Please do not cite or circulate the printed version of this HTML document. Please refer instead to the original online version of this document, which is available online at , or contact the author(s).

Original Research Article

Following UvAMOOC on Twitter

A Network Analysis of a Massive Socio-Technical and Cross-Platform Online Learning Environment

Tessa de Keijser, Fernando N. van der Vlist [alphabetical]

(Graduate School of Humanities,) University of Amsterdam, the Netherlands

Working paper

Published online: 1 February 2014

Abstract

In this article we argue for a conceptualisation of MOOCs as socio-technical and cross-platform learning environments. We explore different competing visions and expectations of a contested space, and observe the kind of actors that are actually attracted by these visions and expectations to grasp what kind of learning environment these actors help constitute. To do this, we trace distance learning and interactivity as two initially separate historical and conceptual lineages of contemporary educational technology, and paid careful attention to the particular visions of learning these technologies carry with them. We argue that MOOCs build on Web 2.0 as a platform (O'Reilly 2005) as well as on online social networks to conjure up a specific socio-technical learning environment. We specifically focus our attention on Twitter, and how this microblogging platform is deployed as part of massive online learning environments. To operationalise this research, we have devised an outline for an analytical framework for “following Twitter”, both as medium and as a network of actors (Latour [2005] 2007). This framework would not just render productive particular issues such as uncertainties and instabilities with which we are presented, but also allow us to “put ahead of us” that which we want to study. Consequently, it enabled us to do a network analysis and describe what is actually happening inside the “data-space” as it has been captured by the DMI-TCAT, a device we deploy to “follow the medium” (Rogers 2009) in our acquisition of Twitter data. Ultimately, we debunk some of the idealised visions and inflated expectations following from Web 2.0 “possibility thinking” in the case of our pilot study of UvAMOOC on Twitter. This pilot study then caused us to reconsider the concept of “context collapse” (Marwick and boyd 2010) and the co-existence of a range of actors involved to different degrees within the same “data-space”.

Keywords

educational technology, learning environment, digital methods, network analysis, Twitter, UvAMOOC

educational technology
learning environment
digital methods
network analysis
Twitter
UvAMOOC

1. Introduction

Massive Open Online Courses (MOOCs) constitute, above all, an idea in development. The first MOOCs emerged from the Open Educational Resources (OER) movement in response to George Siemens' course called Connectivism and Connective Knowledge (CCK08).1 In principle, an MOOC aims for large-scale interactive participation and open access via the Web and thus rely on the support of computer and network technologies. They build upon the Web as a platform (O'Reilly 2005) and online social networks for interactive forms of community building, engaging with open knowledge, and e-learning. Attending such an MOOC does not just entail watching recorded video lectures, but also searching for answers, participating in discussion, helping other students at the designated forums, or peer-reviewing graded assignments. As a digital practice and new kind of learning environment in progress, with accompanying learning techniques and devices inscribed with particular visions about education and knowledge, MOOCs are of special interest to us.

By way of context, it should be mentioned that in the middle of the hype, The New York Times had optimistically exclaimed the year 2012 to be “The Year of the MOOC” (Pappano 2012) and Time magazine expressed that MOOCs would make education more democratic (Ripley 2012). The subsequent year, however, has also seen quite a few critiques. Will Oremus (2013) criticised free online classes for their focus on replacing teachers and classrooms, instead of improving them in the first place. Similarly, Bogost (2013) critiqued the radical innovation MOOCs seem to promise by arguing that the traditional learning system has not been “inverted” or “flipped”, so much as abstracted it. Dan Colman (2013), founder of Open Culture (a “high-quality cultural & educational media for the worldwide lifelong learning community”) identified a number of reasons why readers had not finished their MOOCs, and found as problems that it takes too much time, either assumes too much or too little knowledge, lecture fatigue (relying on formal video lectures), bad communication tools, bad peer review and trolls, hidden costs, and the idea that you are not there for the credential at the end. The “massive” scale in itself also introduces logistical problems related to evaluation and assessment (Head 2013). Brent Chesley, who teaches 17th- and 18th-century British literature at Aquinas College, evaluated MOOCs based on his own experience of teaching such a course. He argues for the combination of conventional structures with online additional resources to enhance the class (Chesley 2013). Finally there are also critiques relating to centralisation. Google and edX created spin off Web sites for “the rest of us”, thus critiquing MOOC providers for only recruiting elite, high-profile institutions and their professors (Ripley 2012; Kolowich 2013). So-called distributed open collaborative courses (DOCCs), developed by feminist professors under the name FemTechNet, emerged in response issues of centralisation by challenging the role of the instructor, money, hierarchy, and scale (i.e. the value of “massive”). For instance, the consequence of some “best” professors becoming worldwide superstars educating the rest of us, and the discouragement of group-learning in MOOCs (Jaschik 2013).

If anything, the critiques and enthusiasm indicate that MOOCs mark an emergent and contested space, filled with competing visions and inflated expectations. In this article, we want to explore these different competing visions and expectations, and suggest further consideration of their implications for what it means to know and learn through these digital practices. At the same time, we want to investigate what kind of actors are actually attracted by these visions and expectations, and what kind of specific socio-technical learning environment these actors help constitute. By means of reducing our broad scope, we narrow our focus to a small-scale pilot study of UvAMOOC on Twitter over the past six months, which we delineate based on the data acquisition methods we will be using.2 Such an investigation allows us to insist that MOOCs should be understood as a socio-technically specific digital practice. Rather than a complete and polished object of study, MOOCs and educational technologies more generally are constantly in-the-making. Moreover, at the very least it will open up consideration of what it means to know and learn in a digital economy (McAuley et al. 2010), and will allow us to put in perspective and perhaps move beyond the “MOOC hype” (Yang 2013; Young 2013; Russell 2013).

In what follows, we will first trace and weave together to separate historical and conceptual lines of descent of contemporary educational technology as they relate to each other. In particular, we will focus on the rhetorics and discursive “work” (Gillespie 2010) that come with certain technologies and which cause these technologies to attract some specific actors or groups and not others. Second, after having established the specificity of MOOCs as the most recent form of educational technology, we will focus our attention on the particular relationship between microblogging platform Twitter and how it is deployed as part of massive online learning environments. Third, we outline our analytical framework for “following Twitter” and render productive the uncertainties and instabilities with which we are presented when doing a kind of real-time research that takes Twitter as an object of study. The techniques and devices we deploy will be followed (to borrow from Bruno Latour) so that we may “put ahead of us” the socio-technical specificity of that which we want to study. Fourth, using network visualisation, we then gain a way to control and render visible the “data-space” as it is captured. At this point, we finally allow ourselves to describe these empirical materials and discuss what actually happens in that data-space in the form of a pilot study of UvAMOOC on Twitter.3

2. Tracing Technology/Education

There is a long-standing debate on how technologies can fundamentally alter modes of learning and teaching. Arguably, contemporary ideas and innovations in the field of educational technology (EdTech) concerning the mutual shaping of technology and education, can be traced back in history through a number of lineages or lines of descent. We want to focus on one of these lineages, an understanding of which we argue is crucial to explain the socio-technical specificity of the most recent forms of educational technology, and which helps us situate and contextualise some of the many terms used to describe, or associate with, broadly synonymous developments. This particular lineage connects two previously separate conceptual and historical lines of descent: the first is distance education and the second is the concept of interactivity as it relates to particular visions of the relationship between technology and education.

