This article has Open Peer Review reports available.
How the study of online collaborative learning can guide teachers and predict students’ performance in a medical course
© The Author(s). 2018
Received: 2 December 2016
Accepted: 17 January 2018
Published: 6 February 2018
Collaborative learning facilitates reflection, diversifies understanding and stimulates skills of critical and higher-order thinking. Although the benefits of collaborative learning have long been recognized, it is still rarely studied by social network analysis (SNA) in medical education, and the relationship of parameters that can be obtained via SNA with students’ performance remains largely unknown. The aim of this work was to assess the potential of SNA for studying online collaborative clinical case discussions in a medical course and to find out which activities correlate with better performance and help predict final grade or explain variance in performance.
Interaction data were extracted from the learning management system (LMS) forum module of the Surgery course in Qassim University, College of Medicine. The data were analyzed using social network analysis. The analysis included visual as well as a statistical analysis. Correlation with students’ performance was calculated, and automatic linear regression was used to predict students’ performance.
By using social network analysis, we were able to analyze a large number of interactions in online collaborative discussions and gain an overall insight of the course social structure, track the knowledge flow and the interaction patterns, as well as identify the active participants and the prominent discussion moderators. When augmented with calculated network parameters, SNA offered an accurate view of the course network, each user’s position, and level of connectedness. Results from correlation coefficients, linear regression, and logistic regression indicated that a student’s position and role in information relay in online case discussions, combined with the strength of that student’s network (social capital), can be used as predictors of performance in relevant settings.
By using social network analysis, researchers can analyze the social structure of an online course and reveal important information about students’ and teachers’ interactions that can be valuable in guiding teachers, improve students’ engagement, and contribute to learning analytics insights.
Over the past few decades, the use of technology-enhanced learning (TEL) has become increasingly ubiquitous in the health education sector. TEL has the power to transcend the boundaries of space and time, offering convenience, efficiency, and cost-effectiveness. It also facilitates networked learning by means of computer-supported collaborative learning (CSCL)—features that have been demonstrated to positively enhance learning when coupled with properly designed resources .
The benefits of collaborative learning have for long been recognized. It may support learners, diversify understanding among students and educators, and provide a stage for cooperation in a positive atmosphere that fosters the building of learning communities . Well implemented collaborative learning can facilitate knowledge construction, and encourage involvement and motivation of learners in the learning process [3–8].
Social constructivists see humans as social creatures who grow up by developing knowledge and skills through interactions with different communities. They assume that learning is a social byproduct of conversation and negotiation with peers and that learners acquire knowledge by participating in relevant social activities or working collaboratively in groups [9, 10]. Connectivism—a theory developed to address learning in the digital age—asserts that knowledge and learning exist in a multiplicity of viewpoints and that learning as a process occurs by connecting sources of information. It values the role of communication and information appraisal skills in the development of learning and staying up-to-date [8, 11].
When implemented in stimulating learning environments using authentic activities, interaction with other students may help learners make deeper and more meaningful knowledge construction. It becomes particularly effective when learners are encouraged to respond to arguments, negotiate concepts, debate points of views, contribute to ideas, and share insights and alternative perspectives to the discussed topics [3, 4, 12–17].
An online asynchronous discussion board (forum) is a tool for CSCL that offers the opportunity for students to interact and cooperate in online communities . Forums establish a platform for dialogue among peers and educators that facilitates reflection and exchange of ideas. In that way, learners can build on ideas posted by their peers and learn collaboratively [17, 19–21].
Forums have become a standard feature of all the key LMSs. However, the built-in analytics dashboards of most major LMSs offer limited insight into studying interactions among students. Those limited insights are in the form of statistics and frequency of participation. For instance, the Moodle™, and Blackboard™ LMSs report only the number of views and posts by each participant in each forum, whilst lacking the capabilities for studying the patterns of interactions and the structure of communication. Such features might be better analyzed through visualization and social network analysis, which can be obtained only through external applications or plugins .
Social network analysis
Social network analysis (SNA) is a distinct type of analytics that can be used to map relations and interactions between actors within groups in participatory environments [22–26]. Advocates of social network analysis emphasize the role of the social structure, one’s position in the community, and one’s relations and interactions as important factors that shape one’s behavior and performance, in addition to attributes such as age, gender, and disposition . SNA has been used across a wide variety of disciplines. For example, in criminology, SNA has been used to study collaboration between offenders, patterns of criminal behavior, and gang rivalry ; in management, SNA has helped assess organizational communication hierarchies, the flow of information, and the decision-making process ; and in medicine, SNA has facilitated the study of how infectious diseases propagate  and the exploration of connections in human gene networks . In academia, SNA methods are well established in the study of scientific collaboration, co-authorship, and citation analysis . Despite this widespread use in various fields, the use of SNA in education is still limited [23, 25, 30]. The analysis of social networks is commonly performed through two types of methods: network visualization and quantitative analysis.
