Computers & Education 157 (2020) 103963 Contents lists available at ScienceDirect Computers & Education journal homepage: http://www.elsevier.com/locate/compedu Understanding the determinants of learner engagement in MOOCs: An adaptive structuration perspective Yongqiang Sun a, Yanping Guo a, Yiming Zhao b, * a b School of Information Management, Wuhan University, China Center for Studies of Information Resources, Wuhan University, Wuhan, 430072, Hubei Province, China A R T I C L E I N F O A B S T R A C T Keywords: MOOC Adaptive structuration theory Consensus of appropriation Learner engagement As an innovative educational paradigm, the massive open online course (MOOC) has received extensive attention from both academia and industry. Extant literature has made much effort to investigate the technological supports of MOOCs but how learners understand and master these technologies is still unclear. Moreover, given the collaborative features (e.g. discussion forum and peer assessment) of MOOCs, it is vital to ascertain the mechanism through which group learners reach a consensus on the technology usage and thereby engage in MOOCs. To advance this line of research, the study proposes a theoretical model leveraging adaptive structuration theory. Spe­ cifically, we identify three contextualized factors (i.e., collaborative spirit, task interdependence, and social interaction ties) as the antecedents of consensus of appropriation. Through an online survey recruiting 374 Chinese University MOOC learners, this study demonstrates that collabo­ rative spirit, task interdependence and social interaction ties are positively related to consensus of appropriation, which will facilitate commitment and learner engagement. In addition, commit­ ment is also confirmed to enhance learner engagement. Implications and limitations are further discussed. 1. Introduction The recent years have witnessed a flourish of Massive Open Online Courses (MOOCs) (Almatrafi et al., 2018; Chen & Chen, 2015; Joo et al., 2018). In 2012, many MOOC platforms, such as Coursera, edX and Udacity, were launched, giving rise to “the year of MOOC”. Till the end of 2018, more than 900 universities had created 11,400 MOOCs involving over 101 million learners around the world. As a novel mode leveraging the technology affordances in higher education (Joo et al., 2018; Zhu et al., 2018), the MOOC renders education resources accessible to anyone anywhere and thus raises the potentials for learner engagement. Although MOOCs were held up as an educational innovation (Joo et al., 2018), the problem of high dropout rate also emerged (Hew & Cheung, 2014). As argued by Reich and Ruiperez-Valiente (2019), the low completion rates of MOOCs had never been improved over 6 years. Such issues raised serious concerns (e.g., Hew & Cheung, 2014; Hone & El Said, 2016) and many researchers demon­ strated that participants with a lower level of engagement were more likely to drop out (Wang et al., 2019). Hence, there is an urgent need for investigating how to improve learner engagement in MOOCs. Extant scholars have made initial effort to investigate engagement issues within MOOCs (Li & Baker, 2018). Nevertheless, they still regarded technological tools as the primary drivers for learner engagement. This is evident in that many existing MOOC studies * Corresponding author. Center for Studies of Information Resources, Wuhan University, Wuhan, 430072, Hubei Province, China E-mail addresses: sunyq@whu.edu.cn (Y. Sun), guoyanping@whu.edu.cn (Y. Guo), zhaoyiming@whu.edu.cn (Y. Zhao). https://doi.org/10.1016/j.compedu.2020.103963 Received 2 January 2020; Received in revised form 21 June 2020; Accepted 23 June 2020 Available online 26 June 2020 0360-1315/© 2020 Elsevier Ltd. All rights reserved. Computers & Education 157 (2020) 103963 Y. Sun et al. advocate the technological determinism, and they discuss how the Internet and computer technologies can maximize the effectiveness of MOOCs and support participant learning (Xie, 2019; Zhu et al., 2018). However, the opinion that technologies naturally drive certain actions may hide designers’ intention and it cannot account for how people learn to use technologies (Oliver, 2011). That is, learners exercise a high level of agency for setting forth their ideas (Cacciamani et al., 2012). In this respect, more attention should be paid to how learners utilize technologies rather than how technologies govern learner behaviors. Furthermore, although a few scholars have emphasized the importance of collaboration in MOOCs (Razmerita et al., 2018), the mechanisms underlying this collaboration process are yet unclear. It should be noted that in this paper, the collaboration pertains to a broad concept that includes any collaborative activities in the MOOC context, such as participants’ attending peer assessments and exchanging their ideas on the discussion forum in MOOCs (Formanek et al., 2017; Wise et al., 2017). Generally, MOOCs can provide technological supports to broadcast centralized materials like pre-recorded video lectures, slide presentations, e-textbooks assign­ ments, and automatic assessments. In addition, MOOCs can also provide collaborative tools like discussion board that enables question & answering and opinion debating for students to grasp a deep understanding of the material (Almatrafi et al., 2018). Moreover, the peer assessment function in MOOCs allows learners to rate their classmates in a reciprocal learning manner. From this perspective, learning in MOOCs is regarded as a social process in which group learners are responsible to construct knowledge, solve problems, and cultivate skills together (Razmerita et al., 2018). Hence, understanding the mechanism through which learners reaches a consensus on the MOOC system usage is crucial. On these accounts, our research question is: How will group learners reach a consensus on the tech­ nology usage and subsequently engage in MOOCs? To address this issue, the study draws upon adaptive structuration theory (AST) because it is suitable for capturing the features of learners’ collaborative process. Originated from the organizational research, AST illustrates how an advanced information technology (IT) is appropriated by group members (Desanctis & Poole, 1994). Central to AST is the concept consensus of appropriation, which represents the degree of agreement among group members regarding how the technology should be used and how it fits the group’s work (DeSanctis et al., 2008). Moreover, AST offers a comprehensive framework to interpret three sources of technology appropriation including IT structures, task features, and the group’s internal system (Desanctis & Poole, 1994). Contextualized to the MOOC setting, we identify collaborative spirit, task interdependence, and social interaction ties as the antecedents for consensus of appropriation. Then we establish a research model to illuminate the mechanism of learner engagement and test it using the survey data from Chinese University MOOC learners. This study can yield several implications. First, we clarify the mechanism of group learners’ engagement in the MOOC context. Given the fact that most previous scholars pay attention to the technological influences on learner engagement, we offer both theo­ retical and practical insights on how to boost MOOC learners’ collaboration. Second, we provide new understanding on MOOC