A first line of descent to be traced can be located in the history of distance education. As described by Holmberg (2008), “Organised distance education in the form of correspondence instruction can . . . be dated back to the eighteenth and nineteenth centuries, but letter writing for the purpose of teaching is probably as old as the art of writing itself.” (13). In fact, one of the earliest examples of distance education comes from Isaac Pitman, who in 1844 delivered a series of shorthand courses, delivered on a weekly basis by correspondence and thus using the new postal system, which was in turn enabled by quicker travelling and communication afforded by the new rail system (Tait 2003; Holmberg 2008, 13). Crucially, Pitman would correct the work of his students and send it back to them (13), thus maintaining a cyclical learning procedure. Later in the late 1930s, as educational sociologist Lloyd Allen Cook (1938) has written, television and radio would be used as “master teachers” for more effective (that is mechanised) one-to-many distribution of broadcast lessons. This is already quite similar to some aspects of more recent forms of distance education and their particular use of technology. Then during the 1960s, the British Open University was founded, which embodied a similar vision of independence and of developing new technologies to improve part-time and/or distant learning programmes in higher education (cf. Nasseh 1997). The Open University helped popularise the idea of distant learning in its particular form across the world.

A second line of descent is found in the history of the concept of interactivity. In his self-published manifesto Computer Lib/Dream Machines ([1974] 2003), Ted Nelson described his vision of the potential of the personal computer for educational application. He compared and criticised ordinary teaching and the then-emerging idea of computer-assisted instruction (CAI) for preventing the student to directly interact and engage with subject matter (308–310). In the former case, the teacher would be the obstacle between the student and the subject matter, and in the latter case the computer becomes that obstacle. He therefore modestly proposed the idea of “responding resources” – of “facilities” and “hyper-media” – to give free play to the student's initiative, permitting the student to control the system and find his own way through the resources available to him (312–313). This is, of course, in stark contrast to “devising elaborate systems permitting the computer or its instructional contents to control the situation.” (313). This is one example of how interactivity, as a concept, thus gradually became positively associated with autonomy and control outside of the realm of cybernetics. Following the trail of breadcrumbs, roughly thirty years later Andrew Barry (2001) analysed the “politics of interactivity” and criticised the “moral preoccupation with the importance of scientific and technological citizenship” (127, emphasis in original) through his observations of how the concept of interactivity was used for educational purposes by museums of science. Specifically, this analysis reflects the political belief of these museums in “active” modes of learning, which “promises to turn the unfocused visitor-consumer into the interested, engaged and informed technological citizen.” (129). Furthermore, Barry calls upon Deleuze to argue that “Today, interactivity has come to be a dominant model of how objects can be used to produce subjects. In an interactive model, subjects are not disciplined, they are allowed.” (129). Interactivity is more than just a technical feature of a system, and associates particular ideals of active scientific and technological citizenship.

From the late 1990s to the early 2000s, a wide array of largely similar notions was introduced which are now often categorised under virtual education (Stallings 1997; Starr 1998; Farrell 1999, 2001; Kurbel 2001; O'Donoghue et al. 2001; Kurbel and Pokhomov 2002; Barbera 2004) or e-learning, and which starts weaving together the two previously separate lineages. On Wikipedia, we find the following list of alternative names: “multimedia learning, technology-enhanced learning (TEL), computer-based instruction (CBI), computer-based training (CBT), computer-assisted instruction or computer-aided instruction (CAI), internet-based training (IBT), web-based training (WBT), online education, virtual education, virtual learning environments (VLE) (which are also called learning platforms), m-learning, and digital educational collaboration.” (Wikipedia, “E-Learning”). As is noted, these alternatives are broadly synonymous and overlap in their use, but they do put emphasis on method of distribution, or particular components or aspects. In e-learning, for instance, a distinction between synchronous or asynchronous learning is generally used (Wallace 1991; Starr 1998; Brown 2001; Farrell 2001; O'Donoghue et al. 2001; Ebner 2007; Hrastinski 2008). Where former puts emphasis on learning and collaborating in real-time, the latter emphasises self-discipline and self-paced learning, as a result of which one is less dependent on other participants. In a similar vein, computer-based training (CBT) is based on a linear model of learning (Knowles 1975; Tough 1971). Perhaps, as Vera Cerbantez (2011) argues, “due in part to the general shift in education from primarily direct instruction to more constructivist online learning environments, early linear models … have been joined by later models involving more interactivity” (63; see also Brockett and Heimstra 1991; Danis 1992; Garrison 1997). In practice, each of these particular forms of virtual education refers to instructional courses distributed with network technology via the Internet, and their “virtuality” lies in the substitution of the physical classroom and the accompanying face-to-face experiences in cyberspace. As such, the cyber-libertarian discourse of the same period strongly resonates in this particular phase of development. This can be seen in Dees Stallings' speculative comments on the Virtual University (1997) and the surrounding discussion of whether the new education environment will be non-profit or for-profit, the answer of which was thought to depend in part on compatibility of the profit motive with the concept of quality education, and how well corporate cultures and academic cultures would be able to work together (271).

From the early to late 2000s, this discussion was absorbed within the broader turn towards Web 2.0, the Web as platform (O'Reilly 2005), and online social networks. Within the broader category of computer-supported collaborative learning (CSCL), “classroom 2.0” and “e-learning 2.0” in particular have emerged from this turn characterised by “possibility thinking” and a new interest from the corporate world (Downes 2005; Ebner 2007; Karrer 2007). In both cases, the professor is not the single and principle source of knowledge and skill anymore, and the new forms of collaborative learning that are materially supported by computer and network technology emphasise learning from and with each other through participation. In many ways, this notion is rhetorically deployed in a similar way as interactivity was before it, but importantly it turns the concept from a technical into a social direction. With the introduction of the idea of “social learning” with software, the emphasis more strongly shifted from what to how we are learning: “The emphasis on building communities of students and scholars, as much as on providing access to educational content.” (Brown and Adler 18, 26). Moreover, John Seely Brown and Richard Adler also argue that the notion is “based on the premise that our understanding of content is socially constructed through conversations about that content and through grounded interactions, especially with others, around problems or actions.” (2008, 18, emphasis in original; see also Johnson 2007). It is in this tradition of educational technology and “possibility thinking” that MOOCs have emerged as well. Similar to the examples above, MOOCs build on top of existing computer and network technologies, the Web as platform and online social networks. But as “platforms”, they also position themselves in a specific way, as revealed through the discursive “work” (Gillespie 2010) that the new name does. Alexander McAuley et al. (2010) write that the MOOC environment is “open and distributed, iterative and participatory” (46). This is also reflected in their writing about MOOC environments as “self-guided”, “that a participant is willing to engage”, “the ability to self-evaluate”, autodidactic, “connecting with others”, “extending beyond one's own cultural context”, “diversity of ideologies”, “cross-pollination”, and other similar terms. Others have characterised it with terms such as “inversion” or “flipped classrooms” (Mazur 2009; Bruff et al. 2013; Haggard 2013; Oremus 2013), “condensed classrooms” (Bogost 2013) “openness”, “access”, “connectivism” and “sharing” (Mackness et al. 2010; McAuley et al. 2010; de Waard et al. 2011; Cheverie 2013; Purser et al. 2013; Carr 2013; Haggard 2013). There are, in short, quite a few different and sometimes competing rhetorics that are used next to each other, indicating that within these learning environments, some forms of learning are privileged over others.