A social network is represented by what is commonly known as a “sociogram” or a graph. A sociogram is a graphical mapping of relations and interactions between actors in a network. Each actor—a student or a teacher in the learning context—is denoted by a node in the graph, and a relationship or interaction between actors is denoted by an edge . Sociograms can be directed (where each interaction is mapped from one node to another) or undirected (where there is no certain direction of the interaction) .
Using SNA to visualize learning networks, educators can have an easy-to-understand bird eye overview of all discussions and social interactions in a course , active and inactive students [32, 33], instructor role and learning design , the flow of information, and efficiency of group work .
Monitoring online interactions can reveal patterns that are amenable to meaningful intervention. An instructor who uses SNA to monitor a discussion thread can have an outline of interactions and might be able to stimulate an inactive discussion by promoting a collaborative dialogue [36, 37]. Isolated students who are at risk of poor performance can be identified [38, 39], and facilitated through inclusive online environments, well-designed collaboration scripts, improving networking skills, increased social capital, and rewarding collaborative group learning [35, 40–43]. In a similar way, instructor dominated networks can be identified. These networks are characterized by instructor-centered interactions, few student-to-student interactions, and low level of knowledge construction [23, 33]. Intervention might help promote interactions among students and encourage teachers to facilitate rather than dominate. Using SNA can assist in determining those non-participatory patterns and possibly monitor the intervention [23, 25, 43]. Insights generated by SNA in a course can also support course redesign in subsequent iterations .
Network quantitative analysis
Network quantitative analysis is a mathematical method to quantify the connectedness and relations of actors in a network. Centrality is the construct used to indicate how prominent a particular node is in a network or how important is that node to the communication of information . Interest in using centrality measures as predictors of student achievement or group performance has arisen with the emergence of learning analytics as a discipline [37, 45]. Romero et al.  used in-degree (total number of incoming interactions) and degree centrality (total number of incoming and outgoing interactions) measures to predict students’ performance and found that most students who passed the course had obtained high centrality scores. Hommes et al.  found a positive association between degree centrality and student learning; that association was stronger than previous grades, social integration, and academic motivation. Gašević et al.  found closeness centrality (how reachable or close a student is to his peers) to positively correlate with higher grades. Ángel et al.  found a correlation between centrality measures and performance in some courses and negative in others. They concluded that these mixed results do not suggest that SNA predictors are useless, but rather call for further research to identify the context in which SNA centrality measures might work as reliable predictors. In a trial to assess the role of network position in predicting performance, Joksimović et al.  found weighted degree (total number of interactions, taking into account the quality or strength of interactions) centrality to be the most significant factor. Similar to Ángel et al. , Joksimović et al.  attributed the differences between their findings and those of others to the context in which the study was based.
Peer to peer interactions are important in promoting engagement and enhancing the learning process by enabling the student to establish functional relationships, construct meaning, and understand concepts through discourse and reflection [14, 36]. Increasing interactivity among peers has been reported to promote higher achievement . Thus, the study of social interactions in online collaborative setting may be of potential value in predicting academic performance. Factors such as social capital, prominence, and students’ roles as information-brokers might also add to the available indicators and help obtain more accurate predictive learning analytics in relevant contexts [13, 41, 49].
The quality of collaboration and interactions in a course have been found to affect the whole course learning environment as well as students’ performance [8, 45]. Using visual analytics has the potential to further enhance our understanding of the status and dynamics of online collaborative clinical discussions, ensure that the learning environments are collaborative and engaging for learners, as well as allow us to identify the factors that enhance participatory behavior or situation where intervention is needed [3, 4, 12, 22, 50].
Although social network analysis techniques have been used for a broad range of disciplines and purposes , the field’s applications are still rare in education in general, and medical education in particular [5, 23, 51]. The aim of this work is to assess the potential of visual and statistical social network analysis to study clinical case discussions and to evaluate how the course’s social structure and user parameters might help predict student performance.
The research questions of this study were
What information can the study of the social structure provide about the status of online collaborative learning on the course, discussion, and individual levels?
Which network parameters correlate best with students’ performance?
How can student’s position, interactions, and relations in a network be used to predict his or her final performance?
This study was designed as a case study that applied social network analysis on students’ interactions within a blended medical course. The course was a Surgery Course of the second term 2015 at Qassim University, College of Medicine, Saudi Arabia. This was a term-long clinical course that used clinical case discussion (clinical case scenarios or patient problems) as a teaching strategy to enhance clinical competence and engage students . The discussions used the Moodle learning management system (LMS) forum module as a platform for interaction, and they were moderated by the instructor. The cases were designed to help achieve the objectives and intended learning outcomes of the course, which were: using clinical reasoning to understand the patient’s problem to reach a diagnosis and be able to create an appropriate management plan, as well as improve one’s capability to work efficiently in a team and use information and communication technology for learning and evidence-based practice of medicine.
