Keywords

1 Introduction

Online courses lend themselves well to social constructivist instruction by providing students with opportunities to discuss ideas, work in teams to solve cases, problems, projects, and even assess themselves and their peers, which is part of the reason why online courses are as critical to the long-term strategy of higher education institutions around the world, as face-to-face courses. Furthermore, learning management systems and their user activity tracking and content archiving capabilities allow researchers to study online interaction among students in a relatively inexpensive way, technically speaking.

In this vein, what is the best way to orchestrate discussion forums that foster interaction in an online course? A current conundrum both in undergraduate and graduate online courses is interaction among students that leads to social construction of knowledge. In spite of a myriad of studies related to student-to-student interaction in online discussion forums, there is inadequate literature [1] about the orchestration of discussion forums that foster interaction aimed at generating social construction of knowledge.

Social construction of knowledge is a phenomenon that can be defined as a function of interaction [2], which is understood as a reciprocal influence among individuals that engage in online interaction. Like a patchwork quilt, interaction is the collection of unique messages sewn together, resulting in socially constructed knowledge.

There are three themes in the literature about student-to-student interaction in online discussion forums, namely: (1) studies focused on the process of knowledge construction [3], (2) social networks [4], and (3) a combination of both [5]. However, most studies offer basic explanations of student-to-student interaction or do not provide practical solutions to the orchestration of discussion forums that promote interaction.

In U.S. Higher Education, around one in four students (28%) took at least one online course in 2015. Online students equaled a total of 5,828,826 students, which represented an annual increase of 3.9% compared to the 3.7% rate recorded in 2014. The total of 5.8 million online students included 2.85 million that took all of their courses online and 2.97 million that took some online courses. Public universities have the largest proportion of online students, with 72.7% of all undergraduate and 38.7% of all graduate-level students [6].

The 13th annual report of the state of online learning in U.S. Higher Education, reported even though the proportion of university leaders that say online courses are critical to their long-term strategy fell from 70.8% in 2014 to 63.3% in 2015, the proportion that rate the learning outcomes in online courses as the same or superior to those in face-to-face courses was at 71.4% in 2015. Furthermore, only 29.1% of university leaders reported that their faculty accept the value and legitimacy of online courses, as defined by Allen and Seaman’s (2016) survey, and colleges with the largest online enrollments 60.1% reported faculty acceptance while only 11.6% of the colleges without online enrollments reported so.

In online courses “to have discussion for discussion’s sake is not good instructional design. The discussions within an online distance education course must be well orchestrated to enable the learner to meet the learning outcomes, and build knowledge and insights” [7].

In the past scholars recognized online communication had the potential to represent a new generation of distance education [8] and paved the way for many studies on online and asynchronous group communication. For example, there was a study [9] about questions related to online interaction, particularly, related to the effects of the frequency of interaction, types of students, subject matter, alignment of interaction and learning objectives, and the effects of interaction on student satisfaction.

Later other scholars [10] published what would become a standard textbook on distance education in the USA, in which they devoted a chapter to technologies and media that included a section about learning management systems where they state online instructors “…have found the most valuable feature to be the asynchronous threaded discussion forum in text format. A discussion forum allows students and instructors to interact by posting and reading messages, while each has the flexibility regarding when they do it.”

There is a scholarly reference on online distance education [11] that includes a chapter on interaction in the context of online courses, which presents a revamped version of a seminal idea of modes of interaction, namely: student-teacher interaction, student-to-student interaction, and student-content interaction. This reference states “although interaction among students has been studied most frequently, the various the [sic] forms and combinations of interaction discussed here would benefit from systematic and rigorous research using a variety of research tools and methodologies.”

Thus, it is worth pointing out the following conundrum: what is the best way to orchestrate discussion forums that foster interaction in an online course? This is still a challenge both in undergraduate and graduate online courses as interaction among students may lead to social construction of knowledge.

1.1 Purpose of the Study

This mixed methods research project examined the relationship between social construction of knowledge and student centrality in three online discussion forums, which were part of a graduate online course on web conferencing. The purpose of the study was to identify student-to-student interaction patterns by analyzing discussion forum posts, measuring student centrality, and generating social network diagrams in order to explain characteristics of posts that lead or contribute to social construction of knowledge.

