Predicting Actual Use of mLearning Systems (2020) - AQA

Interactive Learning Environments ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/nile20 Predicting the actual use of m-learning systems: a comparative approach using PLS-SEM and machine learning algorithms Muhammad Alshurideh , Barween Al Kurdi , Said A. Salloum , Ibrahim Arpaci & Mostafa Al-Emran To cite this article: Muhammad Alshurideh , Barween Al Kurdi , Said A. Salloum , Ibrahim Arpaci & Mostafa Al-Emran (2020): Predicting the actual use of m-learning systems: a comparative approach using PLS-SEM and machine learning algorithms, Interactive Learning Environments, DOI: 10.1080/10494820.2020.1826982 To link to this article: https://doi.org/10.1080/10494820.2020.1826982 Published online: 01 Oct 2020. Submit your article to this journal View related articles View Crossmark data Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=nile20 INTERACTIVE LEARNING ENVIRONMENTS https://doi.org/10.1080/10494820.2020.1826982 Predicting the actual use of m-learning systems: a comparative approach using PLS-SEM and machine learning algorithms a,b Muhammad Alshurideh , Barween Al Kurdic, Said A. Salloumd, Ibrahim Arpaci e and Mostafa Al-Emran f a Department of Management, College of Business Administration, University of Sharjah, Sharjah, UAE; bMarketing Department, School of Business, The University of Jordan, Amman, Jordan; cMarketing Department, Amman Arab University, Amman, Jordan; dResearch Institute of Sciences & Engineering, University of Sharjah, Sharjah, UAE; e Department of Computer Education and Instructional Technology, Tokat Gaziosmanpasa University, Tokat, Turkey; f Applied Computational Civil and Structural Engineering Research Group, Faculty of Civil Engineering, Ton Duc Thang University, Ho Chi Minh City, Vietnam ABSTRACT ARTICLE HISTORY Despite the plethora of m-learning acceptance studies, few have tackled Received 20 December 2019 the importance of examining the actual use of m-learning systems from Accepted 18 September the lenses of social inﬂuence, expectation-conﬁrmation, and satisfaction. 2020 Additionally, most of the prior technology adoption literature tends to KEYWORDS use the structural equation modeling (SEM) technique in analyzing the Mobile learning; higher structural models. To address these limitations, this study extends the education; TAM; ECM; PLS- technology acceptance model (TAM) with the expectation-conﬁrmation SEM; machine learning model (ECM) and social inﬂuence to predict the actual use of m- algorithms learning systems. A comparative approach using the partial least squares-structural equation modeling (PLS-SEM) and machine learning algorithms was employed to test the proposed model with data collected from 448 students. The results revealed that both techniques have successfully provided support to all the hypothesized relationships of the research model. More interesting, the J48 classiﬁer has performed better than the other classiﬁers in predicting the dependent variable in most cases. The employment of a comparative analytical approach is believed to add a signiﬁcant contribution to the information systems (IS) literature in general, and the m-learning domain in speciﬁc. 1. Introduction Mobile learning (m-learning) is deﬁned as “learning that occurs when learners have access to infor- mation anytime and anywhere via mobile technologies to perform authentic activities in the context of their learning” (Martin & Ertzberger, 2013). M-learning oﬀers an exclusive opportunity to leverage learners’ experiences in formal and informal learning (Joo et al., 2016). The portability and ﬂexibility of mobile computing devices allow learners to situate their learning in an informative way, experi- ence learning with real-world problems, and personalize their learning (Traxler, 2009). Since the appearance of the m-learning concept until this moment, information system (IS) and educational scholars have examined how to incorporate m-learning in instructional practices. The continuous insistence of those scholars stems from the fact that m-learning systems enable students to access their learning materials at “anytime anywhere” settings using wireless networks (Al-Emran et al., 2019; Sarrab et al., 2018). Despite this enthusiasm, the investment in m-learning systems requires the understanding of students’ lack of motivation to use these systems for instructional CONTACT Mostafa Al-Emran [email protected] © 2020 Informa UK Limited, trading as Taylor & Francis Group 2 M. ALSHURIDEH ET AL. activities (Tan et al., 2014). The existence of m-learning platform does not guarantee that learners would use it for educational practices; students need to be aware of its advantages and adopt it in their academic lifestyle (Nguyen et al., 2015). While several studies were conducted in the past to understand the determinants inﬂuencing the adoption of m-learning (Al-Emran et al., 2018a), various issues still need to be discussed. First, the employment of m-learning in the higher education context is still in its early stages, and its theor- etical basis is not yet matured (Kumar & Chand, 2019). Second, the students’ use of m-learning does not always meet the expectations of several educational environments (Aburub & Alnawas, 2019). This stems from the inﬂuential factors which mainly rely on the variation in the context, infra- structure, and students’ readiness. Besides, there are several technical and non-technical issues regarding the use of mobile computing devices in classrooms (Adegbija & Bola, 2015). Third, there is a little debate concerning the factors inﬂuencing the continued use of m-learning systems (Al-Emran et al., 2020). The identiﬁcation of these factors would facilitate the understanding of students’ actual use of m-learning. Fourth, by analyzing the studies published on m-learning systems in a recent systematic review (Al-Emran et al., 2018a), it has been noticed that there is a limited number of studies concerning the continued use of these systems. It has been also observed that there is a knowledge gap in explaining the impact of social inﬂuence on perceived ease of use and perceived usefulness and how this, in turn, would aﬀect the continued use of m-learning systems. Fifth, despite the fact that most of the relevant m-learning studies have employed structural equation modeling (SEM) techniques in explaining the causal-eﬀect among the theoretical model constructs, there is a scarce of knowledge regarding the use of other analytical techniques, such as machine learning algorithms (Arpaci, 2019) and neural networks (Al-Shihi et al., 2018). To handle the aforementioned issues, this research aims to