Data collection and analysis of this study followed the data mining method of Romero et al. , which is divided into four steps:
Structured Query Language SQL was used to extract interaction data from the Moodle LMS database and export it to a table (spreadsheet). The data extracted were user ID, forum ID, parent forum, author of the post (source), author of the reply (target), time created, time modified, subject, post content, and student group. Data also included students’ final grades and midterm grades.
The data were cleaned by removing corrupted records (two records were deleted due to missing the target of interaction), participants’ names were anonymized, students were coded as S1 to S35, and data were converted to a format compatible with the SNA application “Gephi” for import . Two datasets were created: the first covered the whole course duration and the second covered the first midterm.
Data mining and analysis
Data were visualized and analyzed using Gephi 0.91 software. Gephi is an open source SNA application that can be used for network visualization and analysis . It has multiple algorithms for network visualization, of which Forced Atlas 2 was used. It uses a force directed layout that draws each node based on its relations and connections with other nodes. As such, structurally related nodes are rendered closed to each other. This technique allows a better visualization and interpretation of the overall network structure. Gephi’s main advantage over other applications is its “dynamic mode,” which enables researchers to visualize network evolution and time of events and reflects changes in node position in real-time .
Students’ final grades were used to measure final course performance, and midterm grades were used to measure performance up to the midterm. Objective Structured Clinical Exam grades were used to measure clinical performance, and multiple-choice questions (MCQs) were used to measure knowledge comprehension, analysis, and application. Students were classified either as underachievers who are at risk of failing (lowest 1/3) or achievers who are relatively safe from failing (top 2/3).
Since network data are known to violate the traditional assumptions of conventional statistics (normal distribution and independence) , we chose Kendall’s Tau-b test to measure the correlation coefficient between ranked variables. The test was performed using permutation methods by PAST (Paleontological statistics software package for education and data analysis); the permutation test was based on 9999 random replicates. SPSS software version 24 was used to perform automatic linear regression (ALM) and binary logistic regression. ALM offers improvement in key areas, namely better variable selection, handling of extreme values (outliers), as well as merging of similar predictors and conducting ensemble methods . SPSS was also used to perform a stepwise backward binary logistic regression model for the prediction of underachievers.
Network quantitative analysis
Network size: the number of nodes.
Average degree: the mean degree centrality of all group members. The average degree is a quantification of the average level of interactivity of participants .
Network density: the ratio of actual interactions between peers to the total possible; in contrast to the simple quantification provided by the average degree or the network size, network density is a relative indicator that increases as more members participate, and thus points to the diversity of participation, collaborative behavior, and group cohesiveness .
Average clustering coefficient: the average clustering coefficient of the group members; an indicator of the tendency of group members to interact together .
Because there are different criteria for node importance or prominence, centrality can be calculated in various ways depending on the context . In this study, we calculated centrality measures that reflect three groups of constructs:
The quantity of interactions
In-degree centrality is the number of interactions an actor receives, and it is considered an indicator of popularity or prestige. Students who have high in-degree centrality usually have influence or prestige .
Out-degree centrality is a measure of outgoing interactions from an actor. It is a quantification of the interactions a student makes and is an indication of how active the student is in the network .
Role in moderating and relay of information
Betweenness centrality is a measure of the actor’s involvement in moderating interactions; it is measured by counting the times a participant comes “in-between” others. By doing so, the participant connects the unconnected peers and thus facilitates communications and acts as a bridge or broker of information exchange [44, 58].
Information centrality is a measure of the importance of a node in information flow and network cohesion. Information centrality of an actor is defined as the relative drop of network communication efficiency if this actor was removed. A student with high information centrality often has a prominent role in information exchange and communications .
Closeness centrality is a measure of how close an actor is to the other collaborators in a network. It is calculated as the inverse of distance between the participant and all other peers in the networks. Close actors can quickly interact with others and are easy to reach [44, 57].
Eigenvector centrality measures the importance of an actor taking into consideration how well connected the neighbors of the actor are. Connections to well-connected or important actors in the network translate to higher values of Eigenvector centrality . Eigenvector centrality is one of the methods used to estimate the social capital and the influence of one’s ego network.
Eccentricity measures how far an actor is from other actors in the network and can be viewed as an indication of isolation. Students with high eccentricity scores are expected to be less connected to others in the network and difficult to reach .
Clustering coefficient measures the overall tendency of a student to work with peers in the group; it is calculated as the proportion of actual edges between a node and its neighbor peers to the total possible edges that can be achieved [26, 44].
Interpretation and evaluation of the results
The results were analyzed using two different methods:
Visualization of students’ interactions and interaction patterns on three levels:
Course level: to have an overall view of the status of collaborative learning in a course, information-giving, and information-receiving networks, and the role of the teacher and students in the discussions.
Discussion thread level (discussion thread is a group of interactions under the same topic): to have an outline of interactions in individual discussion threads.
Learner level (ego networks): to map the social profile of students.
Network analysis: extraction and interpretation of network metrics and how they are linked to performance.
Research question 1
What information can the study of the social structure provide about the status of online collaborative learning on the course, discussion, and individual level?