learners’ behaviors through the lens of AST. In particular, current work is among the first to expound the process of collaboration in MOOCs ingeniously using the concept consensus of appropriation. Such a theorization presents a clear picture about how group learners collaboratively harness technologies to facilitate their learning. Third, current study also identifies the contextualized characteristics of the MOOC system as the antecedents for consensus of appropriation. This complements the deficiencies of the context-specific factors among the MOOC research (Hood et al., 2015), and supplies practitioners with concrete suggestions on improving the completion rate of MOOC. 2. Literature review 2.1. Learner engagement in MOOCs MOOCs are an online learning environment available for anyone via the Internet (Hood et al., 2015). In MOOCs, instructors deliver content to learners through pre-recorded video lectures, courseware resources, and interactive activities on the discussion forum, and learner performance are in general assessed through quizzing and peer assessment (Formanek et al., 2017). By virtue of the Internet and computer technologies, MOOCs expand traditional education to a global scope. Millions of learners around the world have registered on MOOC platforms, like edX, Coursera, and Udacity, to share knowledge and create collaborative learning opportunities (Cohen et al., 2019). However, the issue of high dropout rate also annoyed many scholars (e.g. Hew & Cheung, 2014; Hone & El Said, 2016). Since recent literature concluded that learners with higher levers of engagement were less likely to drop out (Wang et al., 2019), more investigations on the mechanism of learner engagement are needed. Learner engagement is defined as the amount of physical and psychological energies that student devotes to the academic expe­ rience (Oh et al., 2017). Following the definition, learner engagement embodies the psychological state that learners are emotionally and cognitively active in their course learning and the behavioral effort that learners expend in MOOCs to master the knowledge and pursuit high-quality performance (Sun et al., 2019). Conceptualizing learner engagement as the combination of psychological and behavioral dimensions is reasonable because this study investigates how learners reach agreement on the system usage, during which learners not only need to understand and accept the system design on the psychological level, but also make endeavor to participate in course-related activities via the MOOC system. Specifically, psychological engagement refers to the level of an individual’s positive, fulfilling, and action-related state of mind that is characterized by vigor, dedication, and absorption in MOOCs (Oh et al., 2017); while behavioral engagement pertains to learners’ sustained behaviors that extend beyond typical or expected in-role actions in MOOCs (Oh et al., 2017). On the whole, having a higher degree of engagement indicates that learners may have more enthusiasm for using technological tools to communicate and learn in MOOCs. Most extant works have focused on the patterns and measurements of learner engagement in MOOCs (Hew & Cheung, 2014). Another proportion of empirical research employs technology acceptance model (TAM) and task-technology-fit (TTF) to examine 2 Computers & Education 157 (2020) 103963 Y. Sun et al. learners’ adoption and usage of MOOCs (Joo et al., 2018). These studies position technological tools as the primary drivers of learner behaviors in MOOCs. However, the technology determinism cannot explain how people learn to use technologies because people are not simply mechanical conveyers of animating external sources of influence (Oliver, 2011). Instead, learners possess a high level of agency and they are responsible for constructing new skills and knowledge (Cacciamani et al., 2012). In this regard, MOOCs are learner-centered in which learners decide how they should use the technologies to aid their learning. Furthermore, researchers have pointed out that education should pay attention to not only the use of new technological solutions, but also the collaborative methods for students’ skill development (Sung & Hwang, 2013). Existing studies have also explored MOOC learners’ engagement from a social interaction perspective (Hone & El Said, 2016; Wang et al., 2019), including the discussion forum (Wise et al., 2017) and peer assessment (Formanek et al., 2017). In the online environment, learner interaction becomes particularly important because the traditional learning mode are shifting to the learner-centered approach that emphasizes collaboration (Zhao, Wang, & Sun, 2020). Nevertheless, prior conceptual works fail to theorize and empirically examine the collaborative process in MOOCs. Hence, understanding how learners can jointly handle the technologies to finish their assignments and subsequently engage in MOOCs is crucial, which calls for an integrated theoretical perspective. 2.2. Adaptive structuration theory (AST) Since MOOCs can take advantage of the Internet resources and technological tools to facilitate learners’ communications (Xie, 2019; Zhu et al., 2018), investigating learner engagement requires a thorough consideration of the technologies, tasks, and social interactions among learners. AST is a suitable theory to capture these features and portray learners’ behavioral mechanisms. Origi­ nated from the organizational research, AST was proposed as a blend thought of decision-making and institutional schools positing that group members’ task performance hinged on the appropriation of information technologies (ITs) under the social structures (Desanctis & Poole, 1994). Such social structures comprise of rules and resources provided by the characteristics regarding the technologies, tasks, and group norms (Desanctis & Poole, 1994; Kang et al., 2012). The process of creating such structures is so-called structuration (Kang et al., 2012). Structuration is not consequentially a reactive and passive adaptation to the environment, but it involves creative and proactive aspects that group members persistently refine and renovate throughout their membership careers (Kang et al., 2012). Due to its remarkable explanative power, AST has been used by previous scholars who adopt multi-faceted perspectives to investigate factors influencing system use and user performance across various contexts including organizational systems (Kang et al., 2012; Liang et al., 2015), health information technologies (Romanow et al., 2018), and customer relationships management (Johnson et al., 2012). For instance, Liang et al. (2015) drew on AST to examine how system complexity, task characteristics (i.e., job autonomy and task variety), and innovation climate jointly affect employees’ exploration and extended use of enterprise systems. Similarly, AST can provide an integrated framework to investigate how the features of the system, tasks, and group learners facilitate their consensus on technology usage and subsequent engagement, as elaborated next. AST concentrates on the impact of social structures, including the structures of ITs (e.g., structural features and spirits), other sources of structures (e.g., tasks and the organizational environment), and the group’s internal system, on the appropriation of structures which in turn leads to decision outcomes (e.g., efficiency, quality, and