investigate the determinants inﬂuen- cing the students’ continued use of m-learning systems. More speciﬁcally, the role of social inﬂuence would be examined in relation to the perceived usefulness and perceived ease of use of m-learning. In doing so, this study develops a theoretical model based on the integration of the Technology Acceptance Model (TAM) (Davis, 1989) and Expectation-Conﬁrmation Model (ECM) (Bhattacherjee, 2001). To validate the developed model, this study employs a comparative analytical approach using the partial least squares-structural equation modeling (PLS-SEM) alongside machine learning algorithms to develop a high-performance predictive model. PLS-SEM is a widespread multivariate analysis technique that is used to empirically test theoretical models. On the other hand, the use of machine learning algorithms and data mining has become an innovative technique for developing predictive models that explain, predict, and describe the social/human behavior (Osoba & Davis, 2019). This study used the PLS-SEM approach to identify the causal relationships between the endogenous (independent) and exogenous (dependent) variables. Further, the study employed the classiﬁer model to predict dependent variables based on the independent variables. Although the presentation of the two analyses (PLS-SEM vs. machine learning algorithms) seems to produce two separate ﬁndings regarding the operationalization of the same variables in both techniques, it is in fact, considered a hybrid modeling approach. 2. Research model and hypotheses development In line with the analysis of m-learning studies provided in a recent systematic review (Al-Emran et al., 2018a), it has been observed that there is little empirical research on students’ continuous intention towards m-learning systems with a little explanation of how social inﬂuence would aﬀect the per- ceived usefulness (PU) and perceived ease of use (PEOU) of these systems. Therefore, this study develops a theoretical model to examine the role of social inﬂuence grounded on the integration of TAM and ECM. Those two theories are among the most cited theories in the contexts of technol- ogy acceptance and post-adoption behavior. The rationale behind the integration of those two the- ories stems from the reason that TAM involves the pre-adoption expectation, while ECM accounts for the post-adoption expectation (Bhattacherjee, 2001; Davis, 1989). As such, the incorporation of both INTERACTIVE LEARNING ENVIRONMENTS 3 theories would better explain the students’ continued use of m-learning and how that would inﬂuence their actual use. While the integration of those two theories has been well-supported in the previous literature (Al-Emran et al., 2020; Joo et al., 2016; Luo et al., 2017), it is believed that adding the social inﬂuence factor would have a signiﬁcant value in explaining the interrelationships among the constructs in those theories. It is suggested that social inﬂuence would have signiﬁcant eﬀects on both PEOU and PU of m-learn- ing systems. Besides, it is anticipated that the expectation-conﬁrmation would have signiﬁcant eﬀects on PU and satisfaction. Further, it is proposed that the continuous intention to use m-learning systems would be aﬀected by PEOU, PU, and satisfaction. It is also assumed that continuous intention would aﬀect the actual use of m-learning systems. Figure 1 depicts the proposed theoretical model. 2.1. Social inﬂuence Social inﬂuence (SI) refers to “the degree to which an individual perceives that important others believe he or she should use the new system” (Venkatesh et al., 2003). Through the use of several technologies, prior research showed that social inﬂuence has a signiﬁcant eﬀect on the PEOU (Wamba & Queiroz, 2019; Zhang et al., 2020) and PU (Hassan et al., 2020; Nikou & Economides, 2017; Vanduhe et al., 2020; Wamba & Queiroz, 2019; Zhang et al., 2020). By extending this relation- ship to the context of m-learning, we hypothesize the following: H1: Social inﬂuence would predict the PEOU of m-learning systems. H2: Social inﬂuence would predict the PU of m-learning systems. 2.2. Expectation-conﬁrmation Expectation-conﬁrmation refers to “users’ perceptions of the congruence between the expectation of information system usage and its actual performance” (Bhattacherjee, 2001). Previous research suggested that expectation-conﬁrmation has a signiﬁcant impact on satisfaction and PU of several mobile technologies (Al-Emran et al., 2020; Le et al., 2020; Nascimento et al., 2018). Hence, we suggest the following: H3: Expectation-conﬁrmation would predict the PU of m-learning systems. Figure 1. Research model. 4 M. ALSHURIDEH ET AL. H4: Expectation-conﬁrmation would predict the satisfaction of m-learning systems. 2.3. Perceived ease of use Perceived ease of use (PEOU) is deﬁned as “the degree to which a person believes that using a par- ticular system would be free of eﬀort” (Davis, 1989). Previous studies revealed that PEOU has a sig- niﬁcant impact on the continuous intention to use m-learning systems (Al-Emran et al., 2020; Joo et al., 2016). Thus, we propose the following: H5: PEOU would predict the continuous intention of using m-learning systems. 2.4. Perceived usefulness Perceived usefulness (PU) is deﬁned as “the degree to which a person believes that using a particular system would enhance his or her job performance” (Davis, 1989). Earlier studies pointed out that PU has a signiﬁcant eﬀect on the continuous intention to use various mobile technologies (Joo et al., 2016; Le et al., 2020; Nascimento et al., 2018). Consequently, we suggest the following: H6: PU would predict the continuous intention of using m-learning systems. 2.5. Satisfaction Satisfaction is deﬁned as “the aﬀective attitude towards a particular computer application by an end user who interacts with the application directly” (Doll et al., 1998). Prior research found that satisfac- tion has a signiﬁcant eﬀect on the continuous intention to use diﬀerent mobile technologies (Le et al., 2020; Nascimento et al., 2018; Tam et al., 2020). Accordingly, we hypothesize the following: H7: Satisfaction would predict the continuous intention of using m-learning systems. 2.6. Continuous intention Continuous intention (CI) is deﬁned as “users’ intention to continue using the information system” (Bhattacherjee, 2001). Earlier studies showed that CI has a signiﬁcant inﬂuence on the actual use of m-learning (Al-Emran et al., 2020; Joo et al., 2016). Thus, this leads to the following: H8: Continuous intention would predict the actual use of m-learning systems. 