Content uploaded by Irina Kuznetcova
Author content
All content in this area was uploaded by Irina Kuznetcova on Dec 10, 2018
Content may be subject to copyright.
1
Abstract
Some emergent forms of online educational tools support the formation of a non-
hierarchically distributed communication. Theories posit that this form of communication can
harvest a more optimal learning environment compared to traditional classroom environments,
but evidence supporting the theoretical assumption is scarce. The current study examined
whether a Multi-User Virtual Environment (MUVE) tool called Second Life (SL) in conjunction
with blogging and traditional lecturing (MUVE condition) can facilitate the formation of a
distributed learning environment as compared to only blogging and traditional lecturing (active-
control condition). Two classes of students (N = 56, 53.6% female) enrolled in an undergraduate-
level, semester-long psychology course participated in the study. The students in both
experimental and control conditions had a lecture and blogging component, and in the
experimental condition the students also used SL during class. Student interactions recorded on
the blogging platform in both conditions were analyzed by social network analysis. This was
supplemented by student interviews early and later in the semester. The results revealed that the
network in the MUVE condition was more connected and more distributed than in the active-
control condition. The findings suggest that MUVE enables students to play a more active role in
the classroom learning community.
Keywords: Computer-mediated communication; Interactive learning environments; Multi-User
Virtual Environments; Social network analysis.
2
1. Introduction
Within the past decades, more and more educational researchers have started to recognize
and investigate the importance of the social context of teaching and learning (Hausfather 1996).
The role of communication in the learning process is now acknowledged in many educational
theories (e. g., social learning theory; Bandura 1977; Bandura & Walters 1963; sociocultural
theory; Vygotsky 1962; 1978; self-determination theory; Deci et al. 2016; Ryan 1991;
transformational education; Mezirow 1997). However, with fast developing technology, the
means, norms and nature of communication and socialization have started to shift (Glassman,
2016). The Internet and the steadily decreasing cost of devices such as smartphones and laptops
have led to the emergence of new types of learning environments and educational tools,
including online or virtual learning ecologies, bulletin and image-based discussion boards, blogs,
Learning Management Systems (LMS), online communities, 3D virtual environments, video
conferencing, etc.
The Internet was originally envisioned as a democratic platform with a free flow of
information unconstrained by any single overarching authority (Glassman 2016). Research has
shown that at least some of the abovementioned online educational platforms and tools can
support a similar type of communication – distributed, non-hierarchical, and with an
unconstrained flow of ideas (BArtholomew, Jones, & Glassman 2012; Caballé et al. 2010). At
the same time, some educational theories posit that the most optimal learning environments are
non-hierarchical in nature. Environments where the teacher recedes as the central information
hub allow students to make connections with each other so that they can rely on multiple social
connections with others to advance their learning (Glassman, Bartholomew, & Jones 2011;
3
Mikami et al. 2012; Serdiouk et al. 2015; Siemens 2014). This begs to question: Can we harvest
the potential of online educational platforms to promote a distributed, egalitarian, non-
hierarchical type of learning environment? In particular, can online educational platforms serve
as contexts for what Dewey (1916) refers to a democratic education, and Mezirow (1997) refers
to as transformative education: preparing students for the types of vital activities they will
encounter over their lifetimes, especially outside of traditional school environments, and learning
to be autonomous thinkers capable of creating collaborative efforts in response to unique
problems, democratic citizens for a new information age (Glassman 2016)? The current study
investigates this question in relation to one such type of platform – Multi-User Virtual
Environment (MUVE) from a social network analysis perspective.
2. Theoretical background
2.1 Social networks from the social network analysis perspective
Social network analysis may be an important emerging tool for understanding the role(s)
social interactions play in teaching/learning processes. Although the existence of an
acknowledged theoretical framework that synthesizes all concepts captured by social network
analysis is still debated (Scott 2017), general social network concepts can help us further our
understanding of how social interactions unfold in learning environments – in the case of this
paper virtual learning environments.
Social networks are viewed as a set of people – participants – and the relationships or
interactions that connect the participants (Liu et al. 2017; Wassermann & Faust 1994). The
participants of the networks are called “nodes” (also known as actors) and the
4
relationships/interactions between the participants are referred to as “links” or “ties” (also known
as edges) (Wassermann & Faust 1994). The links between nodes are directional, including
outgoing (a node reaches out to another node), incoming (a node receives a link), and reciprocal
(two nodes reach out to and receive a link from each other) links. Networks that capture the
directionality of links are called directional networks (Wassermann & Faust 1994).
Some nodes might have more influence over the information flow in the network; they
have significantly more incoming links than other nodes in the network and are called hubs
(Wassermann & Faust 1994). Based on how the power and influence are distributed among
nodes, networks can be categorized as centralized, decentralized or distributed (Baran 1964).
Centralized networks (Fig. 1) are characterized by one distinct hub, while other nodes
will have few connections with each other. If the hub is removed, individual nodes become
vulnerable (Baran 1964). The centralized network is often reflected in traditional classroom
models in which the teacher represents the information authority and the hub and students’
learning mainly relies on their links to the hub.
Fig. 1 Centralized network graph
5
Decentralized networks (Fig. 2) usually have a few hubs connected to subgroups of
nodes. Just like centralized networks, decentralized networks are still vulnerable to the influence
of the hubs, although to a lesser extent. For example, students working in groups with their group
leaders are more likely to keep learning when the teacher is around.
Fig. 2 Decentralized network graph
Distributed networks (Fig. 3) do not have hubs. All nodes are interconnected with several
other nodes; the flow of information is least affected by a hub’s censure compared to the other
types of networks. Moreover, even if a few nodes are removed from the network, there is a
greater chance that the network will still remain connected. In the context of learning, one
example could be creating a wiki as a whole class when students equally contribute to document
development and work collaboratively to produce the content.
6
Fig. 3 Distributed network graph
Different network types determine different levels of access to information and different
types of information flow. Centralized networks are more vulnerable to the disruption of its free
flow of information because all nodes are connected through a single hub. Distributed networks
allow for more egalitarian interactions and equal access to resources. Decentralized networks are
in-between centralized and distributed networks in terms of the stability of information flow.
2.2 Social networks from the educational perspective
Some educational researchers agree that hierarchical social networks in the classroom can
lead to negative consequences for students. For example, Serdiouk et al. (2015) found that social
status inequality in the classroom predicts more peer aggression, peer victimization and rejection
– especially when teachers put little effort into decreasing this inequality. Mikami et al. (2012)
reported similar results, extending this idea to the instructional context: students in the
classrooms where teachers use hierarchy-supporting instructional strategies (e.g., supporting
competitive behaviors as opposed to cooperation) tend to have lower peer acceptance.
7
Some educational theories have implied the role of using new tools of the information
age to challenge the hierarchical nature of traditional classroom models; these tools are theorized
to offer new ways of looking at the teacher-student and student-student classroom interactions
(for example, Glassman et al., 2011; Siemens 2014, Downes 2010). One of such theories is
connectivism (Kop & Hill 2008; Siemens 2014). It was developed based on the propositions of
connectionism, which posited that knowledge is not developed through a defined cognitive
processing unit but rather through the connections among pieces of information from experience
(Bereiter 1991; Glassman & Kang 2010). Connectionism views knowledge as a non-hierarchical,
non-linear, dynamic process that cannot be passed from the teacher as authority to student as
neophyte – it should be developed within experiential, distributed interactions. Internet
applications can allow for these types of interactions by providing greater opportunities to
remove or deemphasize the authority/hub from dominating the learning process through
asynchronous (e. g., blogs, wikis) and synchronous (e. g., instant messaging) communication
(Glassman & Kang 2010).