commitment) (Desanctis & Poole, 1994). To be specific, the structures of an advanced IT can be described in terms of its spirit, which refers to the general intent with respect to values and goals underlying a given set of structural features. The unconditional access of MOOCs grants global learners bountiful courses, facilitating them to share knowledge and collaborate (Cohen et al., 2019). Moreover, previous studies have shown that the MOOC’s functions, such as the discussion forum (Wise et al., 2017) and peer assessment (Formanek et al., 2017), can promote learners’ collaboration. Therefrom, in the MOOC context, collaborative spirit can be identified as the design philosophy of this learning system. Besides, the contents and constraints of a given task are another major source of structures. MOOCs remind learners the importance of joint effort and encourage them to complete the learning tasks through collaboration. Thus, task interdependence can be viewed as an important source of structures. Moreover, a certain structure may be appropriated differently relying on the group’s internal system, which refers to the nature of members and their relationships inside the group (DeSanctis et al., 2008). Since learners in MOOCs are favored to have a stable construction of their own knowledge through interactions with peers, social interaction ties are suitable to represent the structure of group’s internal system. 這個研究只做年輕⼈ 延伸研究可以做其他年齡層 Fig. 1. Research model. 3 Computers & Education 157 (2020) 103963 Y. Sun et al. Further, whether group members can leverage these structures to achieve learning outcomes largely relies on the consensus of appropriation, which is the extent of agreement among group members concerning how the technology should be used and how it fits the group’s work (DeSanctis et al., 2008). In the context of collaborative technology use, the benefits can only be realized once the entire learning group or an adequately large number of members collectively adopt the technology (Kang et al., 2012). At first, members may have divergent opinions until they reach a consensus on how to utilize the Internet and computer technologies to accomplish their tasks (Desanctis & Poole, 1994). They gradually settle such disagreements by learning and communicating with each other. Subsequently, improvements in these decision processes are expected to result in desirable outcomes like strong commitment and high degree of engagement. 3. Research model and hypotheses According to the arguments above, we draw upon adaptive structuration theory (AST) to develop a research model as depicted in Fig. 1. Next, we will provide detailed justifications about the proposed hypotheses. 3.1. Structures of appropriation AST claims that in a specific information system, structures reflect the beliefs, values, and goals as understood by the designers. In other words, they are designed to reflect a spirit (Desanctis & Poole, 1994). By extension, the spirit is the “official line” that the technology presents to the users regarding how to behave when using the system, how to interpret its characteristics, and how to fill in gaps in the process which are not explicitly specified. Learners are unconditionally supported by MOOCs to attend courses, share knowledge and collaborate (Cohen et al., 2019). Prior researchers have indicated several means that the MOOC system can afford to support learners’ collaboration including the discussion forum (Wise et al., 2017) and peer assessment (Formanek et al., 2017). Thus, we identify collaborative spirit as an essential concern for the MOOC system. In this study, we define collaborative spirit as the general intent with respect to collaborative values and goals underlying the system design of MOOCs. The spirit provides the “legitimation” for the technology by supplying a normative frame with regard to behaviors that are appropriate in a specific system. Meanwhile, spirit can function as a means of expression, because it assists users in understanding and interpreting the meaning of the technology (Desanctis & Poole, 1994). Specific to our study, the more collaborative spirit it demonstrates, the more group learners will stick to the goals and values offered by the MOOC system. In this way, they are more likely reaching an agreement on the system usage. The first hypothesis is put forward as follows: H1. Collaborative spirit is positively related to consensus of appropriation In addition to IT structures, group members can utilize the structural potentials offered by other sources, among which an important element is the contents or constraints of the group’s tasks (DeSanctis et al., 2008). MOOCs improve communications and interactions that can drive learners to consider other’ demands and to evaluate the impacts of their learning patterns on other learners (Zhu et al., 2018). In this regard, group members are more likely to realize the importance of collaborative tasks (Sharma & Yetton, 2007). That is, the completion of learning tasks in MOOCs is dependent on the group’s joint effort. We thus consider interdependence as a critical constraint for the MOOC tasks. Task interdependence is defined as the extent to which group members rely on one another to perform and complete their works (Van de Ven et al., 1976). In the MOOC context, as the dependence on one person shifts to multiple individuals, technologies are supposed to accommodate different needs for these task variations (Razmerita et al., 2018). Hence, the features of tasks can affect which technological structures are chosen for using, how the results are explained, and how they are applied. Thus, a high level of task interdependence will lead to a more consistency on how the Internet and computer technologies should be used by MOOC learners. H2 is formulated below: H2. Task interdependence is positively related to consensus of appropriation For MOOCs, building relations with these learners who share the same interests is a key to their success. Further, extant scholars also argue that learning in MOOCs is a social process where individuals take the responsibility for constructing their own under­ standing about knowledge through social interactions (Chen & Chen, 2015). Therefore, apart from collaborative spirit and task interdependence, the third premise for adaptive structuration is the social interaction ties within the group’s internal system, defined as the strength of the relationships, the amount of learning time spent, and the communication frequency among members in MOOCs (Chiu et al., 2006). Early research has suggested that the relationships among group members may appropriate given structures inside their internal system (DeSanctis et al., 2008). During the learning process in MOOCs, this adaptation may happen in two ways. For one thing, the increasing frequency of contact can strengthen the trust and attachment among group members. Learners will be more open for others’ opinions and suggestions on the MOOC system usage. For another, enhanced communications and interactions will breed the sense of belonging to the tasks and courses. In such a condition, learners are willing to discuss how to maximize the utilities of the technologies and they tend to reach an agreement easily. Therefore, we propose the next hypothesis: H3. Social interaction ties are positively related to consensus of appropriation. 