3. Research methodology 3.1. Context and subjects The participants of this study are the students enrolled at the British University in Dubai and the Uni- versity of Fujairah, in which both located in the United Arab Emirates (UAE). In both universities, the students used the m-learning system to register their courses, receive notiﬁcations, and communi- cate with their colleagues and instructors through the learning management system (LMS). The stu- dents are also able to access their lectures, download and upload their assignments, and check the similarity score of their work using the similarity checking system in each university. The data were collected between October and November 2018 through the use of self-administrated surveys. The participants volunteered to ﬁll out the surveys without getting any compensation for their involve- ment. This research uses the convenience sampling approach in gathering the data. While this approach is commonly used in quantitative studies (Etikan et al., 2016), it also raises some issues con- cerning the subjective nature in selecting the respondents and data bias. To handle these issues, we have employed the common method bias (CMB) test (Podsakoﬀ et al., 2003). Out of 500 surveys INTERACTIVE LEARNING ENVIRONMENTS 5 distributed, a total of 448 students have successfully ﬁlled out the entire survey with a response rate of 89.6%. Of those, there were 230 females and 218 males. 56% of the participants were aged between 18 and 29 years old. Further, 56% of the participants were bachelor’s degree students, fol- lowed by master’s degree students (20%), PhD students (19%), and diploma students (5%). 3.2. Research instrument The research instrument of this study consists of two parts. The ﬁrst part is devoted to collect the demographic data of the participants, whereas the second one is designed to collect responses regarding the factors in the conceptual model. The indicators in the second part were measured using a “5-point Likert scale”. The indicators used to evaluate the PEOU and PU were adopted from Davis (1989). The items used to evaluate expectation conﬁrmation, satisfaction, and continuous intention were adopted from Bhattacherjee (2001). The social inﬂuence items were adopted from Venkatesh et al. (2003), while the indicators of actual use were adopted from Mohammadi (2015). The constructs and their underlying items are listed in the Appendix. 3.3. Data analysis This study employs two diﬀerent techniques for evaluating the developed research model. Concern- ing the ﬁrst approach, this research uses the partial least squares-structural equation modeling (PLS- SEM) via the SmartPLS software (Ringle et al., 2015). The main reason behind the employment of PLS- SEM in this research stems from the fact that PLS-SEM provides concurrent analysis for both measurement and structural model, which in turn, leads to more precise results (Barclay et al., 1995). In terms of the second technique, this study uses the machine learning algorithms through Weka to predict the dependent constructs in the theoretical model (Arpaci, 2019). 3.4. Common method bias To ensure that the collected data do not have CMB, the Harman’s single-factor has been carried out with seven factors (social inﬂuence, expectation-conﬁrmation, PEOU, PU, satisfaction, continuous intention, and actual use) (Podsakoﬀ et al., 2003). The seven factors were then loaded into a single factor. The analysis shows that the largest variance explained by the newly created factor is 23.21%, which is below the threshold value of 50% (Podsakoﬀ et al., 2003). Hence, there were no concerns regarding the CMB in the collected data. 4. Results 4.1. Measurement model assessment The measurement model is assessed through testing the reliability and validity (Hair et al., 2016). For reliability testing, the Cronbach’s alpha and composite reliability (CR) measures were used. The rec- ommended values for each of these measures should be ≥ 0.70 (Hair et al., 2016). According to the results in Table 1, the values of both measures are considered satisfactory, and thus, the reliability is conﬁrmed. Concerning validity testing, Hair et al. (2016) suggested evaluating the convergent and discrimi- nant validities. For convergent validity, the average variance extracted (AVE) and factor loadings were tested. The values of AVE should be ≥ 0.50 (Fornell & Larcker, 1981), whereas the values of factor loadings should be ≥ 0.70 (Hair et al., 2010). As per the results in Table 1, the values of both measures are accepted, and therefore, the convergent validity is ascertained. For discriminant validity, Henseler et al. (2015) suggested testing the “Heterotrait-Monotrait ratio (HTMT)” of 6 M. ALSHURIDEH ET AL. Table 1. Reliability and convergent validity results. Constructs Items Factor Loadings Cronbach’s Alpha CR AVE Actual Use AU1 0.928 0.852 0.856 0.931 AU2 0.938 Continuous Intention CI1 0.889 0.849 0.877 0.906 CI2 0.871 CI3 0.858 Expectation-Conﬁrmation EXP1 0.890 0.891 0.971 0.922 EXP2 0.933 EXP3 0.952 Perceived Ease of Use PEOU1 0.909 0.831 0.848 0.899 PEOU2 0.799 PEOU3 0.883 Perceived Usefulness PU1 0.796 0.731 0.758 0.850 PU2 0.906 PU3 0.814 Satisfaction SAT1 0.857 0.787 0.789 0.877 SAT2 0.905 SAT3 0.750 Social Inﬂuence SI1 0.904 0.837 0.849 0.890 SI2 0.826 SI3 0.798 SI4 0.742 correlations. The values of HTMT should be < 0.85. As per the readings in Table 2, all the values are accepted, and hence, the discriminant validity is established. 4.2. Hypotheses testing using PLS-SEM Table 3 describes the beta (β) values, t-values, and p-values for each of the developed hypotheses based on the generated results through the PLS-SEM technique. The ﬁrst hypothesis describes the relationship between social inﬂuence and PEOU (β = 0.460, t = 11.079). The result of this hypothesis indicates that social inﬂuence has a signiﬁcant positive impact on the PEOU of m- learning systems. Thus, H1 is supported. The second hypothesis shows the correlation between social inﬂuence and PU (β = 0.175, t = 2.274). The result of this hypothesis reveals that social inﬂuence has a signiﬁcant positive eﬀect on the PU of m-learning systems. Therefore, H2 is supported. The third hypothesis demonstrates the relationship between expectation-conﬁrmation and PU (β = 0.578, t = 14.434). The result of this hypothesis suggests that expectation-conﬁrmation has a sig- niﬁcant positive inﬂuence on the PU of m-learning systems. Hence, H3 is supported. The fourth hypothesis characterizes the correlation between expectation-conﬁrmation and satisfaction (β = 0.458, t = 9.952). The result of this hypothesis indicates that expectation-conﬁrmation has a signiﬁ- cant positive inﬂuence on satisfaction. Thus, H4 is supported. The ﬁfth hypothesis exhibits the relationship between