predictive relevance (Q2), were conducted. f2 assesses the relative contribution of each exogenous construct to endogenous constructs, with thresholds of 0.02, 0.15, and 0.35 indicating small, medium, and large effects, respectively [69,70,76]. In this study, all f2 values exceeded 0.02, ranging from FC (lowest) to EE (0.592), followed by ST (0.322), BI (0.145), SI (0.038), and PE (0.03), indicating effects from small to large. Hence, these results reflect a range of effect sizes from small (FC, SI) to large (EE), underscoring the differentiated but meaningful contribution of each predictor within the structural model. Further, Q2, assessed via blindfolding, evaluates predictive relevance, where values above zero confirm predictive capability [74]. Thresholds of 0.02, 0.15, and 0.35 indicate small, medium, and large relevance, respectively [69,70,76,79]. The results show strong predictive performance, with BI (Q2 = 0.494) and PE (Q2 = 0.367) demonstrating large predictive relevance, and ST (Q2 = 0.238) indicating moderate relevance. Therefore, the f2 and Q2 results reinforce the structural model's robustness, confirming that the independent constructs not only exert statistically significant effects but also carry meaningful explanatory and predictive weight within the proposed framework.
Further, robustness checks were conducted to ensure that the findings are stable and not influenced by model assumptions or methodological biases [79,80]. These assessments included linearity, unobserved heterogeneity, and endogeneity. Linearity refers to the assumption that relationships between independent and dependent variables are linear [33]. In this study, linearity was examined using the quadratic effect approach, which tests whether the inclusion of squared terms significantly improves explained variance [80]. The results (Table 8) show that most quadratic effects are not significant (p > 0.05), indicating no evidence of non-linearity. As per Tiwari et al. [81], unobserved heterogeneity is the presence of hidden subgroups within data that exhibit different structural relationships but are not directly observed. In this study, it was assessed using FIMIX PLS-SEM with five segments. Heterogeneity is confirmed when minimum values across model selection criteria occur in different segments [33,70,79,82]. The results show varying minima across criteria such as AIC, BIC, CAIC, HQ, and MDL, along with LogLikelihood and entropy indices, indicating latent subgroups and confirming unobserved heterogeneity. Similarly, endogeneity, referring to correlations between predictors and the error term that may bias estimates, was assessed using the Gaussian Copula approach in PLS-SEM [80]. The results indicate that most paths, including PE → BI, EE → PE, EE → BI, FC → BI, SI → BI, and SI → ST, are not significant, suggesting no endogeneity. However, the ST → BI path (p = 0.038) is significant, indicating potential endogeneity in this relationship. Overall, the robustness checks confirm that the findings are largely stable and generalizable for explaining Gen Z’s behavioral intention toward Society 6.0 in Nepal.
Artificial Neural Network (ANN) Analysis.
PLS-SEM is a useful method for analyzing relationships between variables, but it mainly assumes that these relationships are linear [75,76,83]. Because of this limitation, it may overlook non-linear patterns, which can sometimes lead to less accurate results [84,85]. To address this issue, Artificial Neural Networks (ANN) are introduced as a complementary approach [67]. The two methods are combined because they serve different purposes: PLS-SEM tests theory-based hypotheses about significant relationships, while ANN, a non-parametric method, checks the relative predictive importance of those predictors and their out-of-sample predictive performance [78,85,86]. ANN is an artificial intelligence technique inspired by the human brain, capable of learning from data through repeated training and storing knowledge for prediction [68,87]. Such brain-inspired prediction models are now widely used to decode outcomes from complex input patterns [88]. Because ANN works in a way that does not clearly show how inputs are converted into outputs, it cannot be used to explain or confirm cause-and-effect relationships between variables [65,66]. Therefore, it is not suitable on its own for testing hypotheses. To overcome this limitation, researchers use a two-step hybrid approach. First, PLS-SEM is used to identify and test which relationships between variables are statistically significant [67]. Then, only those significant variables are used as inputs in the ANN model to assess their predictive accuracy and importance [83,84].
The results indicate that, among the tested paths, only the impact of PE on BI (Table 8) shows a significant non-linear relationship, which may not be fully captured by PLS-SEM as it mainly assumes linear effects [75,83]. The remaining paths were consistent with linear relationships, so the ANN is used as a complementary check on the predictive importance of the constructs of behavioral intention of Gen Z toward Society 6.0, not as evidence of widespread non-linear effects. Based on the conceptual model and PLS-SEM results, three ANN models were developed (Table 9). In Model 1, the inputs are PE, FC, SI, and ST, while BI is the output, excluding EE as it was found to be insignificant in PLS-SEM results. Model 2 uses EE as the input with PE as the output, and Model 3 uses SI as the input with ST as the output. The automatically determined architecture of Model 1 comprised four input neurons, two hidden layers with three and two neurons respectively, and one output neuron (4-3-2-1 structure), as shown in Fig 5. Similarly, the architectures of Models 2 and 3 comprised one input neuron, two hidden layers with with one neuron each, and one output neuron (1-1-1-1 structure).