3. Twitter and Massive Online Learning Environments

We needed to develop a lineage in order to be able to historicise and understand why MOOCs are the way they are, that is their specificity, and to now make our case of why we need to focus our attention specifically on MOOCs as socio-technical and cross-platform4 learning environments. McAuley et al. (2010) have observed that “a MOOC integrates the connectivity of social networking, the facilitation of an acknowledged expert in a field of study, and a collection of freely accessible online resources.” (4). More importantly, Timo van Treeck and Martin Ebner (2013) argue that “in this context, MOOCs are usually spread across the world through existing social networks, mostly Facebook and Twitter. Consequently, these platforms are not only used before, but also during and after a lecture session.” (412). Continuing their argument, they claim that Twitter, and microblogging platforms in general are used in education for various reasons. Ebner (2013) points out the following six uses:

Enhancing interaction in mass education through the use of Twitter walls; Discussion beyond face-to-face lectures by using a specific Twitter hashtag; Exchanging lecture content by collecting Internet resources using a defined Twitter hashtag; Documentation and information retrieval, with the help of specific Web applications that collect tweets automatically; Enhancing academic conferences by using Twitter as an online backchannel; Connecting with researchers, teachers, and learners with similar interests based on Twitter's recommendations. (qtd. in van Treeck and Ebner 412)

These uses, and the ability to bring users together into communities and subcommunities through platform-specific features such as hashtags, mentions and following, make Twitter a particularly interesting component of massive online learning environments. In fact, van Treeck and Ebner's study analyses two particular MOOCs to evaluate the usefulness of Twitter as a microblogging platform for learning in massive communities. Using hashtag analysis, they were able to show that 70% of tweets were directly related to the course, 39% of those refer to specific topics, and 31% to affiliation of the course.

The particular relationship between MOOCs and social media platforms has sparked new interest in researching the implications of this socio-technical environment for learning. Emily Purser, Angela Towndrow and Ary Aranguiz (2013) explored the realisation of peer-to-peer learning by embracing networked social media to “tame” MOOCs and completely bypass traditional notions and tools of online learning support. Twitter hashtags (#edcmooc), Google Plus and blogs were suggested as useful for connecting with one another, and from there on students could create further subgroupings (potentially also in other social media). Sui Fai John Mak, Roy Williams and Jenny Mackness (2010) studied specifically the role of blogs and forums as facilitators for communication and learning tools in an MOOC, and found that the use of these platforms to a large extent correlated with specific individual learning styles and media affordances. Use of blogs was associated the ability to create personal space for personal learning, quiet reflection and personal relationships, and forums were associated with fast paced challenging interaction, building learning communities, open discussion, and more links to broader themes and the bigger picture (275). Derek Bruff et al. (2013) investigated student perceptions of attempts to blend MOOCs with conventional models into “blended learning”, and advocate for drawing from multiple MOOCs as well as other online sources (see also van Weigel 2001). As also shines through in the reflection of Brent Chesley (2013), the success is dependent on the course design, and so naturally, MOOCs and the particular platforms they require to support them will not work the same or with the same success for every discipline or study. Peter Tiernan (2012) argued for the importance of interaction and engagement for learning and investigated issues relating to the use of Twitter to increase such interaction and engagement for students in lecture discussion such as technological barriers and lack of integration with regular courses. The study concluded that Twitter indeed has vast potential for continuing engagement in ongoing conversations during lectures. Moreover, he found that students who would ordinarily remain silent are now given an opportunity to express their opinions and experiences. Relating to the characteristics of MOOCs introduced earlier, Jenny Mackness, Sui Fai John Mak and Roy Williams (2010) observe a paradox: “The more autonomous, diverse and open the course, and the more connected the learners, the more the potential for their learning to be limited by the lack of structure, support and moderation normally associated with an online course, and the more they seek to engage in traditional groups as opposed to an open network.” (275). Finally, Martin Ebner et al. (2010) have analysed tweets around the #edmedia10 hashtag, which was used during a specific e-learning conference. Trying to find out whether these tweets could be taken as a basis for further semantic analysis, they show that it is indeed possible to get meaningful outcomes for filtering important content or users.

It should be clear by now that how Twitter is actually used can be very different: some use it only for following particular accounts or as information resource, others use it only for posting excessively about their own lives, and probably most users are somewhere in between or use it in still other ways. Based on this observation, Kate Crawford (2009) has suggested listening as metaphor for paying attention online and has identified “background listening” (528), “reciprocal listening” (529), and “corporations ‘listening in’” (531) as modes of listening on Twitter and has considered how these are experienced and performed by individuals, politicians and corporations each with their own interests. The purpose of using Twitter, then, may also be quite different: for example, a politician might want to influence the reception of a story and shift public opinion, teens may want to share details of their personal lives in a cry for attention, and educational institutions may want to distribute valuable quality resources, or mine Twitter data to conduct “learning analytics” (Duval 2011; Long and Siemens 2011; Knight et al. 2013). Studies like these can be situated in what has been called Twitter Studies, and of which Richard Rogers (2013) has suggested a three-step periodisation based on how Twitter has been “de-banalised” and conceived as an object of study. Starting from its initial development in 2006 until roughly 2009, Twitter research could mostly be characterised as an ambient friend-following and messaging utility, or an “urban lifestyle tool” with many tweets on mundane (that which is “of the earth”) activities such as lying in bed or taking a walk in the park. Research on Twitter in this first tradition (“Twitter I”) is often directed at determining the value of tweets and the degree to which messages are “banal”, thus focussing mostly on the content (357). Characterised by a change of the tagline from “What are you doing?” to “What's happening?” in 2009, a second phase or tradition (“Twitter II”) conceives of Twitter as an “event-following tool”, a way to monitor, for instance, elections or disasters (359). But this stage of Twitter research is now shifting towards a third phase or tradition (“Twitter III”) that is concerned with Twitter as it is settling into a massive data set (362). As data set it is of value to itself but is also an object of cultural heritage as indicated by the fact that as of April 2010, their public tweets are now being archived and preserved by the Library of Congress (“Twitter Donates Entire Tweet Archive to Library of Congress”) so as to “preserve and provide access to a universal collection of knowledge and the record of America's creativity for Congress and American people.” (“Update on the Twitter Archive At the Library of Congress”, 5, emphasis added).

4. Following Twitter and the Actors Themselves

Uncertainties such as those mentioned above concerning our object of study should prevent us from artificially imposing categories on an object of study that is “behind us”. Instead, drawing from Bruno Latour's work on actor-network-theory (ANT) and the work on “digital methods” (Rogers 2013), we suggest to “follow the medium” (Rogers 2009, 10) and “the actors themselves” (Latour [2005] 2007, 12). This may be the only way to “put ahead of us” – and suspend judgement about – that which want we observe or study (cf. 171). In the following, we establish an analytical framework for “following Twitter”, which not only allows us to work alongside or with these problems of uncertainty, ephemerality and instability introduced in some detail above, but which even more fundamentally allows us to empirically study the actual deployment of Twitter in massive open online learning environments. Thus, we are not only following the medium, but also the social space it conjures up.