To approach said relationship, the Interaction Analysis Model [2]—commonly referred to as IAM—was utilized to determine if students constructed knowledge through interaction in discussion forums. In addition, SNA [12] was used to measure student centrality in order to account for the social aspect of knowledge construction. Graphing the structure of the social network that emerges from a discussion forum with social network diagrams is a way of “x-raying” interaction patterns with the ultimate purpose of identifying posts that provide potential paths to higher levels of knowledge construction.

This study was aimed at advancing the academic study of social construction of knowledge in online discussion forums previously reported [13,14,15], which demonstrated the adequacy of combining the Interaction Analysis Model and SNA. The relevance of supplementing the Interaction Analysis Model with measures of student centrality and social network diagrams that depict interaction patterns lies on the ability to advance previous studies not only by accounting for the social aspect of knowledge construction in social network terms, but by examining empirical data in Spanish within the Mexican sociocultural context.

Online instructors and instructional designers who develop online courses may find suggestions on the application of social constructivist principles to the design of discussion forums capable of fostering interaction. Also, this study may offer some clarification for university leaders on the alignment of discussion forums as a learning activity with the expected level of social construction of knowledge set by course and/or learning objectives as they relate to substantive and frequent interaction quality standards.

1.2 Research Question

How does social construction of knowledge relate to student centrality in online discussion forums?

2 Online Interaction and Social Network Analysis

There have been some research efforts to study interaction in online discussion forums, as it relates to both construction of knowledge and student centrality, from both a quantitative and qualitative perspective because textual data does not seem to be enough to explain the discussion process in a more visual manner and vice versa. For example, in studies where quantitative results were limited, several researchers [13,14,15,16,17,18,19] conducted mixed methods research to carry out supplemental analyses that explained social construction of knowledge and student centrality.

Taking into consideration the studies above, the main advantage of a mixed methods approach seems to be the ability for researchers to supplement their analysis with two or more perspectives, as opposed to being restricted to analysis techniques typically associated with qualitative research or quantitative research [20].

While the Interaction Analysis Model offers researchers a qualitative research technique that is subjective by nature to examine interaction in online environments (mediated by computer communication), SNA offers researchers different quantitative research techniques that are objective by nature to examine interaction in a variety of environments. Furthermore, the Interaction Analysis Model is an abstract way of outlining the process of social construction of knowledge and it is rooted in a theoretical framework based on social constructivism, on the other hand, SNA is a perspective rooted in sociology and social psychology, both of which focus on relationships or interactions among social entities and their patterns.

In addition, while researchers who use the Interaction Analysis Model argue for a complete post as the unit of analysis, social network analysts have developed unique techniques to analyze relation-based data, so they take a post as the unit of analysis in conjunction with the individual because though a post or an individual can fundamental units of analysis, in SNA they are not primary on their own because it is not theoretically sound to rely on separate units from this perspective, which requires researchers to operationalize concepts relationally.

To reiterate, the unit of analysis in SNA is also the post, but in connection to the student interaction, which occurs between members of the social network [21]. It is worth highlighting the fact interaction itself is the conceptual overlap between the Interaction Analysis Model and SNA that allows researchers to mix the two approaches because even though the attributes of these posts (e.g., the author, the message content) are primary to the first approach, they are secondary to SNA, but from a mixed methods perspective these attributes are key to the interpretation of the interaction patterns that are revealed by SNA.

The Interaction Analysis Model and SNA are similar in that both perspectives can be used to explain interaction and consider it equally relevant to analyze interaction patterns of independent relations as well as the totality of interconnected relations among social entities. Interaction is key in this study, because social construction of knowledge can be defined as a function of online interaction, which requires an information flow that coexists with a social relation among students.

2.1 The Interaction Analysis Model

The Interaction Analysis model [2] was created to examine knowledge construction in an online environment mediated by computer communication. The model’s theoretical framework is based on social constructivist principles, so it considers knowledge construction as a function of interaction. The authors of this model put forward a definition of interaction that considers “the entire gestalt formed by the online communications among the participants” and presented an analogy between knowledge construction and a patchwork quilt as an organized whole with many unique messages sewn together. This definition of interaction is different than other definitions in that it does not focus only on individual relations, but on the totality of interconnected relations that emerge from online communication, so the authors argue for considering an entire message/post as the unit of analysis.

