Keywords

1 Introduction

Massive open online courses (MOOCs) are educational provisions that enable access to professional development and lifelong learning for millions of individuals across the globe. Instead of formal credentials, MOOCs provide free or low-cost structured opportunities to learn for self-fulfillment or upskilling. Such unaccredited learning opportunities are intentional from the learner perspective, and structured through the learning objectives, course time and support. The context and characteristics of MOOCs are similar to other non-formal educational offerings, that is non-accredited courses whereby “learners opt to acquire further knowledge or skill by studying voluntarily with a teacher who assists their self-determined interests, by using an organised curriculum, as is the case in many adult education courses and workshops” [1].

The non-formal nature of MOOCs influences the social context generated through forum participation. By “social context,” we refer to interpersonal peer-to-peer and peer-to-instructor interactions characterised by a relational quality such as emotional attachment. By operationalising interpersonal interactions as “social context,” we emphasise that they are a part of the group communication history and include personal understanding of this communication. In other words, a social context is comprised from interpersonal interactions but qualitatively reaches beyond simple information exchange.

In the context of formal education, where a bounded group of individuals interact continuously from day 1 of the course, interpersonal interactions and the emergent social context are tightly coupled. That is, the set-up of a formal course creates the necessary structure and boundaries for learners and teachers to interact in discussion forums, or obligatory assignments and tutorials. As a result, for formal education offerings, interpersonal interactions can readily evolve beyond simple information exchanges to shared understanding and history to enable social support, deeper learning, and motivation at later stages. In MOOCs, however, such emergent quality of interpersonal interactions cannot be assumed alongside aggregated patterns of interpersonal interactions. In MOOCs, the group boundaries are ill-defined, and participation in social interactions is intermittent at best. Given these differences, not all interpersonal interactions contribute to the quality of the emergent social context. Furthermore, analysis of the social context in non-formal educational settings needs to reflect the peculiarities of the MOOC context, rather than replicate approaches previously applied in formal education.

Current study examined the social context in four edX MOOC forums. Instead of including all forum interactions, analysis was limited to the interactions between individuals who regularly posted on the forums as the course progressed. The networks of regular MOOC posters were modelled through exponential random graph modelling (ERGM) using the network patterns of direct, indirect and triadic-level reciprocity. Results suggested that reciprocity patterns differentiated the interaction structures in the four analysed courses. The study also highlighted reciprocity patterns associated with the teaching staff that allows hypothesising their role in the network formation. Suggested network indicators of reciprocity can be used to describe MOOC forums and compare networks of aggregated interactions between the learners by the end of the course.

2 Structure of MOOC Networks

This section explains how reciprocity network patterns can model the structure of MOOC regular posters. We review the literature on (1) conceptualization of informal knowledge-exchange online communities; (2) reciprocity, with the focus on indirect reciprocity as reflecting altruistic motivation; and (3) network structures in both Electronic Networks of Practice (ENP) and MOOC forums. Through the literature review, we argue that patterns of direct reciprocity (knowledge exchange), indirect reciprocity (altruistic contributions) and triadic-level reciprocity (amplification of network flow) can describe different structures of MOOC forum networks.

2.1 Electronic Networks of Practice (ENP)

The formation of informal online communities has been previously explained through individuals’ altruistic motivation to contribute to the collective benefit [2]. Faraj, Wasko and Johnson examined the so-called electronic network of practice defining this as “a self-organizing, open activity system that focuses on a shared interest or practice and exists primarily through computer-mediated communication” [3]. Wasko and Faraj [2] conceptualised ENP as sustained by voluntary participation and knowledge exchange. Participants of these informal online communities shared knowledge about particular interests that comprised group purpose.