By means of a first step, there are some important limitations that arise from conceptualising Twitter as “data provision machine” (Rogers 362). Most notably, there are issues for real-time research caused by Twitter's ephemerality (Back, Lury, and Zimmer 2013; Elmer 2013), and issues of limited accessibility to some aspects of the data especially for researchers that are not collaborating or affiliated with Twitter. Regarding these limitations, Noortje Marres and Esther Weltevrede (2013) offer an operational analytical framework as they investigate the device of scraping. In particular, they show how scraping imports analytical presumptions into the research, and that scrapers import data that is already pre-formatted. Consequently, they raise the question of whether their research helps “to bring into view the dynamics of media platforms or also those of the social phenomena they enable?” To operationalise that question, they suggest to distinguish between two kinds of real-time research: “those dedicated to monitoring live content (which terms are current?) and those concerned with analysing the liveliness of issues (which topics are happening?).” (356, emphasis in original). Where live social research “seeks to render analytically productive the formatted, dynamic character of digital networked data” (15), which means that it makes possible the study of changing “natively” digital objects, research on liveliness aims to capture what is currently happening in a particular social space; which actors are the most active and make things happen in the “eternal now”. By “following the medium”, and repurposing its devices and techniques or “natively digital” objects, it is better prepared to handle such unstable objects. According to Rogers,

Twitter is particularly attractive for research owing to the relative ease with which tweets are gathered and collections are made, as well as the in-built means of analysis, including RT (retweets) for significant tweets, #hashtags for subject matter categorization, @replies as well as following/followers for network analysis and shortened URLs for reference analysis. Given its character limit and the fact that each tweet in a collection is relatively the same length, it also lends itself well to textual analysis, including co-word analysis. (2013, 363; see also Marres and Weltevrede 2013)

Each of these elements are “native” or device-specific, and can to a specified degree be accessed through application programming interfaces (APIs). This is part of what Cornelius Puschmann and Jean Burgess (2013) dub the “politics of Twitter data”: the negotiations of the platform provider with the end users through terms of service and application programming interfaces. This device-specificity of the data has implications for our research because it means that it is pre-formatted (which is at the same time what makes it specific) in an open data-interchange standard called JavaScript Object Notation (JSON).

To leverage or rather follow this specificity, we relied on the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT), an open-source toolset for the capturing of tweets and that allows for multiple types of analysis (e.g. hashtags, mentions, users, search). As Erik Borra and Bernhard Rieder explain, the Toolset enables epistemic plurality and provides “robust and reproducible data capture and analysis, and interlinks with existing analytical software.” (“Programmed Method”, 1). Moreover, it calls attention to methodological, epistemological, logistical, legal, ethical, political issues arising from the use of software in the study of (proprietary) data extracted from social media platforms (2). Thus, while the DMI-TCAT offers us a way around some of the limitations and restraints of similar tools, it is at the same time sensible to methodological repercussions. For data acquisition, the DMI-TCAT is built on top of the Twitter Streaming and REST APIs and as such is bound to their possibilities and limitations (7–8). But because Borra and Rieder have remained as close as possible to the “native” formats of the data, so as to introduce as little additional bias as possible, their toolset perfectly aligns with our present aim to “follow Twitter”. From the 515,140,404 tweets that have been captured and archived by the Toolset so far, the particular dataset we explore is composed of the metadata of 406,464 tweets containing either “mooc” or “uvamooc” since 7 March 2013. Within that set, we queried [2013-06-03] as the starting data, and [2013-12-03] as the end date to get a smaller subset spanning over the past six months leading up to our analysis. The subset contains 106,316 distinct users and 278,685 tweets, of which 72.6% (202,302) contain links and 27.4% (76,383) do not. Looking at this data with the built-in timeline feature presents us with a regular pattern of peaks and troughs, with a smaller overall number of tweets during the summer holidays roughly from June up until the end of August (Figure 1a). Filtering this dataset further, we queried [uvamooc] so that only tweets containing that keyword would remain. Only a miniscule portion of the complete dataset remained, and we are presented with a completely different pattern, as well as with a surprisingly much lower number of tweets containing links (Figure 1b). This subset contains only 79 distinct users and 151 tweets, of which 21.2% (32) contain links and 78.8% (119) do not (Table 1).

Fig. 1a. Overview of the “mooc” dataset selection generated by the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT). Screenshot created on 13 Dec. 2013, 16:56:16.

Fig. 1b. Overview of the “uvamooc” data subset selection generated by the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT). Screenshot created on 13 Dec. 2013, 16:56:03.

Table 1. DMI-TCAT output comparison.

Dataset KeywordsNumber of Distinct UsersNumber of TweetsTweets Containing LinksTweets Containing No Links
mooc, uvamooc106,316278,685202,30276,383
uvamooc7915132119

Data: DMI-TCAT (4a05476c87).

Although helpful for an overview like this, the built-in analytics tools of the DMI-TCAT fall short if we actually want to find out more about the actors and their associations, the groups, chains and conversations in which they take part. For further tracing, we therefore relied on Gephi, which is an open-source software package for network visualisation and analysis. Network analysis allows us to continue following Twitter by empirically looking at the captured data-space from the ground up and to observe what emerges as significant from or within that space. Consequently, we use it for a kind of interactive or dynamic probing: the iterative process of changing parameters and observing what happens when applying various structuring algorithms that alter the relative positions of vertices or nodes and their edges; thus exploring how dimensions are generated and maintained on a “flatland” (cf. Latour 172). On such a “flatland”, space does not come from volume, but rather from differences such as sizes and distances, as these are captured by the data (hence the hyphen in “data-space”). This is the main reason why it is has been so important that we follow the medium. Continuing from here, those differences relating to spatialisation are generated by so-called layout algorithms, for which we relied on Mathieu Jacomy's built-in ForceAtlas2, which is a continuous force-vector graph layout algorithm based on a linear-linear model (Gephi.org, “ForceAtlas2, the new version of our home-brew Layout”). Because force-directed algorithms are actually physical simulations, it is able to run homogeneously while displaying the “live” result of the spatialisation and attraction and repulsion between nodes are proportional to the distance between these nodes (Jacomy et al. 2012, 3; see also Kobourov 2013). For our visualisations and analysis, this means that the edges between the nodes are assigned forces relative to their positions, thus allowing us in particular to distinguish and compare between subcommunities and the strength of “ties”. It also means that these visualisations are compromised by each single irrelevant node that does not actually belong in the dataset but ended up there simply because it matched the initial parameters. In short, network visualisation offers us a way to “put ahead of us” the actors as they assemble on a flat space, which ensures that their elasticity – their “bending, stretching, squeezing” (Latour 173) – will be rendered visible.

