1. INTRODUCTION
The rapid development of information technology (IT) helps provide rich educational content over the Internet, which is known as online learning. This refers to the integration of education and the learning process with modern technologies. Flexibility and accessibility are the main advantages of this learning technology (Wu et al., 2010). Moreover, online learning facilitates the learning process, reduces costs, and provides accessible education (Mahande & Malago, 2019). Besides these advantages, several issues and challenges need more effort to achieve the desired goals of online learning. In its early application, limited access and communication represented its key drawbacks (Chin, 1999). With time, this was replaced with other issues such as isolation, lack of motivation and direct guidance, and lack of experience (Abbad, 2021; Bouhnik & Marcus, 2006; Dutton et al., 2001).
Despite the spreading use of online learning technologies around the world, developing countries still face many challenges. This is due to the low rate of technology acceptance, the unavailability of adequate technical and human infrastructure, and the lack of information sharing and institutional cooperation (Kim & Park, 2018). Accordingly, online learning systems are not gaining popularity in developing countries (Farid et al., 2015). In Iraq, for example, after the partial and full shutdown due to the COVID-19 pandemic, all educational institutions adopted distance and blended education. However, many obstacles still prevent the successful application of this technology in Iraq (Al-Azawei et al., 2016; Al‐Radhi, 2008). To address such issues, it is necessary to understand what can affect users’ acceptance and adoption of online learning technology (Ashraf et al., 2016). Thus, identifying factors that may affect online learning adoption can help educational institutions apply specific strategies to attract students towards this technology (Park, 2009).
Consequently, it is necessary to make further efforts to expose the common factors that may influence learners’ decisions in adopting online learning technology, particularly in developing countries. This present study, therefore, aims to (1) investigate learners’ adoption of online learning in Iraq as a case of developing nations, (2) extend the Unified Theory of Acceptance and Use of Technology (UTAUT) model to understand the influence of other theories on technology acceptance, (3) improve the predictability of UTAUT by integrating new variables, and (4) highlight constructs that can help predict effort expectancy and performance expectancy, as this has been neglected in the original UTAUT. The research draws upon UTAUT (Venkatesh et al., 2003). However, it proposes an integrated model based on four theories, namely UTAUT, compatibility, online self-efficacy, and knowledge sharing and knowledge acquisition. Two motivations are behind this extension. First, this can address the deficit of UTAUT by investigating the influence of the integrated theories on online learning adoption. Second, these theories can complement each other, as they look towards technology acceptance from different angles, particularly in online learning systems.
2. PREVIOUS WORK AND THE PROPOSED MODEL
Implementing IT relies on user acceptance (Davis, 1989). The acceptance may refer to user satisfaction with this particular technology to accomplish the activities and tasks for which the technology was intended (Al-Azawei, 2019; Walldén et al., 2016; Wixom & Todd, 2005). Ignoring learners’ perceptions, on the other hand, can negatively affect the acceptance of educational technologies (Alowayr & Al-Azawei, 2021).
To reach the desired acceptance, many efforts have been made in several domains including psychology, sociology, and information system (IS), to develop theoretical models for predicting and explaining user acceptance of IT or IS (Chao, 2019). The technology acceptance model (TAM) of Davis (1986) is a widely cited theory in this field. However, some studies have pointed out many disadvantages of TAM related to external variables of its perceived usefulness and perceived ease of use (Chao, 2019; Tsai et al., 2018). Therefore, Venkatesh et al. (2003) proposed another theory which is the so-called UTAUT. It has attracted considerable attention in online learning acceptance research (Abbad, 2021; Alowayr & Al-Azawei, 2021; Mahande & Malago, 2019).
Although UTAUT has been widely used and adopted, there is controversy and doubt about its capability to explain users’ technology acceptance. This may indicate that the factors and variables used in UTAUT may not be sufficient to determine the required level of technology adoption. Accordingly, previous research modified and extended UTAUT (Chao, 2019; Cimperman et al., 2016; Khalilzadeh et al., 2017; Mtebe et al., 2016). Such literature suggested that adding several different external variables could improve the model’s predictability. In online learning, Alowayr and Al-Azawei (2021) extended the model based on the self-determination and expectation-confirmation theories. Based on UTAUT, Kim and Lee (2020) built a conceptual framework of effective information communication technologies-based instruction.
To address some of UTAUT’s limitations, this present study extends it by constructing an integrated framework based on four well-known theories. Here, four external variables are integrated with UTAUT, which are compatibility, online self-efficacy, knowledge sharing, and knowledge acquisition. This study, therefore, investigates the effect of these four external constructs as depicted in the proposed research model (see Fig. 1).
2.1. Unified Theory of Acceptance and Use of Technology
UTAUT is a research theory based on psychology and sociology, developed from previous models (Venkatesh et al., 2003). Practically, UTAUT is used for the explanation of user perception and acceptance behavior. The original UTAUT model consists of four essential variables and four moderators. Performance expectancy is the belief of users that the system (specific technology) can help to improve job performance, while effort expectancy refers to users’ beliefs that technology does not need a high mental effort. Social influence means the social pressure on users’ decisions and their perceptions when other parties important to them believe that they should use the technology. Finally, facilitating conditions refer to users’ beliefs that technical and organizational infrastructures are available to support the use of technology (Venkatesh et al., 2012). Moreover, UTAUT proposed four key moderators, namely the voluntariness of use, experience, gender, and age.
Aliaño et al. (2019) highlighted that effort expectancy, performance expectancy, and social influence were determinants of behavioral intention to use online learning. Furthermore, Davis (1986) confirmed that effort expectancy is a predictor of performance expectancy. In UTAUT, Venkatesh et al. (2003) found that behavioral intention and facilitating conditions were predictors of actual use. Accordingly, the hypotheses proposed in this study are:
-
H1: Effort expectancy significantly affects behavioral intention
-
H2: Effort expectancy significantly affects performance expectancy
-
H3: Performance expectancy significantly affects behavioral intention
-
H4: Social influence significantly affects behavioral intention
-
H5: Facilitating conditions significantly affect the actual use
-
H6: Behavioral intention significantly affects the actual use
2.2. Compatibility
Compatibility refers to the degree to which innovations are perceived as in agreement with the current values, needs, and past experiences of probable adopters (Rogers, 1995). In online learning technology adoption, the compatibility theory has emerged as one of the important factors that may affect the behavioral intention of users to adopt modern technology (Cheng, 2015; Isaac et al., 2019; Ozturk et al., 2016). Chang et al. (2005) showed that there is a need for more effort to demonstrate the importance of this factor in e-learning systems. Moreover, a significant relationship between real usage of mobile learning adoption and compatibility was found (Cheng, 2015). In this research, we assume that the compatibility factor positively affects both behavioral intention and actual use.
2.3. Knowledge Acquisition and Knowledge Sharing
A large number of enrolled students in educational institutions could raise challenges and issues related to knowledge acquisition and knowledge sharing (Al-Emran et al., 2019). Such issues could be addressed by online learning systems (Al-Emran & Teo, 2020). Nevertheless, a few efforts demonstrate the impact of these two factors on the adoption of online learning. In this context, knowledge acquisition and knowledge sharing might have an influential effect on learners’ intention to adopt online learning technology and/or performance expectancy (Al-Emran & Teo, 2020; Lau & Tsui, 2009).
Knowledge acquisition means how a learner gains new knowledge by extracting, structuring, and organizing knowledge from one source (Al-Emran & Teo, 2020; Huang, 2020). Al-Emran et al. (2018) revealed that knowledge acquisition was a determinant of performance expectancy and effort expectancy. Knowledge sharing means the spread of diverse resources among individuals involved in particular activities. Previous studies indicated a positive relationship between knowledge sharing and performance expectancy (Al-Emran et al., 2018; Cheung & Vogel, 2013). According to this discussion, the current study suggested the following hypotheses:
2.4. Online Self-efficacy
Self-efficacy is another widely used cognitive factor that is related to users’ motivational beliefs. It refers to users’ evaluation of their personal ability to complete tasks and goals well (Bandura, 1986). In an online learning context, self-efficacy is “a student’s self-confidence in his or her ability to perform certain learning tasks using the e-learning system” (Tarhini et al., 2014).
In line with the above, self-efficacy shows a significant effect on adopting online learning (Qiao et al., 2021; Zhang & Liu, 2019). Thus, the motivation of students to use online learning systems is highly related to their successful adoption (Wang & Newlin, 2002). Earlier literature showed that self-efficacy was a determinant of both performance expectancy and effort expectancy (Wu, 2017; Yilmaz, 2016; Zhang & Liu, 2019). According to this discussion, we assumed that
3. RESEARCH METHODOLOGY
This study adopts a survey research design to examine the cause and effect relationships among the proposed research model. One of the advantages of this research is that the subjectivity issues are not found, as researchers rely solely on the statistical findings to understand the possible causality associations between different constructs.
3.1. The Research Instrument
The distributed online questionnaire included the main page that presented the key aims of the questionnaire, its filling out time, ethical considerations, and a few instructions to answer its questions correctly. This was followed by general questions to collect demographic information from the research participants. The third part consisted of items that were designed to measure students’ perceptions and intention to use online learning. Overall, 38 questions were designed based on previous literature to cover the ten constructs of the proposed research model, namely behavioral intention, actual use, knowledge acquisition, knowledge sharing, performance expectancy, online self-efficacy, effort expectancy, facilitating conditions, social influence, and compatibility (see Appendix). A five-point Likert scale was adopted, ranging from one which means ‘strongly disagree,’ whereas five refers to ‘strongly agree.’
3.2. The Research Participants
This study recruited undergraduate and postgraduate students from a public university in one of the center governorates in Iraq. Of about 400 students, 264 responded voluntarily to the online questionnaire with a response rate of 66%. The research subjects agreed to take part in this study based on submitting the questionnaire as an indicator of their consent. It was also illustrated in the questionnaire that all collected data would be used for research purposes only and would not be shared with a third party.
Out of the 264 participants, 93 (35.2%) were man and 171 (64.8%) were woman. Regarding the age class, 202 (76.5%) were aged from 18 to 22, while only 62 (23.5%) were 23 years old or over. In terms of the study type, 182 (68.9%) were from the morning study, whereas 82 (31.1%) were from the evening study. Table 1 shows the key features of the research participants.
Table 1
Participants’ information
| Variable | Value (n=264) |
|---|---|
| Year | |
| First | 107 (40.5) |
| Second | 44 (16.7) |
| Third | 31 (11.7) |
| Fourth | 80 (30.3) |
| Postgraduate | 2 (0.8) |
| Gender | |
| Man | 93 (35.2) |
| Woman | 171 (64.8) |
| Age (yr) | |
| 18-22 | 202 (76.5) |
| 23 or more | 62 (23.5) |
| Study type | |
| Morning | 182 (68.9) |
| Evening | 82 (31.1) |
3.3. Data Collection
The research data were collected in the second semester of 2020-2021. The participation was voluntary, and the authors provided a general view of the purpose of the study before filling out the research questionnaire. The questionnaire was distributed online via Google Classroom, as it was the online learning platform used by the university. All received responses were valid because all questions were required to prevent submitting incomplete answers. The filling procedure of the research questionnaire took from 10 to 15 minutes. Here, the convenience sampling approach was adopted. The research data were analyzed using SmartPLS software package version 3.0 (Ringle et al., 2015). The p-value was set to 0.05.
4. RESULTS
The proposed theoretical model was investigated using partial least squares (PLS). In comparison to traditional statistics, such as regression, this method has many advantages. First, it can be used to examine the association among a series of constructs (Al-Azawei, 2017). Moreover, according to Chin (1998), structural equation modeling (SEM) is a superior method for theory development and prediction. Finally, previous research on predicting users’ behavior has widely adopted this technique (Al-Azawei & Alowayr, 2020; Ameen et al., 2019; Shin & Kang, 2015). The collected data were analyzed in two steps. The first was validating the research survey, whereas the second was examining the predictability of the independent variables to the dependent constructs.
4.1. Validating the Research Questionnaire
The significance of confirming reliability and validity comes from the influence of both measurements on the quality of the gathered data (Pallant, 2013). Moreover, the impact of reliability and validity cannot be about the quality of data only, as those characteristics can also affect the research findings and recommendations (Al-Sabawy, 2013). In SEM, examining reliability and validity is a step prior to investigating the structural model. This is to confirm the questionnaire’s validity. In this research, convergent, construct, and discriminant validities were validated.
The reliability can be evaluated based on measuring the unidimensionality of each factor in the research model. Based on the threshold proposed by Hulland (1999), the unidimensionality of the research questionnaire was supported because the outer loadings of all indicators used to measure the research constructs were greater than 0.7. Moreover, Cronbach’s alpha and composite reliability (CR) were all greater than 0.7 as assumed by Pallant (2013). Based on Tables 2 and 3, it is clear that the outer loadings of all indicators used to measure a particular variable are greater than 0.7; as well, Cronbach’s alpha (internal consistency) and CR are more than the threshold of 0.7. Thus, the research questionnaire has confronted all thresholds recommended and this, in turn, clearly supports its reliability and validity. The Cronbach’s alpha and CR of all constructs ranged from 0.824 to 0.940 and from 0.896 to 0.957 for both measurements respectively.
Table 2
Outer loading
| Item | AU | BI | Comp | EE | FC | KA | KS | OSE | PE | SI |
|---|---|---|---|---|---|---|---|---|---|---|
| BI1 | 0.935 | |||||||||
| BI2 | 0.897 | |||||||||
| BI3 | 0.946 | |||||||||
| BI4 | 0.904 | |||||||||
| COMP1 | 0.913 | |||||||||
| COMP2 | 0.948 | |||||||||
| COMP3 | 0.915 | |||||||||
| EE1 | 0.879 | |||||||||
| EE2 | 0.890 | |||||||||
| EE3 | 0.887 | |||||||||
| EE4 | 0.895 | |||||||||
| FC1 | 0.888 | |||||||||
| FC2 | 0.885 | |||||||||
| FC3 | 0.808 | |||||||||
| KA1 | 0.892 | |||||||||
| KA2 | 0.905 | |||||||||
| KA3 | 0.876 | |||||||||
| KA4 | 0.895 | |||||||||
| KA5 | 0.836 | |||||||||
| KS1 | 0.752 | |||||||||
| KS2 | 0.822 | |||||||||
| KS3 | 0.860 | |||||||||
| KS4 | 0.900 | |||||||||
| KS5 | 0.858 | |||||||||
| OSE1 | 0.873 | |||||||||
| OSE2 | 0.917 | |||||||||
| OSE3 | 0.910 | |||||||||
| PE1 | 0.878 | |||||||||
| PE2 | 0.910 | |||||||||
| PE3 | 0.861 | |||||||||
| PE4 | 0.873 | |||||||||
| SI1 | 0.923 | |||||||||
| SI2 | 0.926 | |||||||||
| SI3 | 0.926 | |||||||||
| USE1 | 0.910 | |||||||||
| USE2 | 0.912 | |||||||||
| USE3 | 0.910 | |||||||||
| USE4 | 0.746 |
Convergent validity refers to a type of measurement validity for multiple indicators according to the notion that indicators of one construct will act alike or converge, whereas construct validity means how well the indicators of one variable converge or how well the indicators of different variables diverge (Bernard, 2012). Average variance extracted (AVE) between 0.5 and 1 is a good indicator to confirm the convergent validity (Fornell & Larcker, 1981). Table 3 confirms the convergent validity of the research questionnaire. On the other hand, the discriminant validity of a research measurement was also proved as discussed by Alowayr and Al-Azawei (2021). Based on Table 4, the discriminant validity has also been supported. Finally, construct validity can be evaluated based on the goodness of fit (Al-Sabawy, 2013) as indicated in Table 5.
Table 3
Construct reliability and validity
| Factor | Cronbach’s alpha | rho_A | Composite reliability | Average variance extracted |
|---|---|---|---|---|
| Actual use | 0.893 | 0.908 | 0.927 | 0.761 |
| Behavioral intention | 0.940 | 0.940 | 0.957 | 0.848 |
| Compatibility | 0.916 | 0.917 | 0.947 | 0.857 |
| Effort expectancy | 0.910 | 0.911 | 0.937 | 0.788 |
| Facilitating conditions | 0.824 | 0.826 | 0.896 | 0.741 |
| Knowledge acquisition | 0.928 | 0.928 | 0.946 | 0.777 |
| Knowledge sharing | 0.895 | 0.897 | 0.923 | 0.705 |
| Online self-efficacy | 0.883 | 0.888 | 0.928 | 0.810 |
| Performance expectancy | 0.904 | 0.905 | 0.933 | 0.776 |
| Social influence | 0.916 | 0.916 | 0.947 | 0.856 |
Table 4
Discriminant validity
| Factor | AU | BI | Comp | EE | FC | KA | KS | OSE | PE | SI |
|---|---|---|---|---|---|---|---|---|---|---|
| AU | 0.872 | |||||||||
| BI | 0.797 | 0.921 | ||||||||
| Comp | 0.812 | 0.832 | 0.926 | |||||||
| EE | 0.673 | 0.666 | 0.716 | 0.888 | ||||||
| FC | 0.695 | 0.702 | 0.723 | 0.727 | 0.861 | |||||
| KA | 0.759 | 0.772 | 0.752 | 0.719 | 0.679 | 0.881 | ||||
| KS | 0.673 | 0.715 | 0.653 | 0.691 | 0.735 | 0.815 | 0.840 | |||
| OSE | 0.735 | 0.737 | 0.755 | 0.691 | 0.764 | 0.696 | 0.679 | 0.900 | ||
| PE | 0.782 | 0.793 | 0.750 | 0.739 | 0.746 | 0.819 | 0.773 | 0.730 | 0.881 | |
| SI | 0.647 | 0.688 | 0.707 | 0.677 | 0.692 | 0.697 | 0.656 | 0.677 | 0.687 | 0.925 |
Table 5
Model fit
| The goodness of fit | Saturated model | Estimated model |
|---|---|---|
| SRMR | 0.046 | 0.084 |
| d_ULS | 1.581 | 5.288 |
| d_G | 1.150 | 1.298 |
| Chi-square | 1,666.087 | 1,746.200 |
| NFI | 0.843 | 0.835 |
4.2. Findings of the Proposed Model
Table 6 shows that ten out of twelve hypotheses were supported based on the path association among the proposed variables, whereas Fig. 2 illustrates R2 and β values of the model after performing the PLS. According to this analysis, the results indicate that performance expectancy (βPE→BI=0.371, p<0.001) and compatibility (βComp→BI=0.514, p<0.001) were significant predictors of behavioral intention to explain 76.1% (R2=0.761) of the variance of this construct, supporting hypotheses H3 and H7. On the other hand, effort expectancy (βEE→BI=-0.043, p=0.428) and social influence (βSI→BI=0.100, p=0.055) were not determinants of behavioral intention, and so hypotheses H1 and H4 were rejected.
Table 6
Research findings
| Hypothesis | β | Mean | SD | T-value | p-value | Findings |
|---|---|---|---|---|---|---|
| H1: Effort expectancy -> Behavioral intention | -0.043 | -0.043 | 0.055 | 0.793 | 0.428 | Rejected |
| H2: Effort expectancy -> Performance expectancy | 0.193 | 0.193 | 0.062 | 3.117 | 0.002 | Supported |
| H3: Performance expectancy -> Behavioral intention | 0.371 | 0.371 | 0.067 | 5.547 | 0.000 | Supported |
| H4: Social Influence -> behavioral intention | 0.100 | 0.100 | 0.052 | 1.923 | 0.055 | Rejected |
| H5: Facilitating conditions -> Actual use | 0.152 | 0.156 | 0.057 | 2.683 | 0.007 | Supported |
| H6: Behavioral intention -> Actual use | 0.345 | 0.341 | 0.067 | 5.147 | 0.000 | Supported |
| H7: Compatibility -> Behavioral intention | 0.514 | 0.513 | 0.054 | 9.538 | 0.000 | Supported |
| H8: Compatibility -> Actual use | 0.416 | 0.416 | 0.070 | 5.951 | 0.000 | Supported |
| H9: Knowledge acquisition -> Performance expectancy | 0.392 | 0.393 | 0.074 | 5.321 | 0.000 | Supported |
| H10: Knowledge sharing -> Performance expectancy | 0.186 | 0.185 | 0.070 | 2.661 | 0.008 | Supported |
| H11: Online self-efficacy -> Effort expectancy | 0.691 | 0.692 | 0.040 | 17.358 | 0.000 | Supported |
| H12: Online self-efficacy -> Performance expectancy | 0.198 | 0.197 | 0.068 | 2.922 | 0.003 | Supported |
Regarding the predictors of actual online learning use, the findings show that facilitating conditions (βFC→AU=0.152, p=0.007), behavioral intention (βBI→AU= 0.345, p<0.001), and compatibility (βComp→AU=0.416, p<0.001) were determinants of this construct to explain 71.8% (R2=0.718) of its overall variance. Such results supported hypotheses H5, H6, and H8.
The results also confirm that four constructs were predictors of performance expectancy, namely effort expectancy (βEE→PE=0.193, p=0.002), knowledge acquisition (βKA→PE=0.392, p<0.001), knowledge sharing (βKS→PE=0.186, p=0.008), and online self-efficacy (βOSE→PE=0.198, p=0.003). These variables explained 75.1% (R2=0.751). Thus, hypotheses H2, H9, H10, and H12 were confirmed. Finally, online self-efficacy (βOSE→EE=0.691, p<0.001) was also a significant predictor of effort expectancy to explain 47.7% of its variance. This confirms hypothesis H11.
5. DISCUSSION
This research extends UTAUT to investigate the acceptance of online learning. It takes into account compatibility as a direct predictor of behavioral intention. The study also assumes that knowledge acquisition and knowledge sharing are determinants of performance expectancy, whereas online self-efficacy is proposed as a predictor of effort expectancy and performance expectancy. Thus, this research addresses some of the weaknesses in the original UTAUT, as it did not explain what can affect effort and performance expectancies. It also ignored the role of compatibility in examining online learning adoption. Table 7 summarizes the key findings of the proposed hypotheses. From Fig. 2, it is clear that knowledge acquisition was the strongest predictor of performance expectancy (βKA→PE=0.392). Compatibility, on the other hand, was the best predictor of both behavioral intention (βComp→BI=0.514) and actual use (βComp→AU=0.416). This reveals that extending UTAUT with the new constructs such as knowledge acquisition and compatibility has a significant influence on improving its predictability.
Table 7
Summary of the research hypotheses and findings
| Hypothesis | Findings |
|---|---|
| Hypotheses for predicting performance expectancy | |
| H2: Effort expectancy -> Performance expectancy | Supported |
| H9: Knowledge acquisition -> Performance expectancy | Supported |
| H10: Knowledge sharing -> Performance expectancy | Supported |
| H12: Online self-efficacy -> Performance expectancy | Supported |
| Hypotheses for predicting behavioral intention | |
| H1: Effort expectancy -> Behavioral intention | Rejected |
| H3: Performance expectancy -> Behavioral intention | Supported |
| H4: Social influence -> Behavioral intention | Rejected |
| H7: Compatibility -> Behavioral intention | Supported |
| Hypotheses for predicting actual use | |
| H5: Facilitating conditions -> Actual use | Supported |
| H6: Behavioral intention -> Actual use | Supported |
| H8: Compatibility -> Actual use | Supported |
| Hypotheses for predicting effort expectancy | |
| H11: Online self-efficacy -> Effort expectancy | Supported |
As for the predictors of performance expectancy, all proposed variables, namely effort expectancy, knowledge acquisition, knowledge sharing, and online self-efficacy, were determinants of this construct, supporting hypotheses H2, H9, H10, and H12. These four constructs explained 75.1% of the variance of performance expectancy. In agreement with our results, earlier research revealed that effort expectancy is a predictor of performance expectancy (Al-Emran et al., 2020; Davis, 1989), indicating that users can perceive the usefulness of technology if its use does not require a high mental effort. Moreover, other studies also found that both knowledge acquisition and knowledge sharing were predictors of performance expectancy (Al-Emran & Teo, 2020). This means that online courses should be designed in a way that allows participants to share knowledge and improve its acquisition. Thus, when students are capable of sharing and acquiring knowledge through online learning technology, their perceptions about technology usefulness in improving their overall performance can be enhanced as well. The research analysis also shows that online self-efficacy had a direct and positive relationship with performance expectancy. This may indicate that when learners are self-confident about their personal abilities in using online learning technology, their perceptions regarding its usefulness will be increased accordingly.
Online self-efficacy as a reference of learners’ self-expectation was a predictor of effort expectancy to explain 47.7% of its variance, confirming hypothesis H11. Bandura (1986) assumes that people’s expectations of efficacy can determine subsequent behavior in terms of the beginning and continuous coping behavior. According to Lin (2021), the self-efficacy of people may reveal the motivational level of effort invested in an endeavor and this, in turn, could indicate that an individual with a low self-efficacy level will not persist in adopting or using technology. Based on our analysis, it is obvious that when students have a high level of self-efficacy, a low effort will be required to perform a particular action.
It was also found that two constructs out of four were determinants of behavioral intention, namely performance expectancy and compatibility, confirming hypotheses H3 and H7 to explain 76.1% of the variance of behavioral intention towards online learning use. On the other hand, this analysis rejects hypotheses H1 and H4. Overall, our findings are in agreement with earlier research that performance expectancy is a determinant of behavioral intention (Alowayr & Al-Azawei, 2021; Wang & Lin, 2021). This means that students are willing to continue using online learning because of its usefulness in improving their performance. The research analysis also shows that compatibility had a significant effect on intention to use. Such a finding is consistent with the outcomes of Wang and Lin (2021). This may indicate that students had a high level of compatibility in online learning settings and this, in turn, reflected their positive attitudes towards online classes. On the contrary, social influence and effort expectancy had no significant influence on behavioral intention. This may indicate that students had positive perceptions of the advantages of online learning regardless of the effort that is required to adopt it or the perspectives of other people who are important to the students. In an earlier research study (Al-Azawei & Alowayr, 2020), it was also found that both effort expectancy and social influence did not affect students’ decision or perception to use mobile learning technology in the Iraqi context. A possible explanation for the weak influence of effort expectancy is that as students are part of the youth, they use online technologies daily, so their experience in dealing with them is great. Moreover, Ameen and Willis (2018) also revealed that the social influence of Iraqi students did not affect behavioral intention.
The research findings suggest that facilitating conditions, behavioral intention, and compatibility were determinants of online learning’s actual use, which explained 71.8% of the variance of this construct and supported hypotheses H5, H6, and H8. In UTAUT, Venkatesh et al. (2003) assume that facilitating conditions could be a predictor of technology’s actual use. This assumption was confirmed in our research to indicate that educational institutions have to provide reliable information and communication technologies infrastructure, and support all users to ensure the successful application of online learning. Furthermore, based on previous literature, behavioral intention towards a particular action can indicate that people will actually perform it (Ajzen & Fishbein, 1980). Finally, in this research, compatibility was the strongest predictor of online learning usage. This means that students might perceive the consistency of this technology with their individual needs, particularly after the spread of the COVID-19 pandemic worldwide. It also suggests that when learning technology is compatible with learners’ individual needs and practice styles, their actual use will increase accordingly. Such an outcome is in line with the findings of earlier literature that real usage of mobile learning can be predicted by compatibility (Cheng, 2015). In agreement with the UTAUT hypotheses (Venkatesh et al., 2003), technology’s actual use can be positively enhanced if users feel that there is direct organizational support for technical issues that they may face.
Based on the research findings and analysis, many theoretical and practical implications can be drawn. Theoretically, this research extends previous work on online learning adoption, particularly in the case of developing countries. Furthermore, it extends current technology acceptance theories according to the integration of UTAUT with variables from the social cognitive theory (self-efficacy), compatibility theory (compatibility), and knowledge management topic (knowledge sharing and knowledge acquisition). Accordingly, this research is among the early literature that investigates the extension of UTAUT in an online learning context. From the practical view, this research provides clear guidance for decision-makers, teachers, and online technology designers to apply innovative means in online learning to enhance knowledge sharing and acquisition among learners. Moreover, educational institutions should ensure that both academic staff and students have high confidence in their personal abilities to use the technology, as this could affect their decision to adopt it. Finally, as compatibility was a predictor of behavioral intention and actual use, online learning courses should be designed in a manner that responds to learners’ individual expectations and preferences.
6. CONCLUSION
This research aimed at extending UTAUT to understand online learning acceptance. To achieve this aim, four constructs were integrated with UTAUT, namely compatibility, knowledge acquisition, knowledge sharing, and online self-efficacy. Although UTAUT suggested that effort expectancy and performance expectancy are determinants of adoption behavior, it does not explain what can affect these factors. Accordingly, the present study confirmed that knowledge acquisition and/or sharing as well as online self-efficacy had a significant role in predicting such constructs. Moreover, this research supports the necessity of considering compatibility in predicting both behavioral intention and actual behavior.
In regard to future research directions, it is recommended to carry out a longitudinal research design to understand the phenomenon under investigation over a longer period. Furthermore, choosing a sample from different disciplines and cultures should be another aim of future research. Finally, researchers also need to analyze the role of knowledge sharing and knowledge acquisition in other contexts of the learning process.
REFERENCES
(2017) (Doctoral dissertation) Modelling e-learning adoption: The influence of learning style and universal learning theories, https://centaur.reading.ac.uk/77921/
, , (2018) Students' perceptions towards the integration of knowledge management processes in m-learning systems: A preliminary study International Journal of Engineering Education, 34(2A), 371-380 https://scholar.google.com/citations?view_op=view_citation&hl=da&user=J-p9z_gAAAAJ&citation_for_view=J-p9z_gAAAAJ:VaXvl8Fpj5cC.
(2013) (Doctoral dissertation) Measuring e-learning systems success, https://eprints.usq.edu.au/27422/
(1998) Commentary: Issues and opinion on structural equation modeling MIS Quarterly, 22(1), vii-xvi https://www.jstor.org/stable/249674.
(1986) (Doctoral dissertation) A technology acceptance model for empirically testing new end-user information systems: Theory and results, http://hdl.handle.net/1721.1/15192
(2009) An analysis of the technology acceptance model in understanding university students' behavioral intention to use e-learning Educational Technology & Society, 12(3), 150-162 https://www.jstor.org/stable/jeductechsoci.12.3.150.
, , (2015) SmartPLS 3 http://www.smartpls.com
, (2021) The relationships among students' personal innovativeness, compatibility, and learning performance: A social cognitive theory perspective Educational Technology & Society, 24(2), 14-27 https://www.jstor.org/stable/27004928.