4 Computers & Education 157 (2020) 103963 Y. Sun et al. 3.2. Outcomes for appropriation Promotion in decision outcomes will emerge only if the group members come to an agreement that the focal technology is suitable for the tasks at hand (Desanctis & Poole, 1994). Commitment is a typical outcome for appropriation that can be defined as the learner’s sense of belonging and positive feeling toward the MOOC system (Ellemers et al., 1999). Basically, consensus of appropriation plays dual roles. On one hand, it lessens the uncertainty about which structures of the technological interventions are appropriate for a given task. At this point, users can obtain utilitarian benefits and the feelings of reliability from the MOOC system, which may prompt the establishment of affective bonds between them. On the other hand, the consensus is also associated with less ambiguity and conflict over technology using patterns (Im, 2014). Resolving this concern can improve learners’ efficiency and thus facilitate their sense of involvement in the MOOC system. On these accounts, group members will develop the tendency to emotionally attach themselves to the MOOC system, which is the foundation of commitment (Allen & Meyer, 1990). Therefore, we hypothesize: H4. Consensus of appropriation is positively related to commitment In addition to commitment, the process of technology appropriation can also contribute to straightforward participation and system usage (Im, 2014; Kang et al., 2012). Specific to the MOOC context, consensus may serve as a social pressure among the learners in a group that arouses their intense enthusiasm and motivations to fulfill the learning tasks. Moreover, the social learning theory (SLT) (Bandura, 1977) can provide a theoretical explanation for our prediction from another angle. SLT posits that individuals are more likely to adopt a modelled behavior with functional values if the model is similar to the observer’s and has an admired status (Bandura, 1977). Generally, people are exposed to alternative learning opportunities by observing others’ experiences. When the system usage leads to positive results, other members are likely to imitate that behavior once there is strong consensus among the group (Im, 2014). On account of these justifications, if MOOC learners experience less uncertainty and ambiguity concerning appropriations, they will be more prone to engagement in the MOOC system. Hence, H5 is proposed: H5. Consensus of appropriation is positively related to learner engagement Lastly, we assume that affective outcomes have a direct relationship with the psychological and behavioral ones. That is, commitment will also render the learning experiences more meaningful and thus facilitate learner engagement. Commitment em­ bodies an individual’s emotional attachment, identification, and involvement toward the object (Allen & Meyer, 1990). Rather than driven by interests, commitment is based on the affective bonds between a focal person and the target in general (Meyer et al., 2002). As a consequence, people with this sense of belonging often demonstrate the tendency of maintaining long-term relationships (Hashim & Tan, 2015). Correspondingly, engagement has been postulated as a psychological process and behavioral manifestation that can lead to the formation of loyalty (van Doorn et al., 2010). Specific to the MOOC context, engagement is viewed as co-occurring with the construction of group knowledge, which can facilitate task accomplishment and goal realization (Fang et al., 2019). Owing to these benefits, committed group learners are willing to develop longstanding association with the MOOC system through a joint regulation on the task comprehending (i.e., psychological engagement), as well as the endeavor investment and on-task persistence (i.e., behavioral engagement) (Fang et al., 2019). Therefore, we put forward the last hypothesis: H6. Commitment is positively related to learner engagement 4. Methodology 4.1. Research setting To test the proposed hypotheses, our data were collected using an online survey from the users of the Chinese University MOOC (http://www.icourse163.org/). The platform embraces functions such as bulletin board, discussion board, courseware, test and assignment, final exam, peer assessment, and sharing courses to social media. Chinese University MOOC was launched in 2014 and it has been one of the most authoritative MOOC platforms in China with more than 11 million active learners. Moreover, it involves 612 outstanding universities and institutions to provide thousands of high-quality courses that cover a broad range of disciplines. Considering these merits, Chinese University MOOC is suitable to be used in this investigation. 4.2. Measures All the measures except for collaborative spirit were adapted from prior studies and had been confirmed to be reliable and valid. To fit the MOOC context, we made some slight wording modifications. Seven-point Likert scales anchored with “strongly disagree” to “strongly agree” were adopted for multiple items of all latent constructs. Specifically, learner engagement was taken as a second-order reflective construct composed of psychological and behavioral dimensions, both of which were measured with the items adapted from Stumpf et al. (2013). In addition, the items measuring social interaction ties and commitment were adapted from Chiu et al. (2006). Besides, consensus of appropriation was measured with five items adapted from Kang et al. (2012) while task interdependence was measured with four items adapted from Sharma and Yetton (2007). Further, because there were no validated items of collaborative spirit among previous studies, we developed the initial instruments including three items based on its definition that the general intent with respect to collaborative values and goals underlying the system design. The face validity of the instruments was first assessed by several PhD students majoring in education and then by several actual users of Chinese University MOOC. Statistical validity was 5 Computers & Education 157 (2020) 103963 Y. Sun et al. further assessed and reported in the next section. All the constructs and measures were elaborated in Table 1. 4.3. Data collection procedure Our survey was conducted via wjx.cn, the largest online questionnaire distribution and collection platform in China. This platform has been widely adopted by more than 30,000 companies and over 90% of Chinese universities. On average, more than one million people fill in questionnaires on it every day. The platform is effective and powerful, with abundant sources of respondents, and it supports online viewing, screening, as well as data transformation in a variety of formats. Given these advantages, we adopted wjx.cn to collect our survey data. We first did a pilot test among 20 random-selected students who had experience for learning on Chinese University MOOC platform and we collected their comments on the length and wording of the instruments, which helped to improve the quality of the ques­ tionnaire and to ensure its clarity and validity. Functions that reflect the nature of collaboration, such as the discussion forum and peer assessment, are inherent in the MOOC platform. The questionnaires were sent to students in a certain course taught by us via the e-mail addresses they provided at the Chinese University MOOC platform. The purpose of