5. Pilot Study: Following UvAMOOC on Twitter

In what follows, we will be investigating the second key component of our study, which concerns a small-scale pilot study of UvAMOOC on Twitter. What kind of actors are actually attracted by these visions and expectations, and what kind of specific socio-technical learning environment do these actors thus help constitute? For this analysis, we have used three simple yet powerful sets of techniques that we will describe through a rather “dry telling” of settings and their results, followed by a somewhat broader discussion of the implications, structured into two themes: user affiliations, and the “field” in which these users are active in relation to the MOOC. A first set of techniques was used to find the important nodes, clusters and subclusters in this dataset. We exported a spreadsheet file from the DMI-TCAT containing metadata on the users in that group (Table 2) so that we could identify the most active users by looking at the number of tweets these users have posted. Subsequently, by looking at the user profiles and following the trail of information that users put on there, as well as following the hyperlinks on these profiles and looking up these users Google Search or on LinkedIn, we were able to find out which cause or affiliation these user are affiliated with, if any, and what kind of relation they have to UvAMOOC (Table 2–4). A second set of techniques concerns the analysis of groups of users. We exported different network files from the DMI-TCAT with metadata on users and their mentions so that we could produce a directed social graph by mentions based on those interactions between users. Whenever one user mentions another user with @reply, a directional relation is created. The more often this happens, the stronger the link becomes (link weight). Furthermore, nodes were scaled according to their number of mentions (“no_mentions”), and colours follow a heat map scheme based on the number of tweets (Figure 2a; Table 5). We then ran the PageRank statistics algorithm to calculate a recursive status or authority ranking (cf. Rieder 2012) on top of this, which gives us a kind of measured account of socially acknowledged authority based on those @replies (Figure 2b; Table 5). This was then followed by running the network diameter statistics algorithm to calculate betweenness centrality for the users in the same set (Figure 2c; Table 5). This measure calculates which users are of highest value for bringing together different clusters of users, which is a powerful technique to find those few users that connect otherwise closed networks to each other. A third set of techniques was used to create an overview of tweets that have been retweeted using the DMI-TCAT built-in option to list each individual retweet. Furthermore, here we have looked up the original Twitter users behind the retweeted content manually (Table 6). This way we could see whether or not there are some highly influential users whose information is retweeted by other users.

Fig. 2a. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours represent the number of tweets, which is set according to a heat map schema. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 2b. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours represent PageRank, which is set according to a heat map schema. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 2c. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours represent betweenness centrality, which is set according to a heat map schema. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Here, we will first describe the results relating to affiliations of the users in the data subset. These results are derived from categorising the different users in the dataset from the overall user statistics spreadsheet (Table 2) in terms of the affiliation they belong to, as well as the social graph by mentions through which an overview of the actors and their relations was created. These results show that the UvAMOOC data-space is comprised of user groups with quite different affiliations. In table 2 we give a plain overview of the 79 users that appear in total in the UvaMOOC dataset, with our addition of the categories of the country to which the users can be linked according to their profiles, their affiliation, type of affiliation (e.g. academic or corporate), and the field in which they work or present themselves in relation to UvAMOOC. In table 3, these users were then grouped in terms of their affiliations. As can be seen, from a total of 79 users, 46 are affiliated with an educational institution. The University of Amsterdam has 20 related users, the Hogeschool van Amsterdam has 4, other Dutch educational institutions account for 10, while non-Dutch and educational technology websites each account for 6 users. Another 11 users belong to corporations and 11 to other affiliations. 11 Users could not be classified because their affiliation was unknown (Figure 3). Again, what stands out here is that the majority of users is linked to an educational affiliation (58.2%). From the total number of educational affiliations (46), 52.2% is based in Netherlands, which is the country from which this specific MOOC originates. Moreover, there is a strong presence of parties related to the University of Amsterdam (UvA), the very institution that provides the MOOC. The UvA accounts for 25,3% of the total number of affiliations. These are either “official” Twitter accounts, employees or students at the UvA (note that these are not MOOC participants). Perhaps not too surprisingly, this institution is also highly present in the social graph by mentions, with 6 affiliated accounts in the top 10 mentioned users (Figure 2). Not only is there a Twitter account for the UvAMOOC itself (“UvA MOOC”), which appears in the middle as the biggest node here with 121 mentions (@replies), but also for the UvA itself (“uva_amsterdam”, 18 mentions), the helpdesk for students (“uva_student”, 9 mentions), which is the Dept. of the university concerned with developments in digital learning practices (“ictinnovatieuva”, 4 mentions), and for Frank Benneker (“frankbenneker”, 7 mentions) who is one of the employees responsible for innovation in education (Table 5).5 These accounts seem to form a subcluster or community, tweeting and mentioning each other in a seemingly self-referential manner. Thus, a cluster is formed around the University of Amsterdam, but importantly, this this group is not closed off from other actors in the network.

Fig. 3. Pie chart visualisation of our analysis of user affiliations. Data from DMI-TCAT (4a05476c87) (see Table 3).

The results from the retweet analysis appears to support the hypothesis that the University of Amsterdam not only forms a cluster with its own affiliated accounts, but that it also constitutes a strong influence within the UvaMOOC data-space compared to the other users in that data-space. All 8 “retweet chains” originate from a tweet initially sent by an account affiliated with the UvA. 4 chains originate from the “uvamooc” account, and 1 from “uva_amsterdam”, “uva_student”, “ictinnovatieuva” and “frankbenneker” respectively (Table 6). These chains do indicate that links exist between the MOOC provider and other users that choose to retweet. There is thus at least some degree of communication happening between the MOOC provider and its participants. Following up on this, the betweenness centrality analysis for the social graph by mentions shows that “uvamooc”, “onleon”, “frankbenneker”, “adenboon”, “carmenmanea”, “uva_amsterdam” and “ictinnovatieuva” are the most important users for bridging different subcommunities (Table 7). The top 10 nodes with the highest PageRank (Table 8, “uvamooc”, “onleon”, “carmenmanea”, “weijinti”, “julianamarques”, “frankbenneker”, “uva_amsterdam”, “uva_student”, “ictinnovatieuva”, and “patriciavdl”) also show similar results adding further to this point, as both analyses present us with users affiliated with the UvA as well as UvAMOOC participants. Importantly, however, it should be mentioned here that the UvAMOOC participants are not as strongly connected amongst themselves, thus debunking our expectation that these participants would form learning communities with their peers. Other mentioned users besides those affiliated with the UvA can for the most part be divided into either MOOC participants or educational technology commentaries (such as blog posts commenting on particular issues related to MOOCs), yet neither of these form strong subcommunities in the social graph by mentions. Therefore, based on the analysis of affiliations, we observe that UvAMOOC on Twitter is comprised of users representing different stances in the MOOC debates, but also that the MOOC provider is the most influential voice in this network.

After looking at patterns in the affiliations of the users through categorisations, we now look for patterns in the “fields” of activity of these users, while still using the same data subset. These “fields” were grouped and categorised based on our investigations of the users themselves (Table 4; Figure 4). Rather than imposing a preconfigured set of categories on these users, we thus followed the trails of information these users left behind on Twitter and on the Web to mould these categories. The categories themselves may be divided into three subgroups. First, there are results related to the UvAMOOC and the subject matter of particular courses (communications); Second, there are users that engage with educational technology as a field in itself; And third, there is a group of users that could not easily be categorised, and which are categorised here as either “Other” or “Unknown”. The first group of categories contains 39 users, which together accounts for 49.4% of the total number of users. This set contains the UvAMOOC participants, and communication theorists and advisers who provide UvAMOOC with a subject-related context (Figure 4, bluetones). The second and third groups of categories account for 21 and 19 users, respectively accounting for 26.6% and 24% of the total set. This second set in particular seems to position UvAMOOC within the broader field of developments and innovations in educational technology (Figure 4, greentones). It is worth observing, then, the specific range of activity that is occurring here. Providers of MOOCs, MOOC participants, and other users concerned with developments, commentary or research on educational technology all seem to be acting within the same Twitter data-space.

Fig. 4. Pie chart visualisation of our analysis of “fields” of user activity. Data from DMI-TCAT (4a05476c87) (see Table 4).

6. Beyond Collapsing Contexts and Inflated Expectations

In this final part, we progress by raising some of the implications relating to our main observations that the UvA itself maintains an influential presence on Twitter, and that multiple categories of actors exist and operate within the same data-space. Earlier, we mentioned Crawfords suggestion to use listening as a metaphor for paying attention online. This made us aware of how different actors or even networks of actors may be interested in the same data-space, but that these actors are not necessarily interested for the same reasons. Similarly, the notion of “peripheral participants” was introduced by McAuley et al. (2010, 41), which alludes to the notion that extra-MOOC practices are in fact blending together with actual MOOC practices. Here, we draw from and problematise Alice Marwick and danah boyd's concept of “context collapse”, which refers to the question of how content providers “navigate” the “imagined audiences” on Twitter (2010, 1). It is useful to us because it offers a solution to the problem of multiple contexts overlapping or blending within the same space, at the same time generating and solving conflicts between personal and the public modes of expression. We want to argue, however, that the “collapsing” of context is not as destructive as the term suggests. Rather than flattening out any sense of differential context into a more-or-less homogeneous space, we maintain that “context” results from differences of distance between actors. This is perhaps best illustrated with a study conducted by Ebner (2010) who investigated how useful microblogging posts (including those on Twitter) originating from a conference are to audiences of non-participants that are “listening” through these same online platforms. What he found is that in fact only a few of these microblogs were actually of interest to these non-participants. Actions are only understood and relevant within certain specific networks. Perhaps we can thus argue that the particular way in which one is involved or connected to such an event matters for the formation and delineation of communities and subcommunities. For our study, this means that MOOCs (or any other kind of similar programme for that matter), through their integration with online social networks, add yet another “sphere of activity” to those already part of one's stream. The crucial point is, then, that even while contexts are “collapsing” or flattening out, they are also re-instantiated in a different manner: not through volume but by distinctions. Furthermore, while the notion of “collapsing” contexts suggests that social networks are representative of someone's life outside of that network, we argue that it is a cross-platform space in and of itself, where socio-technically specific communities are instantiated.

Building further on this, we observed different kinds of knowledge being extracted from the same data-space. Although we reckon that our data set is limited, contrary to the statements of Brown and Adler that knowledge is socially constructed through conversations about content (2008, 18) to which we referred earlier, we in fact do not observe much conversation about content. Within the data-space we analysed, we found convincing evidence for a second kind of knowledge related to MOOC programmes themselves. Improvements to educational technology and “learning analytics” are occurring roughly at the same time as the actual courses are taught. This indicates a particular real-time kind of research practice, constituted by a feedback loop between activity of participants and the taught MOOC. This also brings us to a final point of critique. Returning to the notion of “possibility thinking” so strongly present in many Web 2.0 enterprises, it strikes us as odd that the idealised visions and expectations that so strongly characterise the “MOOC hype” seem to appear much less than we initially hypothesised. Most notably, as McAuley et al. claim, “[MOOC] facilitators need participants who create resources and share their opinions. Each act of creation is a potential node for connection.” (2010, 46). Such connectivity is thus not just technical, nor just social. Fitting with our characterisation of MOOCs as socio-technically specific and cross-platform learning environments, we conclude, then, by claiming that just because something is technologically possible does not mean that it also fits socially.

7. Conclusions

We set out to explore different competing visions and expectations of a contested space, and observe the kind of actors are actually attracted by these visions and expectations to grasp what kind of specific socio-technical learning environment these actors help constitute. We have argued for a conceptualisation of MOOCs as socio-technical and cross-platform learning environments, and debunked some of the idealised visions and inflated expectations following from Web 2.0 “possibility thinking” in the case of our pilot study of UvAMOOC on Twitter. To make such claims, we have started by tracing distance learning and interactivity as two initially separate historical and conceptual lineages of contemporary educational technology, and paid careful attention to the particular visions of learning these technologies carry with them. From this we were able to understand the relationship between MOOCs, Web 2.0 and online social networks as a specific set of qualities conjuring up an evenso specific learning environment. Here we narrowed our focus on microblogging platform Twitter, or more specifically how Twitter is deployed as part of massive online learning environments. To operationalise our research interest, we have devised an outline for an analytical framework for “following Twitter”, both as medium and as a network of actors. This would not just render productive particular issues such as uncertainties and instabilities with which we are presented, but also allow us to “put ahead of us” our object of study so as to suspend judgement and introduce as little as possible additional bias or judgement during the analysis. Our aim has been to describe what is actually happening inside (or rather “on”) the “data-space” we captured. In line with these methodological decisions, we used the DMI-TCAT for this capturing process, which allowed us to follow the technical specificity of pre-formatted data. By means of network analysis and visualisation we could then finally render visible that data-space as it was captured, which ensured that we could distinguish between differences of distance and size.

Most notably, we found that the UvA itself maintains an influential presence on Twitter, and that multiple categories of actors exist and operate within the same data-space, prompting us to reconsider in particular the concept of “context collapse” (Marwick and Boyd 2010). Rather than a “collapse” or flattening out of context, we argued that it is re-instantiated not through volume but through distinctions of distance. This pointed us to the important realisation that being connected to other users, MOOC participants or providers, does not necessarily mean that these nodes form a homogeneous network. Distance presents itself as the key to understanding how different networks, communities, subcommunities or simple chains of conversations, co-exist and overlap within the same data-space. Contexts are thus not spheres characterised by their “volume”, but rather the elasticity of relations between an actor and his networks. Following the socio-technical specificity of massive open online learning environments has helped us hold against the light the MOOC hype and debunk some of its claims and expectations for our narrow pilot study. Still, it seems a promising approach that could also be applied to larger MOOC programmes or perhaps even completely different “hyping” phenomena. The present stage in educational technology in particular maintains less of a distinction between platforms or activities of participants and their development through a cyclical process of teaching and analysis, each of which is the result of differently involved actors. We like to conclude by stating that even though this study, to some degree, has been an exercise in Big Data research, a reading of this data on a closer and smaller scale, and with simple yet powerful techniques, has helped us develop an understanding of a much broader phenomena. But at the same time, this understanding could not have been developed with only traditional forms of content analysis.

8. Endnotes

1. Well-known American providers include Udacity, Coursera and edX. European providers include Eliademy.com, Futurelearn, Open Univeristy, Iversity and OpenupEd. Open2Study and OpenLearning are well-known in Australia, EduKart in India, and Veduca in Brazil. (Wikipedia, “Massive open online course”).

2. Here, we are referring to the selection of our DMI-TCAT data subset. As explained in section 4, the subset contains metadata on tweets captured between June 6th and December 6th of this year with the keyword “uvamooc”.

3. The initiative currently offers a course called “Introduction to Grid Computing”, which started on 18 Nov. 2013. After “Introduction to Communication Science”, which started on 20 Feb. 2013, it is their second course in Communication Science. Official page: http://mooc.uva.nl/, UvA MOOC on Twitter: https://twitter.com/UvAMOOC, University of Amsterdam MOOC on Facebook: https://www.facebook.com/uvamooc.