PEOU and continuous intention (β = 0.363, t = 4.552). The result of this hypothesis suggests that PEOU has a signiﬁcant positive inﬂuence on the Table 2. HTMT results. AU CI EXP PEOU PU SAT SI AU CI 0.476 EXP 0.576 0.566 PEOU 0.459 0.426 0.795 PU 0.646 0.751 0.768 0.662 SAT 0.750 0.514 0.561 0.551 0.454 SI 0.410 0.481 0.589 0.537 0.440 0.394 INTERACTIVE LEARNING ENVIRONMENTS 7 Table 3. Hypotheses testing results through PLS-SEM. H Relationship Beta t-value p-value Remarks H1 Social Inﬂuence → Perceived Ease of Use 0.460 11.079 0.000 Supported H2 Social Inﬂuence → Perceived Usefulness 0.175 2.274 0.013 Supported H3 Expectation-Conﬁrmation → Perceived Usefulness 0.578 14.434 0.000 Supported H4 Expectation-Conﬁrmation → Satisfaction 0.458 9.952 0.000 Supported H5 Perceived Ease of Use → Continuous Intention 0.363 4.552 0.001 Supported H6 Perceived Usefulness → Continuous Intention 0.477 8.598 0.000 Supported H7 Satisfaction → Continuous Intention 0.396 6.663 0.000 Supported H8 Continuous Intention → Actual Use 0.852 58.592 0.000 Supported continuous intention to use m-learning systems. Therefore, H5 is supported. The sixth hypothesis describes the correlation between PU and continuous intention (β = 0.477, t = 8.598). The result of this hypothesis reveals that PU has a signiﬁcant positive eﬀect on the continuous intention to use m-learning systems. Hence, H6 is supported. The seventh hypothesis describes the relationship between satisfaction and continuous intention (β = 0.396, t = 6.663). The result of this hypothesis suggests that satisfaction has a signiﬁcant positive inﬂuence on the continuous intention to use m-learning systems. Thus, H7 is supported. The eighth hypothesis shows the correlation between continuous intention and actual use (β = 0.852, t = 58.592). The result of this hypothesis indicates that the continuous intention to use m-learning systems has a signiﬁcant positive impact on its actual usage. Therefore, H8 is supported. Concerning the coeﬃcient of determination (R2) results in Figure 2, it can be observed that social inﬂuence explains 43.7% of the variance in PEOU. Moreover, social inﬂuence and expectation-conﬁr- mation together explain 49% of the variance in PU. Besides, expectation-conﬁrmation explains 44.6% of the variance in satisfaction. Further, PEOU, PU, and satisfaction together explain 39.4% of the var- iance in continuous intention. Additionally, the continuous intention explains 32.7% of the variance in actual use. In comparison with the proposed R2 values (Chin, 1998), the values observed in this study are regarded to be acceptable. 4.3. Hypotheses testing using machine learning algorithms This research employs machine learning classiﬁcation algorithms by applying a wide range of meth- odologies, including Bayesian networks, decision trees, if-then-else rules, and neural networks, to Figure 2. PLS algorithm results. 8 M. ALSHURIDEH ET AL. Table 4. Predicting the PEOU by social inﬂuence. Classiﬁer CCI1 (%) TP2 Rate FP3 Rate Precision Recall F-Measure BayesNet 72.79.728.523.683.728.674 Logistic 71.12.711.610.690.711.585 LWL 72.79.728.523.683.728.674 AdaBoostM1 72.79.728.523.683.728.674 OneR 71.84.718.540.606.718.650 J48 73.03.730.538.698.730.665 1 CCI: Correctly Classiﬁed Instances. 2TP: True Positive, 3FP: False Positive. predict the relationships in the proposed theoretical model (Arpaci, 2019, 2020). Weka (ver. 3.8.3) was used to test the predictive model based on several classiﬁers, including Bayesian classiﬁer (BayesNet), meta classiﬁer (AdaBoostM1), lazy classiﬁer (LWL), logistic regression classiﬁer (Logistic), decision tree (J48), and rule learner (OneR) (Frank et al., 2009). Based on the results in Table 4, it can be observed that J48 performs better than the other classiﬁers in predicting the PEOU of m-learning systems. The J48 predicted the PEOU with an accuracy of 73.03% for the 10-fold cross-validation. Hence, H1 is supported. This classiﬁer had a better performance in terms of the TP rate (.730), pre- cision (.698), and recall (.730) as compared to the other classiﬁers. The results also suggested that the J48 had a better classiﬁcation performance than the other classiﬁers in predicting the PU, as shown in Table 5. The J48 predicted the PU by the attributes of social inﬂuence and expectation-conﬁrmation with an accuracy of 78.29%, and thereby, H2 and H3 were both supported. As demonstrated in Table 6, the results showed that both OneR and J48 classiﬁers performed better than the other classiﬁers in predicting the satisfaction by expectation-conﬁrmation. The OneR and J48 classiﬁers predicted the satisfaction with an accuracy of 63.72%. Accordingly, H4 is supported. The results presented in Table 7 indicated that the J48 performed better than the other classiﬁers in predicting the continuous intention by the attributes of PEOU, PU, and satisfaction. The J48 pre- dicted the continuous intention with an accuracy of 88.31%. Therefore, H5, H6, and H7 were supported. The results also indicated that Logistic performed better than the other classiﬁers in predicting the actual use by the continuous intention, as shown in Table 8. The Logistic classiﬁer predicted the actual use with an accuracy of 78.99%. Accordingly, H8 is supported. 5. Discussion 5.1. Research hypotheses ﬁndings The present study proposed a conceptual model to examine the students’ actual use of m-learning. To validate the proposed model, this research employed a comparative analytical approach, includ- ing PLS-SEM and machine learning algorithms. In terms of predicting the PEOU, the PLS-SEM results suggested that social inﬂuence has a signiﬁ- cant impact on the PEOU of m-learning systems. Further, the results of machine learning indicated that the J48 classiﬁer showed better performance than the other classiﬁers in predicting the PEOU of m-learning systems with an accuracy of 73.03%. These ﬁndings seem to be consistent with those found in the previous literature (Wamba & Queiroz, 2019; Zhang et al., 2020). This indicates that the positive opinions and reports of other students concerning the use of m-learning systems would signiﬁcantly aﬀect the PEOU of these systems. With respect to predicting the PU, the PLS- SEM results showed that the PU of m-learning systems was signiﬁcantly and positively aﬀected by social inﬂuence and expectation-conﬁrmation together. Besides, the results of machine learning indi- cated that the J48 showed better performance by predicting the PU with an accuracy of 78.29%. These results also accord with the observations reported in previous literature, which suggested INTERACTIVE LEARNING ENVIRONMENTS 9 Table 5. Predicting the PU by social inﬂuence and