Connectivism echoes these ideas by recognizing learning as a process of making
connections among ideas and resources made available through non-hierarchical social ties with
other interested learners (Siemens 2014). The information flow comes from the social
connections learners share with peers; forming stable, non-hierarchical social connections is,
therefore, essential ongoing learning processes. The best way to accomplish continuous
information flow is by establishing egalitarian, peer to peer networks of learning. Information
flow can travel freely from node to node, without hubs completely taking control over it. In other
8
words, connectionism and connectivism both support the idea of distributed, non-centralized
networks as optimal learning environments.
Similar to connectivism, the Open Source Educative Processes framework (OSEP;
Glassman et al. 2011; Glassman & Kang 2016) posits that online communities have become a
driving force in knowledge production and dissemination. Through establishing and sustaining
online communities, individuals work toward solving problems relevant to the community and
learn through the collaborative problem-solving and decision-making process. This type of
interaction, however is much more difficult in hierarchical (centralized) networks because most
nodes will not have access to the information required to collaboratively make decisions; these
processes should be situated in distributed social networks.
These theories are further supported by ideas underpinning of social capital theory (Lin
2002; Lin, Cook, & Burt 2001). This framework views social networks from the perspective of
nodes’ access to network resources. Social capital is defined as establishing connections within
one’s network that have resources for achieving one’s personal goals (bridging social capital).
The resources themselves are part of the network structure. Nodes with higher centrality tend to
situate closer to the center of the network with greater access to potential connections, and by
extension, to potential resources. Nodes will have most access to network resources in a highly
connected (highly dense), more distributed (with low centralization) network where individual
nodes have high individual centrality. In the educational settings, what the theory implies would
be that students who have connections only to the teacher will not be able to benefit from all
other potential peer connections available in the classroom network, which in theory minimizes
their learning potential over time. To increase possibilities of learning within the information
9
network, students need to be given opportunities and support to link to other students in the
classroom.
The importance of distributed networks offers unique opportunities for what Mezirow
(1997) has referred to as transformative learning. Mezirow has suggested a goal of education in
the service of life long thinkers should be to transform their thinking through critical reflection
on their working assumptions. Individuals need to be able to quickly change their thinking
(especially dysfunctional thinking) in coherent and cogent ways. This can only be accomplished
through effective discourse. For Mezirow,
Effective discourse depends on how well the educator can create a
situation in which those participating have full information; are
free from coercion; have equal opportunity to assume the various
roles of discourse (to advance beliefs, challenge, defend, explain,
assess evidence, and judge arguments); become critically reflective
of assumptions; are empathic and open to other perspectives; are
willing to listen and to search for common ground or a synthesis of
different points of view; and can make a tentative best judgment to
guide action. (1997, p.10)
We suggest this type of discourse is best, and perhaps only, achieved
within more distributed networks, especially networks where learners are capable
of transforming their thinking through their own agency.
While theories point to the benefits of distributed learning networks, little research
provided direct evidence showing how to achieve distributed learning in formal learning
contexts. Previous research on such online education tools as blogs and discussion boards
showed that they can provide unique opportunities to design a more distributed, non-hierarchical
learning process (Bartholomew et al. 2012; Caballé et al. 2010; Farmer et al. 2008; Halic, Lee et
al. 2010; Johnson 2007). In gaming research there is evidence that MMORPGs (Massively
10
Multiplayer Online Role-Playing Games) facilitate meaningful and long-lasting interactions
between players and foster collaborative learning (Steinkuehler 2006; Thorne et al. 2009).
Moreover, Zhong (2011) found that collective action had a positive effect on players’ social
capital and online civic engagement – as well as offline civic engagement. Despite the promise
of these learning platforms, none of these studies adopted a social network approach that could
clearly demonstrate whether such environments help create distributed social networks.
In the current study, we focus on a learning platform known as Multi-User Virtual
Environment (MUVE) to investigate its influence on the structure of social networks formed in
the learning process.
2.3 MUVEs in education
The term MUVEs refers to a specific type of virtual environment in which the user,
represented by a digital avatar, has a capability to interact with the environment and other users,
also represented by avatars (Dede et al. 2017). Most MUVEs simulate interactive multi-user 3D
environments that can be navigated through a keyboard, mouse and/or joystick and displayed on
the computer screen. They can be used on any computer that has Internet access and a relatively
moderate amount of processing power and bandwidth. Some examples of educational MUVEs,
as mentioned by Dieterle and Clarke (2006), include such initiatives as Quest Atlantis (Barab et
al. 2007; Barab et al. 2005) and Whyville (Kafai 2010) – all designed specifically for educational
purposes. Some educators tried to adopt commercial open-ended MUVEs such as ActiveWorlds,
OpenSim, Minecraft and Second Life (SL) for educational activities (e. g., Andreas et al. 2010;
Corbit 2000; Dickey 2005; Ryoo et al. 2011; Wang & Burton 2013). Among these MUVE
11
platforms, Second Life has received most attention – perhaps, due to having (arguably) the most
advanced functionality, graphics and communication channels. In SL, users can change the
appearance of personalized avatars, shop for clothes and other materials, navigate the space by
moving their avatars, instantly teleport to different locations, engage in text and voice chat with
other users, explore a variety of communities, build objects and (at a high level of expertise)
program avatar and object interactions.
The available research on SL and other MUVEs suggests that these platforms can have
significant positive impacts on students’ learning outcomes (Merchant et al. 2014; Vogel et al.
2006). Most of this research has focused on students’ affective and motivational response to the
environment, learning gains and achievement (Hew & Cheung 2010) in a broad variety of
domains, from arts and social sciences to STEM area (Wang & Burton 2013; Yu et al. 2017).
However, almost no studies investigated the impact of MUVEs on classroom social network
structure.
Meanwhile, MUVEs have a number of features that make them stand out among other
online teaching tools. This includes: (1) an open-ended, sandbox type of environment:
participants can set their own goals and rules (Merchant et al. 2014). (2) Synchronous and
egalitarian communication channels, such as text and audio chat. (3) Experientially meaningful
environments in which communication becomes part of an organic collective problem-solving
process. (4) Opportunities for creative activities – creative in the sense of (re)creating artifacts
and/or the entire environment, which creates a sense of ownership over the space and the
processes unfolding within that space. (5) Opportunities for anonymity and identity exploration
12
(both real and virtual). This warrants the investigation of MUVEs’ impact on classroom social
networks – a topic that has been tackled in only one study so far (Lorenzo et al., 2012).