4. Although broadly synonymous with the term multi-platform, we use this one because we are not able to clearly delineate the one from the other platform. Instead of just multiple platforms, these platforms are in fact also integrated in each other.

5. Noteworthy for his lack of appearance in the UvAMOOC data subset is Rutger de Graaf, the actual instructor of the course. He does not seem to be tweeting about the course at all.

9. References

Andrew, Barry. “On Interactivity.” Political Machines: Governing a Technological Society. London: Athlone Press, 2001. 127–152. Print.

Back, Les, Celia Lury, and Robert Zimmer. Doing Real Time Research: Opportunities and Challenges. National Centre for Research Methods Methodological Review paper. Goldsmiths College, University of London, 2013. Print.

Barbera, Elena. “Quality in Virtual Education Environments.” British Journal of Educational Technology 35.1 (2004): 13–20. Print.

Bogost, Ian. “The Condensed Classroom.” The Atlantic. The Atlantic Monthly Group, 27 Aug. 2013. Web. 3 Dec. 2013. <http://www.theatlantic.com/technology/archive/2013/08/the-condensed-classroom/279013/>.

Borra, Erik K., and Bernhard Rieder. “Programmed Method: Developing a Toolset for Capturing and Analyzing Tweets.” Unpublished Manuscript.

boyd, danah m., Scott Golder, and Gilad Lotan. “Tweet, Tweet, Retweet: Conversational Aspects of Retweeting on Twitter.” Proceedings of the 43rd International Conference on System Sciences (HICSS) (2009): 1–10. Print.

Brockett, Ralph G., and Roger P. Hiemstra. Self-Direction in Adult Learning: Perspectives on Theory, Research, and Practice. New York: Routledge, 1991.

Brown, John Seely, and Richard P. Adler. “Minds on Fire: Open Education, the Long Tail, and Learning 2.0.” EDUCAUSE Review Magazine 43.1 (2008): 16–32. Print.

Brown, Ruth E. “The Process of Community Building in Distance Learning Classes.” Journal of Asynchronous Learning Networks 5.2 (2001): 18–35. Print.

Bruff, Derek O., et al. “Wrapping a MOOC: Student Perceptions of an Experiment in Blended Learning.” Journal of Online Learning and Teaching (JOLT) 9.2 (2013): n. pag. Web. 4 Dec. 2013. <http://jolt.merlot.org/vol9no2/bruff_0613.htm>.

Carr, David F. “Udacity Hedges On Open Licensing For MOOCs.” InformationWeek. UBM Tech, 20 Aug. 2013. Web. 4 Dec. 2013. <http://www.informationweek.com/software/udacity-hedges-on-open-licensing-for-moocs/d/d-id/1111226>.

Cervantez, Vera A. “The Influence of Classroom Community and Self-Directed Learning Readiness on Community College Student Successful Course Completion in Online Courses.” Dissertation, University of North Texas, Aug. 2011. Print.

Chesley, Brent. “My Problem With MOOCs.” Inside Higher Ed. Inside Higher Ed, 6 Aug. 2013. Web. 4 Dec. 2013. <http://www.insidehighered.com/views/2013/08/06/essay-mooc-debate-and-what-really-matters-about-teaching>.

Cheverie, Joan. “MOOCs and Intellectual Property: Ownership and Use Rights.” EDUCAUSE Blog. EDUCAUSE, 8 April 2013. Web. 4 Dec. 2013. <http://www.educause.edu/blogs/cheverij/moocs-and-intellectual-property-ownership-and-use-rights>.

Colman, Dan. “MOOC Interrupted: Top 10 Reasons Our Readers Didn't Finish a Massive Open Online Course.” Open Culture. 5 Apr. 2013. Web. 3 Dec. 2013. <http://www.openculture.com/2013/04/10_reasons_you_didnt_complete_a_mooc.html>.

Cook, Lloyd Allen. Community Backgrounds of Education: A Textbook and Educational Sociology. New York: McGraw-Hill, 1938. 249–250. Print.

Crawford, Kate. “Following You: Disciplines of Listening in Social Media.” Continuum 23.4 (2009): 525–535. Print.

Danis, C. “A Unifying Framework for Date-Based Research Into Adult Self-Directed Learning.” Eds. Huey B. Long, et al. Self-Directed Learning: Application and Research. Norman: Oklahoma Research Center for Continuing Professional and Higher Education, University of Oklahoma, 1992. Print.

Downes, Stephen. “E-Learning 2.0.” eLearn Magazine. Association for Computing Machinery, Inc., Oct. 2005 Web. 15 Dec. 2013. <http://elearnmag.acm.org/featured.cfm?aid=1104968>.

Duval, Erik. “Attention please! Learning analytics for visualization and recommendation.” Proceedings of the 1st International Conference on Learning Analytics and Knowledge (LAK ’11), 2011. print.

Ebner, Martin, et al. “Getting Granular on Twitter: Tweets From a Conference and Their Limited Usefulness for Non-Participants.” IFIP Advances in Information and Communication Technology 324 (2010): 102–113. Print.

Ebner, Martin, Thomas Altmann, and Selver Softic. “@Twitter Analysis of #Edmedia10 - Is the #Informationstream Usable for the #Mass.” Form@re 74.11 (2010): 36–45. Print.

Ebner, Martin. “E-Learning 2.0 eLearning 1.0 + Web 2.0?” Proceedings of the 2nd International Conference on Availability, Reliability and Security (ARES'07) (2007): 1–5. Print.

———. “The influence of Twitter on the academic environment.” Social Media and the New Academic Environment: Pedagogical Challenges. Eds. Bogdan Pătruţ, Monica Pătruţ, and Camelia Cmeciu. Hershey: IGI-Gobal, 2013. Print.

“E-Learning.” Wikipedia, the free encyclopedia. Wikimedia Foundation, Inc., 14 Dec. 2013. Web. 14 Dec. 2013. <http://en.wikipedia.org/w/index.php?title=E-learning&oldid=586022815>.

Elmer, Greg. “Live Research: Twittering an Election Debate.” New Media & Society 15.1 (2013): 18–30. Print.

Farrell, Glen M., ed. The Changing Faces of Virtual Education. Vancouver: The Commonwealth of Learning, 2001. Print.

———, ed. The Development of Virtual Education: A Global Perspective. Vancouver: The Commonwealth of Learning, 1999. Print.

“ForceAtlas2, the new version of our home-brew Layout.” Gephi.org. The Gephi Consortium, 6 June 2011. Web. 5 Dec. 2013. <https://gephi.org/2011/forceatlas2-the-new-version-of-our-home-brew-layout/>.

Garrison, D. R. “An Analysis of the Control Construct in Self-Directed Learning.” Eds. Huey B. Long, et al. Emerging Perspectives of Self-Directed Learning. Norman: Oklahoma Research Center for Continuing Professional and Higher Education, University of Oklahoma, 1993. 27–44. Print.

———. “The Politics of ‘Platforms’.” New Media & Society 12.3 (2010): 347–364. Print.

Haggard, Stephen. “The Maturing of the MOOC: Literature Review of Massive Open Online Courses and Other Forms of Online Distance Learning.” Department for Business Innovation and Skills (BIS), 2013. Print.