expectation-conﬁrmation. Classiﬁer CCI (%) TP Rate FP Rate Precision Recall F-Measure BayesNet 69.45.695.247.715.695.683 Logistic 68.74.687.223.713.687.680 LWL 68.26.683.265.737.683.667 AdaBoostM1 59.19.592.321.599.592.665 OneR 67.78.678.263.706.678.663 J48 78.29.783.160.791.783.780 that the PU is positively aﬀected by social inﬂuence (Hassan et al., 2020; Nikou & Economides, 2017; Vanduhe et al., 2020; Wamba & Queiroz, 2019; Zhang et al., 2020) and expectation-conﬁrmation (Al- Emran et al., 2020; Le et al., 2020; Nascimento et al., 2018). This means that the usefulness of using m- learning systems is increased when the students perceive that these systems are socially accepted by their colleagues. The usefulness of m-learning systems can also be increased if the students have positive expectations that these systems would enhance their performance. Overall, the reason why social inﬂuence is being such a key driver in aﬀecting the PU and PEOU of m-learning systems is that the study was conducted in the UAE, which characterized by collectivistic culture. In collectivistic cultures, the perceptions of individuals are more likely to be aﬀected by others’ decisions and opinions (Zhou & Li, 2014). Hence, social inﬂuence is expected to have a more substantial impact on m-learning acceptance in the UAE culture compared with other cultures. Concerning the prediction of satisfaction, the PLS-SEM ﬁndings pointed out that expectation- conﬁrmation has a positive impact on satisfaction. The results of machine learning also indicated that both OneR and J48 classiﬁers performed better than the other classiﬁers in predicting the sat- isfaction by expectation-conﬁrmation. These ﬁndings also correspond to those observed in prior research (Al-Emran et al., 2020; Le et al., 2020; Nascimento et al., 2018). It can, therefore, be suggested that the students would be satisﬁed with using the m-learning systems if they impose positive expectations that these systems would enhance their learning performance. With regard to predicting the continuous intention, the PLS-SEM results revealed that the con- tinuous intention to use m-learning systems was signiﬁcantly aﬀected by PEOU, PU, and satisfaction. Moreover, the results of machine learning revealed that the J48 showed better performance by pre- dicting the continuous intention with an accuracy of 88.31%. Similarly, these ﬁndings match those noticed in earlier investigations, which reported that continuous intention is positively aﬀected by PEOU (Al-Emran et al., 2020; Joo et al., 2016), PU (Joo et al., 2016; Le et al., 2020; Nascimento et al., 2018), and satisfaction (Le et al., 2020; Nascimento et al., 2018; Tam et al., 2020). These results suggest that students’ intentions will continue to use m-learning systems if these systems keep satisfying them, speciﬁcally when the systems impose free eﬀorts and enhance their learning performance. With respect to predicting the actual use, the PLS-SEM results pointed out that the actual use of m-learning systems was signiﬁcantly and positively inﬂuenced by its continued use. Further, the results of machine learning indicated that the Logistic classiﬁer showed better performance by pre- dicting the actual use with an accuracy of 78.99%. These results were also in agreement with those reported in the previous m-learning literature (Al-Emran et al., 2020; Joo et al., 2016). This indicates that the continued use of m-learning was directly and positively associated with the students’ actual Table 6. Predicting satisfaction by expectation-conﬁrmation. Classiﬁer CCI (%) TP Rate FP Rate Precision Recall F-Measure BayesNet 62.77.628.341.654.628.609 Logistic 59.19.592.381.605.592.566 LWL 61.34.613.344.624.613.597 AdaBoostM1 55.85.558.422.510.558.496 OneR 63.72.637.341.687.637.618 J48 63.72.637.341.687.637.618 10 M. ALSHURIDEH ET AL. Table 7. Predicting the continuous intention by PEOU, PU, and satisfaction. Classiﬁer CCI (%) TP Rate FP Rate Precision Recall F-Measure BayesNet 69.69.697.241.703.697.697 Logistic 69.69.697.235.705.697.698 LWL 73.03.730.218.737.730.730 AdaBoostM1 65.63.656.368.759.656.778 OneR 71.84.718.204.731.718.722 J48 88.31.883.071.891.883.884 Table 8. Predicting the actual use by continuous intention. Classiﬁer CCI (%) TP Rate FP Rate Precision Recall F-Measure BayesNet 76.85.768.246.776.768.750 Logistic 78.99.790.253.787.790.778 LWL 72.55.726.336.716.726.797 AdaBoostM1 72.55.726.336.716.726.797 OneR 77.80.778.276.774.778.760 J48 77.80.778.280.775.778.758 use. This means that students who used the m-learning system with a positive or negative experi- ence did not change their usage intentions after their initial trial. In this sense, the students either continued to use the m-learning system in their instructional activities, or they did not use it consistently. 5.2. Research implications We have validated the proposed model through a comparative method using machine learning algorithms and PLS-SEM. While the use of data mining techniques in analyzing theoretical models has been well-perceived in the previous literature (Berrado et al., 2013; Verbraken et al., 2014; Wu, 2011), the employment of a comparative analytical approach (i.e. machine learning and PLS-SEM) is believed to add a signiﬁcant contribution to the IS literature in general, and the m-learning domain in speciﬁc. It is important to report that PLS-SEM can be used for examining the dependent construct and validating the conceptual model based on the extension of an existing theory (Al- Emran et al., 2018b). Likewise, supervised machine learning algorithms (i.e. having a pre-deﬁned dependent construct) can be used for predicting the dependent variable based on independent vari- ables (Arpaci, 2019). It is also essential to indicate that this study has employed several classiﬁcation algorithms with diﬀerent methodologies, including association rules, decision trees, if-then-else rules, Bayesian networks, and neural networks. More speciﬁcally, the ﬁndings suggested that J48 (a decision tree) performed better than other classiﬁers in most cases. It is worth mentioning that the decision tree (nonparametric) was used for classifying both continuous (numerical) and categori- cal variables by splitting the sample into homogeneous sub-samples based on the most signiﬁcant independent variable (Arpaci, 2020). On the other hand, PLS-SEM (a nonparametric procedure) was used to test the signiﬁcant coeﬃcients with replacements from the sample to draw a large number of sub-samples randomly. Although the presentation of the two analyses (PLS-SEM vs. machine learning algorithms) seems to produce two separate ﬁndings regarding the operationalization of the same variables in both techniques, it is in fact, considered a hybrid modeling approach. 