Lorenzo et al. (2012) investigated how the use of MUVEs (which they referred to as
MMOL platforms, Massively Multiuser On-line Learning platforms) impact the social network
structure of the classroom based on the student-student and student-tutor interactions in the
virtual space. They compared students’ and tutors’ centrality indices and density of social
networks formed in a MUVE (through synchronous discussions) as opposed to a Learning
Management System (LMS) called Moodle (through asynchronous discussions). The authors
found that students in the MUVE conditions had more links with each other than in the LMS
condition. Centrality scores indicated that tutors had a central position in both networks (they
were hubs); however, the quality of relationship among students in the MUVE condition was
significantly better. Moreover, participants in the MUVE-mediated network had more
homogeneous centrality scores, which means that they engaged in peer interactions more
frequently, had fewer isolates (nodes with no links), and had more access to information from
their peers. In short, the MUVE helped establish a more distributed, egalitarian social network in
the classroom compared to the LMS medium.
Although these results align with the general theoretical frameworks discussed above and
used in this paper, that study had a few limitations. First, the short-term MUVE intervention
prevented the study from investigating how online social network might change over time.
Second, the MUVE and LMS social networks were compared based on communication logs
from the MUVE and LMS; however, these logs represent different types of data - synchronous
communication data in the MUVE (including gazes and gestures) and asynchronous
13
communication data in the LMS. No information about the standardization of the communication
frequency based on the medium was provided. One could argue that synchronous
communication can prompt more interactions simply because students can talk more – and more
easily – in real time. Therefore, this comparison may not be fair. Finally, the authors themselves
note that it is worth investigating how the results might be different when considering using a
different pedagogical framework and different student training for navigating the MUVE. The
following section explains how the current study will address these limitations.
3. The present study
The current study investigated the impact of MUVEs on classroom social networks from
the social network analysis perspective. We focused on several structural measures of classroom
social networks formed in the blogging platform (Blogger) as students experienced MUVE
(Second Life) or traditional lecturing during the spring semester of the year. The findings would
contribute to the nascent field of the literature on MUVEs in education by aligning social
network and educational theories and providing insight into whether and how classroom social
networks are effected through use of Second Life.
We sought to address the following research questions:
Research question 1: Does the use of lecturing/blogging based curricula combined with
a MUVE intervention (Second Life) lead to a more connected classroom social network as when
compared with a lecturing/blogging without a MUVE intervention?
Hypothesis 1: The use of a MUVE intervention will lead to more social interactions
among students as compared to using just lecturing and blogging.
14
Research Question 2: Does the use of a lecturing/blogging with a MUVE intervention
contribute to the formation of a more distributed social network over time as opposed to a
lecturing/blogging without a MUVE intervention?
Hypothesis 2: The network of the students in the class with the MUVE intervention will
be more distributed than the network of students lecturing/blogging class without a MUVE
intervention.
4. Method
4.1 Study Design
The study was conducted in a large Mid-Western university in the United States over the
course of two semesters (Fall and Spring). The Fall semester served as a pilot testing of the
Spring intervention. The current study only report findings from the Spring intervention.
The participants in the Spring semester were undergraduate students enrolled in a general
education course focused on adolescent psychology in school context. The course had two
sections taught by the same instructor who had taught the course for five years. One section of
the course was randomly chosen to implement Second Life together with teacher-centered
lectures and blogging (experimental condition, also referred to as the MUVE class). The second
section served as a control condition, in which students received only the teacher-centered
lectures and blogging (also referred to as the active-control class). The curriculum (readings and
topics) for both sections was identical. Both classrooms used the same blogging platform
(Blogger.com). The weekly assignment for both classes was to post to the community blog once
a week. The students in both classes explored the same topics and the same assignments and
15
their grades were based on the same expectations (a combination of attendance and posting).
During class time, the instructor gave the same lecture in both sections, but the lectures in the
active-control class was given more scheduled time for in-class discussions and additional media
information (e.g. videos) associated with the course curricula to account for instructional time
differences incurred by implementing the MUVE intervention in the experimental class.
No previous experience in Second Life was required of students in the MUVE class, and
inworld activities were not graded. Students were asked to engage in Second Life activities
during class time, but were not forbidden to use SL outside of class. There were four reasons for
this study design: 1) The practical reason was use of Second Life can require a great deal of
support. Having students engage in SL activities by themselves would create too much variance
between students based on available Internet/bandwidth and/or previous experience with Multi-
user Virtual Ecologies. 2) The teaching concern was to engage students in Second Life activities
with as little anxiety as possible. This is the reason why we did not grade Second Life activities
in any way. 3) Another teaching concern was that if we asked for participation outside of the
confines of the classroom, students would be less likely to sign in and/or spend less time in
activities. 4) The more theoretical consideration was the major thrust looking to explore changes
in classroom social networks. We anticipated that synchronous participation in the classroom
would be more likely to have an impact on student social networks than asynchronous
participation outside class time.
It is important to note that the curriculum in both classes was based on an OSEP
framework, including special considerations about providing students with ownership over the
16
learning process, fostering online collaborative problem-solving in an open-ended, ill-defined
problem context, and providing multiple opportunities for students’ creativity.
4.2 Participants and procedure
Out of 56 students enrolled in the two sections of this course, all of them consented to
participate in the study. Among the 56 participants, 53.6% were female. The median age across
conditions was 20, with students ranging in age from 18 to 34 years old (M = 21.38, SD = 3.37).
Most of the students majored in education or in education-related areas of study. None of the
students in the MUVE condition were familiar with SL prior to the class.
Both classes were conducted in the in-person format with an online element. In the active
control class, the instructor usually gave a lecture on a specific topic followed by some in-class
media and/or activities. After class, students were required to make one blog post about the topic
of the week and provide one link to Internet sources that can help others learn more about the
topic. Students were also required to reply to at least one of their classmates. In the MUVE
condition, the procedures were the same except for implementing a SL component in almost
every class for a certain period of time (usually about half the class time). The Second Life
activities in the MUVE class for the most part replaced the whole-class discussions and media
information in the active-control class. The SL activities were designed to relate directly to the
topic of the class lectures (in the same way media and in-class activities did for the active-control
class. For example, for the topic of adolescent identity formation, the students modified the
appearance of their avatar (both their physique and clothes and accessories). In a subsequent
class, as the topic delved deeper into adolescent identity, student groups worked to adopt a group
17
identity and chose their outfits based on this identity; for other topics, students engaged in such
activities as building objects and structures, buying items to decorate these structures, searching
for places and groups in-world and traveling to different places of interest. During all SL
activities, students could only communicate via text chats in SL – including direct individual
messaging and group chats.
The data for analysis were collected from students’ homework assignments (one blog
post a week). The blogs were a completely separate arena of online communication from the
Second Life environment and were implemented in the exact same ways in both classes. At the
beginning of each class the instructor would display a few students’ posts, comment on them and
discuss them with the class. The purpose was to let students know the blogposts were being read
and appreciated. The instructor provided the same blogging prompts every week, but never
replied to students’ blog posts online and never posted on the blog beyond prompts.
To host Second Life activities, we created a private island in SL available only to the
instructor, research team and the students. To reduce the teacher’s interpretative authority in
order to foster more student-centered conversations (Chinn et al. 2001), the instructor never
interfered student activities on the island and did not take part in their conversations, but had a
house on the island where students could approach the instructor for questions. No student
visited the house throughout the semester.