The course included 35 students with one instructor (36 nodes). The data were drawn from 34 discussion threads, and the data set totaled 1251 interactions (forum posts). These interactions were visually and mathematically analyzed on course level, discussion thread level, and individual learner levels as follows:
At the course level
Interpretation of social networks depends on the context and the design of the course where it occurs , and the interactions can be mapped in different ways. One approach is to visualize the interactions to represent participants’ roles in terms of quantity and influence in order to provide a general idea about the course structure and participants’ roles.
The graph in Fig. 2 shows the instructor (represented by the node T1) with most edges pointing to him, which indicates that he is receiving most interactions (high in-degree). Actors who receive most interactions act as leaders in a network (highest prestige or authority). The instructor also has the largest node size (highest degree), and his node is the darkest (highest betweenness centrality). This pattern where the instructor is the most influential and the target of most interactions is recognized as instructor-centered. Since the design of this course relies on interactions usually started by the instructor, it was intended that students reply to the instructor when trying to solve the clinical cases, so the graph is well aligned with the instructional design of this course.
The presence of interconnections among students is a sign of a considerable amount of debate and interactions among students trying to establish their cognitive and social presence [14, 16]. It is also apparent that some students (e.g., S3 and S21) have high degree centrality represented by larger node sizes (an indication of higher activity). However, S3 has a more influential role with high betweenness centrality (dark color), and S3 is important to the flow of information across the network. There are two outliers (S34 and S35) with small node size and few connections, which indicate low activity and isolation.
Dynamic network changes over time can give insights about the course interactions at various points of time. Additional file 1: Video S1 shows how the community forms during the first week (network formation stage). From the second week until around the mid-course, students are actively engaged and responding to the instructor (engagement stage). The engagement slows down just before the mid-term exam for a brief period of time, but after the exam, the engagement resumes, and this time the network interactions are more mature than before, as more interactions are occurring among students. During the final week, students are expectedly disengaged. The video also clearly shows individual students’ interactions in real-time throughout the course.
At discussion thread level
Similar to visualizing interactions at the course level, visualizing individual discussion threads reveals information about the principal actors and patterns in individual discussion threads. The importance of profiling an individual discussion is that it is the building unit of the course network. Diagnosing gaps or pitfalls in dysfunctional online group collaboration starts at the discussion thread level, and if an intervention would take place, it usually happens at a discussion level.
At the individual level (ego network)
Shows a summary of the user parameters. (For variable definitions see section on Review of Social Network Analysis: Network analysis)
Quantity of participation
Role in information relay
Closeness centrality a
Eigenvector centrality a
Research question 2
Which network parameters correlate best with students’ performance?
We investigated how much each student’s participation, interactions, and network parameters were correlated with the student’s academic performance. Two methods were used: Kendall’s Tau-b correlation between the SNA parameters and students’ grades, and automatic linear regression to see how much social network parameters could predict final grades or account for grade variance.
Correlation between centrality measures and students’ performance
Quantity of participation
Research question 3
How can student’s position, interactions, and relations in a network be used to predict his or her final performance?
Predicting performance using SNA predictors by the end of course
3.1 Final Grade (Model accuracy 41.6%)
3.2 Clinical results (Model accuracy 51.9%)
3.3 MCQ (Model accuracy 21.3%)
Early participation (midterm)
Predicting performance using SNA predictors that were calculated at midterm
4.1 Midterm results (Model accuracy 71.6%)
4.2 Final Grade (Model accuracy 70%)
Can social network analysis contribute to predicting underachievers (at-risk)?
Cross-tabulation of predicted and low achievers, bold numbers denote correctly classified
Using BLR, we were able to successfully classify 85.7% of students (91.7% of high achievers, 72.7% of underachievers) A chi-square test of independence was performed to examine the relation between actual and predicted at-risk students (chi-square = 15.3, p < .001, df = 1) (Cox & Snell R Square = 0.52, Nagelkerke R Square = 0.73, Hosmer and Lemeshow Test = 0.66). Both in-degree centrality (P = 0.02) and previous grade (P = 0.01) were statistically significant predictors.
Due to the novelty of social network analysis as a field, education-oriented SNA research has been very limited, and it has been mostly exploratory by nature. This has prompted for more exploration in varying disciplines using different methods [23, 45]. While most previous research concentrated on engineering or business education, this study is one of the first to study an online medical course using social network analysis [18, 19, 26, 32, 33, 45].
Using SNA visual analytics for analyzing learning can offer valuable insights on three levels: the course level, the discussion level, and the individual student level. On the course level, the mapped interactions identified gaps and pitfalls in the collaborative learning process, such as a dominating teacher, a relatively non-participatory network, and few interactions among students. This information offers opportunities for meaningful interventions to improve the status of collaborative learning in this course. Interventions can aim to raise awareness of the importance of participation , use relevant and flexible collaboration scripts, and train students to develop better social skills and communication practices [7, 33, 35, 39, 41, 42]. Teachers can help by scaffolding, supporting inclusive and supportive environments, stimulating interactive dialogues, and offering authentic problems that motivate argumentations and debate [9, 17, 35, 39, 42]. The course can be re-visualized using SNA to assess how effective the intervention was. Research has shown that teacher intervention might be necessary for certain small group situations, such as with dominating students or dysfunctional communications [7, 35]. On the thread level, monitoring individual discussions by means of SNA can help inform instructors about when to intervene: in the example we demonstrated, the non-participatory discussion was a candidate for such intervention.
