A learning management system is an integrated interactive learning environment which integrates communication tools such as private message, notice board, chatroom and provides supplementary multimedia learning materials such as graphics, videos, audio files, references, self-tests and quiz modules (El-Masri and Tarhini, 2017). LMSs are widely adopted by Hong Kong higher institutions to incorporate video communication tools (such as Zoom, Panopto and Cisco Webex), personal response systems (such as Qualtrics, an online survey software), game-based learning assessment tools (such as Kahoot!), anti-plagiarism tools (such as Turnitin and iThenticate) and presentation tools (such as Mentimeter).

Common web-based and mobile-based LMSs adopted by different universities in Hong Kong include Blackboard, Canvas and Moodle, which offer students with flexibility to complete their assignments at their preferred pace and access to course materials (Al-Adwan and Al-Debei, 2024). It is a mandatory requirement for students to attend synchronous (live-streaming) and asynchronous (pre-recorded) classes, access to the printed and multimedia course materials, complete and submit their academic tasks (e.g. exercises or/and quizzes), as well as receive feedbacks through LMSs. Given the use of LMS is mandatory among most Mainland Chinese postgraduate students who pursue their study, it would be vital to understand the drivers which affect students’ intention to use LMS during the emergence of pandemic.

The Unified Theory on Acceptance and Use of Technology (UTAUT) is a classic model for explaining and predicting technology acceptance in organizational setting. This comprehensive model was compiled from eight models and theories including Theory of Reasoned Action (TRA), Theory of Planned Behavior (TPB), Technology Acceptance Model (TAM), combined TAM and TPB (C-TAM-TPB), Model of PC Utilization (MPCU), Motivational Model (MM), Social Cognitive Theory (SCT) and Diffusion of Innovation Theory (IDT) (Venkatesh et al., 2003). The UTAUT2 was later proposed to study the mobile technology acceptance in the Hong Kong consumer context to improve the generalizability of the model (Venkatesh et al., 2012). The UTAUT2 framework consists of four constructs, including performance expectancy, social influence, facilitating conditions and effort expectancy from the UTAUT model and three extra constructs including habit, price value and hedonic motivation as predicators of intention to use and behavior (Figure 1). Whilst the price value construct is added to highlight the importance of product quality-to-price ratio, the habit and hedonic motivation constructs are included for prior behavior and stressing utility respectively.

Ajzen (2011) defines behavioral intention as an indicator of a person’s readiness to perform a specific behavior. Numerous studies have proven that behavioral intention plays a crucial role in actual behaviors (Burke, 2002). In this study, behavioral intention is associated with the Mainland Chinese postgraduate students’ readiness to use LMSs adopted and used by universities in Hong Kong. It is the only endogenous variable in the model, determined by other exogenous variables.

Performance expectancy is defined as the extent to which a person trusts that adopting the system will assist him or her in improving job performance (Venkatesh et al., 2003). Recent studies have consistently demonstrated the significant effect of performance expectancy on technology adoption among students and teachers (Ahmed et al., 2022; Chao, 2019; Lai et al., 2024; Prasetyo et al., 2021; Raza et al., 2021). In the current study, performance expectancy refers to Mainland Chinese postgraduate students’ beliefs about the usefulness of the LMSs adopted by universities in Hong Kong. Thus, the following hypothesis is developed:

Effort expectancy reflects the level of ease associated with utilizing the system (Venkatesh et al., 2003, p. 450). Existing research has demonstrated that effort expectancy significantly influences students’ willingness to adopt eLearning technologies (Ahmed et al., 2022; Chao, 2019; Raza et al., 2021). However, earlier studies have contended that the mere ease of use of a technology does not necessarily guarantee its acceptance by individuals (Dindar et al., 2021; Çera et al., 2020). Instead, learners are more likely to adopt a technology if they have trust in its ability to provide learning benefits and if it is perceived as user-friendly. In this study, effort expectancy represents Mainland Chinese postgraduate students’ beliefs about the effort or ease related to the LMSs adopted and used by universities in Hong Kong. We assume that students’ behavioral intention to use LMSs is developed through their positive understanding about the LMSs’ easiness. The following hypothesis is proposed:

Social influence represents the degree to which individuals perceive other important people believe if they should adopt the new system (Venkatesh et al., 2003). Extant studies have identified social influence as a significant determinant of eLearning technology and meta-education acceptance (Ahmad et al., 2024; Ahmed et al., 2022; Raman and Thannimalai, 2021). Its effect on behavioral intention may be stronger in the Chinese collectivistic society, where individuals’ tendency to adopt a technology is likely to rely on subjective evaluation of it from other-like-minded peers (Lee et al., 2013). However, recent studies have also indicated insignificant effects of social influence on metaverse technology adoption among Gen Z students (Al-Adwan and Al-Debei, 2024) and on eLearning adoption among doctorate students (Prasetyo et al., 2021). In our study, social influence refers to the impact of other people’s beliefs on Mainland Chinese postgraduate students’ intention to use LMSs in this study. The following relationship is hypothesized:

Facilitating conditions refer to “the degree to which an individual considers that an organization and technical infrastructure exist to support the use of the system” (Venkatesh et al., 2003, p. 453). Technical guidance, system management, training and prompt assistance as well as other logistic arrangements to the users are essential facilitating conditions for using the technology (Mei et al., 2017). Previous studies have shown that facilitating conditions promote students’ intention to learn (Zacharis and Nikolopoulou, 2022) and interact with each other (Wut et al., 2022) on eLearning platforms. Nevertheless, Prasetyo et al. (2021) and Raman and Thannimalai (2021) evidenced no significant relationship between facilitating conditions and behavioral intention in the context of eLearning during the pandemic. In the current research, facilitating conditions relate to the sufficiency of resources, training opportunities, timely assistance and technical support provided to students to facilitate their use of LMSs. The below relationship is proposed:

Hedonic motivation concerns about the pleasure or fun derived from adopting a technology, which is crucial in predicting use intention of a technology (Venkatesh et al., 2012). Hedonic motivation was commonly recognized as a notable driving force behind behavioral intention in educational settings (Al-Gahtani, 2016; El-Masri and Tarhini, 2017). However, recent studies conducted amidst the COVID-19 pandemic have yielded contrasting findings. Specifically, these studies have indicated that hedonic motivation does not have a direct and significant relationship with the acceptance of LMSs (Prasetyo et al., 2021; Raman and Thannimalai, 2021). In this study, hedonic motivation refers to the pleasure or fun experienced by users of LMSs. The following relationship is hypothesized:

Since the price value of the eLearning technologies is difficult to be calculated from the tuition fee, we followed Ain et al. (2016) to replace “price value” with “learning value” in the research model to fill in the research gap in Hong Kong. While Harja et al. (2020) and Ain et al. (2016) have found learning value a strong predictor of eLearning technology and LMS adoption among students respectively, Prasetyo et al. (2021) have evidenced that learning value significantly predicts eLearning platform adoption among medical students during the pandemic.

Students devote time and effort (also part of the tuition fees paid to support the development and maintenance of technologies) to gain benefits from LMSs. The use of LMSs holds a value if knowledge is gained from using them. In this study, learning value refers to the cognitive trade-off between the effort and time spent on LMSs and their perceived value by users. Thus, the hypothesis is proposed as follows:

Habit is considered as an individual’s tendency to act spontaneously for learning (Venkatesh et al., 2012). Earlier research has revealed that individuals’ prior experience with technology usage contributes to their ongoing actions, as evidenced by their behavioral intention (El-Masri and Tarhini, 2017; Masa’deh et al., 2016; Moorthy et al., 2019; Raman and Thannimalai, 2021). In this study, habit relates to students’ habitual or automatic behaviors of using LMSs. Students who consistently utilize LMSs are believed to have higher potential in using them. Therefore, the following hypothesis is suggested:

Trust refers to one’s readiness to accept vulnerability based on positive expectations about the intentions or the acts of others in an environment featured by risk and interdependence (Ennew and Sekhon, 2007). Trust plays a crucial role in promoting information exchange, which is essential in the knowledge transfer process (Cheung and Lai, 2022). Prior literature has shown dichotomous results about the relationship between users’ decisions to adopt eLearning technologies and their perceived reliability and trustworthiness of system. Recent studies have identified significant effect of trust on behavioral intention in eLearning, m-learning and self-regulated learning context (Chao, 2019; El-Masri and Tarhini, 2017; Harja et al., 2020; Lai et al., 2024). Wang and Yu (2015) have showed that the better educated Chinese are more conscious of precautions and therefore have lower tendency to trust. Despite of these, some have found no direct relationship between trust and behavioral intention in the consumer and information system context (Eneizan et al., 2019; Kabra et al., 2017).