4.3 Data sources and data coding
4.3.1 Social Network Analysis
18
From the series of blog posts over a 15-week time span, we chose to analyze blogposts at
a beginning time point (Post 1) and an end-of-the-semester time point (Post 12) to examine
changes in student interactions over time. We did not use the data from Post 13 or Post 14
because by the end of the semester many students chose to skip their posts (students were
permitted to skip one post during the semester), partially due to the final exam workload.
The students and the teacher constituted the nodes of the network, and their blog
interactions constituted the links in the network. Students who did not post at these time points
were removed from later analysis (active-control condition: three students at Post 1, two students
at Post 12; MUVE condition: two students at Post 1, three students at Post 12). The links
between students and the instructor were identified in two ways: (1) students mentioned the
instructor or another student in their blog post; (2) students commented on another student’s blog
post. Both of these data sources were considered important because they directly represent
students’ agency in the communication process. While they were required to post, who or what
they mentioned in their writing as well as who they directed their comments towards were
completely under students’ control. Mention of teacher or student in blog posts can be construed
as a rough measure of whether students were responding to teacher presence or social presence
of other students in the community. Peer to peer commentary is considered one of the most
central and telling characteristics of social engagement in both place-based (Reznitskaya et al.
2009; Smith et al. 2009) and especially online communities (Bereiter & Scardamelia 2003; Stahl
2006). How students use their agency in freely commenting on peers is an important indicator of
how they view and participate in their immediate learning community (e.g. Rientes et al. 2012,
Downes 2010). If a blog post satisfied at least one of these criteria, a link was coded between the
19
post/comment author and the recipient of this link. Ties were coded either as 1 – when the
recipient of the link could be clearly identified from the message – or as 0 – when the interaction
was absent or the recipient was unclear.
4.3.2 Interview
Semi-structured interviews, lasting 20-30 minutes were conducted with ten students from
the MUVE class near the beginning of the semester and near the end of the semester to gauge
changes in student thinking about the use of Second Life as a central part of the class structure.
The same students served as subjects for both sets of interviews (early semester and late
semester). The interviewers asked essentially the same questions at both time points. The
students were a convenience sample as the interviewees were the first ten to volunteer for the
interviews in response to a general call (the lead interviewer took ten minutes at the beginning of
the first class to explain the general procedure and ask for volunteers), in exchange for a five-
dollar Starbucks card at the end of each interview. The interviews were conducted in a separate
location from the class and were video-taped.
5. Analysis
We conducted a social network analysis on the data collected from blog posts using two
statistical packages (statnet; Handcock, Hunter, Butts, Goodreau, & Morris 2003) and igraph
(Csardi & Nepusz 2006) of program R, version 3.4.4 (R Development Core Team 2008). The
networks were analyzed at two levels: classroom level (focus on the network as a whole) and
individual level.
20
The classroom-level indices derived from online blog interactions were centralization
and density. Density is the ratio of the present links to their maximum number possible (Scott
1991; Wassermann & Faust 1994). Ranging from 0 to 1, the density of 0 indicates that there are
no connections in the network, while the density of 1 means that every node is connected to
every other node. The higher the density, the more connected the network is. Freeman’s
centralization is an index measuring the degree of variation among node centralities in the
network (Freeman 1978; Wassermann & Faust 1994). Centralization can be calculated based on
different individual centrality measures, such as in-degree, eigenvector or all- (in- and out-)
degree centrality indices (as described below). It is represented by the ratio of the sum of
differences in the individual centrality scores to their maximum possible sum. On the spectrum
from 0 to 1, a network with a centralization value of 0 has no hubs (distributed network), defined
as the most prominent individual in the network. A network with a centralization value of 1 has a
single dominating hub (centralized network).
Individual centrality refers to the prominence of an individual in the network
(Wassermann & Faust 1994). Several individual centrality measures were calculated. For the
purposes of this study, we focused on in- and out-degree centrality, eigenvector centrality, and
betweenness centrality (Freeman 1978; Wassermann & Faust 1994). Two measures were used to
capture the directional interactions: in-degree centrality (the number of links a node receives)
and out-degree centrality (the number of outgoing links) (Freeman 1978). The higher the in-
degree centrality, the more popular and prominent the individual is. The higher the out-degree
centrality, the more active the individual is in disseminating information in the network.
Eigenvector centrality (Bonacich 2007; Ruhnau 2000) extends the measure of degree centrality
21
by considering not only the number of students an individual has links with but also their value
of centrality. The individual’s centrality increases if he or she is linked to more prominent
students with higher degree centrality in the network. Betweenness centrality measure
(Wassermann & Faust 1994) is based on the idea that an individual has more influence over the
network because he or she lies on the shortest paths between other students. Individuals with
higher betweenness centrality can have higher influence on the communication between dyads of
students. Individual and classroom-level indices were used together with network graphs (also
known as sociograms; Wassermann & Faust 1994) to investigate the structure of social networks
of interest.
A set of statistical analysis were conducted to examine the condition effect (MUVE vs.
Active-control) on the changes in the online blogging network structures over time. Due to the
violation of the assumptions of variance homogeneity and the normality assumption, we ran a
non-parametric analysis alternative of Analysis of Covariance (ANOVA) suggested by Quade
(1967), using SPSS 24.0 (IBM Corp. 2016). Quade’s test does not require data to be normally
distributed. The dependent and independent (covariate) variables were transformed into ranks,
the transformed dependent variable was regressed on the transformed covariate, and the
unstandardized residuals from the regression were saved to perform a one-way ANOVA
(analysis of variance), where the residuals were entered as a dependent variable and condition
was entered as a factor.
Analysis of the pre and post interviews was conducted by three independent coders.
Coders had no contact with the original interviewers and were not involved with the use of
Second Life in the classroom or discussions about curriculum and/or technology. The interview
22
coders received simple instructions to determine whether they were able to detect any differences
in the students’ attitudes towards the class, their general sense of participation in the class, and
the general use of second life in the class. The coders analyzed all 20 interviews (10 early
semesters and 10 late semester) separately. There were two general joint meetings, the first to
discuss the process and the second to discuss the findings. Consensus of the analysis was derived
through the second meeting.
6. Results
6.1 Research question #1: Network connectedness
The sociograms created in R (Fig. 4) showed that the active-control condition had a
denser network of connections at Post 1 than the MUVE condition. At Post 12, the instructor was
an isolate in both conditions. This means that the instructor did not receive any incoming links
from the students; the in-degree centrality value of the instructor was 0.
23
Fig. 4 Network sociograms: circular layout
The network density index (Table 1) revealed that, the active-control condition network
at Post 12 (density = .023) became less dense compared to Post 1 (density = .034), while the
density of classroom social network in the MUVE condition became denser over time (from
density = .028 at Post1 to density = .036 links at Post 12). At Post 12, the MUVE group had a
more connected network than the active-control group. A similar trend was evident in the
number of isolates in both conditions. The number of isolates (network participants without
24
incoming links) from Post 1 to Post 12 increased in the active-control condition (from 2 to 6) and
decreased in the MUVE condition (from 5 to 2).
Table 1
Network descriptives and network density
Index
MUVE P1
MUVE P12
Active-Control
P1
Active-Control
P12
N of nodes
29
27
29
30
N of links
23
25
28
20
N of isolates
5
2
2
6
Density
.028
.036
.034
.023
Note: N = number; P1 = Post one; P12 = Post twelve.