this survey was claimed in the email, and participating in this survey was voluntary and anonymous. In order to motivate more students to answer the questionnaire, we guaranteed participants extra credits by checking their screenshots of the questionnaire completion page. We were informed of all the participants, but we did not ascertain the exact questionnaires corresponding to them. In this way, we could offer the participants credit reward and meanwhile ensure the anonymous of our survey process. Finally, we obtained 374 valid samples and the de­ mographics were shown in Table 2. Since online survey was adopted to collect data, the response bias may exist. In order to solve this problem, we checked the de­ mographic difference between the first third and the last third of respondents (Kim et al., 2016). The results uncovered that there were no significant differences between the two sub-groups in demographics except for the login frequency. Moreover, we compared the construct means for these two sub-groups and found that there was no significant difference. These examinations demonstrated that response bias was not a serious concern for this study. 5. Data analysis Partial least squares (PLS) was used in the data analysis based on several advantages of this technique. As a second-generation structural equation modeling (SEM) approach, PLS can estimate the loadings of indicators on constructs and the causal relation­ ships among constructs (Fornell & Bookstein, 1982). Besides, compared to covariance-based SEM approach, PLS is more appropriate to deal with small sample size (Hair et al., 2011), which fits our study. Moreover, whereas CB-SEM is regarded as being more appropriate Table 1 Measures for the constructs in the research model. Constructs Items Collaborative Spirit (Self-developed) CLS1: The MOOC system is designed with the spirit of facilitating cooperation between students. CLS2: The MOOC system is designed with the spirit of facilitating collaboration between students. CLS3: The MOOC system is designed with the spirit of facilitating students’ teamwork. TID1: The learning tasks in the MOOC system can be performed fairly independently of others.* TID2: The learning tasks in the MOOC system can be planned with little need to coordinate with others.* TID3: It is rarely required to obtain information from others to complete the learning tasks in the MOOC system.* TID4: The learning tasks in the MOOC system is relatively unaffected by the performance of other individuals.* SIT1: I maintain close social relationships with some members in the MOOC system. SIT2: I spend a lot of time interacting with some members in the MOOC system. SIT3: I have frequent communication with some members in the MOOC system. SIT4: I know some members in the MOOC system on a personal level. CAP1: Overall, MOOC members agreed on how we should use the MOOC system for learning. CAP2: MOOC members reached mutual understanding on how we should use the MOOC system to perform our learning tasks. CAP3: There was no conflict in MOOC members regarding how we should use the MOOC system in our learning. CAP4: MOOC members were able to reach consensus on how to apply the MOOC system to learn. CAP5: MOOC members were able to reach consensus on how we should use the MOOC system to perform our learning tasks. CMT1: I feel a sense of belonging towards the MOOC system. CMT2: I am proud to be a member of the MOOC system. CMT3: I have the feeling of emotional attachment to the MOOC system. PENG1: When participating in MOOCs, I am energized by the work I do. PENG2: When participating in MOOCs, I am enthusiastic about my work. PENG3: When participating in MOOCs, my work really interests me. PENG4: When participating in MOOCs, the work that I do is very satisfying to me. PENG5: When participating in MOOCs, my work is personally fulfilling. BENG1: When participating in MOOCs, I often take extra initiative to get things done. BENG2: When participating in MOOCs, I actively seek opportunities to contribute. BENG3: When participating in MOOCs, I often put more effort into my work than is required. Task Interdependence (Sharma & Yetton, 2007) Social Interaction Ties (Chiu et al., 2006) Consensus of Appropriation (Kang et al., 2012) Commitment (Chiu et al., 2006) Psychological Engagement (Stumpf et al., 2013) Behavioral Engagement (Stumpf et al., 2013) Note: * The item was reverse coded. 6 Computers & Education 157 (2020) 103963 Y. Sun et al. Table 2 Demographic statistics. Characteristics Levels Frequency Percentage (%) Gender Male Female �22 23–25 26–28 �29 Below bachelor Bachelor Master or above �2 years 3–4 years 5–6 years 7–8 years >8 years 1 2–3 4–5 �6 199 175 171 68 19 116 33 260 81 20 66 66 58 164 66 173 61 74 53.2 46.8 45.7 18.2 5.1 31.0 8.8 69.5 21.7 5.4 17.6 17.6 15.5 43.9 17.6 46.3 16.3 19.8 Age Education Internet experience Number of courses enrolled in for theory confirmation, PLS does provide a good approximation of CB-SEM in terms of final estimates (Hair et al., 2011). From the above considerations, PLS was used in the analysis. Specifically, SmartPLS was adopted as the analytic tool. 5.1. Measurement model All the constructs were reflectively measured, so the measurement model for the constructs was evaluated by examining their reliabilities, convergent and discriminant validities. The reliability of a construct could be assessed by checking its average variance extracted (AVE), composite reliability, and Cronbach’s Alpha (Fornell & Larcker, 1981). As shown in Table 3, the AVEs, composite reliabilities, and Cronbach’s Alpha coefficients for all the constructs were greater than 0.6, 0.8, and 0.8, exceeding the suggested threshold values of 0.5, 0.7, and 0.7, respectively (Fornell & Larcker, 1981). Therefore, all of these constructs were with suitable reliabilities. Convergent validity can be assessed by checking whether or not the item loadings on their respective constructs were high enough while discriminant validity can be assessed by checking whether or not the item loadings on their respective constructs were higher than the loadings on the other constructs (i.e., cross-loadings). As shown in Table 4, the item loadings on their respective constructs were higher than 0.7 and these loadings were higher than the cross-loadings, suggesting that these constructs had adequate convergent and discriminant validities. Another method to evaluate the discriminant validity is to compare the square root of AVE for a construct and the correlation coefficients related to this construct. As shown in Table 5, the square roots of AVE for all the constructs were greater than the cor­ relation coefficients, proving that these constructs were with good discriminant validities (Bock et al., 2005). As discussed in the preceding paragraph, learner engagement was taken as a second-order reflective construct consisting of psy­ chological engagement and behavior engagement. As shown in Table 6, the loadings for the two first-order constructs were above 0.8 with the significance at p < 0.01 level, proving a good validity of learner engagement. Finally, since the correlations between constructs were relatively high, a multicollinearity concern could arise. However, the regression analysis results indicated that the variance inflation factor (VIF) values for collaborative spirit, task interdependence, social interaction ties, consensus of appropriation, commitment, psychological engagement, and behavioral engagement were 1.338, 1.305, 1.451, 1.529, 1.883, 2.612, and 2.098, respectively, lowering than the suggested threshold value 3.3 (Petter et al., 2007). Therefore, multicollinearity was not a serious problem in this study. Table 3 Construct reliability. Construct AVE Composite Reliability Cronbach’s Alpha BENG CAP CLS CMT PENG SIT TID 0.805 0.792 0.906 0.848 0.684 0.882 0.779 0.925 0.950 0.967 0.943 0.915 0.968 0.933 0.878 0.934 0.948 0.910 0.881 0.955 0.905 Note: BENG¼Behavioral engagement; CAP¼Consensus of appropriation; CLS¼Collaborative spirit; CMT¼Commitment; PEN­ G¼Psychological engagement; SIT¼Social interaction ties; TID ¼ Task interdependence. 