One of the most common problems in ANN analysis is overfitting, which occurs when the model memorizes all training data and loses its ability to generalize, meaning it performs poorly in predicting outcomes for new or unseen data inputs [80,84]. To overcome this issue, the procedure was repeated across ten random 70/30 partitions, and training and testing errors were compared using the RMSE [86]. The RMSE is a standard measure of predictive accuracy of ANN models [85–87], with similar training and testing values indicating good generalization, whereas a markedly lower training error would signal overfitting [67]. The results for both training and testing datasets across all ANN models are presented in Table 9. The RMSE values range from 0.0957 to 0.1389 for both training and testing. Mean training and testing RMSE were close for Model 1 (0.1011 and 0.1017), Model 2 (0.1305 and 0.1281), and Model 3 (0.1323 and 0.1309), with standard deviations of 0.0039 or less across the ten partitions, indicating no material overfitting. These relatively low and consistent RMSE values indicate that all ANN models demonstrate good predictive performance and adequate generalization ability [66,68].
Finally, to assess the relative importance of each predictor in explaining variations in the output, a sensitivity analysis of the ANN models was conducted. The normalized importance of each variable was obtained by dividing its average importance value by the highest importance value [86,87]. For example, the normalized importance of FC was calculated by dividing its average value (0.2156) by the highest value in Model 1 (PE = 0.3298), resulting in 0.6537 or 65.37%, as presented in Table 10. The ANN results indicate that PE is the most influential predictor of BI with the highest normalized importance (100%), followed by ST (80.17%), FC (65.37%), and SI (57.61%). However, this pattern slightly differs from the PLS-SEM findings, where PE also emerged as the strongest predictor (β = 0.345), followed by ST (β = 0.239), SI (β = 0.176), and FC (β = 0.160). As shown in Table 11, in Model 1, PE ranked first in both PLS-SEM and ANN, followed by ST, which ranked second in both approaches. However, the rankings differed for SI, which ranked third in PLS-SEM but fourth in ANN, and FC, which ranked fourth in PLS-SEM but third in ANN. In the single-input ANN models, EE significantly predicted PE, while SI significantly predicted ST, with both relationships showing consistent first rankings across PLS-SEM and ANN. Therefore, both approaches confirm PE as the strongest predictor of BI, while the differences in the rankings of FC and SI indicate variations in their relative predictive importance across the two analytical approaches.
From a practical standpoint, the ANN results should be interpreted as a complementary assessment of the relative predictive importance of the constructs rather than evidence of intervention effects. As mentioned before, normalized importance indicates the contribution of each predictor to the network’s prediction of behavioral intention [86]. It does not represent causal effects or estimate potential gains from specific interventions [87]. The ANN assigns relatively greater importance to FC than SI compared with PLS-SEM, while ST remains the second most important predictor across both approaches. This pattern suggests that trust of the society and enabling conditions can operate jointly, consistent with prior research indicating that the influence of one factor can depend on the presence of other enabling conditions [89]. However, this interpretation remains tentative and requires validation through interaction analysis, longitudinal studies, or experimental designs. In this regard, government agencies and technology providers should consider ST and FC as interconnected drivers, rather than as isolated supporting factors, that can promote the Gen Z’s adoption of Society 6.0.
4. Discussion
The study used the UTAUT model to examine factors influencing the behavioral intention of Gen Z to adopt Society 6.0 in Nepal. The hypothesis testing revealed that eight out of nine proposed direct and indirect relationships were supported, indicating a robust explanatory framework. PE (β = 0.345, p = 0.000) was found to significantly influence BI of Gen Z to adopt Society 6.0 in Nepal, indicating that their adoption decisions are largely driven by perceived benefits. Also, ANN analysis confirmed PE as the strongest independent predictor in this context, with 100% normalized importance, reinforcing its dominant role in shaping adoption intention. This aligns with Rimadias [41], who found that Gen Z adopts AI when a clear functional value is evident. It also aligns with Azhar et al. [32], who identified performance-related beliefs as a key driver of AI-driven hospitality service acceptance, and with Al-Mamary [29], who found perceived usefulness to strongly influence ChatGPT adoption in education. Therefore, in the context of Nepal, where economic security, career mobility, and global competitiveness are major concerns, Gen Z appears to evaluate Society 6.0 technologies based on tangible benefits rather than abstract ideals [34,46,51,77].
EE direct effect on BI (β = 0.038, p = 0.239). This finding aligns with Doven et al. [21], who reported an insignificant relationship between EE and BI in the context of robot co-working adoption under Industry 5.0, while confirming a strong and positive relationship between EE and PE. In the context of Society 6.0, this implies that ease of use by itself is not enough to encourage adoption. This is because Gen Z’s high level of digital familiarity leads them to treat usability as a basic expectation rather than a distinguishing factor [41]. Having grown up with user-friendly digital interfaces, Gen Z is unlikely to perceive usability as a distinguishing feature of new technologies [35]. Instead, ease of use serves as a basic expectation rather than a direct driver of behavioral intention [77]. This finding is consistent with Alquhaif and Al-Mamary [40], who reported that EE influenced adoption intention indirectly through perceived usefulness. Therefore, technologies need to provide clear performance gains and meaningful practical value to effectively influence behavioral intention [33]. The study further found that FC (β = 0.160, p = 0.000) significantly influence Gen Z’s BI to adopt Society 6.0. This result is consistent with prior studies, including Agriculture 5.0 adoption by [33], smart city adoption by [37] and AI adoption by [41]. In the context of Society 6.0, this indicates that robust technical infrastructure, accessible digital technologies, institutional capacity to operate advanced systems, widespread connectivity, and supportive policy frameworks are fundamental drivers of adoption [19,36,43]. In Nepal, where technological resources and internet access remain uneven, particularly in rural areas, strong facilitating conditions are crucial to convert behavioral intention into actual technological engagement.
The study also found that SI (β = 0.176, p = 0.000) and BI of Gen Z to adopt Society 6.0 and ST (β = 0.494, p = 0.000). This finding is consistent with prior research, including smart city adoption by Al-Haraizah et al. [57], green technology adoption by Shahzad et al. [39], and e-government adoption by Hooda et al. [36]. In the context of Society 6.0, this suggests that close family members, peer networks, media, community leaders, and society at large play a pivotal role in shaping perceptions and building trust. For Gen Z, prevailing social norms and strong social endorsement not only influence attitudes but can decisively accelerate or constrain their behavioral intention to embrace transformative technologies, thereby making social validation and trust powerful determinants of adoption [41]. Similarly, ST (β = 0.239, p = 0.000) were found to significantly influence Gen Z’s intention to adopt Society 6.0 in Nepal, emerging as the second most important predictor after PE, with normalized importance of 80.17% as per ANN analysis. This finding aligns with prior research, including Al-Haraizah et al. [57], Hooda et al. [36], Azhar et al. [32] and Alkdour et al. [38]. In the context of developing countries like Nepal, this suggests that factors such as transparency, data privacy, accountability, and ethical governance are fundamental to fostering trust in advanced technologies [36,55]. Without strong societal trust, Gen Z remains cautious, limiting their full engagement with Society 6.0 initiatives. Given the second position as strongest predictor, ST has important policy implications. Government agencies should prioritize transparent AI governance, effective data protection, and clear oversight mechanisms that are communicated to digitally engaged citizens, particularly Gen Z [35,36,41]. In Nepal, integrating accountability and ethical safeguards into initiatives such as the Digital Nepal Framework (2019) and the National AI Policy (2025) could strengthen Gen Z’s trust and support the adoption of Society 6.0 technologies [33,50,51]. Without trust-building efforts, even advanced technologies in Society 6.0 may face limited adoption among Gen Z users concerned about fairness, privacy, and accountability [13,20,35,41].
Moreover, the study demonstrates that PE mediates the relationship between EE and BI (β = 0.210, p = 0.000). This is consistent with Doven et al. [21]. This mediation indicates that perceived ease of use does not directly translate into intention unless it enhances perceptions of functional value, efficiency, and outcome improvement. In the context of Society 6.0, this suggests that Gen Z in Nepal evaluates technological systems not merely on their simplicity but on the extent to which that simplicity contributes to meaningful performance gains. Thus, usability operates as a foundational condition that strengthens utility perceptions, which in turn shape adoption intention. Similarly, ST mediates the relationship between SI and BI (β = 0.118, p = 0.000). This result aligns with Hooda et al. [36], Rimadias [41], and Al-Haraizah et al. [57]. Further, this finding implies that social endorsement from peers, institutions, or media fosters behavioral intention only when it reinforces trust, transparency, and ethical confidence in technological governance. In the context of Nepal, where concerns regarding accountability and data protection persist, trust becomes the critical mechanism through which social influence transforms into sustained intention to adopt Society 6.0 initiatives.
5. Conclusion
This study investigated the behavioral intention of Gen Z toward Society 6.0 in Nepal using a hybrid approach of PLS-SEM and ANN analysis. The PLS-SEM results indicate that the behavioral intention of Gen Z in Nepal towards Society 6.0 is primarily influenced by performance expectancy, societal trust, social influence, and facilitating conditions, while effort expectancy does not have a significant direct effect on intention. However, effort expectancy indirectly affects behavioral intention through performance expectancy, and social influence also exerts an indirect effect through societal trust. The ANN results further confirm that performance expectancy is the strongest predictor of behavioral intention, followed by societal trust, facilitating conditions, and social influence. Therefore, these findings suggest that the readiness of Gen Z to embrace Society 6.0 in Nepal, and potentially in other developing economies, hinges not merely on technological usability but on the interplay of tangible performance value and trustworthy governance. This underscores that sustainable technological transformation, in the era of Society 6.0, must be built on a foundation of both functional benefit and public trust.
5.1. Theoretical implications
The study contributes theoretically by extending the UTAUT model through the integration of societal trust as both a direct determinant and a mediating construct linking social influence and behavioral intention, thereby addressing a gap in prior research that has predominantly emphasized theoretical predictors while overlooking governance-oriented constructs such as trust. Methodologically, this study advances technology adoption research, in an advanced society, by demonstrating the complementary value of integrating PLS-SEM with ANN. While PLS-SEM identifies linear relationships, ANN captures non-linear patterns, as reflected in the differing importance of societal trust and facilitating conditions across the two methods. This hybrid approach provides a more robust and replicable framework for future adoption research. In an empirical context, it provides one of the earliest large-scale, evidence-based assessments of Gen Z’s readiness to adopt Society 6.0, contributing novel insights to an emerging and underexplored area of societal-technology adoption research.
5.2. Practical implications
In addition to that, from a policy perspective, the findings offer practical guidance for national technological advancement strategies, particularly aligning with Nepal’s Digital Nepal Framework (2019) and AI Policy (2025) to support the realization of Society 6.0 in a developing context. The study also offers actionable implications. Educational institutions should redesign curricula to enhance performance expectancy through practical, hands-on technology integration that builds tangible digital competencies. Technology developers should prioritize clear functional value and demonstrable user benefits over usability alone, since ease of use functions merely as a baseline expectation for digitally native users. Moreover, civil society organizations should promote community-based digital initiatives, particularly in underserved and rural areas, to build trust, extend access, and accelerate broader societal acceptance of Society 6.0 technologies.
5.3. Limitations
This study is subject to several limitations that should be acknowledged. Although stratified and random sampling was applied across all seven provinces of Nepal, respondents were predominantly drawn from universities in the Terai region and urban cities, such as Kathmandu Valley. As a result, Gen Z populations from Mountain and Himalayan districts were underrepresented. This is important because these regions differ substantially in terms of digital infrastructure, internet access, and exposure to emerging technologies, factors that may influence behavioral intention differently. Consequently, the findings may not be fully generalizable to Nepal’s remote and infrastructurally disadvantaged communities. In addition, the study does not provide province-wise comparative analysis, limiting understanding of regional differences in Society 6.0 adoption intention. The cross-sectional design captures perceptions at a single point in time and therefore cannot assess how attitudes and intentions may change over time. Finally, the study relies on self-reported survey data, which may be affected by social desirability and acquiescence bias, potentially leading respondents to report more favorable technology attitudes. Although common method bias was assessed through full collinearity testing and was not found to be a significant concern, the limitations associated with self-report measures cannot be entirely eliminated.
5.4. Directions for future research
Future studies should employ more geographically inclusive sampling strategies, particularly targeting Mountain and Himalayan districts, in order to better capture regional and contextual variations in Society 6.0 adoption intention. Province-wise comparative analyses are also recommended to uncover regional differences that a national-level model may obscure. Longitudinal research designs would also help examine how attitudes and behavioral intentions of Gen Z toward Society 6.0 evolve over time. Mixed-methods approaches, incorporating qualitative techniques such as interviews or focus group discussions, could further contextualize findings and mitigate self-report biases. Future research could incorporate behavioral or observational data to complement self-reported measures. Additional variables, such as trust in AI, ethical innovation, sustainable innovation, digital readiness, technological optimism, and innovation resistance, could be examined to develop a more comprehensive understanding of factors shaping adoption behavior. Cross-country comparative studies would offer valuable insights into how Society 6.0 adoption behavior varies across different cultural and economic contexts. Moreover, cross-generational comparative studies could clarify how adoption behavior differs across age cohorts beyond Gen Z.
Supporting information
S1 File. Raw dataset for the study on Society 6.0 in Nepal.
https://doi.org/10.1371/journal.pone.0362300.s001
(XLSX)
References
- 1. Manza J. Sociology for Beginners. Annu Rev Sociol. 2023;49(1):241–61.
- 2.
Sket R, Mulej M, Žižek SŠ. Measurement of development in the well-being society 6.0. Bases for an innovative sustainable socially responsible society volume II: Selected sustainability viewpoints. Cham: Springer Nature Switzerland. 2026. p. 21–68.
- 3. Žižek SŠ, Mulej M, Potočnik A. The sustainable socially responsible society: well-being society 6.0. Sustainability. 2021;13(16):9186.
- 4. Narvaez Rojas C, Alomia Peñafiel GA, Loaiza Buitrago DF, Tavera Romero CA. Society 5.0: A Japanese Concept for a Superintelligent Society. Sustainability. 2021;13(12):6567.
- 5. Ferreira CM, Serpa S. Society 5.0 and social development. Management and Organizational Studies. 2018;5(4):26–31.
- 6. George AS, George AH. Towards a super smart society 5.0: opportunities and challenges of integrating emerging technologies for social innovation. Partners Universal International Research Journal. 2024;3(2):01–29.
- 7. Mahat D. Society 5.0: A bibliometric analysis from management approach. NPRC J Multidis Res. 2024;1(2):1–19.
- 8.
Polat L, Erkollar A. Industry 4.0 vs. Society 5.0. The International Symposium for Production Research. Cham: Springer International Publishing. 2020. p. 333–45.
- 9. Mourtzis D, Angelopoulos J, Panopoulos N. A literature review of the challenges and opportunities of the transition from industry 4.0 to Society 5.0. Energies. 2022;15(17):6276.
- 10. Carayannis EG, Canestrino R, Magliocca P. From the dark side of industry 4.0 to society 5.0: looking “beyond the box” to developing human-centric innovation ecosystems. IEEE Trans Eng Manage. 2024;71:6695–711.
- 11. Deguchi A, Hirai C, Matsuoka H, Nakano T, Oshima K, Tai M, et al. What is society 5.0. Society. 2020;5(0):1–24.
- 12. Ellitan L. Competing in the era of Industrial Revolution 4.0 and Society 5.0. Jurnal Maksipreneur: Manajemen, Koperasi, dan Entrepreneurship. 2020;10(1):1–12.
- 13. Carayannis EG, Posselt T, Preissler S. Toward Industry 6.0 and Society 6.0: the quintuple innovation helix with embedded ai modalities as enablers of public interest technologies, strategic technology management, and road mapping. IEEE Transactions on Engineering Management. 2024;71:11238–52.
- 14.
Turečková K, Nevima J, Vaňová A, Vitálišová K. Society 4.0: general economic implications. 2023. https://dspace.wunu.edu.ua/bitstream/316497/51636/1/TURE%C4%8CKOV%C3%81.PDF Accessed 2025 July 6.
- 15. Huang S, Wang B, Li X, Zheng P, Mourtzis D, Wang L. Industry 5.0 and Society 5.0—Comparison, complementation and co-evolution. Journal of Manufacturing Systems. 2022;64(1):424–8.
- 16. Javaid M, Haleem A, Singh RP, Suman R, Gonzalez ES. Understanding the adoption of Industry 4.0 technologies in improving environmental sustainability. Sustainable Operations and Computers. 2022;3:203–17.
- 17. Hassoun A, Aït-Kaddour A, Abu-Mahfouz AM, Rathod NB, Bader F, Barba FJ, et al. The fourth industrial revolution in the food industry—Part I: Industry 4.0 technologies. Critical Reviews in Food Science and Nutrition. 2023;63(23):6547–63.
- 18.
World Economic Forum. The Future of Jobs Report 2025. 2025. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
- 19. Siddique MR, Devkota N, Lowar DB, Lawaju P. Society 6.0 for climate-resilient development: Bibliometric analysis and conceptual framework for the Global South. Society. 2026.
- 20.
Žižek SŠ, Mulej M, Šket R, Potočnik A, Horvat BK. Artificial intelligence, virtual and real well-being as an important part of “well-being society 6.0” and ISSRS. Bases for an innovative sustainable socially responsible society volume I: Non-technological innovation. Cham: Springer Nature Switzerland. 2026. p. 185–233.
- 21. Doven G, Sezen B, Demir K, Balcioglu Y. Industry 5.0: Are We Going to Accept Robots as Co-Workers in Office Environments? An Empirical Analysis. Applied Sciences. 2025;15(3):1591.
- 22. Gil IA, Ferreira FA, Ferreira NC, Smarandache F, Khanam M, Daim T. Artificial intelligence, society 5.0 and smart city adaptation initiatives for businesses: An integrated approach. Technovation. 2026;150:103377.
- 23. Sudaryanto MA, Sylvana A. Generation Z and entrepreneurship: Understanding the character of entrepreneurs in the midst of industrial revolution 5.0. Innovative: Journal of Social Science Research. 2024;4(6):4344–51.
- 24. Liao S, Lin L, Chen Q. Research on the acceptance of collaborative robots for the industry 5.0 era -- The mediating effect of perceived competence and the moderating effect of robot use self-efficacy. International Journal of Industrial Ergonomics. 2023;95:103455.
- 25.
Aberšek B, Flogie A, Kordigel Aberšek M. Transformation of education: from dehumanization to re-humanization of society. In: Science and technology education: new developments and innovations, 2023. 18–27. https://doi.org/10.33225/balticste/2023.18
- 26. Diananda A, Waspada MM. Developing early childhood character in the digital era: Transition from society 5.0 to society 6.0. Indonesian Journal of Early Childhood Education Studies. 2025;14(1).
- 27.
Carayannis EG, Zotas NR. Technological futures in the age of Industry 6.0 and Society 6.0. Industrial innovation excellence in the 21st century: Theory, policy, practice and politics. 2026. https://doi.org/10.4337/9781839109188.00018
- 28.
Roshid MdM, Yeamin MdB, Waaje A, Karim R, Kaswan MS. Understanding the Nexus of Industry 5.0 and Circular Economy for Enriched Lean and Environment Sustainability. Transformative Lean Six Sigma Techniques for the Quality 5.0 Paradigm. IGI Global Scientific Publishing. 2025. p. 349–80. https://doi.org/10.4018/979-8-3373-0943-9.ch013
- 29. Al-Mamary YH. A comprehensive model for AI adoption: Analysing key characteristics affecting user attitudes, intentions and use of ChatGPT in education. Human Systems Management. 2025;44(6):978–99.
- 30. Chen S, Li X, Xu Y, Law R, Zuo Y. Tourism marketing in multimodal digital content environments: Generation of travel intention and its influencing mechanism. Tourism Management. 2026;117:105467.
- 31. Qiao G, Shao Z, Gu J, Chen X. Accessible tourism for tourists with visual impairment: the role of auditory perception in shaping behavioral attitudes. Tourism Review. 2026;:1–24.
- 32. Azhar M, Akhtar MJ, Siddiqui MS, Khan Z, Khan FA. Smart hospitality: What drives consumers to accept AI services?. Journal of Hospitality and Tourism Insights. 2025;9(2):707–27.
- 33. Mishra N, Bhandari N, Maraseni T, Devkota N, Khanal G, Bhusal B, et al. Technology in farming: Unleashing farmers’ behavioral intention for the adoption of agriculture 5.0. PLoS One. 2024;19(8):e0308883. pmid:39172798
- 34. Aalam MdP, Siddique MR, Kunwar R. Impact of digital complaint management system on banking customer satisfaction: Evidence from Kathmandu valley, Nepal. Econ Rev Nep. 2025;8(2):93–107.
- 35. Adhikari GM, Lamichane S, Parajuli D, Bhattarai G. Gen Z’s Intention to Accept Mobile Banking Services: The Impact of Digitalization and Technologization. SEC. 2024;8(4):31–49.
- 36. Hooda A, Gupta P, Jeyaraj A, Giannakis M, Dwivedi YK. The effects of trust on behavioral intention and use behavior within e-government contexts. International Journal of Information Management. 2022;67:102553.
- 37. Popova Y, Zagulova D. UTAUT model for smart city concept implementation: use of web applications by residents for everyday operations. Informatics. 2022;9(1):27.
- 38. Alkdour T, Almaiah MA, Shishakly R, Lutfi A, Alrawad M. Exploring the success factors of smart city adoption via structural equation modeling. Sustainability. 2023;15(22):15915.
- 39. Shahzad M, Qu Y, Rehman SU, Zafar AU. Adoption of green innovation technology to accelerate sustainable development among manufacturing industry. Journal of Innovation & Knowledge. 2022;7(4):100231.
- 40. Alquhaif AS, Al-Mamary YH. Examining factors influencing the adoption of accounting information systems: An analysis of behavioral intentions and usage behavior. Human Systems Management. 2024;44(3):460–76.
- 41. Rimadias S. Key Drivers of AI utilization for environmental sustainability among indonesia’s educated Gen Z. AJABM. 2025;4(1):211–28.
- 42. Pujiati T, Setiyowati H, Rawat B, Santoso NPL, Ilham MG. Exploring the role of artificial intelligence in enhancing environmental health: UTAUT2 Analysis. Sundara Advanced Research on Artificial Intelligence. 2025;1(1):37–46.
- 43. Siddique MR, Devkota N, Lowar DB, Paudel UR, Kunwar R, Lawaju P. Responsible generative artificial intelligence for sustainable pedagogy systems: A conceptual framework for the Global South. International Journal of Emerging Technologies in Learning. 2026;21(2):4.
- 44. Al-Mamary YHS, Alraja MM. Understanding entrepreneurship intention and behavior in the light of TPB model from the digital entrepreneurship perspective. International Journal of Information Management Data Insights. 2022;2(2):100106.
- 45. Dixit A, Shaw R. Smart Cities in Nepal: The concept, evolution and emerging patterns. Urban Governance. 2023;3(3):211–8.
- 46. Adhikari AP, Bhattarai K. Towards creating smart cities in Nepal. Nepal Journal of Science and Technology. 2021;20(2):1–16.
- 47.
Shrestha SB, Pradhananga R, Bhattarai K. Smart and compact rural settlements for resilient and sustainable development in Nepal: Policy and Implementation Perspectives. Sustainable Development Goals Series. Springer Nature Singapore. 2025. p. 25–41. https://doi.org/10.1007/978-981-96-6149-7_3
- 48.
National Planning Commission. The fifteenth plan (Fiscal year 2019/20–2023/24). 2019. http://elibrary.moest.gov.np/bitstream/123456789/307/1/15.pdf
- 49.
National Planning Commission. The sixteenth plan (Fiscal year 2024/25–2028/29). 2024. http://elibrary.moest.gov.np:8080/bitstream/123456789/308/1/16.pdf
- 50. Giri S. Dimensions of digital Nepal framework and appropriate roadmap. International Journal of Science and Research. 2018;9(1):719–24.
- 51. Dhakal C, Adhikari D, Subba G, Kapadi PR. Reimagining Nepal’s Future: AI for Human Development and Education. Panauti J. 2025;3:121–34.
- 52. Karki YB. Power of Zen-Z demographics: and the next roadmap for Nepal. The Himalayan Times. 2025.
- 53. Rajbhandari S, Devkota N, Khanal G, Mahato S, Paudel UR. Assessing the industrial readiness for adoption of industry 4.0 in Nepal: A structural equation model analysis. Heliyon. 2022;8(2):e08919. pmid:35243054
- 54. Yu S, Zhao X, Fang X, Sun Y-F, Guo Y. How public perceptions of unplanned power outage policies affect energy-saving behavioral intention: Evidence from China’s social media mining. Energy Policy. 2025;206:114727.
- 55. Al-Mamary YHS. Understanding the use of learning management systems by undergraduate university students using the UTAUT model: Credible evidence from Saudi Arabia. International Journal of Information Management Data Insights. 2022;2(2):100092.
- 56. Al-Mamary YHS, Siddiqui MA, Abdalraheem SG, Jazim F, Abdulrab M, Rashed RQ, et al. Factors impacting Saudi students’ intention to adopt learning management systems using the TPB and UTAUT integrated model. JSTPM. 2023;15(5):1110–41.
- 57. Al-Haraizah A, Jawarneh M, Kumar R, Khrisat DA, Al Mrayat OI. An Extended UTAUT Model to Explain Factors Influencing Smart City Technology Adoption. Human Behavior and Emerging Technologies. 2025;2025(1).
- 58. Mandal P, Siddique MR, Kanwar B. Influence of emotional intelligence on employees performance in commercial bank: a mediating role of leadership style. HJNR. 2026;1(1):29–45.
- 59. Venkatesh V, Thong J, Xu X. Unified Theory of Acceptance and Use of Technology: A Synthesis and the Road Ahead. JAIS. 2016;17(5):328–76.
- 60. Chawla U, Mohnot R, Singh HV, Banerjee A. The mediating effect of perceived trust in the adoption of cutting-edge financial technology among digital natives in the post-COVID-19 Era. Economies. 2023;11(12):286.
- 61.
Saunders M, Lewis P, Thornhill A. Research Methods for Business Students. Pearson. 2023.
- 62. Runfola D, Anderson A, Baier H, Crittenden M, Dowker E, Fuhrig S, et al. geoBoundaries: A global database of political administrative boundaries. PLoS One. 2020;15(4):e0231866. pmid:32330167
- 63.
United Nations Development Programme. LDC Graduation, Smooth Transition Strategy. 2024. https://www.undp.org/sites/g/files/zskgke326/files/2024-05/Final%20LDC%20Graduation%20STS.pdf
- 64. Ahmed SK. How to choose a sampling technique and determine sample size for research: A simplified guide for researchers. Oral Oncology Reports. 2024;12:100662.
- 65. Liébana-Cabanillas F, Marinkovic V, Ramos de Luna I, Kalinic Z. Predicting the determinants of mobile payment acceptance: A hybrid SEM-neural network approach. Technological Forecasting and Social Change. 2018;129:117–30.