To measure the distribution of the individual centrality indices, we compared changes in
the in-degree and out-degree centrality, eigenvector centrality and betweenness centrality from
Post 1 to Post 12 between the two conditions. Quade’s test was used as a non-parametric version
of ANCOVA.
Table 2
Quade’s test results
Welch F
(df1, df2)
Mactive-control
(SD)
MMUVE
(SD)
DV: In-degree centrality P12
27.02 (15.05)
31.20 (15.70)
In-degree centrality P1
31.67 (15.68)
27.33 (15.45)
Condition
.71 (1, 50.15)
DV: Out-degree centrality P12
25.63 (14.56)
32.74 (9.99)
Out-degree centrality P1
31.69 (12.69)
27.31 (17.42)
Condition
3.56 (1, 48.62)
DV: Eigenvector centrality P12
23.72 (16.99)
34.87 (9.90)
Eigenvector centrality P1
34.52 (13.38)
24.48 (14.47)
Condition
7.01 (1, 44.46)*
DV: Betweenness centrality P12
23.65 (11.37)
34.94 (16.55)
Betweenness centrality P1
34.72 (16.53)
24.27 (10.58)
Condition
2.49 (1, 44.58)
Note: DV = dependent variable. * = p < .05. The means and standard deviations are based on the rank data. Welch F was used when the Leven’s
test of homogeneity of variances was significant.
25
Results (Table 2) showed that condition was not a significant predictor of any centrality
indices except for eigenvector centrality (Welch F = 7.01, p < .05). Quade’s test can be
interpreted as the probability that a centrality index score randomly chosen from one population
is higher than the score of the other population or the average population. The significant
condition effect on individual eigenvector centrality suggests that any individual eigenvector
centrality randomly chosen from the MUVE condition was likely to be higher than that in the
active-control condition, after controlling for Post 1 eigenvector centrality. This implies that the
information flow in the MUVE condition network was more connected than in the active-control
condition.
6.2 Research question #2: Network centralization
The sociograms (Fig. 5) at Post 1 showed that the classroom social networks in both
conditions were relatively similar in the link distribution. At Post 12, the network in the MUVE
condition seemed to become less centralized and therefore more distributed than at Post 1, as the
sociogram contained fewer hubs. In contrast, the networks in the active-control condition did not
seem to change much from Post 1 to Post 12 in terms of the number of hubs present in the
network.
26
Fig. 5 Network sociograms: Fruchterman-Reingold layout
The classroom centralization indices are reported in Table 3. The most salient trend was a
decrease in eigenvector centralization in the MUVE condition. There was lower eigenvector
centralization in the network at Post 12 than at Post 1. The same index in the active-control
condition increased over time. In-degree and all-degree centralization went down from Post 1 to
Post 12 in the MUVE condition but increased in the active-control condition. The trends in all
network centralization indices indicate that the classroom social network in the MUVE condition
became less centralized and therefore more egalitarian over time than that in the active-control
condition.
Table 3
Network centralization indices
27
Index
MUVE P1
MUVE P12
Active-Control
P1
Active-Control
P12
In-degree
centralization
.156
.123
.075
.083
Eigenvector
centralization
.424
.041
.277
.299
All degree
centralization
.085
.065
.039
.049
Note. P1 = Post one; P12 = Post twelve.
6.3 Pre- and post-interviews data in support of research questions
The interview coders independently identified two important dimensions directly related
to the research questions when comparing students’ pre-interviews to their post interviews.
1. Do the students’ answers to the (same) interview suggest a recognizable shift from more
passive learning process (in the first interview) to a more active learning process (in the
second interview)?
2. Do the students’ answers suggest a recognizable shift from more teacher-dependent
learners to more autonomous learners?
These dimensions while developed by the coders through the processes of analysis offer
insights into both the specific research questions of this paper and the larger theoretical
frameworks of transformative learning (Mezirow 1997) and Open Source Educational Processes
(Glassman & Kang 2016) leading to more autonomous thinking/educational activities. A
transformation to more active learning should theoretically lead to increased student-to-student
communication through shared inworld activities (we believe this was apparent in students’
discussion as suggested by excerpts presented below). Increased autonomy in thinking and action
can lead to students becoming less dependent on authority-based information sources (e.g.
28
instructor) and more dependent for learning on their own and shared peer activities (we believe
this is a more abstract idea but can also be captured in student responses in the interviews).
The coders reached agreement on the qualitative changes along the aforementioned
dimensions in eight out of the ten interview sets. One coder found differences in eight sets of
interviews and one coder found differences in nine sets of interviews. The coders found
consistency between both dimensions (i.e., if they found evidence of change in one, they found
evidence of the other). This is consistent with both Mezirow’s theory of transformative learning
and OSEP. The examples of changes in students’ responses between first interview and second
interviews are provided below.
6.3.1 Responses reflecting shift from more passive learning to more active learning
Interviewee 6 (each subject was numbered which was/is used as their identifier)
Interview 1: I think that SL does a great job. Like part of the time we have lecture and
other time online. The computer does bring more hands-on activity. Actually, I’ve got
big problem connecting SL to actual course content.
Interview 2: At first, it would be hard to see the connection between like what the
course would teach and then application of SL. It feels like a jump. But now, like some
of the group work and things we are doing. Through the questions on the posts
(editorial note: by posts the student is referring to blog posts and comments), I see
there is connection. I think the connection for me comes from the collaboration with
peers. Actual actions of yourself in the games, like going to this country, taking a
picture from the tower.
29
These two responses represent a theme that ran through most of the interview sets. In
the early semester interview students were struggling to understand the connection between
the course content being assigned and taught by the instructor and their activities in the
MUVE. While some students saw the MUVE as a break from lecturing they viewed its
primary purpose as in some way amplifying what the instructor taught them. In the second
interview students shifted to seeing the importance of them and their peers exploring and
engaging in activities together. The peer-based activity rather than content became central
in their learning processes.
The following excerpts seem to represent an even more dramatic shift from passive to
active learning.
Interviewee 9
Interview 1: I haven’t really gained anything beneficial out of second life so it’s a mixed
bag. I don’t know what’s out there but if there was some other program. I think the thing
about SL that has been weird to me is that you’re coming to this world and everything you
do is supposed to be you and I struggle with it and I don’t see that connection.
Interview 2: It has been an interesting learning experience for me like it’s not something I
would have never thought of like oh we’re going to go into this virtual world and learn
things. That’s really interesting, to try and figure how that’s learning in this class, like what
I have learnt from it and what that means for learning in general,
30
In the first interview the student seemed to be waiting for the connection between what they
were learning in class and their SL activities to be made for them and did not see any benefit to
SL without the connection being provided. In the second interview the student described what
theorists like Dewey (1916) and Mezirow (1997) see as central to life-long learning processes:
learning how to learn, even about learning itself.
6.3.2 Responses representing the shift from more teacher dependent to more autonomous
learning
Interviewee 7
Interview 1: I think if a course were to use a platform similar to Second Life, taking the
material. I’m used to a lot of psych classes with lectures and Powerpoint so it’s a
change of pace there. It gives us the material and then we immediately take that
material and apply it to ourselves in second life so I think that’s useful.
Interview 2: I’ve taken a lot of psychology courses and you sit in an auditorium looking at a
Powerpoint. Throwing in virtual reality on top of that came out of nowhere. I’m kind of a
reserved person so incredibly forced social interactions like that could be over the top. I
guess social media more but Second Life is kind of like that. As far as me looking towards
new experiences, it was great in that regard.
The interviewee started out in the same place in both the early semester and late semester
responses, comparing the MUVE class to larger, auditorium classes using Power Point
presentations. However, the answers then diverged in important ways. In the first interview the
subject was concerned with taking the material as presented by the instructor and applying it to
31
their SL activities. SL was seen as a place to engage in activities that amplify what they learned
in the lectures. In the second response SL was thought of more as a shared MUVE which opened
users up to new types of interactions with others and exploring new experiences.
Finally, the change in the responses of the following interviewee captured much of the
change we saw in the interviews as well as our findings through the social network analysis.
Interviewee 3
Interview 1: I would say it was an interesting experience. I don’t know how it is necessarily
say it good or bad. As a nerd, I have played video games before. It’s different from I
expected. It’s hard to control. I’m not used to playing with other people. But that’s ok. That
is me. But overall, it seems an interesting platform.
Interview 2: It is enjoyable and still I don’t know and kind of figuring out how to use
it, move around and control everything. But for the most part, it’s really interesting
and different activities. I value the experiences. It’s cool to work out those concepts.
Sometimes there were moments kind of like we did one thing in the lecture and something
else in the SL. I think they go with each other pretty well. It’s a good way to force us to re-
experience adolescent life.
As with the previous interviewee, student 2 started off his response in essentially the
same place – the “game” is difficult to figure out. In the first interview even though the
student was used to playing games they seemed to be primarily single player rather than
MUVEs. The student seemed to be trying to figure out what he was supposed to be getting
out of the “game,” but not willing to try. In the second interview, while the student still had
32
some trouble figuring out how to use SL he began to see being inworld as offering
something different from the lecture, his own experience.
The interview data offer some possibilities for the changes found through the social
network analysis of class communications. The interviews suggest that many of the students
were beginning to see SL experience as something more active than traditional classrooms,
experiences that are separate from, and at least as valuable, as the top down course content.
At least some students (from the analysis the majority of students) began to see the SL
experiences as their own and their peers’.
7. Discussion
With the emergence and ever-growing popularity of online learning, the understanding of
social interactions in online settings has become an important step in advancing this type of
learning experience. This study presents an attempt to fill this gap by investigating social
networks formed in an online learning environment, particularly in MUVEs settings.
With respect to the two research questions that were driving this study, our hypothesis for
the first research question was confirmed. Sociograms and the density index showed that the
network in the MUVE condition was denser (more connected) than in the active-control
condition. Quade’s test also indicated that MUVE students became more active in taking control
over the information flow in the network over time as compared to the active-control students.
These findings align with the proposed theoretical frameworks. SL seemed to have
created a space for students to form online connections that transcended the immediate MUVE
environment and translated into more interactions on the blogging platform. One possible
33
explanation, taken from the student interviews, is that SL caused students to communicate more
with each other in real time in ways they considered relevant and important. As they become
more active and autonomous in their learning experiences and more willing to engage with other
students in those experiences. This could lead to establishing more contacts inworld that later
translated into blog interactions. From a theoretical perspective this reflects some of Mezirow’s
ideas: learning in the MUVEs has the potential to be transformative creating the learner as a
more autonomous thinker in general.
Our hypothesis for the second research question was also confirmed. Network
sociograms and classroom-level centralization indices revealed the increase in centralization over
time in the active-control condition, but the opposite trend was evident in the MUVE condition.
The decrease in centralization combined with the finding showing that students became more
active in information sharing in the MUVE classroom over time imply that the network in the
MUVE classroom became more distributed over time than the network in the active-control
classroom. Again, educational theorists such as Dewey and Mezirow might see this as the result
of students being more active in the learning process, not just in experimenting with concepts
provided by the instructor but owning/taking control of the process, “re-experiencing
adolescence” as one of the interviewees put it. MUVEs can serve very knew bottles for old
wines.
Second Life created a space in which the interactions were not just more frequent (which
possibly led to a denser network on the blog) but also less hierarchical. This is consistent with
both Mezirow’s transformational learning and the OSEP framework, which advocate for more
autonomous, distributed learning environment. Due to the instructional design that was based on
34
the OSEP premises, the MUVE students were involved in collective problem-solving through
this egalitarian communication channel, which could have prompted more reciprocal student-
student interaction that persisted beyond the SL environment. Moreover, SL activities such as
artifact creation can develop students’ sense of ownership of the learning environments and, by
extension, of the learning process itself, thus developing a sense of autonomy in learning about
and exploring the subject.
John Dewey (1916) suggested that education exists in the ways we communicate to
develop a common purpose. The emphasis of the teaching learning process is not collaboration
to achieve a goal but the way learners communicate with each other in shared agendas, reflected
in one of the interviewees realizations, That’s really interesting, to try and figure how that’s
learning in this class like what I have learnt from it and what that means for learning in general.
Problems will change over time but the ways we learn to engage each other when confronting
problems as a group is at the heart of developing new solutions.
In this paper we argue that the unique attributes of Multi-User Virtual Environments
(MUVEs) might offer an important new avenue for combining specific knowledge goals with the
development of autonomous, non-hierarchical learning groups. Students might be able to use
MUVEs to create virtual “playgrounds” (Kafai et al. 2010) or reflective halls of mirrors
(Kuznetcova, Teeple, & Glassman 2018) for the new ideas they are learning. New contexts that
invite experimentation with limited consequence offer opportunities for students to
simultaneously break free of the traditional classroom while using and exploring the new ideas
they have learned (through their avatars). Students can engage in shared activities related to their
new knowledge that engender or enhance communications for a common purpose among avatars
35
in a shared environment to further their endeavors. Students take not only the information they
learn but the types of shared communication that help them use this knowledge to solve real
world problems.
Overall, the current study suggests MUVEs can be used as an effective tool for fostering
the type of active learning leading to distributed social networks fostering autonomous
thinking/communication. MUVEs can contribute to forming distributed, egalitarian social
interactions among students leading to transformational education that can transcend the
immediate the MUVE boundaries (e.g. blog communications). The ability to create denser and
more distributed networks through the use of MUVEs in shared knowledge building and problem
solving might have far ranging consequences for education that go beyond the immediate
classroom. They might serve as playgrounds of experimentation for in-service teachers, bringing
their school community closer together, giving voice to colleagues (not just the ability to speak
to problems but listen to others), and increasing the collective efficacy of teachers in the school
(Goddard & Goddard 2001). The gains might fall over into other types of communication
between teachers and between administrators and teachers. MUVEs might have abilities to
extend and create communities of practice in every sense of the concept.
8. Limitations
The results of this study should be interpreted in light of its limitations. First, the SL
curriculum was based on the premises of OSEP theoretical framework. It might be that using
other theoretical frameworks for the instructional design might yield different results with
respect to social network structures and dynamics. Second, the coding of the blogging
36
interactions did not consider the quality of these interactions. In other words, any connection a
student made to another student through a reply or mentioning a name was coded as a link
between nodes; however, it might be that the quality of these links (e. g., an intellectual
challenge as opposed to “Nice post!”) also impact how student connect to each other.