7 Computers & Education 157 (2020) 103963 Y. Sun et al. Table 4 Cross-loadings. BENG1 BENG2 BENG3 CAP1 CAP2 CAP3 CAP4 CAP5 CLS1 CLS2 CLS3 CMT1 CMT2 CMT3 PENG1 PENG2 PENG3 PENG4 PENG5 SIT1 SIT2 SIT3 SIT4 TID1 TID2 TID3 TID4 BENG CAP CLS CMT PENG SIT TID 0.867 0.931 0.892 0.262 0.276 0.195 0.243 0.240 0.276 0.264 0.232 0.451 0.482 0.481 0.458 0.687 0.665 0.620 0.522 0.198 0.190 0.237 0.109 0.227 0.252 0.212 0.252 0.255 0.243 0.240 0.834 0.912 0.891 0.930 0.880 0.356 0.291 0.312 0.355 0.327 0.345 0.411 0.289 0.268 0.218 0.258 0.482 0.443 0.447 0.426 0.350 0.317 0.393 0.302 0.250 0.249 0.230 0.260 0.329 0.275 0.301 0.338 0.963 0.953 0.939 0.448 0.383 0.363 0.274 0.243 0.249 0.310 0.315 0.344 0.312 0.314 0.249 0.058 0.045 0.087 0.093 0.478 0.473 0.425 0.292 0.353 0.316 0.351 0.341 0.432 0.409 0.389 0.907 0.924 0.932 0.461 0.505 0.516 0.492 0.539 0.392 0.372 0.416 0.281 0.229 0.248 0.230 0.204 0.643 0.688 0.606 0.324 0.324 0.276 0.296 0.289 0.331 0.333 0.287 0.548 0.561 0.565 0.652 0.887 0.880 0.869 0.826 0.275 0.267 0.321 0.175 0.330 0.318 0.267 0.304 0.278 0.149 0.102 0.452 0.427 0.451 0.438 0.363 0.332 0.278 0.317 0.418 0.282 0.379 0.494 0.227 0.161 0.134 0.206 0.943 0.960 0.948 0.903 0.286 0.265 0.379 0.152 0.253 0.237 0.224 0.370 0.335 0.392 0.333 0.300 0.109 0.055 0.059 0.237 0.215 0.261 0.339 0.311 0.268 0.258 0.263 0.292 0.296 0.349 0.244 0.904 0.921 0.889 0.811 Notes: BENG¼Behavioral engagement; CAP¼Consensus of appropriation; CLS¼Collaborative spirit; CMT¼Commitment; PENG¼Psychological engagement; SIT¼Social interaction ties; TID ¼ Task interdependence. Table 5 Means, standard deviations, and correlations. BENG CAP CLS CMT PENG SIT TID Mean SD BENG CAP CLS CMT PENG SIT TID 5.589 4.378 5.223 5.235 5.384 3.298 4.725 1.122 1.422 1.344 1.340 1.067 1.757 1.562 0.897 0.274 0.271 0.512 0.722 0.198 0.266 0.890 0.337 0.372 0.346 0.481 0.391 0.952 0.432 0.333 0.326 0.080 0.921 0.606 0.391 0.259 0.827 0.288 0.347 0.939 0.315 0.882 Notes: SD¼Standard deviation; BENG¼Behavioral engagement; CAP¼Consensus of appropriation; CLS¼Collaborative spirit; CMT¼Commitment; PENG¼Psychological engagement; SIT¼Social interaction ties; TID ¼ Task interdependence. Boldfaced diagonal elements are the square roots of AVEs. Table 6 Second-order constructs. High-order construct Low-order construct Loading t-statistic Learner engagement Psychological engagement Behavioral engagement 0.954 0.896 146.951 53.582 5.2. Structural model PLS results of the structural model were illustrated in Fig. 2. The results showed that collaborative spirit (β ¼ 0.210, t ¼ 3.770), task interdependence (β ¼ 0.270, t ¼ 5.233), and social interaction ties (β ¼ 0.326, t ¼ 6.683) were all positively related to consensus of appropriation, thus supporting H1, H2, and H3. Additionally, consensus of appropriation was found to be significantly related to commitment (β ¼ 0.372, t ¼ 7.023) and learner engagement (β ¼ 0.128, t ¼ 2.719), confirming H4 and H5. Finally, commitment was proved to have a significantly positive relation with learner engagement (β ¼ 0.583, t ¼ 13.976), lending support to H6. 8 Computers & Education 157 (2020) 103963 Y. Sun et al. Fig. 2. PLS results. 6. Discussion 6.1. Key findings Drawing upon adaptive structuration theory (AST), the study examines how learners adapt themselves to the technology and thus engage in MOOCs. Our results can elicit several crucial findings. First, collaborative spirit, task interdependence, and social interaction ties are all strong predictors of consensus of appropriation. This verifies AST’s basic presumption that group members’ adaptation for an advanced technology depends on the joint effects of IT structures, task characteristics, and the group’s internal system (DeSanctis et al., 2008). Second, consensus of appropriation is confirmed to significantly enhance both commitment and learner engagement, demonstrating that when learners reach an agreement about how to use the technology, emotional bonds and behavioral tendencies are likely to develop (Romanow et al., 2018). Under the circumstances, positive outcomes toward the MOOC system will emerge. Although several extant studies have investigated the antecedents of engagement in MOOCs (Sun et al., 2019), they mostly focused on learners’ characteristics, goals, and motivations. This research is unique in that the construct consensus of appropriation is utilized to theorize learners’ collaborative process in MOOCs and we empirically confirm its positive relation with learner engagement. Lastly, the results also validate commitment as a facilitator for learner engagement. A possible reason lies in that learner engagement denotes persistent participation in the MOOC system, which fits prior conclusions that commitment is conducive to the formation of long-term relationships (Hashim & Tan, 2015). 6.2. Theoretical implications This study contributes to related literature in multiple ways. First, we clarify the mechanism shaping learner engagement in the MOOC context. Although prior scholars have demonstrated the significant role of learner engagement in MOOCs (Fang et al., 2019; Li & Baker, 2018; Wang et al., 2019), they mainly investigate the effects of technologies and intrinsic motivations, nevertheless failing to interpret how learners collaborate with each other in MOOCs. Under the MOOC setting, the degree of productive participation hinges on the quality of collaboration in shared activities. Therefore, ascertaining the mechanism underlying learner collaboration can shed light on how to facilitate effective engagement in the MOOC system. Second, we make a fresh attempt to address the collaboration process in MOOCs through the concept consensus of appropriation. With the paradigm shift from technological determinism to human agency, researchers moreover pay attention to whether and how people can handle the system rather than how the system should be designed (Oliver, 2011). In other words, if users cannot construct sufficient understanding on the using patterns, a well-built system will probably become a waste. Particularly in terms of the MOOC, only when learners collaboratively comprehend the system rules and are able to use the learning tools can the MOOC system realizes its maximum utility (Razmerita et al., 2018). However, a proper theoretical perspective to explain the collaborative mechanism is still lacking among extant research. Consensus of appropriation, a key concept of AST, represents the extent of agreement among group members concerning how the technology should be used to match the group’s works (DeSanctis et al., 2008). Consequently, it fits the process of collaboration in MOOCs and can provide new understanding for related