In this study, trust refers to the Mainland Chinese postgraduate students’ understanding towards the trustworthiness and reliability of LMSs. Users’ perception on security and trust are expected to be important factors of LMS acceptance. Therefore, the following hypothesis is set:

Instructor characteristics is the extent to which the instructors show care, give advice, and accommodate their students’ learning needs (Ibrahim et al., 2017). Existing literature has shown inconsistent influence of instructor characteristics on behavioral intention. While numerous studies have reported the positive impact of instructor characteristics on perceived usefulness of eLearning technology and behavioral intention to use it (Lee et al., 2009; Prasetyo et al., 2021; Selim, 2007), Ibrahim et al. (2017) have indicated no direct or indirect relationship of instructor characteristics on behavioral intention. In this study, we follow the framework proposed by Prasetyo et al. (2021) to include the instructor characteristics construct to explore the importance of instructors’ support during social isolation at the time of pandemic. In this research, instructor characteristics is defined as the instructors’ tendency to encourage, motivate and facilitate Mainland Chinese postgraduate students to use LMSs. Accordingly, the following hypothesis is set:

The research model of the study is proposed as follows (Figure 2):

This set of hypotheses covers the direct relationships (i.e., H1 to H9) between the nine exogenous variables (performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, learning value, habit, trust and instructor characteristics) and the only endogenous variable (behavioral intention) in the proposed research model.

A non-experimental and quantitative research methodology is adopted in this study. Using a SEM approach, the current research tests the relationships among the latent variables adapted and extended from the UTAUT2 framework. First, we conducted a demographic analysis to understand the participant profiles. Then, using IBM SPSS AMOS (version 20) software, we performed the SEM technique in two stages – confirmatory factor analysis (CFA) and path analysis. CFA helps to confirm the uni-dimensionality, reliability and validity of the constructs (MacCallum and Austin, 2000). After that, the structural model was analyzed using path analysis to examine the hypotheses of the research.

An online questionnaire survey including 40 items related to technology acceptance and use is designed to collect descriptive and quantitative samples from Mainland Chinese postgraduate students at different universities in Hong Kong during the pandemic. The measurement instrument represents the theoretical model that relates the latent variables to the manifest indicators, which are measured through questionnaire survey. The survey is divided into two sections. The first section aimed to gather demographic information and the second section then sought to find out participants’ intention to use LMS.

The 40 manifest items and the corresponding descriptions for measuring the 10 latent variables are presented in English in Table 1. Participants were requested to show their levels of agreement towards the descriptions based on a 5-point Likert scale arranging from “strongly disagree” to “strongly agree”.

Before conducting the data analysis, we performed data screening to fulfil the requirement of multivariate analysis. We purified the data by replacing missing values, assessing outliers, data normality and multicollinearity. The common method bias refers to a systematic error in research data which arises from the measurement method itself instead of the constructs being measured (Podsakoff et al., 2003). This bias can occur when data is collected using the same method for all variables in a single study. As all data was collected from self-report questionnaire in our study, Harman’s single-factor test was employed to assess the common method bias.

The survey was conducted to collect samples from postgraduate students at different universities in Hong Kong from March to April 2022. After the pilot study, the revised questionnaire was distributed on a user-friendly survey tool¹ through eleven WeChat App groups (group size from 358 to 500) for Mainland Chinese students who had enrolled to taught postgraduate programs in Autumn 2021. A number of 352 questionnaires were returned online with a response rate of 7.6%. From March to April 2022, all the universities in Hong Kong were closed so the use of web-based and mobile-based LMSs (Blackboard, Canvas and Moodle) was mandatory at the time of study. In order to reach a vast number of samples from different universities, convenience sampling method was adopted in this study. For SEM studies, a sample size between 200 and 400 is considered appropriate (Hair et al., 2019; Rehman et al., 2022). Data were collected from postgraduate students who participated in the survey voluntarily without any incentives. All the responses were collected online as social distancing posed a barrier for face-to-face data collection.

The participants’ demographic profile is shown in Table 5. The findings show that female participants (73%) are significantly more than the males (27%), reflecting the high proportion of female Mainland Chinese postgraduate students studying in Hong Kong. The majority (93.2%) of participants was less than 40 years old from a variety of higher institutions. Participants were from different disciplines, with most of them having enrolled in business (23.6%), arts, journalism and language (17.9%), engineering and information technology (12.8%). Most of them had good (41.2%) to very good (38.1%) internet connectivity in their study location. They generally possessed some eLearning experience, with the majority having less than two years (27.6%), two to four years (42%) and five to seven years (24.7%) experience.

The findings reveal that one factor model explains 25.255% of the total variance, suggesting that the common method variance is not a serious problem (Hoyle, 1995). The results are presented in Table 6.

This study comprises an overall confirmatory factor analysis (CFA) model with discussion of the measurement model development. To ensure the CFA accuracy, Hair et al. (2019) proposed examining goodness-of-fit indicators, convergent and discriminant validity. The first step of CFA is to evaluate the goodness of fit (GOF) indices of the measurement model. The GOF indices of our measurement model and the recommended values from literatures are listed in Table 7. From the values below, we conclude that the measurement model has good fit of data and requires no adjustments.

The second step of CFA is to examine the reliability and convergent validity of each construct. The reliability was assessed through reviewing the Cronbach’s alpha, composite reliability (CR) and average variance extracted (AVE) whereas the constructs’ convergent and discriminant validity were also examined. The results of CFA are presented in Table 8. The factor loadings for all 40 items are more than 0.5, which shows the meanings of the factors have been preserved by these indicators. The AVEs, ranging between 0.610 and 0.763, exceed the cut-off value of 0.5 (Fornell and Larcker, 1981). The CR values, ranging from 0.862 to 0.928, are above the recommended value of 0.6 (Bagozzi and Yi, 1988) so the measurement is considered reliable. The Cronbach’s Alpha values, ranging between 0.862 and 0.927, are above the threshold of 0.7 (Nunnally and Bernstein, 1994) so the manifest indicators of a latent construct are considered as closely related and sufficiently error-free.

Table 9 below shows the discriminant validity of the measurement model. The inter-factor correlations between the ten constructs range from −0.090 to 0.544, which are less than the threshold of 0.85 (Kline, 2010). The correlations are less than the square root of the AVE by the indicators, which reflects a satisfactory discriminant validity.

After evaluating the goodness to fit indices, convergent and discriminant validity of the measurement model, we can conclude the measurement scale is valid and reliable. Figure 3 presents the measurement model with standardized factor loadings for the 40 items.

To judge if all constructs are suitable for analysis, we adopted the covariance matrix method. The composite scores of the variables were computed by parceling the original measurement item scores. Table 10 presents the standard deviations, the means, the minimum and maximum values of constructs, which were measured on a 5-point Likert scale:

The mean is considered as an indicator of central tendency. The mean values in Table 10 reflect that apart from social influence and learning value, participants’ perception towards all other constructs, with mean values greater than the mid-point of 3, is above the average. Performance expectancy has the greatest mean value of 4.07 whereas social influence has the lowest mean value of 2.683.

The standard deviation serves as a dispersion index which shows the extent to which the values deviate from the mean. Among all the variables, performance expectancy has the greatest deviation from its mean (SD = 0.997), suggesting reasonably high variability in participants’ perception toward performance expectancy. Learning value has the smallest deviation from the mean (SD = 0.668), suggesting relatively low variability in participants’ perception toward performance expectancy. Figure 4 presents a graphical illustration of the mean and standard deviations of all constructs.

Table 11 presents the Pearson’s correlation matrix for all variables in the study. The results reveal predominantly positive and statistically significant correlations. On average, the correlations range from −0.095 to 0.489. As no correlation exceeds the threshold of 0.80 (Cheung and Lai, 2023), there is no evidence of multicollinearity among the variables.

Further, Table 12 demonstrates that all variance inflation factors (VIFs) are below 10 (Cheung and Chung, 2022). Consequently, the study does not encounter any issues related to multicollinearity.

After the measurement model had been validated, we applied the SEM technique on the data to evaluate the hypothesized relationships using IBM SPSS AMOS (version 20) software and maximum likelihood estimate (MLE) technique. We examined the causal effects of performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, learning value, habit, trust and instructor characteristics on behavioral intention in the proposed model.

The goodness-of-fit indices reflect that the observed data adequately fits the model: χ2 = 907.938, df = 695, p-value = 0.000, GFI = 0.891, AGFI = 0.871, CFI = 0.976, TLI = 0.973, IFI = 0.977, RMSEA = 0.030 and χ2/df = 1.306. Though the chi-square results are statistically significant, this is acceptable as the sample size is large (Bagozzi et al., 1991). The R2 value for behavioral intention is 0.50, which suggests 50 percent of variations in behavioral intention can be explained by its nine determinants. The overall results show that the scores of R² value meet the condition for the cut off value of 0.10 (Quaddus and Hofmeyer, 2007). Then, the coefficient parameter estimates are reviewed to examine the hypothesized causal relationships of the variables. The hypothesis results of causal relationships are shown in Table 13.

From the above table, except the two paths for hedonic motivation and habit on behavioral intention, all other paths from performance expectancy, effort expectancy, social influence, facilitating conditions, learning value, trust and instructor characteristics on behavioral intention are statistically significant with p-value less than the standardized significant level of 0.05. Thus, the hypotheses H1, H2, H3, H4, H6, H8 and H9 are supported while hypotheses H5 and H7 are rejected. The findings reflect that the strongest predictor of behavioral intention is trust with the standardized path coefficient of 0.204. The second strongest determinant is instructor characteristics with the standardized path coefficient of 0.202. Figure 5 depicts the path analysis results with the standardized coefficients.

From Table 13, the critical ration (c.r.) and p-value of performance expectancy in predicting behavioral intention are 2.719 and 0.007 respectively. It reflects that the probability of having a critical ratio as large as 2.719 in absolute value is 0.007. As the regression weight for performance expectancy in predicting behavioral intention is significantly different from zero at the 0.01 level (two-tailed), H1 is supported. Moreover, the standardized estimate of Beta is 0.181, showing a positive relationship between performance expectancy and behavioral intention.

The regression weight for effort expectancy in predicting behavioral intention is significantly different from zero at the 0.01 level (two-tailed). Thus, H2 is supported. In addition, the standardized estimate of Beta was 0.148, reflecting a positive association between effort expectancy and behavorial intention.

The regression weight for social influence in predicting behavioral intention is significantly different from zero at the 0.05 level (two-tailed). Therefore, H3 is supported. Moreover, the standardized estimate of Beta is 0.115, indicating a positive relationship between social influence and behavorial intention.

The regression weight for facilitating conditions in predicting behavioral intention is significantly different from zero at the 0.001 level (two-tailed). Thus, H4 is supported. Additionally, the standardized estimate of Beta is 0.181, providing evidence about the positive relationship between facilitating conditions and behavorial intention.

The finding indicates that there is no significant association between hedonic motivation and behavioral intention. The path coefficient is −0.016, c.r. is −0.293 and p-value is 0.770. Thus, H5 is rejected.

The regression weight for learning value in predicting behavioral intention is significantly different from zero at the 0.005 level (two-tailed). Thus, H6 is supported. Also, the standardized estimate of Beta is 0.118, suggesting a positive association between learning value and behavioral intention.

The results show that there is no significant association between habit and behavioral intention. The path coefficient is 0.014, c.r. is 0.298, p-value = 0.766. Thus, H7 is rejected.

The regression weight for trust in predicting behavioral intention is remarkably different from zero at the 0.001 level (two-tailed). Thus, H8 is supported. Moreover, the standardized estimate of Beta is 0.204, indicating a positive relationship between trust and behavioral intention.

The regression weight for instructor characteristics in predicting behavioral intention is significantly different from zero at the 0.001 level (two-tailed). Thus, H9 is supported. Besides, the standardized estimate of Beta is 0.202, demonstrating a positive relationship between instructor characteristics and behavioral intention.

The rapid shift of eLearning as well as reliance on available course information and task submission on LMS result in new ethical questions about online security and data privacy (García-Morales et al., 2021). In alignment with the results of Chao (2019), El-Masri and Tarhini (2017), Harja et al. (2020) and Lai et al. (2024), our study finds trust as the most important antecedent which influences the well-educated Mainland Chinese student participants’ intention to use LMSs in this research. China, as the largest data pool in the world, used to have grievous trust crisis related to privacy in different fields, with no exception in online learning context (Wang and Yu, 2015). The demand from Chinese citizens for improved data protection is strong (Roberts, 2021). The results of this study are in line with Wang and Yu (2015), which shows that well-educated adults possess strong awareness of data security and trust has direct effect on behavioural intention. However, our findings are in contrast to Eneizan et al. (2019) and Kabra et al. (2017), which identified no positive association between trust and behavioral intention in the mobile marketing and information system context respectively.

In line with most studies on eLearning technology (Lee et al., 2009; Prasetyo et al., 2021; Selim, 2007), instructor characteristics is identified as the second strongest motivator of behavioral intention among Mainland Chinese participants in this research. These findings differ from those of Ibrahim et al. (2017), which suggest no direct or indirect relationship between instructor characteristics and behavioral intention. During the pandemic, the sudden shift of traditional pedagogical approach to online teaching required much preparation by instructors, ranging from designing audio and visual materials, regular planning on duration of study hours, choosing correct assessment tools, to developing interactive and formative class activities for students’ peer collaboration (García-Morales et al., 2021). Under social isolation at the time of pandemic, instructors’ encouragement and motivation were highly correlated with participants’ LMS acceptance. Instructors’ enthusiasm in teaching course contents, encouraging students’ use of LMSs and inviting students to ask or answer questions via the cyberspace is a crucial factor which contributes to students’ positive perception of LMSs.

The findings reflect that technical support and participants’ technological knowledge are significant factors for intention of LMS use during social distancing. The problems associate with eLearning technologies, including downloading errors, issues with installation, login problems, audio and video problems, can discourage students from using LMSs (Dhawan, 2020). Mainland Chinese students may need to use a Chrome browser to enable full functionality of LMS. However, they may face challenges in installing Google Chrome, as access to Google services is often restricted within China. In alignment with the findings of recent studies (Wut et al., 2022; Zacharis and Nikolopoulou, 2022), our study reveals that technical support by universities, students’ resources, knowledge and skills for solving technical problems are vital for their LMS acceptance. This differs from the results by Prasetyo et al. (2021) and Raman and Thannimalai (2021), which suggest no positive relationship between facilitating conditions and behavioral intention during the pandemic.

Performance expectancy has been consistently found to be a strong motivator of behavioral intention in the educational context (Ahmed et al., 2022; Chao, 2019; Lai et al., 2024; Prasetyo et al., 2021; Raza et al., 2021). In line with prior studies, the correlation between performance expectancy and behavioral intention is found to be significant in this study. The implication is that the Mainland Chinese postgraduates are willing to use LMSs by Hong Kong universities as long as the platforms are perceived to be useful for improving their course results. This is not surprising since the online quizzes, polls, written and multimedia course materials on LMSs can help to accomplish their tasks efficiently. Communication channels such as built-in messaging systems and discussion forums can also facilitate academic discussions between instructors and students, which help to improve students’ learning outcomes (Green et al., 2014).

Some prior studies have argued that there is no direct relationship between effort expectancy and behavioral intention, as individuals’ acceptance of a technology is not solely determined by its ease of use but also its performance expectancy (Dindar et al., 2021; Hu et al., 2003). However, our study aligns with other prior studies which have suggested that effort expectancy has a remarkable impact on students’ behavioral intention to adopt eLearning technologies (Ahmed et al., 2022; Chao, 2019; Raza et al., 2021). The results of this study suggest that students’ behavioral intention is developed upon their positive understanding on LMSs’ easiness during the pandemic.

Prior to the pandemic, the world had already experienced substantial growth and adoption in educational technologies, with the online education market projected to reach USD 350 billion by 2025 (World Economic Forum, 2020). Further, the global COVID-19 crisis compelled universities to shift from traditional pedagogical approaches to eLearning. The onset of the pandemic resulted in a sudden and significant surge in the utilization of LMSs (Veluvali and Surisetti, 2022). This transition has led to rapid advancements in the technological infrastructure of education, and students have become increasingly familiar with the use of LMS. Students’ perception of the low degree of effort required to use the LMS contributes to a higher behavioral intention to utilize it. Given that the majority of survey participants were from the younger generation and resided in areas with acceptable internet connectivity, they generally found the well-developed LMSs implemented by Hong Kong universities to be effortless to use. In contrast, those who encountered difficulties in using the LMSs had a lower intention to adopt the system for learning during the pandemic.

Learning value pertains to the cognitive trade-off between the perceived value of LMSs, and the time and effort spent on exploring and adopting them (Ain et al., 2016). Since Mainland Chinese students devote much time and effort (including tuition fee) to enroll into postgraduate programs in Hong Kong, they have expectation on gaining reasonable benefits from using LMSs. Consistent with recent studies in education context (Ain et al., 2016; Harja et al., 2020; Prasetyo et al., 2021), the findings suggest significant impact of learning value on postgraduate student participants’ intention towards LMS use during pandemic. If they find the LMSs convenient and efficient while leading to the desired learning outcome, they have higher motivation to use them.

Social influence concerns user’s perception about the influence of important people. In line with the study conducted by Ahmed et al. (2022) in five different Asian countries (excluding China), the results reflect social influence is a determinant of LMS acceptance. The results also align with Ahmad et al. (2024) and Raman and Thannimalai (2021) which respectively report the positive influence of social influence on behavioral intention to adopt eLearning platforms and metaverse technology in education. Nonetheless, the predicting effect of social influence is the least significant among other determinants in our study. Traditionally, China has been regarded as a collectivistic society where individuals tend to rely on the opinions of their important peers when making decisions about technology adoption (Lee et al., 2013). However, due to the influence of globalization, the younger generation in China has been exposed to Western cultures and has embraced more individualistic values (Hsu and Barker, 2013). As suggested by Al-Adwan and Al-Debei (2024), there was no significant impact of social influence on the adoption of metaverse technology among young students from Generation Z. From Table 5, the majority of our samples fall within the same age group. Further, Prasetyo et al. (2021) found social influence an insignificant predictor of eLearning platform adoption among doctorate students. It is important to consider that our research was conducted on well-educated, young students during a period of social isolation. The influence of peers is not as significant as other important factors in our research findings.

Previous studies have found prior experience with technology use leads to persistent actions reflecting through behavioral intention of individuals (El-Masri and Tarhini, 2017; Masa’deh et al., 2016; Moorthy et al., 2019; Raman and Thannimalai, 2021). However, our finding suggests a contradictive result – habit plays an insignificant role for student participants’ intention of LMS use during the pandemic. As the majority of Mainland Chinese postgraduate students enrolled into one-year full-time taught postgraduate degree programs, it is reasonable for them to perceive their discontinuation of LMS use after graduation, regardless of the positive experience during study.

Hedonic motivation associates with intrinsic motivation such as fun, pleasure or enjoyment of using the system. Before the pandemic, hedonic motivation was frequently identified as a significant motivator of behavioral intention in educational context (Al-Gahtani, 2016; El-Masri and Tarhini, 2017). However, the finding of this study indicates hedonic motivation has no direct significant relationship with the acceptance of LMSs, which is consistent with other prior studies conducted during pandemic (Prasetyo et al., 2021; Raman and Thannimalai, 2021). One possible explanation is that as the use of LMSs became mandatory for accessing online classes (Chen et al., 2013), the effect of hedonic factors like fun and pleasure on behavioral intention became less significant for Mainland Chinese postgraduate students. When these students were required to use the LMSs consistently and repeatedly, their intention to use the systems was likely driven more by utilitarian factors, such as the systems’ functional capabilities and necessity for class participation, rather than by inherent enjoyment or entertainment value.