Investigating weighted network links might potentially yield different results. Finally, although
the intervention was semester-long and a preliminary study was conducted, no replication studies
were conducted afterwards, and due to the difficulties of implementation, only a limited sample
of students could participate in the study. Future replications with more diverse samples are
needed to further investigate whether and how these findings can generalize to larger
populations.
9. Future directions
The current study is one of the first to address the issue of social network structure
formed by using MUVEs and explicitly tying together different network types (centralized,
decentralized and distributed) to the learning process. This area of inquiry requires a lot more
systematic research to advance our understanding of how MUVEs can be used to support
effective educational processes.
The current study measured how MUVEs impact structural properties and development
of social networks in an online space. However, the question of whether and how the social
network structure predicts or impacts learning outcomes was not addressed. To answer this
question, one should clearly define what learning outcomes are of interest (for example, product-
based outcomes, such as test scores, or process-based outcomes, such as collaboration and
37
critical thinking skills). Moreover, the pedagogy behind using MUVEs implementation should be
clearly established. MUVEs by themselves do not transform learning; they do so when an
appropriate instructional design informs instructional decisions. It is possible that different
instructional designs result in different social network processes and dynamics.
38
References
Andreas, K., Thrasyvoulos, T., Stavros, D., & Andreas, P. (2010). Collaborative learning in
OpenSim by utilizing Sloodle. In 6th Advanced International Conference on
Telecommunications, AICT 2010 (pp. 90–95). https://doi.org/10.1109/AICT.2010.75
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change.
Psychological Review, 84(3), 191–215.
Bandura, A., & Walters, R. H. (1963). Social learning and personality development. New York:
Holt, Rinehart and Winston.
Barab, S., Dodge, T., Tuzun, H., Job-Sluder, K., Jakson, C., Arici, A., … Heiselt, C. (2007). The
Quest Atlantis Project: A socially-responsive play space for learning. In B. E. Shelton & D.
Wiley (Eds.), The Educational Design and Use of Simulation Computer Games, (pp. 159–
186). Rotterdam, The Netherlands: Sense Publishers. Retrieved from
http://itls.usu.edu/~bshelton/simcompgames/barab-159-186.pdf
Barab, S., Thomas, M., Dodge, T., Carteaux, R., & Tuzun, H. (2005). Making learning fun:
Quest Atlantis, a game without guns. Educational Technology Research and Development,
53(1), 86–107.
Baran, P. (1964). On distributed communications networks. IEEE Transactions On
Communications Systems, 12(1), 1–9.
Bartholomew, M., Jones, T., & Glassman, M. (2012). A community of voices: educational blog
management strategies and tools. Techtrends, 56, 19-25.
Bereiter, C. (1991). Implications of connectionism for thinking about rules. Educational
Researcher, 20(3), 10–16. Retrieved from
39
http://edr.sagepub.com/cgi/doi/10.3102/0013189X020003010
Bereiter, C., & Scardamalia, M. (2003). Learning to work creatively with knowledge. In E. De
Corte, L. Verschaffel, N. Entwistle, & J. van Merriënboer (Eds.), Unravelling basic
components and dimensions of powerful learning environments (pp. 55-68); EARLI
Advances in Learning and Instruction Series.
Bonacich, P. (2007). Some unique properties of eigenvector centrality. Social Networks, 29(4),
555–564.
Caballé, S., Xhafa, F., & Barolli, L. (2010). Using mobile devices to support online collaborative
learning. Mobile Information Systems, 6(1), 27–47.
Chinn, C. A., Anderson, R. C., & Waggoner, M. A. (2001). Patterns of Discourse in Two Kinds
of Literature Discussion. Reading Research Quarterly, 36(4), 378–411.
https://doi.org/10.1598/RRQ.36.4.3
Corbit, M. (2000). Building virtual worlds for informal science learning (SciCentr and SciFair)
in the Active Worlds Educational Universe (AWEDU). In Proceedings of the Workshop on
Enabling Technologies: Infrastructure for Collaborative Enterprises, WETICE (Vol. 2000–
Janua, pp. 21–26). https://doi.org/10.1109/ENABL.2000.883699
Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research.
InterJournal, Complex Systems, 1695(5), 1–9.
Deci, E. L., Vallerand, R. J., Pelletier, L. G., Ryan, R. M., Deci, E. L., Vallerand, R. J., … Ryan,
R. M. (2016). Motivation and education: The self-determination perspective, 26(3 & 4),
325–346.
Dede, C. J., Jacobson, J., & Richards, J. (2017). Introduction: Virtual, augmented, and mixed
40
realities in education. In D. Liu, C. Dede, R. Huang, & J. Richards (Eds.), Virtual,
augmented and mixed realities in education (pp. 1–16). Springer Singapore.
Dewey, J. (1916). Democracy and education. New York: McMillan & Co.
Dickey, M. D. (2005). Three-dimensional virtual worlds and distance learning: Two case studies
of Active Worlds as a medium for distance education. British Journal of Educational
Technology, 36(3), 439–451. https://doi.org/10.1111/j.1467-8535.2005.00477.x
Dieterle, E., & Clarke, J. (2006). Multi-user virtual environments for teaching and learning. In
M. Pagani (Ed.), Encyclopedia of Multimedia Technology and Networking (2nd ed., pp.
1033–1041). Hershey, PA: Idea Group, Inc.
Downes, S. (2010). Learning networks and connective knowledge. In H. Yang and C.H. Yin
(Eds.), Collective intelligence and E-Learning 2.0: Implications of web-based communities
and networking (pp. 1-26). Hershey, PA: IGI Global
Farmer, B., Yue, A., & Brooks, C. (2008). Using blogging for higher order learning in large
cohort university teaching: A case study. Australasian Journal of Educational Technology,
24(2), 123–136.
Freeman, L. C. (1978). Centrality in social networks conceptual clarification. Social Networks,
1(3), 215–239. https://doi.org/10.1016/0378-8733(78)90021-7
Glassman, M. (2016). Educational Psychology and the Internet. Cambridge University Press.
Glassman, M., Bartholomew, M., & Jones, T. (2011). Migration of the mind: The emergence of
Open Source education. Educational Technology, 51(4), 26–31.
Glassman, M., & Kang, M. J. (2010). Pragmatism, connectionism and the internet: A mind’s
perfect storm. Computers in Human Behavior, 26(6), 1412–1418. Retrieved from
41
http://dx.doi.org/10.1016/j.chb.2010.04.019
Glassman, M., & Kang, M. J. (2016). Teaching and learning through open source educative
processes. Teaching and Teacher Education, 60, 281–290. Retrieved from
http://dx.doi.org/10.1016/j.tate.2016.09.002
Goddard, R. D., & Goddard, Y. L. (2001). A multilevel analysis of the relationship between
teacher and collective efficacy in urban schools. Teaching and Teacher Education, 17(7),
807-818.
Halic, O., Lee, D., Paulus, T., & Spence, M. (2010). To blog or not to blog: Student perceptions
of blog effectiveness for learning in a college-level course. Internet and Higher Education,
13(4), 206–213.
Handcock, M. S., Hunter, D. R., Butts, C. T., Goodreau, S. M., & Morris, M. (2003). statnet:
Software tools for the Statistical Modeling of Network Data. Retrieved from
http://statnetproject.org
Hausfather, S. J. (1996). Vygotsky and schooling: Creating a social context for learning. Action
in Teacher Education, 18(2), 1–10. Retrieved from
http://www.tandfonline.com/doi/abs/10.1080/01626620.1996.10462828
Hew, K. F., & Cheung, W. S. (2010). Use of three‐dimensional (3‐D) immersive virtual worlds
in K‐12 and higher education settings: A review of the research. British Journal of
Educational Technology, 41(1), 33–55.
IBM, C. (2016). IBM SPSS Statistics for Windows, Version 24.
Johnson, H. (2007). Dialogue and the construction of knowledge in e-learning: Exploring
students’ perceptions of their learning while using Blackboard’s asynchronous discussion
42
board. European Journal of Open, Distance and E-Learning, 10(1). Retrieved from
http://www.eurodl.org/?p=archives&year=2007&&article=251
Kafai Y.B., World of whyville: An introduction to tween virtual life, Games and Culture, 5(1)
2010.3-22.
Kafai, Y. B., Fields, D. A., & Cook, M. S. (2010). Your second selves: Player-designed avatars.
Games and culture, 5(1), 23-42.
Kop, R., & Hill, A. (2008). Connectivism: Learning theory of the future or vestige of the past?
The International Review of Research in Open and Distributed Learning, 9(3). Retrieved
from http://www.irrodl.org/index.php/irrodl/article/view/523/1103
Kuznetcova, I., Teeple, J., & Glassman, M. (2018). The dialectic of the avatar: Developing in-
world identities in Second Life. Journal of Gaming & Virtual Worlds, 10(1), 59–71.
https://doi.org/10.1386/jgvw.10.1.59_1
Lin, N. (2002). Social capital: A theory of sosical structure and action (Vol. 19). Cambridge
University Press.
Lin, N., Cook, K. S., & Burt, R. S. (2001). Social capital: Theory and research. Transaction
Publishers.
Liu, W., Sidhu, A., Beacom, A. M., & Valente, T. W. (2017). Social network theory. In The
International Encyclopedia of Media Effects (pp. 1–12). Retrieved from
http://doi.wiley.com/10.1002/9781118783764.wbieme0092
Lorenzo, C. M., Ángel Sicilia, M., & Sánchez, S. (2012). Studying the effectiveness of multi-
user immersive environments for collaborative evaluation tasks. Computers and Education,
59(4), 1361–1376. Retrieved from http://dx.doi.org/10.1016/j.compedu.2012.06.002
43
Merchant, Z., Goetz, E. T., Cifuentes, L., Keeney-Kennicutt, W., & Davis, T. J. (2014).
Effectiveness of virtual reality-based instruction on students’ learning outcomes in K-12
and higher education: A meta-analysis. Computers and Education, 70, 29–40.
Mezirow, J. (1997). Transformative learning: Theory to practice. In P. Cranton (Ed.),
Transformative learning in action: Insights from practice. New Directions in Adult and
Continuing Education, 74. San Francisco: Jossey-Bass.
Mikami, A. Y., Griggs, M. S., Reuland, M. M., & Gregory, A. (2012). Teacher practices as
predictors of children’s classroom social preference. Journal of School Psychology, 50(1),
95–111.
Quade, D. (1967). Rank analysis of covariance. Journal of the American Statistical Association,
62(320), 1187–1200.
Reznitskaya, A., Kuo, L. J., Clark, A. M., Miller, B., Jadallah, M., Anderson, R. C., & Nguyen‐
Jahiel, K. (2009). Collaborative reasoning: A dialogic approach to group
discussions. Cambridge Journal of Education, 39(1), 29-48.
Rienties, B., Giesbers, B., Tempelaar, D., Lygo-Baker, S., Segers, M., & Gijselaers, W. (2012).
The role of scaffolding and motivation in CSCL. Computers & Education, 59(3), 893-906.
Ruhnau, B. (2000). Eigenvector-centrality - a node-centrality. Social Networks, 22(4), 357–365.
Ryan, R. M. (1991). The nature of the self in autonomy and relatedness. In G. R. Goethals & J.
Strauss (Eds.), Multidisciplinary perspectives on the self (pp. 208–238). New York:
Springer-Verlag.
Ryoo, J., Techatassanasoontorn, A., Lee, D., & Lothian, J. (2011). Game-based InfoSec
education using OpenSim. Proceedongs of the 15th Colloquium for Information Systems
44
Security Education, 101–106.
Scott, J. (1991). Social network analysis: A handbook. London: Sage.
Scott, J. (2017). Social network analysis. Sage.
Serdiouk, M., Rodkin, P., Madill, R., Logis, H., & Gest, S. (2015). Rejection and victimization
among elementary school children: The buffering role of classroom-level predictors.
Journal of Abnormal Child Psychology, 43(1), 5–17.
Siemens, G. (2014). Connectivism: A learning theory for the digital age. International Journal of
Instructional Technology and Distance Learning, 1, 1–8. Retrieved from
http://er.dut.ac.za/handle/123456789/69
Smith, M. K., Wood, W. B., Adams, W. K., Wieman, C., Knight, J. K., Guild, N., & Su, T. T.
(2009). Why peer discussion improves student performance on in-class concept
questions. Science, 323(5910), 122-124.
Stahl, G. (2006). Sustaining group cognition in a math chat environment. Research and Practice
in Technology Enhanced Learning, 1(02), 85-113.
Steinkuehler, C. A. (2006). Massively multiplayer online video gaming as participation in a
discourse. Mind, Culture, and Activity, 13(1), 38–52.
Team, R. D. C. (2008). R: A language and environment for statistical computing. Vienna,
Austria: R Foundation for Statistical Computing.
Thorne, S. L., Black, R. W., & Sykes, J. M. (2009). Second language use, socialization, and
learning in internet interest communities and online gaming. Modern Language Journal.
Vogel, J. J., Vogel, D. S., Cannon-Bower, J., Bowers, C. A., Muse, K., & Wright, M. (2006).
Computer gaming and interactive simulations for learning: a meta-analysis. Journal of
45
Educational Computing Research, 34(3), 229–243.
Vygotsky, L. (1962). Thought and language. Cambridge: M.I.T. Press Massachusetts Institute of
Technology.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes.
Cambridge, MA: Harvard University Press.
Wang, F., & Burton, J. K. (2013). Second Life in education: A review of publications from its
launch to 2011. British Journal of Educational Technology.
Wassermann, S., & Faust, K. (1994). Social network analysis: Methods and applications.
Cambridge, UK: Cambridge University Press.
Yu, F.-Y., Hsieh, H.-T., & Chang, B. (2017). The potential of Second Life for university
counseling: A comparative approach examining media features and counseling problems.
Research and Practice in Technology Enhanced Learning, 12(1), 24. Retrieved from
https://telrp.springeropen.com/articles/10.1186/s41039-017-0064-6
Zhong, Z. J. (2011). The effects of collective MMORPG (Massively Multiplayer Online Role-
Playing Games) play on gamers’ online and offline social capital. Computers in Human
Behavior, 27(6), 2352–2363.
46