scholars on group learners’ behaviors. Third, we identify the antecedents of consensus of appropriation by contextualizing the system-, task-, and learners-related factors in the MOOC context. Notwithstanding extant scholars have been struggling to investigate how to better facilitate group learners’ collaboration (Chen & Chen, 2015), they conducted case study or content analysis but failed to theoretically identify the prerequisites of learners’ collaboration. This leaves a considerable blank to comprehensively understand what leads to learners’ consensus on the use of MOOC system. As argued previously, MOOCs pertain to a multi-faceted schema where group learners take active control of studying and social interactions with the help of the Internet and computer technologies (Xie, 2019). Accordingly, AST suggests that the major sources of structure for groups as they interact with an advanced information technology include the system’s spirit, tasks, 9 Computers & Education 157 (2020) 103963 Y. Sun et al. and members’ interaction in the environment (Desanctis & Poole, 1994). Therefore, AST offers a sound framework to account for the drivers of group learners’ consensus on the MOOC system usage. 6.3. Practical implications The study also provides abundant insights for practice regarding how to encourage learners to reach an agreement about tech­ nology usage and thereafter engage in the MOOC system. First of all, since collaboration has been proven to be an important part of the MOOC system, practitioners should continue to strengthen this advantage and optimize technological structures, thus fostering learners’ collaborative consciousness and improving their participating intention. Second, MOOC practitioners must keep an eye on the task characteristics within the learning system. Specifically, task interdependence is another critical determinant for group learners’ consensus which indicates that learners can only complete their tasks when they rely on each other. Hence, MOOC operators are supposed to enrich the assignments and activities throughout the courses to maintain task interrelations. Finally, given the fact that group internal factor plays an essential role as well, we suggest MOOC designers refine the communication strategies to expedite user’s social interaction in the system and provide functional supports for teamwork, such as a real-time discussion group embedded in each course. 6.4. Limitations and future research Although this study yields several important implications for academia and industry, some limitations still need to be acknowl­ edged for future research. First, to test the proposed hypotheses, our study employed a cross-section survey. Although multicollinearity was not verified to be a serious concern, mixed-methods and longitudinal design are strongly suggested to better capture the causal relationships. Second, it is notable that current study was conducted in the setting of Chinese University MOOC. Since previous research indicated that persons with different cultural background may have different behavioral motives (Hofstede, 1980). Specif­ ically, Chinese people hold the collectivism that may better meet the group learning pattern. Future scholars can take cross-cultural issues into account to reach more interesting findings. Third, to explore why learners engage in MOOCs, we draw on AST to incorporate the features of technologies, tasks, and the group’s internal system. However, other essential factors, such as the interface design, group climate, as well as learners’ motivations and experiences are also worth investigating to explain the mechanism underlying learner behaviors in MOOCs. CRediT authorship contribution statement Yongqiang Sun: Conceptualization, Methodology, Writing - original draft. Yanping Guo: Formal analysis, Writing - original draft. Yiming Zhao: Investigation, Writing - review & editing. Acknowledgments The work described in this paper was partially supported by the grants from the National Natural Science Foundation of China (Grant No. 71974148, 71874130, 71420107026, and 71921002), the Humanities and Social Sciences Foundation of the Ministry of Education, China (Project No. 16YJC870011, 17YJC630157, 18YJC870026), the world-class discipline project “Library, Information and Data Science”, and the China Association for Science and Technology (Project No. 2017QNRC001). References Allen, N. J., & Meyer, J. P. (1990). The measurement and antecedents of affective, continuance and normative commitment to the organization. Journal of Occupational Psychology, 63(1), 1–18. https://doi.org/10.1111/j.2044-8325.1990.tb00506.x. Almatrafi, O., Johri, A., & Rangwala, H. (2018). Needle in a Haystack: Identifying learner posts that require urgent response in MOOC discussion forums. Computers & Education, 118, 1–9. https://doi.org/10.1016/j.compedu.2017.11.002. Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall. Bock, G.-W., Zmud, R. W., Kim, Y.-G., & Lee, J.-N. (2005). Behavioral intention formation in knowledge sharing: Examining the roles of extrinsic motivators, socialpsychological forces, and organizational climate. MIS Quarterly, 29(1), 87–111. https://doi.org/10.2307/25148669. Cacciamani, S., Cesareni, D., Martini, F., Ferrini, T., & Fujita, N. (2012). Influence of participation, facilitator styles, and metacognitive reflection on knowledge building in online university courses. Computers & Education, 58(3), 874–884. https://doi.org/10.1016/j.compedu.2011.10.019. Chen, Y. H., & Chen, P. J. (2015). MOOC study group: Facilitation strategies, influential factors, and student perceived gains. Computers & Education, 86, 55–70. https://doi.org/10.1016/j.compedu.2015.03.008. Chiu, C.-M., Hsu, M.-H., & Wang, E. T. G. (2006). Understanding knowledge sharing in virtual communities: An integration of social capital and social cognitive theories. Decision Support Systems, 42(3), 1872–1888. https://doi.org/10.1016/j.dss.2006.04.001. Cohen, A., Shimony, U., Nachmias, R., & Soffer, T. (2019). Active learners’ characterization in MOOC forums and their generated knowledge. British Journal of Educational Technology, 50(1), 177–198. https://doi.org/10.1111/bjet.12670. Desanctis, G., & Poole, M. S. (1994). Capturing the complexity in advanced technology use: Adaptive structuration theory. Organization Science, 5(2), 121–147. https://doi.org/10.1287/orsc.5.2.121. DeSanctis, G., Poole, M. S., Zigurs, I., DeSharnais, G., D’Onofrio, M., Gallupe, B., & Shannon, D. (2008). The Minnesota GDSS research project: Group support systems, group processes, and outcomes. Journal of the Association for Information Systems, 9(10), 551–608. https://doi.org/10.17705/1jais.00177. van Doorn, J., Lemon, K. N., Mittal, V., Nass, S., Pick, D., Pirner, P., & Verhoef, P. C. (2010). Customer engagement behavior: Theoretical foundations and research directions. Journal of Service Research, 13(3), 253–266. https://doi.org/10.1177/1094670510375599. 