On the individual level, the social capital, ego network, and the sphere of influence are of particular importance for students’ learning. Research has shown that students’ social capital is correlated with better academic performance [5, 41, 46, 47]. Using SNA to map the social profile of students can help identify an isolated student such as the inactive student in Fig. 6. Helping an isolated student improve his social skills might help promote his achievement [7, 37, 38].
The timeline of events (Additional file 1: Video S1) showed that the development of the online community was initiated by the instructor, students reacted enthusiastically along the course, and stopped around exam times. What is interesting was the instrumental role of the teacher presence that kept students engaged most of the course time, in contrast to the irregular and spiky patterns reported in non-moderated environments [14, 15, 60]. The sustained engagement during the course might have been the best advantage of having a moderator [14, 15, 20].
Analysis of network parameters can broaden our understanding of the various centrality measures in education and how they correlate with performance [13, 22, 37, 45]. Among centrality measures that reflect the quantity of participation, degree and out-degree centrality measures showed no correlation to performance, but in-degree centrality was significantly and moderately correlated with grades, as were centrality measures reflecting a role in information relay (betweenness, closeness, and information centrality). In-degree was also the only significant predictor of underachievement using logistic regression after controlling for age and previous performance. This is an indication that what mattered was how students were able to establish their cognitive presence as judged by their peers to be worth discussing [14, 15]. According to Chi et al. , interactions may be considered beneficial only when they generate knowledge that is beyond the presented learning materials and beyond other peer’s contributions, which deserves peers’ replies and discussion. Results also showed that the parameters reflecting group cohesion (high clustering coefficient and low eccentricity) were positively correlated with better performance.
Eigenvector centrality—a measure for the strength of neighbors (ego network) and social capital—were positively and significantly correlated with student final grades. MCQ and clinical performance emphasized the significance of personal networks and were in agreement with previous studies that underscored the value of social networks to academic performance [5, 41, 46, 47]. Regarding the prediction of the final grade, information and Eigenvector centralities were the most important predictors of final grade; MCQ and Eigenvector centrality accounted for 40% of clinical result predictability. Interestingly, early participation was more predictive of performance, with a model accuracy of 70%. It was apparent from the regression results that the weight of centrality measures varied according to the time it was measured.
Results from correlation coefficients and linear and logistic regressions indicate that a student’s position and role in information relay in online case discussions, combined with the strength of that student’s network (social capital), can be used as indicators of performance especially in relevant settings [37, 41, 46, 47]. This finding highlights a rarely studied potential of SNA parameters as early indicators of performance. The incorporation of SNA parameters in multi-dimensional learning analytics models can improve the predictive power of the models designed to identify students at risk of under-achievement in courses that makes use of online collaborative learning [45, 49].
These results are in accordance with similar studies in the field, although predictors vary from a study to another depending on the context and course design [37, 45, 48]. Nonetheless, measures of social capital showed agreement between our results and others [5, 41, 46, 47]. We concur with the previous studies in their belief that context and design play a major role. We believe that the significance of each indicator should be related to the framework of the time it was recorded, the setting of the online collaborative learning, and the social structure of the course (which is best understood by means of SNA). In this study, out-degree centrality was an early indicator of good performance. As the course advanced, in-degree centrality was more significant, as it represented an indirect vote of the quality of the contribution that garnered peers’ responses. In a course where the instructor was prominent (as demonstrated in the visual analytics), a prominent student was likely to be a successful one. The optimum use of SNA in evaluating online collaborative learning should not separate centrality measures from visual analytics, but rather combine them to better understand the context and interpret the inferences of each indicator.
The strength of SNA as a tool lies in the breadth of information it offers, which is relatively quick to produce and easy to interpret. Their speed and breadth of automatic analysis options stand in contrast to traditional content analysis methods, which require lengthy coding and manual analysis that make it impractical to implement it beyond research settings .
These results have many implications; firstly, for LMS designers to create ways that incorporate SNA in their systems so that teachers can monitor the social aspect of their courses; secondly, for administrators to harness the ever-growing field of learning analytics; thirdly, for teachers to seek training on new ways to monitor their work and that of their students to improve the collaborative work they offer; and lastly, for students who might benefit from SNA based intervention adaptive scaffolding.