5.3. Theoretical contributions The theoretical contributions of this research can be summarized in three points. First, this study explored the role of social inﬂuence grounded on the combination of TAM and ECM to examine the actual use of m-learning. It is believed that the proposed model would add value to the ECM INTERACTIVE LEARNING ENVIRONMENTS 11 and TAM at one hand, and the m-learning domain on the other hand. Second, most of the earlier studies in m-learning have investigated the determinants aﬀecting either the continuous intention or actual use. To the best of the authors’ knowledge, this is one of the few attempts that examined the factors aﬀecting the continuous intention and how that would inﬂuence the students’ actual use of m-learning. Third, PU and PEOU have been well-perceived in the previous m-learning literature to be the key drivers of behavioral intention. With the introduction of the proposed model, this research shows that other essential factors need to be considered, whether directly or indirectly, when approaching the continued use of m-learning, such as social inﬂuence, expectation-conﬁr- mation, and satisfaction. 5.4. Practical implications The evidence from this research imposes various practical implications. First, the ﬁndings indicated that PU (β = 0.477) is the most crucial factor which aﬀects the continuous intention to use m-learn- ing. Hence, educators and practitioners are required to ensure that the incorporated content materials in m-learning systems are useful and would enhance the learning performance of students in order to sustain their continued use. While taking other factors into consideration, the decision- makers should focus more on the usefulness when purchasing new m-learning systems or upgrading the existing ones. Second, the ﬁndings revealed that social inﬂuence has a signiﬁcant impact on increasing the ease of use and usefulness of m-learning systems. While this might be due to the characteristics of the studied sample, researchers might explore the impact of social inﬂuence in other non-collectivistic cultural environments. Third, since all the hypotheses were supported, m- learning developers should consider the importance of the studied factors in designing more colla- borative m-learning systems in order to sustain the continued use of these systems. M-learning systems should not require much time and eﬀort from their students to get through the learning material (Amoroso & Lim, 2017). Therefore, m-learning developers should emphasize on designing user-friendly systems with a better reading layout and richer presentation. 6. Conclusion While several studies were conducted in the past to examine the factors aﬀecting the m-learning acceptance, few have studied what impacts the actual use of m-learning from the perspective of social inﬂuence, expectation-conﬁrmation, and satisfaction. At the same time, most of the prior studies have mainly relied on the SEM approach concerning the analysis of structural models. There- fore, this study aimed to address these limitations by extending the TAM with ECM and social inﬂuence to predict the actual use of m-learning systems. To validate the proposed model, this research employed a comparative approach using PLS-SEM and machine learning algorithms. The ﬁndings indicated that both techniques provided support to all the suggested hypotheses. More interesting, the J48 classiﬁer has performed better than the other classiﬁers in predicting the depen- dent variable in most cases. It is, therefore, believed that the use of a comparative analytical approach would add a valuable contribution to the IS literature in general, and the m-learning domain in particular. A number of important limitations need to be reported. First, the generalization of the results to the other higher educational institutions in the UAE or other countries should be treated with caution. This stems from two reasons: (a) the concentration on only two institutions for picking up the samples, and (b) the employment of a convenience sampling technique for selecting the par- ticipants. Further research needs to consider these issues to enhance the opportunity of results gen- eralization. Second, the study focused on examining the actual use of m-learning systems by students only. Future attempts are highly encouraged to measure the instructors’ actual use of m-learning systems in order to get further insights into the aﬀecting determinants and draw a com- prehensive picture of the implementation of these systems. 12 M. ALSHURIDEH ET AL. Acknowledgment This is an extended version of a conference paper published by the International Conference on Advanced Intelligent Systems and Informatics 2019. Disclosure statement No potential conﬂict of interest was reported by the author(s). Notes on contributors Muhammad Alshurideh is working for the College of Business Administration – University of Sharjah in UAE as a full- time faculty member and for the School of Business – University of Jordan in Jordan. Regarding the teaching, he has the responsibilities to teach a wide range of marketing and business topics for both undergraduate and postgraduate stu- dents. He has more than 40 published papers in diﬀerent marketing and business topics mainly CRM and Customer Retention. He has published in good ranked journals such as Journal of Marketing Communications and International Journal of Electronic Customer Relationship Management. Barween Al Kurdi is an Assistant Professor in Marketing and she is working for Amman Arab University – Faculty of Business – Marketing Department. She is a member of large number of committees and mainly the social committee. She used to publish in good ranked journals such as Journal of Marketing Communications and International Journal of Marketing Studies. Said A. Salloum had graduated from The British University in Dubai with a distinction with MSc in Informatics (Knowl- edge and Data Management). He got his bachelor’s degree in Computer Science from Yarmouk University. Currently, he is working at the University of Sharjah “Research Institute of Sciences and Engineering (RISE)” as a researcher on diﬀerent research areas in Computer Science such as data analysis, machine learning, knowledge management, and Arabic Language Processing. Salloum is an Oracle expert since 2013 along with various recognized international certiﬁ- cates that are issued by Oracle. Ibrahim Arpaci is an Associate Professor and Chair of the Department of Computer Education and Instructional Tech- nology at Gaziosmanpasa University, where he is also Director of the Distance