10 Computers & Education 157 (2020) 103963 Y. Sun et al. Ellemers, N., Kortekaas, P., & Ouwerkerk, J. W. (1999). Self-categorisation, commitment to the group and group self-esteem as related but distinct aspects of social identity. European Journal of Social Psychology, 29(2-3), 371–389. https://doi.org/10.1002/(SICI)1099-0992(199903/05)29:2/3<371::AID-EJSP932>3.0.CO;2U. Fang, J., Tang, L., Yang, J., & Peng, M. (2019). Social interaction in MOOCs: The mediating effects of immersive experience and psychological needs satisfaction. Telematics and Informatics, 39, 75–91. https://doi.org/10.1016/j.tele.2019.01.006. Formanek, M., Wenger, M. C., Buxner, S. R., Impey, C. D., & Sonam, T. (2017). Insights about large-scale online peer assessment from an analysis of an astronomy MOOC. Computers & Education, 113, 243–262. https://doi.org/10.1016/j.compedu.2017.05.019. Fornell, C., & Bookstein, F. L. (1982). Two structural equation models: LISREL and PLS applied to consumer exit-voice theory. Journal of Marketing Research, 19(4), 440–452. https://doi.org/10.1177/002224378201900406. 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. Hair, J. F., Sarstedt, M., & Ringle, C. M. (2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19(2), 139–152. https://doi.org/10.2753/ MTP1069-6679190202. Hashim, K. F., & Tan, F. B. (2015). The mediating role of trust and commitment on members’ continuous knowledge sharing intention: A commitment-trust theory perspective. International Journal of Information Management, 35(2), 145–151. https://doi.org/10.1016/j.ijinfomgt.2014.11.001. Hew, K. F., & Cheung, W. S. (2014). Students’ and instructors’ use of massive open online courses (MOOCs): Motivations and challenges. Educational Research Review, 12, 45–58. https://doi.org/10.1016/j.edurev.2014.05.001. Hofstede, G. (1980). Culture’s consenquence: International differences in work-related values. London: Sage. Hone, K. S., & El Said, G. R. (2016). Exploring the factors affecting MOOC retention: A survey study. Computers & Education, 98, 157–168. https://doi.org/10.1016/j. compedu.2016.03.016. Hood, N., Littlejohn, A., & Milligan, C. (2015). Context counts: How learners’ contexts influence learning in a MOOC. Computers & Education, 91, 83–91. https://doi. org/10.1016/j.compedu.2015.10.019. Im, G. (2014). Effects of cognitive and social factors on system utilization and performance outcomes. Information & Management, 51(1), 129–137. https://doi.org/ 10.1016/j.im.2013.10.002. Johnson, D. S., Clark, B. H., & Barczak, G. (2012). Customer relationship management processes: How faithful are business-to-business firms to customer profitability? Industrial Marketing Management, 41(7), 1094–1105. https://doi.org/10.1016/j.indmarman.2012.04.001. Joo, Y. J., So, H. J., & Kim, N. H. (2018). Examination of relationships among students’ self-determination, technology acceptance, satisfaction, and continuance intention to use K-MOOCs. Computers & Education, 122, 260–272. https://doi.org/10.1016/j.compedu.2018.01.003. Kang, S., Lim, K. H., Kim, M. S., & Yang, H.-D. (2012). Research note—a multilevel analysis of the effect of group Appropriation on collaborative technologies use and performance. Information Systems Research, 23(1), 214–230. https://doi.org/10.1287/isre.1100.0342. Kim, S. H., Mukhopadhyay, T., & Kraut, R. E. (2016). When does repository KMS use lift performance? The role of alternative knowledge sources and task environments. MIS Quarterly, 40(1), 133–156. https://doi.org/10.25300/misq/2016/40.1.06. Liang, H., Peng, Z., Xue, Y., Guo, X., & Wang, N. (2015). Employees’ exploration of complex systems: An integrative view. Journal of Management Information Systems, 32(1), 322–357. https://doi.org/10.1080/07421222.2015.1029402. Li, Q., & Baker, R. (2018). The different relationships between engagement and outcomes across participant subgroups in massive open online courses. Computers & Education, 127, 41–65. https://doi.org/10.1016/j.compedu.2018.08.005. Meyer, J. P., Stanley, D. J., Herscovitch, L., & Topolnytsky, L. (2002). Affective, continuance, and normative commitment to the organization: A meta-analysis of antecedents, correlates, and consequences. Journal of Vocational Behavior, 61(1), 20–52. https://doi.org/10.1006/jvbe.2001.1842. Oh, C., Roumani, Y., Nwankpa, J. K., & Hue, H. F. (2017). Beyond likes and tweets: Consumer engagement behavior and movie box office in social media. Information & Management, 54(1), 25–37. https://doi.org/10.1016/j.im.2016.03.004. Oliver, M. (2011). Technological determinism in educational technology research: Some alternative ways of thinking about the relationship between learning and technology. Journal of Computer Assisted Learning, 27(5), 373–384. https://doi.org/10.1111/j.1365-2729.2011.00406.x. Petter, S., Straub, D., & Rai, A. (2007). Specifying formative constructs in information systems research. MIS Quarterly, 31(4), 623–656. https://doi.org/10.2307/ 25148814. Razmerita, L., Kirchner, K., Hockerts, K., & Tan, C. W. (2018). Towards a model of collaborative intention: An empirical investigation of a massive online open course (MOOC). In Paper presented at the proceedings of the 51st Hawaii international conference on system sciences. https://doi.org/10.24251/HICSS.2018.091. Reich, J., & Ruiperez-Valiente, J. A. (2019). The MOOC pivot. Science, 363(6423), 130–131. https://doi.org/10.1126/science.aav7958. Romanow, D., Rai, A., & Keil, M. (2018). CPOE-enabled coordination: Appropriation for deep structure use and impacts on patient outcomes. MIS Quarterly, 42(1), 189–212. https://doi.org/10.25300/MISQ/2018/13275. Sharma, R., & Yetton, P. (2007). The contingent effects of training, technical complexity, and task interdependence on successful information systems implementation. MIS Quarterly, 31(2), 219–238. https://doi.org/10.2307/25148789. Stumpf, S. A., Tymon, W. G., & van Dam, N. H. M. (2013). Felt and behavioral engagement in workgroups of professionals. Journal of Vocational Behavior, 83(3), 255–264. https://doi.org/10.1016/j.jvb.2013.05.006. Sung, H.-Y., & Hwang, G.-J. (2013). A collaborative game-based learning approach to improving students’ learning performance in science courses. Computers & Education, 63, 43–51. https://doi.org/10.1016/j.compedu.2012.11.019. Sun, Y., Ni, L., Zhao, Y., Shen, X.-L., & Wang, N. (2019). Understanding students’ engagement in MOOCs: An integration of self-determination theory and theory of relationship quality. British Journal of Educational Technology, 50(6), 3156–3174. https://doi.org/10.1111/bjet.12724. Van de Ven, A. H., Delbecq, A. L., & Koenig, R. (1976). Determinants of coordination modes within organizations. American Sociological Review, 41(2), 322–338. https://doi.org/10.2307/2094477. Wang, W., Guo, L., He, L., & Wu, Y. J. (2019). Effects of social-interactive engagement on the dropout ratio in online learning: Insights from MOOC. Behaviour & Information Technology, 38(6), 621–636. https://doi.org/10.1080/0144929x.2018.1549595. Wise, A. F., Cui, Y., Jin, W., & Vytasek, J. (2017). Mining for gold: Identifying content-related MOOC discussion threads across domains through linguistic modeling. Internet and Higher Education, 32, 11–28. https://doi.org/10.1016/j.iheduc.2016.08.001. Xie, Z. (2019). Bridging MOOC education and information sciences: Empirical studies. IEEE Access, 7, 74206–74216. https://doi.org/10.1109/ ACCESS.2019.2921009. Zhao, Y., Wang, A., & Sun, Y. (2020). Technological environment, virtual experience, and MOOC continuance: A stimulus–organism–response perspective. Computers & Education, 144. https://doi.org/10.1016/j.compedu.2019.103721. Zhu, M. N., Sari, A., & Lee, M. M. (2018). A systematic review of research methods and topics of the empirical MOOC literature (2014-2016). Internet and Higher Education, 37, 31–39. https://doi.org/10.1016/j.iheduc.2018.01.002. 11