In addition to the attachment to the group’s purpose and knowledge exchange around a shared practice, such communities have been associated with altruistic behaviour. That is, in online communities based on shared purpose or interest, people seek information by relying on the “kindness of strangers” [4]. Specifically, Constant et al. [4] found that the employees of a global computer manufacturer participating in an ENP gave useful advice and solutions to the problems of others despite a lack of personal connections. Some 81% of the providers did not know the seekers, and 10% were barely acquainted. Hence, there was no direct benefit derived for a poster providing a solution. Altruistic motivation can explain why individuals make such contributions. In effect, the solutions are offered for the sake of the community and a wider public good. Similarly, Faraj and Johnson [5] suggested that the information exchanges within an ENP have a social, that is group-oriented, motivation. The authors contended that interactions among the participants were more than just information queries, and provided a social purpose. Their contributions were socially oriented towards a broader community benefit. Simply put, individuals participating in an ENP were aware that the larger group would benefit from their reply to an individual information-seeker. They provided the information in ways accessible to others. The underlying altruistic motives were also observed in a range of contexts demonstrating a group-oriented attitude among dominant motivations of ENP members.

2.2 Interpreting Network Patterns Through Reciprocity Construct

Previous work in network science allows representing altruistic behaviours as local patterns of a network. In particular, altruistic behaviour can be captured through the network forms of reciprocity. For instance, the altruistic “pay-it-forward” behavior, that is providing a public good without an expected return, is one of the underlying forms of reciprocity. Other reciprocity types differ in the extent of altruism, that is as to when an action is expected to be reciprocated, if at all. For instance, when both parties negotiate the terms of a bilateral exchange as in “if you scratch my back, I’ll scratch yours”, such reciprocity is considered “negotiated,” and a return of the service is expected [6]. According to the collective action theory, such an exchange indicates a non-altruistic motivation of self-interest [7]. When the direct reciprocation of one’s action could be delayed, or not offered at all, Molm [8] described such a form of exchange as unilateral. Unilateral reciprocity implies that although the possibility that the other person may reciprocate exists, there is a level of uncertainty as to when or if this may occur. Indirect reciprocity, a type of unilateral reciprocity also referred to as generalised exchange [9], is represented by the network form where individual “A may give to B on one occasion and to C on a different occasion”. Indirect reciprocity can also be represented by the following patterns “A gives to B, B gives to C” as well as “A gives to B, B gives to C, and C gives to A” [6]. Examples of such forms of collective exchange can be found, for instance, in voluntary blood-giving activity or academic peer review practices.

The network configurations associated with indirect reciprocity are more than just structural representations. The forms of reciprocity carry meaning about the socio-emotional quality of the exchange. One such quality is the underpinning trust that characterises the forms of unilateral reciprocity. In a unilateral exchange when a service is reciprocated directly or indirectly, the actor provides material or symbolic goods without knowing “whether, when, or to what extent the other will reciprocate” [6, 8, 10]. Since an individual is unsure about the return of the service, an increased possibility of risk is another characteristic of a unilateral exchange (direct and indirect reciprocation). Due to such embedded risk, such reciprocal forms are also associated with a higher emotional value. By reciprocating without an expectation of return, individuals demonstrate their readiness to pay forward a service and show a willingness to continue the relationship. According to Molm [8], the forms of a unilateral exchange (direct and indirect reciprocity) take place in settings where conflict is of low salience. That is, if conflict is minimal, individuals are more likely to offer services unilaterally. Molm et al. [6] further validated that “paying it forward” to other group members, represented by the indirect reciprocity forms, occurred in contexts with higher levels of social solidarity, than those described by direct or negotiated reciprocity.

2.3 Network Patterns of Reciprocity in ENP and MOOCs

Prior research identified the presence of reciprocity forms in network structures of ENPs and all-poster MOOC networks. Both direct and indirect reciprocity configurations were found pertinent to informal online identity-based communities. Specifically, in modelling the ENP, Wasko et al. [11] described their structure through the dominant features of direct reciprocity over indirect reciprocity. Similarly, the features of direct reciprocity were found prominent in describing the structure of all-poster MOOC networks [11,12,13]. Joksimovic et al. [14] suggested that besides network forms associated with direct and indirect reciprocity, the all-poster networks can be defined by a simmelian tie form. We suggest that this form is interpreted as triadic-level reciprocity, that is the mutual exchange of interactions at the triadic level (A → B, B → A, B → C, C → B, C → A, A → C), and represents a clique-like clustering in a network, and its gradual amplification.