Education Application and Research Center. He was a visiting scholar at Ryerson University, Ted Rogers School of Information Technology Management, Toronto, ON, Canada (2012-2013). He holds a BSc in Computer Education and Instructional Technology (2005) from Anadolu University, an MSc in Information Systems (2009) and a PhD in Information Systems (2013) both from Middle East Technical University. His research interests are in instructional systems design and technology, cyberpsy- chology and behavior, culture, learning and technology. Mostafa Al-Emran obtained his Ph.D. degree in Computer Science from Universiti Malaysia Pahang. He received the MSc degree in Informatics from The British University in Dubai with a distinction level along with the top Academic Excel- lence Award, and the BSc degree with honors in Computer Science from Al Buraimi University College. He has published over 70 research articles in highly reputed journals such as Computers & Education, Computers in Human Behavior, Inter- national Journal of Information Management, Telematics and Informatics, IEEE Access, Technology in Society, Interactive Learning Environments, Journal of Educational Computing Research, and International Journal of Engineering Education, among many others. Most of his publications were indexed under the ISI Web of Science and Scopus. He has also edited a number of books published by Springer. He is a certiﬁed recognized reviewer by several leading journals in Elsevier. His current research interests include mobile learning, knowledge management, technology acceptance, and wearable technology. ORCID Muhammad Alshurideh http://orcid.org/0000-0002-7336-381X Ibrahim Arpaci http://orcid.org/0000-0001-6513-4569 Mostafa Al-Emran http://orcid.org/0000-0002-5269-5380 References Aburub, F., & Alnawas, I. (2019). A new integrated model to explore factors that inﬂuence adoption of mobile learning in higher education: An empirical investigation. Education and Information Technologies, 24(3), 2145–2158. https://doi. org/10.1007/s10639-019-09862-x INTERACTIVE LEARNING ENVIRONMENTS 13 Adegbija, M. V., & Bola, O. O. (2015). Perception of undergraduates on the adoption of mobile technologies for learning in selected universities in Kwara state, Nigeria. Procedia - Social and Behavioral Sciences, 176, 352–356. https://doi.org/ 10.1016/j.sbspro.2015.01.482 Al-Emran, M., Arpaci, I., & Salloum, S. A. (2020). An empirical examination of continuous intention to use m-learning: An integrated model. Education and Information Technologies, 25, 2899–2918. https://doi.org/10.1007/s10639-019- 10094-2 Al-Emran, M., Mezhuyev, V., & Kamaludin, A. (2018a). Technology acceptance model in M-learning context: A systematic review. Computers & Education, 125, 389–412. https://doi.org/10.1016/j.compedu.2018.06.008 Al-Emran, M., Mezhuyev, V., & Kamaludin, A. (2018b, September 3–5). PLS-SEM in information systems research: A com- prehensive methodological reference. 4th International Conference on Advanced Intelligent Systems and Informatics (AISI 2018) (pp. 644–653). Al-Emran, M., Mezhuyev, V., & Kamaludin, A. (2019). An innovative approach of applying knowledge management in M- learning application development: A pilot study. International Journal of Information and Communication Technology Education (IJICTE), 15(4), 94–112. https://doi.org/10.4018/IJICTE.2019100107 Al-Shihi, H., Sharma, S. K., & Sarrab, M. (2018). Neural network approach to predict mobile learning acceptance. Education and Information Technologies, 23(5), 1805–1824. https://doi.org/10.1007/s10639-018-9691-9 Amoroso, D., & Lim, R. (2017). The mediating eﬀects of habit on continuance intention. International Journal of Information Management, 37(6), 693–702. https://doi.org/10.1016/j.ijinfomgt.2017.05.003 Arpaci, I. (2019). A hybrid modeling approach for predicting the educational use of mobile cloud computing services in higher education. Computers in Human Behavior, 90, 181–187. https://doi.org/10.1016/j.chb.2018.09.005 Arpaci, I. (2020). What drives students’ online self-disclosure behavior on social media? A hybrid SEM and artiﬁcial intel- ligence approach. International Journal of Mobile Communications, 18(2), 229–241. https://doi.org/10.1504/IJMC. 2020.10017999 Barclay, D., Higgins, C., & Thompson, R. (1995). The partial least squares (pls) approach to casual modeling: Personal com- puter adoption ans use as an illustration. Berrado, A., Elfahli, S., & El Garah, W. (2013). Using data mining techniques to investigate the factors inﬂuencing mobile payment adoption in Morocco. 2013 8th International Conference on Intelligent Systems: Theories and Applications (SITA) (pp. 1–5). https://doi.org/10.1109/SITA.2013.6560791. Bhattacherjee, A. (2001). Understanding information systems continuance: An expectation-conﬁrmation model. MIS Quarterly, 25, 351–370. https://doi.org/10.2307/3250921 Chin, W. W. (1998). The partial least squares approach to structural equation modeling. Modern Methods for Business Research, 295(2), 295–336. Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. https://doi.org/10.2307/249008 Doll, W. J., Hendrickson, A., & Deng, X. (1998). Using Davis’s perceived usefulness and ease-of-use instruments for decision making: A conﬁrmatory and multigroup invariance analysis. Decision Sciences, 29(4), 839–869. https://doi. org/10.1111/j.1540-5915.1998.tb00879.x Etikan, I., Musa, S. A., & Alkassim, R. S. (2016). Comparison of convenience sampling and purposive sampling. American Journal of Theoretical and Applied Statistics, 5(1), 1–4. https://doi.org/10.11648/j.ajtas.20160501.11 Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. https://doi.org/10.2307/3151312 Frank, E., Hall, M., Holmes, G., Kirkby, R., Pfahringer, B., Witten, I. H., & Trigg, L. (2009). Weka-A machine learning work- bench for data mining. In Data mining and knowledge discovery handbook. https://doi.org/10.1007/978-0-387-09823- 4_66 Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2010). Multivariate data analysis. Prentice Hall. Hair Jr., J. F., Hult, G. T. M., Ringle, C., & Sarstedt, M. (2016). A primer on partial least squares structural equation modeling (PLS-SEM). Sage Publications. Hassan, F., Amin, M. K., Khan, T., Emon, M. M. H., & Amin, A. (2020). Roles of social inﬂuence in expediting online learning acceptance: A preliminary study on Bangladeshi learners. Proceedings of the International Conference on Computing Advancements (pp. 1–6). https://doi.org/10.1145/3377049.3377087 Henseler, J., Ringle, C. M., & Sarstedt, M. (2015). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43(1), 115–135. https://doi.org/10. 