Due to the predominance of discussion forums as a fundamental ingredient for knowledge construction over other types of learning activities in online courses, it is worth listing the phases of the Interaction Analysis Model, which describes in detail five phases of knowledge co-construction, generally described as follows: Phase (I) sharing, comparing, Phase (II) dissonance, Phase (III) negotiation, co-construction, Phase (IV) testing tentative constructions, and Phase (V) agreement, application of new knowledge.

2.2 Social Network Analysis and Centrality

SNA [12] is a perspective that offers researchers both a set of algorithms and analysis techniques, which allows them to develop specific ways to measure phenomena and analyze relation-based data. Relation-based data is paramount in the operationalization of social networks because it is not sound to rely only on analytical techniques that consider separate individuals as primary. Studying phenomena from a network perspective requires that at least one theoretically significant concept be defined relationally e.g., social construction of knowledge—a function of interaction—involves an information flow that coexists with a social relation among students.

Researchers who study phenomena from a network perspective think about what kinds of networks are caused by different activities, such as interaction, which requires mapping sociological concepts onto particular network forms. Thus, when the effects of phenomena on networks is studied, the results are sociologically significant, in addition if something causes a network to be fractured so that there is a lack of relation or interaction between actors, the fracture matters because of the social effects it may have.

Social network analysts decide what kinds of networks and what kinds of relations they will study before collecting data [22]. There are two kinds of networks from which analysts must choose before starting to delimit the boundaries of their studies, namely: whole vs ego networks, and one-mode vs two-mode networks. Whole networks take a bird’s-eye view of social structure, focusing on all actors rather than any particular one. These networks begin from a list of actors and include data on the presence or absence of relations between every pair of actors, for example, the network that emerges from students who interact in an online discussion forum. In contrast, ego networks focus on the network surrounding one actor, known as the ego.

Analysts use the whole networks approach to explain characteristics of social networks such as density, the average path length necessary to connect pairs of nodes, the average tie strength, the extent to which the network is dominated by one central actor (centralization) or the extent to which the network is composed of similar nodes (homogeneity) or of nodes with particular characteristics (composition), such as the proportion of network members who are women [23].

Most of the time, researchers who examine whole networks collect data on a single type of actor in networks where every actor could conceivably be connected to any other actor, therefore most of the networks they examine are one-mode networks. In contrast, two-mode networks, also referred to as affiliation networks, involve relations based on co-membership. In addition, researchers have to choose how to measure relations after selecting the kinds of networks they want to study and defining a theoretically significant concept relationally, and this choice is between directed or undirected and binary or valued relations [22]. Directed relations go from one actor to another and may be reciprocated, while undirected relations exist between actors in no particular direction. Both directed and undirected relations can be measured as binary relations that either exist or not within each pair of actors, or as valued relations that can be stronger or weaker.

Centrality. Only In SNA there is a group of metrics known as centrality measures, which quantify the relevance or influence of an individual in a social network based on her relations with other individuals. Central individuals or “actors are those that are extensively involved in relationships with other actors. This involvement makes them more visible to the others” [12], thus what is appealing for researchers studying interaction in online discussion forums is the relationship of students with higher centrality and social construction of knowledge.

With regards to social relations in online discussion forums, the question of who talks to who has important implications for information flow, so it is relevant to analyze interaction patterns of both independent relations and the totality of interconnected relations. Thus, student centrality is a concept that accounts for the social aspect of knowledge construction in that it serves as an indicator of student influence on other students. As I explained in my problem statement, the centrality of different individuals in a social network that emerges from a discussion forum can be analyzed with centrality measures and social network diagrams that depict interaction patterns.

From the SNA perspective, actors (also known as nodes) and their actions are viewed as interdependent rather than independent autonomous units, so the actors in this study will be students. Second, relational ties (linkages also known as arcs or edges) between students are interaction channels for transfer or “flow” of information through posts in discussion forums. Third, social network diagrams can represent patterns of interaction among students. Fourth, each student interacts with other students, each of whom interacts with a few, some, or many others, and so on. Therefore, the concept of social network refers to the finite set of students and their interactions in one discussion forum.

While centrality measures quantify the relevance or influence of an individual in a social network, there is a holistic measure of a social network that takes into consideration the totality of interactions named density [24], which defines “density, d, of a network” is the number of ties (interactions) in the network divided by the possible by number of ties (interactions). Thus, a well-connected social network—with high density—is one where everybody interacts with everybody else, enabling the flow of information in the presence of key students with high centrality (more influential), also known as “information brokers”.

Social network diagrams provide visual representations of interaction in discussion forums that would otherwise be hidden to researchers, online instructors, instructional designers, and even students themselves, as demonstrated by some researchers [5, 25,26,27] who have used SNA to produce social network diagrams as a way of “x-raying” or mapping interaction patterns of online discussion forums to illustrate social construction of knowledge.

3 Methodology

3.1 Research Design

This was a sequential mixed methods study to examine interaction patterns of graduate students that participated in three discussion forums, which were part of an online course on web conferencing in a learning technologies master’s degree at a Mexican university. The first stage of the analysis involved the application of Interaction Analysis Model to transcripts of discussion forums to find occurrences of social construction of knowledge by identifying qualitative characteristics of posts published by students. Next, centrality measures such as number of posts, in-degree, out-degree, and betweenness were taken to derive the degree of student centrality using SNA. Then, a comparison and contrasting of results from both methods was done, highlighting occurrences of social construction of knowledge of students with high centrality, and looking for the nature of the relationship between social construction of knowledge and student centrality.

3.2 Participants

Twenty-one graduate students between the age of 23 and 65 generated a dataset of discussion forum posts and gender was equally represented. This web-based secondary dataset of three online discussion forums contained de-identified authors, title, date, time, and posts extracted from a graduate course on web conferencing, which was part of a master’s in learning technologies in a large public university in western Mexico. The discussion forums were deployed through the Moodle LMS, there was one discussion forum in the beginning, other in the middle, and another by the end of Spring 2015. These three forums were archived when the semester concluded in the university’s LMS. The main inclusion criterion for this study was graduate students should have participated in discussion forums of the selected online course. There was not any sensitive information to be removed from any discussion transcript that could have compromised the identity of a student.

Due to the modular structure of the online course students were expected to study the content and participate in learning activities frequently as they had deadlines, so student-to-student interaction occurred primarily as required participation in discussion forums. At the beginning of the online course, students were studying factual information that introduced them to the subject, then as the course progressed gradually towards more analytical learning activities students were expected to engage in thought provoking discussions, and by the end of the course students worked in small groups preparing to host an educational web conference as a final project.

3.3 Unit of Analysis

The identification of the unit of analysis had to be reliable and encompass the phenomenon under study, so the post was chosen as the unit of analysis because it is objectively identifiable, meaning multiple coders can agree consistently on the total number of units; it produces a clearly delimited set of observations; and it has parameters determined by the author of the post. This choice addressed the lack of uniformity in the choice of the unit of analysis and inadequacies in reliability found in the literature. In addition, by concentrating on the post as the unit of analysis it was possible to report the intercoder reliability level in a straightforward fashion because coders did not need to argue about what a post is, as it is clearly defined by its author. Furthermore, the Interaction Analysis Model argues for a complete post as a unit of analysis.

In the application of the Interaction Analysis Model to examine transcripts of discussions, a post is taken as the unit of analysis and coded for as many occurrences or phases of social construction of knowledge as it contains, as opposed to mutually exclusive categories utilized in content analysis. When conducting SNA, a post can also be taken as the unit of analysis, but the post has to be taken in conjunction with the student because this perspective requires a relational concept such as the concept of interaction. Thus, the post in conjunction with the student become an actor (node) that may be connected to other students who interacted with each other in a discussion forum.

3.4 Data Collection and Analysis

The Interaction Analysis Model required discussion forum transcripts be extracted from a web-based secondary dataset archived in Moodle and exported both as PDF files and web archives, which offer great readability to human coders working with PDF readers or web browsers. Also, PDF files and web archives allow human coders to keep color highlights, annotations, and comments, keeping data safe in password protected computers with encrypted hard/flash drives. For example, posts were copied from said PDF files or web archives and pasted on a coding spreadsheet in order to have the text in the first column and then code with 1 or 0, as a way to improve precision.

SNA required network data, which had to be derived from the coding spreadsheets of the three discussion forums and processed using Microsoft Excel with NodeXL, a SNA plug-in [28], which facilitated entering posts as actors (nodes) with the actor labels being pseudonyms of students, and graph interaction(s) as edges or arcs. For example, if student A replies to student B, a directed edge (depicted as an arrow) was graphed from A to B. Directed edges were added with labels containing posting sequence number as well as the Interaction Analysis Phase of the post. NodeXL was also used to calculate the centrality measures and produce a social network diagram of interaction patterns. In the context of discussion forums, it is valuable to look at social network diagrams that show different interaction patterns and reveal student centrality. The preliminary step to generate these diagrams was to obtain the centrality measures of each student that published a post or replied to another student.

Interaction and social network analyses were conducted first to determine occurrences of social construction of knowledge and student centrality respectively. Then, results from both analyses were compared and contrasted, which involved the use of diagrams of interaction patterns in discussion forums that illustrate the centrality of different students. Finally, to explain the characteristics of the posts published by students with higher centrality, post excerpts were identified as textual evidence to complete the mixed methods design.

4 Results

4.1 Occurrence of Social Construction of Knowledge

Two coders who used the Interaction Analysis Model determined knowledge construction did occur through student-to-student interaction. To reiterate, the model’s analysis procedure consists of reading every post from a discussion transcript and assigning them one or more codes for the purpose of identifying different phases of social construction of knowledge. When two coders use this type of model to code it becomes necessary to report the intercoder reliability level, which can be calculated using the percentage of agreement or Holsti’s method [29]. In general, a Holsti’s percent agreement higher than 90% or 0.90 is considered to be a high level of intercoder reliability and a percent agreement lower than 80% or 0.80 is considered doubtfully reliable [30].

As shown in Table 1 two coders concurred in determining a total of 24 occurrences of knowledge construction in forum 1 with a level of intercorder reliability of 87%.

Table 1. Occurrence of social construction of knowledge in forum 1.
Full size table

Table 2 shows how the coders concurred in determining a total of 36 occurrences of knowledge construction in forum 2 with a level of intercorder reliability of 86%

Table 2. Occurrence of social construction of knowledge in forum 2.
Full size table

Table 3 shows coders concurred in determining a total of 33 occurrences of knowledge construction in forum 3 with a level of intercorder reliability of 70%,

Table 3. Occurrence of social construction of knowledge in forum 3.
Full size table

4.2 The Relationship Between Social Construction of Knowledge and Student Centrality

Student centrality accounts for interaction dynamics in the sense it can be a measure of the influence of a student in the social network that emerges from a forum due to an information flow that co-exists with a social relationship among students. In other words, students with high centrality have the quality of being at the core of the discussion as they are key to information flow.

Table 4 shows that the most prestigious student in forum 1 and the one with more potential access to information was S21, who reached a maximum phase of I by sharing/comparing information. The most influential student was S07, who reached a maximum phase of III by negotiating meaning/co-constructing knowledge. Although S02 did not have prestige or influence, she/he also reached a maximum phase of III. S03 and S17 did not have prestige or influence either, in addition to having the same potential access to information as everybody else, with the exception of S21, and they reached a maximum phase of I.

Table 4. Students with higher centrality in forum 1.
Full size table

Table 5 shows in forum 2 no student was more prestigious or influential than others. S03 had more potential access to information, followed by S16, the first student reached a maximum phase of III by negotiating meaning/co-constructing knowledge while the latter a maximum phase of II experiencing dissonance. S02 and S12 had the same potential access to information, which was less than that of S03 and S16, nevertheless S02 reached a maximum phase of III while S12 only shared/compared information reaching only phase I.

Table 5. Students with higher centrality in forum 2.
Full size table

Table 6 shows in forum 3 no student was more prestigious or influential than others. S15 had more potential access to information and reached a maximum phase of IV by testing tentative constructions of knowledge. S04, S17 and S21 had the same potential access to information, which was less than S15, still S04 reached a maximum phase of IV, S17 a maximum phase of IV, but S21 a maximum phase of II by experiencing dissonance.

Table 6. Students with higher centrality in forum 3.
Full size table

In the social network diagrams shown in Figs. 1, 2, and 3, edges are depicted with arrows showing the direction of information flow. Edges are labeled with an Arabic numeral indicating the sequence of the post and a roman number in parenthesis indicating the Interaction Analysis Model’s phase reached by the student with that post.

Fig. 1.
figure 1

Social network diagram of interaction patterns in forum 1

Full size image
Fig. 2.
figure 2

Social network diagram of interaction patterns in forum 2

Full size image
Fig. 3.
figure 3

Social network diagram of interaction patterns in forum 3

Full size image

5 Findings

Interaction in online discussion forums was the main focus of this study as it is a construct that emerges from the conceptual overlap between the Interaction Analysis Model and SNA because interaction involves an information flow that coexists with a social relation among students. This conceptual overlap allowed the authors of this paper to mix both methods in the sense SNA supplemented the Interaction Analysis Model by accounting for the social aspect of knowledge construction with evidence of the basic generation of knowledge arising in and out of interaction within social networks of students that result from online discussion forums.

The model allowed two coders to measure occurrences of social construction of knowledge in three different discussion forums. The model’s phases or coding categories were simple for coders to apply them to the discussion forum transcripts, which confirms the model’s flexibility that has been appealing to many researchers who needed to identify the characteristics of posts that contribute or lead to higher levels of social construction of knowledge in discussion forums.

Knowledge construction occured through student-to-student interaction as evidenced by aggregate results of the three forums, which show a clear pattern of social construction of knowledge. There were a total of 93 occurrences in all forums spread out across forums as follows: 24 in forum 1, 36 in forum 2, and 33 in forum 3.

Almost half of occurrences in all forums, i.e., 46 out of 93 (49.46%) reached phase I, which is the lowest level where students share or compare information. Around one fifth of occurrences in all forums, 18 out of 93 (19.35%) reached phase II, which is a low level where students experience the discovery and exploration of dissonance or inconsistency among ideas, concepts or statements. Around one fifth of occurrences in all forums, 20 out of 93 (21.50%) reached phase III, which was arbitrarily set in this study as the standard to determine posts as higher level occurrences of social construction of knowledge because it is the one where students experience negotiation of meaning or co-construction of knowledge. Nine occurrences in all forums (9.67%) reached phase IV, where students experience the testing of tentative constructions of knowledge. There were not any occurrences that reached phase V, which is the highest level, where students experience agreement or application of new knowledge.

With regards to the trustworthiness of the Interaction Analysis Model’s results from forum 3, an intercoder reliability level of PA = 70% can be considered doubtfully reliable based on a conservative interpretation of Holsti’s percent agreement [30] to address the historical criticism to Holsti’s method of being too liberal, statistically speaking. It is worth mentioning what other researchers who used the same model along with Holsti’s percent agreement have reported, for instance some have reported a PA = 70% [31], while others [3] have reported a PA = 78% implicitly adhering to a more liberal interpretation of this statistic.

This insight into the results confirms a meta-analysis on literature about the Interaction Analysis Model [1], which reported its results were “quite similar to the results obtained in the original study [2]: there are low levels of complex thinking as the majority of operations coded remained in PhI. There is some evidence of operations in PhII and III, but they are almost non-existent in PhIV and V.”

With regards to the sociocultural context, a previous study [32] attributed its finding of students making a leap from lower phases of social construction of knowledge to higher phases, without passing through intermediate phases to a lack of open disagreement, i.e., dissonance (phase II) was not evident in the data as open disagreement with ideas expressed by others might not to be appropriate or at least not a necessary element in the Mexican sociocultural context. The results of this study are inconclusive on this aspect because on the one hand previous findings [32] can be confirmed in forum 1, but not in the other two forums, where dissonance (phase II) accounted for a fifth (19.35%) of occurrences in all forums.

The centrality of different students in a social network that emerges from a discussion forum can be analyzed with centrality measures and social network diagrams that depict interaction patterns. Student centrality proved to be a concept that accounts for the social aspect of knowledge construction in that it serves as an indicator of student prestige and influence on other students as well as the degree of potential access to information as it flows through the discussion forum.

In-degree, out-degree, and betweenness account for student overall degree of centrality. The in-degree measure counts inbound posts with other students while out-degree counts outbound posts. These measures, when considered separately, are indicators of network “prestige” (in-degree) and influence (out-degree). “Prestige” results from the number of replies directed to a student’s post and represents the degree to which other students seek out that student for interaction, thus students with high in-degree are notable because their information may be considered more important than others in the discussion forum. In contrast, students with high influence are in contact with many other students, as evidenced by the large number of discussion posts that they send to others, therefore students with low influence post fewer messages and do not contribute with information flow as much as other students. Betweenness reflects an individual’s potential access to information as it flows through the network. The characteristics of a social network that emerges from an online discussion forum can be explained in terms of centrality measures obtained through SNA. Although the post is the most fundamental input required to take SNA measures, is not a centrality measure per se, but it is expected to be accounted for in centrality measures tables associated with online discussion forums as observed in the review of literature.

The characteristics of posts published by students with higher centrality in a given forum, can be explained in terms of social construction of knowledge by combining the Interaction Analysis Model and SNA. Social construction of knowledge involves phases such as sharing/comparing of information, dissonance, negotiation/co-construction of knowledge, testing tentative constructions of knowledge, and agreement/application of new knowledge, all of which require an information flow that coexists with a social relation among students. This information flow can in turn be explained with SNA in centrality measures terms, which reveal student centrality.

Whether higher student centrality contributes to a higher level of social construction of knowledge or not, is a question that can also be addressed by combining the Interaction Analysis Model and SNA as social network diagrams provide supplemental visuals related to the way information flows through the network that emerges from a discussion forum. One possible solution to establish what constitutes a higher level of social construction of knowledge is to use the Interaction Analysis Model to set the bar at phase III to determine posts at or above that phase as higher level occurrences of social construction of knowledge. In short centrality measures provide sound indicators of a student’s ability to transfer information and exert influence over other students.

5.1 Limitations

One of the limitations of this study was the lack of access to other sources of data, which limited the scope of the study to the analysis of pre-existing and de-identified transcripts only. Even though this lack of access did not make it possible to ask follow up questions to any of the 21 participants or triangulate information, the research design did not require access to participants, the graduate program, or other type of documents, so in a way it would be appropriate to say it was a specialized analysis that required purposeful sampling, as opposed to a large number of participants, but the tradeoff was that as a specialized analysis it yielded a very specific answer to the research problem, like an “x-ray” that shows visual details, but still a visual from one angle and a given point in time.

Results should not be generalized because statistical tests were not conducted in this study due to the sample size of 21 participants. However, results should have a degree of transferability to similar contexts and settings. The sample size was the result of purposeful sampling, the rationale behind purposeful sampling was to select a set of participants that represented a typical case and it was not intended to make generalizations, which requires random sampling or selecting a large number of participants, as typically found in quantitative studies. In contrast, sample sizes are typically smaller in qualitative research, but sample sizes that are too small cannot adequately support claims of having achieved valid conclusions and sample sizes that are too large do not permit the deep, naturalistic, and inductive analysis that defines qualitative inquiry. Therefore, a sample size of 21 participants was a sound number the addressed the need of the researchers to reach middle ground through mixed methods.

Principles that guide SNA also limited the scope of the study in the sense the authors had to make certain assumptions to explain social construction of knowledge in social network terms. Again, the authors looked at relational data such as a social relation-information flow, not attributes of people such as age or income, in that, the social network approach was used to examine networks within a group of people not the group of people as a whole, which made it possible to make sense of people’s centrality within networks, but not of people’s centrality within the group. For example there were three social networks within the selected group of graduate students because there were three discussion forums in the dataset, therefore students may have had different centrality across the three forums and it is not appropriate to attribute an overall measure of centrality within the group.

Furthermore, relations were examined in a relational context, meaning the authors examined interaction patterns of a social network, not just relations between pairs or triads, which made it possible to account for the broader patterns of ties within the network to address the totality of interconnected relations that emerge from online interaction in a discussion forum. This strategy limited the study in that social networks had to be operationalized in a very specific way—carefully selected from a myriad of possibilities available to researchers—that addressed the phenomenon appropriately vis-à-vis the Interaction Analysis Model. Social networks were operationalized by focusing on whole networks, as opposed to ego networks, and on one-mode data, as opposed to two-mode data, and on directed ties. Other ways to operationalize social networks fall outside the boundaries of this study.

There was an abundance of literature on social construction of knowledge associated with the Interaction Analysis Model, SNA, and mixed methods about discussion forums carried out in English in undergraduate online courses from developed countries, but there was a scarcity of prior research reports on the same topic in connection to discussion forums in graduate courses from a different sociocultural context. This scarcity prompted the authors to limit the study to a graduate online course on web conferencing in Spanish within the Mexican sociocultural context.

The intercoder reliability level of forum 3, which can be considered doubtfully reliable with a PA = 70%, is not to be confused with negative results, but researchers are advised against adopting a more liberal position on the interpretation of this percent agreement statistic.

The aforementioned limitations matter in the sense that they point to the need for researchers to move forward aiming to address the relationship of social construction of knowledge and student centrality by taking into consideration the future research ideas.

5.2 Future Research

Future research should further investigate the association between social presence and the higher levels of knowledge construction according to the Interaction Analysis Model. Furthermore learning analytics can assist qualitative researchers in the application of techniques like data scraping, statistics, programming, and visualization to qualitative data, particularly when guided by models such as the Interaction Analysis Model to produce more robust findings.

5.3 Conclusion

“To have discussion for discussion’s sake is not good instructional design. The discussions within an online distance education course must be well orchestrated to enable the learner to meet the learning outcomes, and build knowledge and insights” [7]. So, what is the best way to orchestrate discussion forums that foster interaction, which in turn can lead to social construction of knowledge. Researchers, online instructors, instructional designers and university leaders, need to gain insight into the orchestration of discussion forums to inform their decisions as they relate to online course offerings and substantive and frequent interaction quality standards.

The objective of this study was to identify student-to-student interaction patterns by analyzing discussion forum posts, measuring student centrality, and generating social network diagrams in order to explain characteristics of posts that lead or contribute to social construction of knowledge. The objective was met as different interaction patterns were identified and explained both in social construction of knowledge and social network terms. Furthermore, it was possible to explain the nature of the relationship between social construction of knowledge and student centrality and support this explanation with diagrams and information that put forward the idea of paths to higher levels of knowledge construction.

It is clear that interaction patterns in discussion forums have important implications for information flow, but the fact of the matter is, student-to-student interaction is important not because of the amount of posts, its frequency or timeliness, but because of its intent and form, which can be explained in terms of social construction of knowledge.

Having established a positive relationship between student centrality and the occurrence of higher levels of social construction of knowledge the authors put forward the notion that social interaction is as important as individual knowledge construction in a discussion forum, therefore there should not be trade-off between quantity of interaction and quality of information in student lead discussion forums, which suggests the balance of interaction lies on the proper alignment of student learning outcomes, specific learning objectives, materials, learning activities, but most important on providing students with explicit information such as grading rubrics, examples of posts, and other resources designed to set interaction expectations before students post as well as to make the social construction of knowledge explicit in a debate oriented forum, otherwise “student’s won’t know what they don’t know,” and knowledge construction will remain an hidden esoteric goal that only exists in abstract form in the online instructor’s mind.

This study contributed to new knowledge about social construction of knowledge by explaining its relationship with student centrality and the same time it advanced the academic study of social construction of knowledge in online discussion forums previously reported [13,14,15] not only by accounting for the social aspect of knowledge construction in social network terms, but by examining data from a graduate level online course’s discussion forums carried out in Spanish within the Mexican sociocultural context. In addition, this study supports the idea that the relationship between social construction of knowledge and student centrality helps researchers gain a better grasp of the characteristics of discussion postings and the degree of student centrality associated with potential paths to higher levels of knowledge construction. In sum, social network diagrams make the social dynamics of online learning tangible which extends the IAM analysis beyond its typical capacity of focusing on cognitive processes [33]).