2.4 Research Questions

Due to the prominence of the features of direct and indirect reciprocity observed in informal online identity-based communities, the study examined the role of reciprocity features in non-formal educational online communities for the network formation through the following research questions:

  1. 1.

    How do network patterns of reciprocity describe the social structure of emergent networks of regular posters in MOOCs?

  2. 2.

    Were course staff, instructor, and assistants embedded in a particular network representation of reciprocity?

3 Methods

To address research questions, we have applied ERGM to four networks of regular posters in MOOCs. To address the second research question, we have removed teaching staff and assistants from the networks and applied ERGMs to the new networks. It is evident that removing the teaching staff does not exclude the effect of their presence on the other ties. As such our interest was to identify the network structures where the teaching staff were situated. This section elaborates on the network construction and the features of ERGMs.

3.1 Network Construction

3.1.1 Defining the Edges

Network ties were non-valued and directed. The direction of the social tie represented the collective orientation of communication flow. That is, posts to the forum were considered as a contribution offered to the collective group engaged in the discussion rather than a reply or discrete message to an individual poster. Consequently, network ties were directed to all the posters in a single thread who preceded the actor based on timestamps. If A posted, B replied, and C commented, and D replied, each of the subsequent actors would have a directed tie to everybody else prior to them. That is, B → A, C → B, C → A, D → C, D → B, D → A. Such representation of interactions through one-to-many ties reflects the principle of collective orientation where an individual contributed information to the posting group (or rather collective discourse within a particular thread), rather than provided a service to an identified individual (as replying to a known person is often not feasible in a MOOC discussion thread).

3.1.2 Defining the Nodes

The purpose of this study was to model the structure of the evolving communities in MOOC forums. Therefore, the inclusion of nodes into the network was guided by the potential of the learner (i.e. node) to develop interpersonal relationships that are typical for communities. As a result, not all forum posters were included as nodes in the constructed networks. In our previous work [15,16,17], we argued that forum interactions in MOOCs are qualitatively different from one another, depending on the regularity of forum posting activity by the learner contributing them. As elaborated in Poquet [15], an average of 2.9% of forum posters have potential to establish interpersonal relationships with one another exclusively through in-forum interactions. These learners establish a familiarity with others in the group due to their continuous/regular posting on the forums. In essence, as these individuals regularly post and receive more timely replies, there is increased opportunity to co-occur with one another over the course duration more so than more irregular and intermittent posters. Therefore, these “regularly posting” learners experience interactions that are effectively embedded within a shared history (in contrast to other posters) that emerges through wide-ranging topics of discussion covering content- and non-content-related issues [15]. We posit that the non-valued edges between the regular posters can be defined by such parameters, whereas the edges between the other posters may not. We refer to this group as the forum core (FC), and assume that due to the quality of the interactions the individuals constituting FC may evolve into a community.

3.1.3 Temporal Boundaries

Network boundaries were set around the course delivery times: between the week the first video lecture was released, and the week when the last lecture in the course was provided. As communication between actors can potentially continue well past a course’s lectures due to exams or assignments, any established closing date for the forum analysis would have been arbitrary. The discrete temporal limits applied in this study made the comparison of courses more feasible.

3.2 Exponential Random Graph Modelling

The network properties describing the structure of forum cores’ networks were used by applying exponential random graph modelling (ERGM). ERGM is a probability model (p*) for network analysis [18] that predicts the presence of the ties and local structures in networks. ERGM estimates the probability of a network G, given the sum of network statistics weighted by parameters inside an exponential [18]. In ERGMs, network structure is understood as emergent from (1) endogenous structural parameters associated with social processes, such as, for example, preferential attachment or reciprocity, and (2) exogeneous parameters, such as individual actor attributes or the impact of interpersonal relations in another network. The network statistics used to reproduce endogenous structural parameters represent social processes expressed in sets of two actors (dyads), three actors (triads) or larger combinations, such as several triads sharing a tie (social circuits). An ERG model estimates the likelihood that each selected configuration, associated with a theoretically formulated social process, will occur beyond chance in a random network generated based on the modelled constraints.

3.3 Modelled Configurations and Effects

3.3.1 Structural Configurations

The study examined the propensity of three structural configurations to form the structure of forum core’s networks, listed along with the correspondent statnet label [19]:

• Direct reciprocity (A → B, B → A; mutual label)

• Indirect reciprocity (A → B, B → C; cyclicalties label)

• Reciprocal exchange at the triad level (A → B, B → A, B → C, C → B, A → C, C → A; simmelianties label)

These network configurations structurally control for the network reciprocity, degree and closure. In our previous work [17], we have modelled MOOC forum core networks without the indirect reciprocity pattern. Although this study demonstrates somewhat similar results, we argue that including indirect reciprocity feature is essential to understanding why the network forms. Moreover, from a network point of view including an indirect reciprocity pattern is a must as it controls for degree distribution within the structural components of ERGM. Table 1 presents the counts of modelled network configurations for each of the networks.

Table 1 Descriptive statistics of structural network configurations
Full size table

3.3.2 Node Attributes

In addition to controlling for the structural parameters of reciprocity, the models also controlled for the effect of learner super-posting activity on the propensity to send and receive ties. The distinct levels of activity across online networks have been previously supported in empirical studies. Studies of Internet communities captured the individuals who represent a large proportion of the overall activity [20]. Similarly, in MOOCs, several studies have shown the presence of hyperposting activity [21, 22]. For this reason, modelling a forum core needed to account for a very active participation level of a select few, and a relatively low level for the overall majority. The nodefactor term was used for modelling the effect of participation level, that is no distinction was made as to whether the learner sent or received ties.

To identify the overall level of activity, instead of relying on the number of messages posted, we used measures of learner activity (1) as devised by Hecking et al. [23]. Hecking and colleagues established in-reach and out-reach metrics of individual activity occurring in a MOOC network [23, 24]. These measures calculate the diversity of outgoing and ingoing relations for a node i, where w(e i,j ) represent the weight of the tie between them, od(i) and id(i) represent out-degree and in-degree, respectively.

$$ {\displaystyle \begin{array}{cc}\mathrm{outreach}(i)& =\mathrm{od}(i)-{\sum}_{j\in \mathrm{outneigh}(i)}w\left({e}_{i,j}\right)\times \log \left(\frac{w\left({e}_{i,j}\right)}{od(i)}\right)\\ {}\mathrm{inreach}(i)& =\mathrm{id}(i)-{\sum}_{j\in \mathrm{inneigh}(i)}w\left({e}_{i,j}\right)\times \log \left(\frac{w\left({e}_{i,j}\right)}{id(i)}\right)\end{array}} $$
(1)

This study replicated this measure in order to identify the in- and out-reach of an individual’s activity across all-posters in MOOC forums. By using k-means clustering of in-reach and out-reach measures per person, all forum posters were divided into three groups: to represent highest, moderate and low forum participation activity in the entire network. The posting activity attributes were used to control for tie propensity formation in forum core ERGMs. Table 2 presents the descriptive statistics of actors within each cluster (counts), along with their average message posting activity. Cluster 1 refers to learners with low posting activity. Cluster 2 refers to the posters with moderate posting activity, and Cluster 3 represents hyperposters.

Table 2 Descriptive statistics of the three clusters representing forum core posting activity levels in forums
Full size table

Labelling of activity levels (Table 2) was relative to the course, and heterogeneous in measures across the courses. That is, a hyperposter in one course may have similar activity numerically as a moderate poster in another course. As shown in Table 2, the Water course had the lowest level of moderate and hyperposting activity (40 posts per person and 218 posts per person, respectively) and FP had rather high level of posting activity for low level and moderate level of participation within its network (20 posts per person, and 125 posts per person, respectively). However, on average, the proportion of activity across the three clusters within the courses appeared to be relatively comparable.

3.3.3 Goodness of Fit

Several steps to ensure goodness of fit were conducted [18, 25]. That is, converged networks with the best fit were checked for degeneracy. For this study’s analysis, R packages statnet and ergm were employed [19, 26]. A Monte Carlo Markov Chain (MCMC) algorithm was used to check for network degeneracy and to examine the goodness of fit. Estimated models demonstrated reasonable goodness of fit (1) as the mean number of networked configurations was similar to configurations in the observed network; (2) the MLE estimation produced non-degenerate networks; and (3) the AIC coefficient demonstrated better fit than the null model (Table~3). Goodness of fit of the modelled network was lower in forum cores with the teaching assistant removed. This was particularly pronounced for the Excel course, where forum core network without staff was also lacking all the nodes defined by hyperposting activity.

Table 3 Outputs for final ERG models in four edX MOOCs
Full size table

4 Results

The first research question inquired if patterns of direct reciprocity associated with information exchange and of indirect reciprocity associated with altruistic motivations can describe networks of regular posters in MOOCs. The networks were modelled through the baseline density of networks, direct reciprocity of ties, indirect reciprocity of ties, and reciprocal exchange at the triadic level. The models also controlled for the propensity of the formation of ties sent and/or received by the posters with different levels of network in-reach and out-reach, reflective of their overall posting activity. The levels were distinguished between individuals with low, moderate and high forum participation behaviour. All models were fitted using the described network configurations.

The estimates for the network parameters of modelled MOOC FC networks are outlined in Table 3. There was a significant and positive effect for direct reciprocity in all four FC networks, namely [Water: 1.27(0.009), Solar: 1.55(0.12), Excel: 2(0.15), FP: 1.02(0.28)], with estimates reported for each course and the standard errors in parentheses. There was a significant and positive effect for indirect reciprocity in all four forum core networks [Water: 0.2 (0.03), Solar: 0.7(0.003), Excel: 0.81(0.006), FP: 1.54(0.24)]. There was also a positive and significant effect for reciprocal exchange at triad level [Water: 0.53 (0.004), Solar: 0.54(0.05), Excel: 0.22(0.05), FP: 0.83(0.13)]. All the estimates for the four models were significant, with a p-value of <0.001.

The results supported the premise that features of reciprocity can reflect the structure of forum core networks, much like in online public electronic groups. The forum cores’ networks in the three analysed courses were more defined by a higher propensity for direct dyadic exchange [Water: 1.27(0.009), Solar: 1.55(0.12), Excel: 2(0.15)] in contrast to a lower propensity over tie formation of indirect exchanges [Water: 0.2 (0.03), Solar: 0.7(0.003), Excel: 0.81(0.006)] or triadic-level exchanges [Water: 0.53 (0.004), Solar: 0.54(0.05), Excel: 0.22(0.05)]. In sum, this reflects the expected network structure observed in informal online communities and all-poster MOOC networks. Such findings generally support the presence of hypothesised social processes of reciprocity in a forum core. In general terms, a lower indirect reciprocity indicates a lower social solidarity, and potentially a lower level of maturity for an identity-based community within a forum core.

However, for the FP course such dynamics were not replicated. The FP course had a higher propensity of indirect exchanges [1.54(0.24)] to form ties than a direct reciprocity configuration [1.02 (0.28)]. Also, FP’s estimate for mutual exchanges at a triadic level was higher than in other courses [Water: 0.53 (0.004), Solar: 0.54(0.05), Excel: 0.22(0.05), FP: 0.83(0.13)]. FP’s higher propensity for network closure can be interpreted as a more evolved network development and FP’s higher level of indirect reciprocity can be interpreted as a network form of social solidarity. The propensity for these two levels of reciprocity to describe the forum core’s structure in the FP MOOC suggests that this social group had a different social dynamics underpinning its network formation.

The second research question inquired into the types of network structures that embedded the teaching staff. Three networks that originally included the staff and teaching assistants were removed, and the networks were re-modelled to highlight the role of the instructors in observed network structures. All additional models also converged on the modelled parameters and were not degenerate. The estimates for the network parameters of modelled MOOC FC networks where staff and teaching assistants were removed are outlined in Table 4. There was a significant and positive effect for direct reciprocity in all four FC networks, namely [Solar: 2.09(0.11), Excel: 2.46(0.12), FP: 1.37(0.18)], with estimates reported for each course and the standard errors in parentheses. There was a significant and positive effect for indirect reciprocity in all four forum core networks [Solar: 0.52(0.004), Excel: 0.6(0.004), FP: 0.57(0.07)]. There was also a positive and significant effect for reciprocal exchange at triad level [Solar: 0.36(0.05), Excel: 0.29(0.06), FP: 0.63(0.08)]. All the estimates in four models were very significant, with a p-value of <0.001. Water FC was not modelled as the network contained no staff teaching assistants that interacted with students regularly, and the forums were unmoderated.

Table 4 ERGM outputs for the remodelled three forum cores, where staff and teaching assistants were removed from the network
Full size table

To aid interpretation, Fig. 1 plots the odds of structural patterns occurring in network structures with teaching staff and without. Figure 1 demonstrates that the teaching staff are largely embedded within indirect reciprocity and triadic-level reciprocity structures. Such results may suggest that the instructor and teaching assistants play a role in promoting altruistic behaviour and clustering of the network leading up to the more evolved social context. Yet, further analyses more fine-tuned to investigating the role of teaching staff are required to draw further conclusions.

Fig. 1
figure 1

Cross-course comparison of the social processes driving the formation of a forum core (FC). The black solid line indicates odds of the network configurations in FC if the staff and teaching assistants were removed

Full size image

5 Discussion

The study examined the structure of four MOOC networks of regular posters using the patterns of reciprocity as indicators of the emergent social context in MOOC forums. We argued that the network configurations can capture the different behaviors that comprise social context in MOOC forums. Specifically, we argued that MOOC networks of regular learners could be described through direct reciprocity to capture dyadic knowledge exchange, indirect reciprocity to capture altruistic behaviour, and network closure comprised of direct and indirect reciprocity. The first research question demonstrated that these reciprocity patterns could describe the structure of MOOC networks of regular posters. Three analysed networks were defined by a higher propensity for direct dyadic exchange, in contrast to a lower propensity over tie formation of indirect exchanges, or triadic-level exchanges. This reflects the expected network structure observed in informal online communities and all-poster MOOC networks. Such dynamics were not replicated in one network that was defined by a higher propensity for network closure and higher level of indirect reciprocity. This network’s structure is indicative of a more evolved social context, that is defined by a high social solidarity indicator, represented by indirect reciprocity configurations, and a high group cohesion indicator, represented by triadic-level closure.

The second research question inquired about the structural patterns that embedded teaching staff. Analysis demonstrated that the teaching staff was embedded within the indirect reciprocity structures and triadic-level reciprocity. However, current analysis did not address the effect that the presence of instructors had on the formation of ties among learners. Further work is needed to determine the extent to which teaching staff is instrumental to the network’s clustering and subsequent amplification.

The social structures observed in FP can be described as the course with the highest level of social context development among the four analysed MOOCs, given the following interpretation of the forms of indirect reciprocity and reciprocal closure. Molm [8] proposed three conditions that describe environments with indirect reciprocity. First, the risk of vulnerability is heightened, that is individuals are vulnerable in offering a service that holds no guarantee of reciprocation. Second, the expressive value of giving is higher, as the act of “paying forward” demonstrates a commitment to the shared practice by contributing without expecting anything back. Third, indirect reciprocity takes place when a conflict is less salient, that is the social tensions do not prevent contributors to “pay forward.” Reciprocal closure, or reciprocation at the triadic level, is more representative of group-like formations that are taking over a network. An amplification of information flow, less control over information within a selected few and a higher integration of participants are aspects associated with such a structure. In sum, the combination of higher indirect reciprocity and triadic reciprocity could be interpreted as social solidarity and group cohesion.

Analyses suggest that there may be a positive relationship between hyperposting activity and direct reciprocity, as well as a negative relationship between hyperposting activity and indirect reciprocity. For the former, the pattern of a high propensity of hyperposters to send and receive ties, high direct exchanges and low triad-level exchanges is observed in three of the four examined courses (Water, Solar and Excel). For the latter, two courses (Water and Excel) with a higher direct reciprocation and higher hyperposting impact on tie formation have low indirect reciprocation. Both relationships can be interpreted as the impact of hyperposters impeding group formation within the networks, despite hyperposters’ effort to stimulate discussions, or help those seeking answers. Higher levels of indirect reciprocity would indicate that learners feel compelled to reciprocate a perceived “favour” they may have received. However, if hyperposters assume responsibility for actively taking on questions, the overall group may not feel compelled to reciprocate to unknown strangers, or develop a common identity to learn and help others learn. In other words, hyperposting individuals may be diminishing the opportunities associated with a “pay it forward” behaviour.

What seems to be a negative relationship between social solidarity and high hyperposting impact does not indicate that moderation is detrimental for an evolving relational quality of interpersonal interactions. In line with prior research and practice of group formation, the role of a teacher, a moderator or a leader is critical at early stages of group formation. In initial course stages, moderation (e.g. Excel, Solar, FP) could stimulate direct reciprocity, as well as indirect reciprocity. The trend of a higher propensity to reciprocate is observed across all moderated courses that were examined in this study, as opposed to the unmoderated one (i.e. Water). However, while a critical mass of moderate posters may be willing to pay effort and actions forward (in their posting activity), they still require opportunities to do so. If a moderator continues addressing all possible queries posted in a MOOC forum, for example in the Excel course, at an incredibly high propensity, affordances for network closure are limited. In other words, as much as moderation is needed at the onset of a course, a moderator’s activity may need to be less involved further along a course’s timeline, in order to allow for reciprocal closure.

Much of the interpretations and their associated implications remain speculative. Although we have highlighted that the network structures of the regular posters in MOOC forums can be defined by the features of reciprocity, the sample examined in the study is rather small. Further analyses are required to both validate the set of reciprocity indicators as telling of MOOC forum social context, as well as better understand the relationships hypothesised in the Discussion section. Moreover, the second research question suggested that teaching staff plays an instrumental role in the formation of the network, while some aspects of the initial analysis demonstrate how teaching staff may impede such formation. Understanding the turns within the network formation and the role of the staff within them, both require more advanced SNA analyses, as well as other methodologies complementing the insights.

6 Conclusions

The study examined if network patterns of reciprocity can define the social structure of MOOC forum posters. Findings support the premise that these network structures can be described by the various reciprocity patterns and, in turn, could be integrated as network-level analytics of evolving social context in MOOC forums. Indicators of direct reciprocity represent dyadic information exchange, indirect reciprocity represents social solidarity and reciprocal closure indicates amplification of a network and a shift to a more egalitarian structure from a centralised one. The study demonstrated that courses where a designated moderator was involved exhibited higher indirect reciprocity than direct reciprocity in network formation. The formation of a forum core network in a highly moderated course was also mostly defined by an amplification of information flow reciprocally, at the triadic level. The study also suggested that hyperposting activity may have prevented networks from clustering, and hence impede a more egalitarian information flow.