1007/s11747-014-0403-8 Joo, Y. J., Kim, N., & Kim, N. H. (2016). Factors predicting online university students’ use of a mobile learning manage- ment system (m-LMS). Educational Technology Research and Development, 64(4), 611–630. https://doi.org/10.1007/ s11423-016-9436-7 Kumar, B. A., & Chand, S. S. (2019). Mobile learning adoption: A systematic review. Education and Information Technologies, 24(1), 471–487. https://doi.org/10.1007/s10639-018-9783-6 Le, T. T., Pham, H. M., Chu, N. H., Nguyen, D. K., & Ngo, H. M. (2020). Factors aﬀecting users’ continuance intention towards mobile banking In Vietnam. American Journal of Multidisciplinary Research & Development (AJMRD), 2(4), 42–51. 14 M. ALSHURIDEH ET AL. Luo, Y., Ye, Q., & Meng, L. (2017). An empirical research on inﬂuence factor of college students’ continued intentions of online self-regulated learning based on the model of ECM and TAM. 2017 3rd International Conference on Social Science and Higher Education (pp. 521–528). Atlantis Press. https://doi.org/10.2991/icsshe-17.2017.132 Martin, F., & Ertzberger, J. (2013). Here and now mobile learning: An experimental study on the use of mobile technol- ogy. Computers & Education, 68, 76–85. https://doi.org/10.1016/j.compedu.2013.04.021 Mohammadi, H. (2015). Social and individual antecedents of m-learning adoption in Iran. Computers in Human Behavior, 49, 191–207. https://doi.org/10.1016/j.chb.2015.03.006 Nascimento, B., Oliveira, T., & Tam, C. (2018). Wearable technology: What explains continuance intention in smart- watches? Journal of Retailing and Consumer Services, 43, 157–169. https://doi.org/10.1016/j.jretconser.2018.03.017 Nguyen, L., Barton, S. M., & Nguyen, L. T. (2015). IPads in higher education - hype and hope. British Journal of Educational Technology, 46(1), 190–203. https://doi.org/10.1111/bjet.12137 Nikou, S. A., & Economides, A. A. (2017). Mobile-based assessment: Investigating the factors that inﬂuence behavioral intention to use. Computers & Education, 109, 56–73. https://doi.org/10.1016/j.compedu.2017.02.005 Osoba, O., & Davis, P. K. (2019). An artiﬁcial intelligence/machine learning perspective on social simulation: New data and new challenges. In P. K. Davis, A. O’Mahony, & J. Pfautz (Eds.), Social-behavioral modeling for complex systems (pp. 443–476). John Wiley & Sons, Inc. Podsakoﬀ, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoﬀ, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903. https:// doi.org/10.1037/0021-9010.88.5.879 Ringle, C. M., Wende, S., & Becker, J. (2015). SmartPLS 3. Bönningstedt: SmartPLS. http://www.smartpls.com Sarrab, M., Al-Shihi, H., Al-Manthari, B., & Bourdoucen, H. (2018). Toward educational requirements model for mobile learning development and adoption in higher education. TechTrends, 62(6), 635–646. https://doi.org/10.1007/ s11528-018-0331-4 Tam, C., Santos, D., & Oliveira, T. (2020). Exploring the inﬂuential factors of continuance intention to use mobile Apps: Extending the expectation conﬁrmation model. Information Systems Frontiers, 22(1), 243–257. https://doi.org/10. 1007/s10796-018-9864-5 Tan, G. W. H., Ooi, K. B., Leong, L. Y., & Lin, B. (2014). Predicting the drivers of behavioral intention to use mobile learning: A hybrid SEM-neural networks approach. Computers in Human Behavior, 36, 198–213. https://doi.org/10.1016/j.chb. 2014.03.052 Traxler. (2009). Current state of mobile learning. In M. Ally (Ed.). Mobile learning transforming the delivery of education and training (pp. 9–24). AU Press. Vanduhe, V. Z., Nat, M., & Hasan, H. F. (2020). Continuance intentions to use gamiﬁcation for training in higher edu- cation: Integrating the technology acceptance model (TAM), social motivation, and task technology Fit (TTF). IEEE Access, 8, 21473–21484. https://doi.org/10.1109/ACCESS.2020.2966179 Venkatesh, V., Morris, M., Davis, G., & Davis, F. (2003). User acceptance of information technology: Toward a uniﬁed view. MIS Quarterly, 27(3), 425–478. https://doi.org/10.2307/30036540 Verbraken, T., Goethals, F., Verbeke, W., & Baesens, B. (2014). Predicting online channel acceptance with social network data. Decision Support Systems, 63, 104–114. https://doi.org/10.1016/j.dss.2013.08.011 Wamba, S. F., & Queiroz, M. M. (2019). The role of social inﬂuence in blockchain adoption: The Brazilian supply chain case. IFAC-PapersOnLine, 52(13), 1715–1720. https://doi.org/10.1016/j.ifacol.2019.11.448 Wu, W. W. (2011). Mining signiﬁcant factors aﬀecting the adoption of SaaS using the rough set approach. Journal of Systems and Software, 84(3), 435–441. https://doi.org/10.1016/j.jss.2010.11.890 Zhang, T., Tao, D., Qu, X., Zhang, X., Zeng, J., Zhu, H., & Zhu, H. (2020). Automated vehicle acceptance in China: Social inﬂuence and initial trust are key determinants. Transportation Research Part C: Emerging Technologies, 112, 220–233. https://doi.org/10.1016/j.trc.2020.01.027 Zhou, T., & Li, H. (2014). Understanding mobile SNS continuance usage in China from the perspectives of social inﬂuence and privacy concern. Computers in Human Behavior, 37, 283–289. https://doi.org/10.1016/j.chb.2014.05.008 INTERACTIVE LEARNING ENVIRONMENTS 15 Appendix: Constructs and items Social inﬂuence SI1: People who inﬂuence my behavior think that I should use the m-learning system. SI2: People who are important to me think that I should use the m-learning system. SI3: The institution management has been helpful in the use of the m-learning system. SI4: In general, the institution has supported the use of the m-learning system. Expectation conﬁrmation EXP1: My experience with using m-learning was better than what I expected. EXP2: The service level provided by m-learning was better than what I expected. EXP3: Overall, most of my expectations from using m-learning were conﬁrmed. Perceived ease of use PEOU1: M-learning is easy to use. PEOU2: Interaction with m-learning is clear and understandable. PEOU3: M-learning is convenient and user-friendly. Perceived usefulness PU1: M-learning enhances my eﬃciency. PU2: M-learning enables me to accomplish tasks more quickly. PU3: M-learning improves my performance. Satisfaction SAT1: I am satisﬁed with using m-learning as a learning assisted tool. SAT2: I am satisﬁed with using m-learning functions. SAT3: I am satisﬁed with multimedia instructions. Continuous intention CI1: I intend to continue using m-learning rather than discontinue its use. CI2: I intend to continue using m-learning than other alternative means. CI3: If I could, I would like to continue my use of m-learning. Actual use AU1: I use the m-learning system on daily basis. AU2: I use the m-learning system frequently.

Predicting Actual Use of mLearning Systems (2020) - AQA

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue