Acceptance of E-learning among university students in UAE: A practical study

Received Oct 11, 2019 Revised Feb 2, 2020 Accepted Feb 10, 2020 The objective of the study is to find out the factors which affect the acceptance of E-learning among students and how the students’ purpose of using E-learning is determined by these factors. The research was based on the Technology Acceptance Model (TAM). The researcher developed a questionnaire to gather data from a sample of 366 university students who actively used the E-learning system. According to the results of the study, subjective norms, perceived usefulness, perceived ease of use, enjoyment, and accessibility” are the vital predicators behind the intention of students for using E-learning systems. This indicates that extended TAM is applicable in the UAE. The results show that policymakers and education developers should take E-learning system seriously since it can be opted not just as a technological solution but also as a learning platform for students belonging to a distant area. The results present practical implications for education developers, policymakers and practitioners in devising useful plans to implement E-learning systems.


INTRODUCTION
As the Information technology continues to evolve at a very fast pace [1], World Wide Web, also called as the Web, has emerged as an influential source through which learning and teaching have become possible at a distance [2]. Through the Internet, users can get a lot of information and can get better chances to interact and cooperate with others. By relying on Internet, students can get access to several learning tools for devising new techniques and means for the implementation of the gathered knowledge or creating new ones [3]. Different terms like E-learning, online learning, and web-based learning are used to explain the teaching and learning experiences gained via the Internet and Web technology [4,5]. E-learning refers to a virtual class through which different kinds of information can be achieved through the internet where, unlike an actual classroom environment, the teachers or students are not present [6]. Students can rely on Internet as a source of information through which they can learn whenever and wherever they want [7]. The course materials available on the Internet can be accessed at any time which means that students can study the same material repeatedly [5]. The flexibility and ease that E-learning offers are very fascinating for numerous students involving adults as well as professionals who might not have been able to get advanced degrees if the only option they had was the conventional studying technique. Several top organizations are now implementing E-learning into their present business operations [5]. Furthermore, many high education institutions and corporate training are adopting E-learning for sorting out different learning and performance related issues [5,8]. E-learning is among the most important modern advancements that occurred in the IS industry which has placed a new challenge for schools as well as the industry. It is important to study this topic as a number of schools are educating their students through E-learning

Subjective norm (SUB)
Subjective norm and social influence is the idea that significant people believe that the system should either be used or not [41]. The subjective norm refers to "the person's perception that most people who are important to him or her think he or she should or should not perform the behavior in question" [42]. Under some circumstances, rather than concentrating on their personal emotions and beliefs, people may adopt a system just to adapt to the needs of others [43]. Many researchers have been conducted on how E-learning adoption or acceptance is affected by subjective norms. A research conducted by [22] showed that perceived ease of use [9,22,23,25,36,37,[44][45][46] and perceived usefulness of E-learning system [23,37,[45][46][47][48] are considerably affected by subjective norms. Hence, the subsequent hypothesis has been developed: H3a: Subjective norm has a significant influence on the perceived ease of use (PEOU). H3b: Subjective norm has a significant influence on perceived usefulness (PU).

The technology acceptance model and user beliefs
The perceived usefulness and perceived ease of use aid the acceptance of new technology. TAM and other relevant studies show that these factors considerably affect the behavior intention to use. The views of a user concerning the ease of comprehending a system significantly aids in recognizing the effectiveness and the degree to which the user shows adoption to new technology [49,50]. A previous study demonstrated that perceived usefulness is considerably affected by the perceived ease of use (PEOU) [29,40,45,46,[51][52][53][54][55].
There are many studies that show a direct as well as an indirect positive relationship between the behavioral intention to use E-learning system (IU) and perceived ease of use (PEOU) [29,40,46,52,54,[56][57][58][59]. It was stated by [51] that the intention to directly or indirectly approve of and adapt to an E-learning system will be affected by perceived usefulness (PU). The past studies show that there exists a positive connection between the perceived usefulness (PU) and intention to use the learning system (IU) [23,54,56,[60][61][62][63][64][65][66]. While keeping in view the literature, the research states the following hypotheses, along with showing a significant positive relationship between IU, PU, and PEOU. H4: Perceived ease of use (PEOU) has a significant influence on the perceived usefulness (PU). H5: Perceived usefulness (PU) has a significant influence on the intention to use the E-Learning system (IU). H6: Perceived ease of use (PEOU) has a significant influence on the intention to use the E-Learning system (IU). The research model is based on these hypotheses as demonstrated in Figure 1. The theoretical model is first given the form of a structural equation model and then it is assessed by using scientific methods.

RESEARCH METHODOLOFY 3.1. Sample and data collection
The data was gathered from two renowned educational institutions in the United Arab Emirates (UAE) who had been using E-learning systems. The online survey involved 400 respondents. Two web-based E-learning systems created by two E-learning platform/system providers were being used in these universities. These E-learning platforms were created and organized about 4 years before and were already being used by the students regularly. The study was carried out through Smart PLS Version T3.2.7 i.e. Structural equation modeling or SEM. SEM was used to assess this measurement model and final path model was managed in the later stages. A detailed image of the collected data has been shown in Table 1. Thirty-four incomplete responses questionnaires had to be discarded from the gathered responses. After that, 366 questionnaires were available that were completed by the respondents which indicate a response rate of 91.5%. A total of 366 responses which involved only valid responses were considered and transformed into a sample size used by [67]. For a population of 300, a sampling size of approximately 169 respondents is used. The conceptual model was then used to examine the responses. It is acceptable to analyze the data through structural equation modeling and create a sample size; hence, the sample size of 366 used in the study was way more than the unnecessary requirements used for hypotheses testing [68]. It must be remembered that the stated hypotheses were developed in the light of present theories but were taken in the context of E-learning.

Study instrument
As mentioned before in the research, the hypothesis was tested through a survey instrument. The number of sources of the constructs involved in the questionnaire is shown in table 2. The survey involves seventeen items with the goal of assessing the six constructs in the questionnaire. Some questions were also taken from the previous studies and included in this study as an attempt to make the study more logical and relatable. Though the questions which the researcher took from the past studies were changed and adjusted according to the study context.

Instrumentation
A tough phase of pre-testing and pilot testing of measures was started once some E-learning users and experts were chosen. The genuine measures of the research are shown in table 3. In the questionnaire, there are three items for each construct and a five-point Likert scale was used to measure these constructs. The scale goes from "strongly disagree" to "strongly agree". The respondents were asked to choose the response they agree to while keeping in view the given statements. In addition to this, demographic information was also required by the respondents.

Pre-test of the questionnaire
University of Fujairah (UOF) and The British University in Dubai (BUiD), two very renowned educational institutions, took part in the study. The researcher took these universities as the target population consisting of a group of E-learning users who have good experience and knowledge about online learning system. For the pilot study, the researcher used 10% of the total sample size of the research survey i.e. 400 respondents. The sample size was chosen in light of the standard research practices. In addition to this, all the questions involved in the questionnaire were pre-tested where 40 students were used who were chosen on random basis. Moreover, as per [72], the reliability of the study was found through Cronbach's Alpha  Table 3. The results indicate the questionnaire used for the study is very reliable. It was ensured that all the respondents understood the final questionnaire as an attempt to further increase the overall reliability and survey quality.

Students' personal information/demographic data
The detailed information regarding the respondents can be seen in Table 4. Nearly the same features were shared by all the participants of the study. A total of 198 females were included which makes 54% of the total participants where 168 males were included which makes 46% of the total participants. Majority of the respondents were between the ages of 18-29, where 65% of the respondents were between the ages of 30-39, 27% were between the ages of 40-49, 5% between the ages of 50-59. The detailed demographic information of the respondents is demonstrated in Table 4.

FINDINGS AND DISCUSSIONS
Smart PLS software, which was developed by [73], is used in this study for Partial Least Squares Structural Equation Modeling (PLS, SEM). It is a renowned software and is extensively used in academia [74,75]. The software is easily accessible to academics and researchers. In addition to this, the software involves an extremely user-friendly interface and modern reporting characteristics. Due to these characteristics, the software has been gaining a lot of popularity since its introduction in 2005 [76]. The two classes of validities i.e. convergent validity and discriminate validity are usually used to assess any measurement model [77]. The measurement model describes the link between the indicators and latent construct that is being assessed.

Data analysis
The Reliability of indicators can be assessed easily by examining the extent to which indicators show internal consistency in a relationship with any construct. Hence, the reliability of indicators can be measured through Cronbach coefficient alpha [78] and the composite reliabilities coefficient [79]. The minimum acceptable limit of composite reliabilities is 0.70 [80,81]. The values of all the coefficients and Cronbach coefficient alpha levels are shown in Table 5. All the values were greater than 0.70, which verifies the quality of research [82]. The data was analyzed by employing a two-stage methodology in which at first the measurement model is developed and then assessed independently from the complete structural equal model [83]. Due to that, every item, convergent and discriminate validity of constructs were checked for reliability as an initial step. Each item is checked for individual reliability through loadings or correlations that exist between the items and construct. The convergent validity per construct is appropriate in case of a loading greater than 0.505 [84]. Table 5 shows that there is a strong agreement between the loadings for each item and established conditions. Factor loading, variance extracted and reliability is used as indicators by following [72] in order to guess the convergent validity's relative value. The composite reliability (CR) and reliability coefficient go above 0.7 for each construct which shows that different measurements of a construct are internally consistent with each other [72]. Cronbach's alpha values go beyond the acceptable value of 0.7 [80,81] whereas composite reliabilities of constructs go from 0.7444 to 0.874. This has been shown in Table 5. Moreover, the standard of explaining a minimum of 50% of variance extracted from a collection of items triggering the latent construct is fulfilled by all average variance extracted (AVE) values which vary from 0.636 to 0.778 [84]. Therefore, it is believed that the scales through which constructs are measured have convergent validity.
It was recommended to consider two criteria i.e. Cornell-Barker criterion and the Hetero traditionalism proportion (HTMT) for achieving the measurement for discriminate validity [85]. Table 6 presents that Cornell-Barker measure verifies that the square roots of all AVEs are more important than its correlation with the remaining constructs [86]. Table 7 presents the HTMT ratio which clearly shows that each construct's value does not exceed the threshold value of 0.85 [87]. This, therefore, verifies the HTMT ratio value. The existence of discriminant validity is confirmed by these results. The data analysis indicates that reliability and validity of the measurement model were measured easily and no issues were encountered. Hence, the collected data can be used to assess the structural model.

Hypothesis testing and coefficient of determination
The structural equation modeling (SEM) approach [88][89][90][91][92][93][94][95][96] was used to test the aforementioned nine hypotheses together. The variance described (R2 value) by each path and every hypothesized connection's path significance in the research model were assessed. The standardized path coefficients and path significances are demonstrated in Figure 2 and Table 3. Table 8 shows that the R2 values for the intention to use the E-Learning system, perceived ease of use and perceived usefulness ranged between 0.466 and 0.537. Therefore, these constructs appear to have Moderate predictive power [88]. Generally, the data supported eight out of nine hypotheses. According to previous studies, all TAMs' constructs were verified in the model (PEOU, PU, and BI). Based on the data analysis hypotheses H1a, H1b, H2a, H2b, H3a, H4, H5, and H6 were supported by the empirical data, while H3b was rejected. The results in table 9

CONCLUSION AND FUTURE WORKS
The study aims to extract factors through which the views concerning E-learning are affected. The study is conducted on the students of UAE. In order to present the proposed research model relevance and hypothesis used for analyzing the behavioral intention to accept E-learning system, the outcomes of data analysis are paid considerable attention. Structure equation modeling (PLS-SEM) was used to analyze the research hypotheses. Figure 2 shows the structural model which is evaluated by examining the structural paths, variance explained (R-squared value) and t-statistics. Table 9 shows the results of data analysis. PLS technique has been used to test the aforementioned nine hypotheses. Upon the path significance of every hypothesized association which is included in the research model and the variance explained (R2) by each path, an assessment was done. Out of all nine hypotheses, eight hypotheses were supported. The hypotheses which were achieved through TAM model (H1a, H1b, H2a, H2b, H3a, H4, H5, and, H6) have been confirmed. The results show that factors like perceived ease of use, subjective norm, accessibility, enjoyment, and perceived usefulness can improve the behavioral intention to embrace E-learning system. The literature review supports these results. Accordingly, it can be assumed that E-learning is quite important when it comes to proficiency and high confidence levels while using online learning platforms.
The study has some limitations which have been mentioned in this part. The research was conducted using only two universities in UAE which is the biggest limitation since the selection was quite limited to study how E-learning system acceptance is affected by different factors. The study could have become more appropriate had the researcher included more universities in UAE in the research. By conducting more research and developing a more genuine understanding of the E-learning system, the factors which affect the actual E-learning system can be understood in a much better way. Another limitation was taking only 366 students as the participants of the study. A survey questionnaire was used to collect the data. In order to achieve more accurate results, a better instrument and sampling method could be used and more educational institutions from other areas such as the Arab Gulf region in which countries like KSA, Kuwait, and Bahrain could be included. In addition to this, in future research, more students will be asked to participate in the study. For better results, focus group, as well as interviews, will be performed. We will also encourage the Arab universities that have participated in the study to incorporate E-learning systems.