In this research, we have shown ways to understand the dynamics of interactions in a clinical course in a medical college. This study is limited to a single course, which narrows the generalizability of the results. However, a broader understanding of complex collaborative environments requires studying the phenomena in a variety of contexts. Another limitation of this study is that we used SQL queries to extract interaction data from the database. This method was used due to two reasons. First, the LMS lacked built-in methods to extract the information at a sufficient detail. Second, SNA is still only a research endeavor, which has not found wide-scale adoption on the commercial level. Our study is a step into a new and growing field, which can only grow by extending research to new and unexplored areas.
By using social network analysis, we were able to analyze a large number of interactions and discussions, gain an overall insight of the course social structure, and track knowledge flow and transfer. Mapping the networks of information giving and information receiving uncovered how information flows in the course, and identified important mediators in each network. The information centrality plot clearly demonstrated the influential actors in the network. When augmented by the calculated network parameters, it offered a precise view of the whole course network, each user’s position, level of connectedness, and participation. We were also able to estimate how much these interactions explained variance in learners’ grades, and what can improved in the course design.
Such insights are not possible using traditional methods, which only count hits or replies but ignore the importance of structure, relations, and interactions. This study demonstrates how SNA can expose invisible sides of online collaborative learning and how much social interaction affects learning.
Because this study is based on a case study of a medical education course, its generalizability is to be tested in terms of results and methodology. Future research should also focus on harnessing the power of SNA to improve learning through, for example, monitoring and intervention, early prediction of outliers, and improving social skills.
The authors would like to acknowledge the support given by Prof Hani Al-Shobaili and Prof Abdullah Alghasham Qassim College of Medicine for the endless support and help given during the concept and preparation of this manuscript.
There was no funding source to report for this research.
Availability of data and materials
MS: Made substantial contributions to conception, design, acquisition of data, analysis, and interpretation of data and drafting of the manuscript. Moreover, agree to be accountable for all aspects of the work. UF and MT: Made substantial contributions to the design of the study, analysis of the data, drafting and revising it critically for important intellectual content; they also have given final approval of the version to be published and agree to be accountable for all aspects of the work.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Liu Q, Peng W, Zhang F, Hu R, Li Y, Yan W. The effectiveness of blended learning in health professions: systematic review and meta-analysis. J Med Internet Res. 2016;18(1):e2.View ArticleGoogle Scholar
- Laal M, Ghodsi SM. Benefits of collaborative learning. Procedia Soc Behav Sci. 2012;31:486–90.View ArticleGoogle Scholar
- Jonassen D, Davidson M, Collins M, Campbell J, Haag BB. Constructivism and computer-mediated communication in distance education. Am J Dist Educ. 1995;9(2):7–26.View ArticleGoogle Scholar
- Rivera G, Cox AM. A practice-based approach to understanding participation in online communities. J Comput-Mediat Commun. 2016;21(1):17–32.View ArticleGoogle Scholar
- DeClute J, Ladyshewsky R. Enhancing clinical competence using a collaborative clinical education model. Phys Ther. 1993;73(10):683–9.View ArticleGoogle Scholar
- Schellens T, van Keer H, Valcke M, de Wever B. Learning in asynchronous discussion groups: a multilevel approach to study the influence of student, group and task characteristics. Behav Inform Technol. 2007;26(1):55–71.View ArticleGoogle Scholar
- Johnson DW, Johnson RT. Social interdependence theory and cooperative learning: the Teacher's role. In: Gillies RM, Ashman AF, Terwel J, editors. The Teacher’s role in implementing cooperative learning in the classroom. Boston: Springer US; 2008. p. 9–37.View ArticleGoogle Scholar
- Woo Y, Reeves TC. Meaningful interaction in web-based learning: a social constructivist interpretation. Internet High Educ. 2007;10(1):15–25.View ArticleGoogle Scholar
- Liu CH, Matthews R. Vygotsky's philosophy: constructivism and its criticisms examined. Int Educ J. 2005;6(3):386–99.Google Scholar
- Siemens G. Learning and knowing in networks: changing roles for educators and designers. ITFORUM Discuss. 2008;27:1–26.Google Scholar
- Siemens G. Connectivism: a learning theory for the digital age. Int J Instr Technol Dist Learning. 2005;2(1):3–10.Google Scholar
- Lapadat JC. Written interaction: a key component in online learning. J Comput-Mediat Commun. 2002;7(4):0.View ArticleGoogle Scholar
- Cho H, Gay G, Davidson B, Ingraffea A. Social networks, communication styles, and learning performance in a CSCL community. Comput Educ. 2007;49(2):309–29.View ArticleGoogle Scholar
- Garrison DR, Arbaugh JB. Researching the community of inquiry framework: review, issues, and future directions. Internet High Educ. 2007;10(3):157–72.View ArticleGoogle Scholar
- Shea P, Hayes S, Vickers J, Gozza-Cohen M, Uzuner S, Mehta R, Valchova A, Rangan P. A re-examination of the community of inquiry framework: social network and content analysis. Internet High Educ. 2010;13(1–2):10–21.View ArticleGoogle Scholar
- Joksimović S, Gašević D, Kovanović V, Riecke BE, Hatala M. Social presence in online discussions as a process predictor of academic performance. J Comput Assist Learn. 2015;31(6):638–54.View ArticleGoogle Scholar
- Hew KF, Cheung WS, Ng CSL. Student contribution in asynchronous online discussion: a review of the research and empirical exploration. Instr Sci. 2009;38(6):571–606.View ArticleGoogle Scholar
- Pena-Shaff JB, Nicholls C. Analyzing student interactions and meaning construction in computer bulletin board discussions. Comput Educ. 2004;42(3):243–65.View ArticleGoogle Scholar
- Yusof N, Rahman AA. Students' interactions in online asynchronous discussion forum: a social network analysis. In Education Technology and Computer, 2009. ICETC'09. Int Conf. Singapore: IEEE; 2009. p. 25–29.Google Scholar
- Garrison D, Anderson T, Archer W. Critical inquiry in a text-based environment: computer conferencing in higher education. Internet High Educ. 2000;2-3(2–3):87–105.Google Scholar
- Conde MÁ, Hérnandez-García ÁJ, García-Peñalvo F, Séin-Echaluce ML. Exploring Student Interactions: learning analytics tools for student tracking. In: Zaphiris P, Ioannou A, editors. Learning Collab Technol. Springer International Publishing; 2015. p. 50–61.Google Scholar
- Borgatti SP, Mehra A, Brass DJ, Labianca G. Network analysis in the social sciences. Science. 2009;323(5916):892–5.View ArticleGoogle Scholar
- Cela KL, Sicilia MÁ, Sánchez S. Social network analysis in e-learning environments: a preliminary systematic review. Educ Psychol Rev. 2015;27(1):219–46.View ArticleGoogle Scholar
- Martínez-López B, Perez A, Sánchez-Vizcaíno J. Social network analysis. Review of general concepts and use in preventive veterinary medicine. Transbound Emerg Dis. 2009;56(4):109–20.View ArticleGoogle Scholar
- Sie RL, Ullmann TD, Rajagopal K, Cela K, Rijpkema MB, Sloep PB. Social network analysis for technology-enhanced learning: review and future directions. Int J Technol Enhanc Learn. 2012;4(3/4):172–90.View ArticleGoogle Scholar
- Grunspan DZ, Wiggins BL, Goodreau SM. Understanding classrooms through social network analysis: a primer for social network analysis in education research. Cell Biol Educ. 2014;13(2):167–78.View ArticleGoogle Scholar
- Carolan BV. The social network perspective and educational research introduction. In: Social network analysis and education: theory, Methods & Applications. Thousand Oaks: SAGE Publications, Inc; 2014. p. 3–22.View ArticleGoogle Scholar
- Barabasi AL, Gulbahce N, Loscalzo J. Network medicine: a network-based approach to human disease. Nat Rev Genet. 2011;12(1):56–68.View ArticleGoogle Scholar
- Newman MEJ. The structure of scientific collaboration networks. Proc Natl Acad Sci. 2001;98(2):404–9.View ArticleGoogle Scholar
- Cambridge D, Perez-Lopez K. First steps towards a social learning analytics for online communities of practice for educators. In: Proceedings of the 2nd international conference on learning analytics and knowledge - LAK ‘12; 2012. p. 69–72.View ArticleGoogle Scholar
- Bakharia A, Dawson S. SNAPP: a Bird’s-eye view of temporal participant interaction. In: Proceedings of the 1st international conference on learning analytics and knowledge - LAK ‘11; 2011. p. 168.View ArticleGoogle Scholar
- Rabbany R, Takaffoli M, Zaïane O. Analyzing participation of students in online courses using social network analysis techniques. Eindhoven: Edm: 2011; 2011. p. 21–30.Google Scholar
- Lockyer L, Heathcote E, Dawson S. Informing pedagogical action: aligning learning analytics with learning design. Am Behav Sci. 2013;57(10):1439–59.View ArticleGoogle Scholar
- Dawson S, Tan JPL, McWilliam E. Measuring creative potential: using social network analysis to monitor a learners’ creative capacity. Australas J Educ Technol. 2011;27(6):924–42.View ArticleGoogle Scholar
- Webb NM. The teacher's role in promoting collaborative dialogue in the classroom. Br J Educ Psychol. 2009;79(1):1–28.View ArticleGoogle Scholar
- Abrami PC, Bernard RM, Bures EM, Borokhovski E, Tamim RM. Interaction in distance education and online learning: using evidence and theory to improve practice. J Comput High Educ. 2011;23(2–3):82–103.View ArticleGoogle Scholar
- Romero C, López M-I, Luna J-M, Ventura S. Predicting students’ final performance from participation in on-line discussion forums. Comput Educ. 2013;68:458–72.View ArticleGoogle Scholar
- Rabbany R, Elatia S, Takaffoli M, Zaïane OR. Collaborative learning of students in online discussion forums: a social network analysis perspective. In: Peña-Ayala A, editor. Educational data mining: applications and trends. Cham: Springer International Publishing; 2014. p. 441–66.View ArticleGoogle Scholar
- de Janasz SC, Forret ML. Learning the art of networking: a critical skill for enhancing social capital and career success. J Manag Educ. 2007;32(5):629–50.View ArticleGoogle Scholar
- Janssen J, Bodemer D. Coordinated computer-supported collaborative learning: awareness and awareness tools. Educ Psychol. 2013;48(1):40–55.View ArticleGoogle Scholar
- Rizzuto TE, LeDoux J, Hatala JP. It’s not just what you know, it’s who you know: testing a model of the relative importance of social networks to academic performance. Soc Psychol Educ. 2008;12(2):175–89.View ArticleGoogle Scholar
- Fischer F, Kollar I, Stegmann K, Wecker C. Toward a script theory of guidance in computer-supported collaborative learning. Educ Psychol. 2013;48(1):56–66.View ArticleGoogle Scholar
- Saqr M, Alghasham H, Alghasham A, Kamal H. The study of online clinical case discussions with the means of social network analysis and data mining techniques. Milan: AMEE; 2014.Google Scholar
- Golbeck J. Chapter 3 - network structure and measures. In: Analyzing the social web. Boston: Morgan Kaufmann; 2013. p. 25–44.View ArticleGoogle Scholar
- Hernández-García Á, González-González I, Jiménez-Zarco AI, Chaparro-Peláez J. Applying social learning analytics to message boards in online distance learning: a case study. Comput Hum Behav. 2015;47:68–80.View ArticleGoogle Scholar
- Hommes J, Rienties B, de Grave W, Bos G, Schuwirth L, Scherpbier A. Visualising the invisible: a network approach to reveal the informal social side of student learning. Adv Health Sci Educ. 2012;17(5):743–57.View ArticleGoogle Scholar
- Gasevic D, Zouaq A, Janzen R. “Choose your classmates, your GPA is at stake!”: the Association of Cross-Class Social Ties and Academic Performance. Am Behav Sci. 2013;57(10):1460–79.View ArticleGoogle Scholar
- Joksimović S, Manataki A, Gašević D, Dawson S, Kovanović V, de Kereki IF. Translating network position into performance. In: Proceedings of the sixth international conference on Learning Analytics & Knowledge - LAK ‘16: 2016. New York: ACM Press; 2016. p. 314–23.View ArticleGoogle Scholar
- Saqr M, Fors U, Tedre M. How learning analytics can early predict under-achieving students in a blended medical education course. Med Teach. 2017;39(7):1–11.View ArticleGoogle Scholar
- Whatley J, Bell F. Discussion across borders: benefits for collaborative learning. Edu Med Int. 2003;40(1-2):139–52.Google Scholar
- Ferguson R, Shum SB. Social learning analytics. In: Proceedings of the 2nd international conference on learning analytics and knowledge - LAK ‘12; 2012. p. 23.View ArticleGoogle Scholar
- Romero C, Ventura S, García E. Data mining in course management systems: Moodle case study and tutorial. Comput Educ. 2008;51(1):368–84.View ArticleGoogle Scholar
- Heymann S, Grand BL. Visual Analysis of Complex Networks for Business Intelligence with Gephi. 2013 17th Int Conf Inf Vis. London; 2013. p. 307–12.Google Scholar
- Bastian M, Heymann S, Jacomy M. Gephi: an open source software for exploring and manipulating networks. ICWSM. 2009;8:361–2.Google Scholar
- Isba R, Woolf K, Hanneman R. Social network analysis in medical education. Med Educ. 2017;51(1):81–8.View ArticleGoogle Scholar
- Yang H. The case for being automatic: introducing the automatic LINEAR modeling (LINEAR) procedure in SPSS statistics. Mult Linear Regression Viewpoints. 2013;39(2):27–37.Google Scholar
- Rochat Y. Closeness centrality extended to unconnected graphs: the harmonic centrality index. Zürich: Asna; 2009. p. 1–14.Google Scholar
- Salter-Townshend M, White A, Gollini I, Murphy TB. Review of statistical network analysis: models, algorithms, and software. Stat Anal Data Min. 2012;5(4):243–64.View ArticleGoogle Scholar
- Latora V, Marchiori M. A measure of centrality based on network efficiency. New J Phys. 2007;9(6):188.View ArticleGoogle Scholar
- Panzarasa P, Kujawski B, Hammond EJ, Roberts CM. Temporal patterns and dynamics of e-learning usage in medical education. Educ Technol Res Dev. 2016;64(1):13–35.View ArticleGoogle Scholar
- Chi MTH, Wylie R. The ICAP framework: linking cognitive engagement to active learning outcomes. Educ Psychol. 2014;49(4):219–43.View ArticleGoogle Scholar
- De Wever B, Schellens T, Valcke M, Van Keer H. Content analysis schemes to analyze transcripts of online asynchronous discussion groups: a review. Comput Educ. 2006;46(1):6–28.View ArticleGoogle Scholar