- 66. Sharma SK, Sharma H, Dwivedi YK. A Hybrid SEM-neural network model for predicting determinants of mobile payment services. Information Systems Management. 2019;36(3):243–61.
- 67.
Walczak S. Artificial Neural Networks. Advances in Computer and Electrical Engineering. IGI Global. 2019. p. 40–53. https://doi.org/10.4018/978-1-5225-7368-5.ch004
- 68. Kalinić Z, Marinković V, Kalinić L, Liébana-Cabanillas F. Neural network modeling of consumer satisfaction in mobile commerce: An empirical analysis. Expert Systems with Applications. 2021;175:114803.
- 69. Sunar DK, Devkota N, Paudel UR. Understanding local perspectives on micro hydropower management in galkot municipality, nepal: a structural equation modeling approach. J Econ Concerns. 2025;16(1–2):1–21.
- 70. Chan SH, Lay YF. Examining the reliability and validity of research instruments using partial least squares structural equation modeling (PLS-SEM). JBSE. 2018;17(2):239–51.
- 71. Herawati I, Hoesni SM, Manap J, Mohd Khatib NA. A qualitative study: exploring marital readiness among generation Z. IJARBSS. 2023;13(12).
- 72. Karna MK. Tracing some challenges of higher education historically in Nepal through Tribhuvan University. Hist J. 2024;15(2):62–73.
- 73. Henseler J, Ringle CM, Sarstedt M. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J of the Acad Mark Sci. 2014;43(1):115–35.
- 74. Kock N. Common method bias in PLS-SEM: A full collinearity assessment approach. International Journal of E-Collaboration. 2015;11(4):1–10.
- 75. Jr. JFH, Matthews LM, Matthews RL, Sarstedt M. PLS-SEM or CB-SEM: updated guidelines on which method to use. IJMDA. 2017;1(2):107.
- 76. Hair JF, Risher JJ, Sarstedt M, Ringle CM. When to use and how to report the results of PLS-SEM. EBR. 2019;31(1):2–24.
- 77. Lawaju P, Adhikari SU, Devkota J. Impact of AI in Education: An Evidence from Use of ChatGPT in Management Education in Nepal. QJMSS. 2024;6(3):621–42.
- 78. Hair JF Jr, Howard MC, Nitzl C. Assessing measurement model quality in PLS-SEM using confirmatory composite analysis. Journal of Business Research. 2020;109:101–10.
- 79. Sarstedt M, Ringle CM, Cheah J-H, Ting H, Moisescu OI, Radomir L. Structural model robustness checks in PLS-SEM. Tourism Economics. 2019;26(4):531–54.
- 80. Kukreja H, Bharath N, Siddesh CS, Kuldeep S. An introduction to artificial neural network. Int J Adv Res Innov Ideas Educ. 2016;1(5):27–30.
- 81. Tiwari DR, Devkota N, Lawaju P. Tourist satisfaction and sustainable homestay management: insights from six rural communities in Western Nepal. QJMSS. 2025;7(1):182–200.
- 82. Fei Z, Wang Y. How does delayed childbearing promote child gender diversity? - Causal evidence and mechanism analysis from China. Acta Psychol (Amst). 2026;267:107191. pmid:42235292
- 83. Sohail MT, Yang M, Maresova P, Mustafa S. An SEM-ANN approach to evaluate public awareness about COVID, A pathway toward adaptation effective strategies for sustainable development. Front Public Health. 2022;10:1046780. pmid:36339186
- 84. Sternad Zabukovšek S, Bobek S, Zabukovšek U, Kalinić Z, Tominc P. Enhancing PLS-SEM-enabled research with ANN and IPMA: Research Study of Enterprise Resource Planning (ERP) Systems’ Acceptance Based on the Technology Acceptance Model (TAM). Mathematics. 2022;10(9):1379.
- 85. Dastres R, Soori M. Artificial neural network systems. International Journal of Imaging and Robotics. 2021;21(2):13–25.
- 86. Hew J-J, Leong L-Y, Tan GW-H, Lee V-H, Ooi K-B. Mobile social tourism shopping: A dual-stage analysis of a multi-mediation model. Tourism Management. 2018;66:121–39.
- 87. Leong L-Y, Hew T-S, Ooi K-B, Lee V-H, Hew J-J. A hybrid SEM-neural network analysis of social media addiction. Expert Systems with Applications. 2019;133:296–316.
- 88. Jin J, Hu Y, Wang Z, Pan Y, Ma B, Dong B, et al. Brain-inspired large model mindreading. Neuroimage. 2026;338:122108. pmid:42413890
- 89. Wu S, Cheng P, Wang J. How relational networks influence team innovation: Dual pathways and synergistic mechanism in knowledge activities. European Journal of Innovation Management. 2026;1–30.

관련 뉴스