Head, Karen. “Inside a MOOC in Progress.” The Chronicle of Higher Education. The Chronicle of Higher Education, 21 June 2013. Web. 3 Dec. 2013. <http://chronicle.com/blogs/wiredcampus/inside-a-mooc-in-progress/44397>.

Holmberg, Börje. The Evolution, Principles and Practices of Distance Education. Vol. 11. Oldenburg: Bibliotheks- und Informationssystem der Universität Oldenburg, 2008. Print.

Hrastinski, Stefan. “Asynchronous & Synchronous E-Learning.” EDUCAUSE Quarterly 31.4 (2008): 51–55. Print.

Johnson, Henry. “Dialogue and the Construction of Knowledge in E-Learning: Exploring Students' Perceptions of Their Learning While Using Blackboard's Asynchronous Discussion Board.” European Journal of Open, Distance and E-Learning (2007): n. pag. Web. 15 Dec. 2013. <http://www.eurodl.org/materials/contrib/2007/Henry_Johnson.htm>.

Jacomy, Mathieu, Sebastien Heymann, Tommaso Venturini, and Mathieu Bastian. “ForceAtlas2, A Graph Layout Algorithm for Handy Network Visualization.” Sciences Po Médialab, 2012. Print.

Jaschik, Scott. “Feminist Professors Create an Alternative to MOOCs.” Inside Higher Ed. Inside Higher Ed, 3 Dec. 2013. Web. 20 Aug. 2013. <http://www.insidehighered.com/news/2013/08/19/feminist-professors-create-alternative-moocs>.

Karrer, Tony. Understanding E-Learning 2.0. Learning Circuits, 2007. Web. 15 Dec. 2013. <http://www.learningcircuits.org/2007/0707karrer.html>.

Kinsley, Samuel. “Bernard Stiegler on MOOCs and Education.” samkinsley.com. 7 Oct. 2013. Web. 15 Dec. 2013. <http://www.samkinsley.com/2013/10/07/bernard-stiegler-on-moocs-and-education/>.

Knight, Simon, Simon Buckingham Shum, and Karen Littleton. “Collaborative Sensemaking in Learning Analytics.” CSCW and Education Workshop (2013): Viewing Education as a Site of Work Practice, Co-Located with the 16th ACM Conference on Computer Support Cooperative Work and Social Computing (CSCW ’13), San Antonio, Texas, USA, 2013. Print.

Knowles, M. S. Self-Directed Learning. New York: Association Press, 1975. Print.

Kobourov, Stephen G. “Force-Directed Drawing Algorithms.” Handbook of Graph Drawing and Visualization. Ed. Roberto Tamassia. Boca Raton: CRC Press, 2013. 383–408. Print.

Kolowich, Steve. “Google and edX Create a MOOC Site for the Rest of Us.” The Chronicle of Higher Education. 10 Sept. 2013. Web. 3 Dec. 2013. <http://chronicle.com/blogs/wiredcampus/google-and-edx-create-a-mooc-site-for-the-rest-of-us/46413>.

Kurbel, Karl. “Virtuality on the Students' and on the Teachers' sides: A Multimedia and Internet based International Master Program.” Proceedings on the 7th International Conference on Technology Supported Learning and Training - Online Educa (2001): 133–136. Print.

Kurbel, Karl, and Alexei Pokhomov. “Production and Delivery of Multimedia Courses for Internet Based Virtual Education.” World Congress “Networked Learning in a Global Environment” Challenges and Solutions for Virtual Education (2002): 1–44. Print.

Latour, Bruno. Reassembling the Social: An Introduction to Actor-Network-Theory. 2005. Oxford: Oxford University Press, 2007. Print. Clarendon Lectures in Management Studies.

Long, Phil and George Siemens. “Penetrating the Fog: Analytics in Learning and Education.” EDUCAUSE Review Magazine 46.5 (2011): 31–40. Print.

Mackness, Jenny, Sui Fai John Mak, and Roy Williams. “The Ideals and Reality of Participating in a MOOC.” Proceedings of the 7th International Conference on Networked Learning (2010): 266–275. Print.

Mak, Sui Fai John, Roy Williams, and Jenny Mackness. “Blogs and Forums as Communication and Learning Tools in a MOOC.” Proceedings of the 7th International Conference on Networked Learning (2010): 275–285. Print.

Marres, Noortje S., and Esther Weltevrede. “Scraping the Social?” Journal of Cultural Economy 6.3 (2013): 313–335. Print.

“Massive open online course” Wikipedia, the free encyclopedia. Wikimedia Foundation, Inc., 16 Dec. 2013. Web. 18 Dec. 2013. <http://en.wikipedia.org/w/index.php?title=Massive_open_online_course&oldid=586359638>.

Mazur, Eric. “Farewell, Lecture?” Science 323.5910 (2009): 50–51. Print.

McAuley, Alexander, et al. The MOOC Model for Digital Practice. Massive Open Online Courses: Digital Ways of Knowing and Learning, 2010. Print.

Mott, Jon. “Envisioning the Post-LMS Era: The Open Learning Network.” EDUCAUSE Quarterly. 33.1 (2010): 1–9. Print.

Nasseh, Bizhan. “A Brief History of Distance Education.” SeniorNet. 1 Jan. 1997. Web. 14 Dec. 2013. <http://www.seniornet.org/edu/art/history.html>.

Nelson, Theodor H. “From Computer Lib/Dream Machines.” The New Media Reader. Eds. Noah Wardrip-Fruin and Nick Montfort. Cambridge, MA: The MIT Press, 2003. 301–338. Print.

O'Donoghue, John, Gurmak Singh, and Lisa Dorward. “Virtual Education in Universities: A Technological Imperative.” British Journal of Educational Technology 32.5 (2001): 511–523. Print.

O'Reilly, Tim. “What Is Web 2.0.” O'Reilly. O'Reilly Media, Inc., 30 Sept. 2005. Web. 15 Dec. 2013. <http://oreilly.com/web2/archive/what-is-web-20.html>.

Oremus, Will. “Free Online Classes Are an Unsustainable Gimmick. Here's a Better Idea.” Slate Magazine. The Slate Group, 18 Sept. 2013. Web. 3 Dec. 2013. <http://www.slate.com/articles/technology/technology/2013/09/spocs_small_private_online_classes_may_be_better_than_moocs.html>.

Pappano, Laura. “The Year of the MOOC.” International New York Times. The New York Times Company, 2 Nov. 2012. Web. 4 Dec. 2013. <http://www.nytimes.com/2012/11/04/education/edlife/massive-open-online-courses-are-multiplying-at-a-rapid-pace.html>.

Purser, Emily, Angela Towndrow, and Ary Aranguiz. “Realising the Potential of Peer-to-Peer Learning: Taming a MOOC with Social Media.” eLearning Papers 33 (2013): 1–5. Print.

Puschmann, Cornelius B. A., and Jean Burgess. The Politics of Twitter Data. Alexander von Humboldt, Institut Für Internet und Gesellschaft, 2013. Print.

Rieder, Bernhard. “What Is in PageRank? A Historical and Conceptual Investigation of a Recursive Status Index.” Computational Culture 2 (2012): n. pag. Web. 4 Dec. 2013. <http://computationalculture.net/article/what_is_in_pagerank>.

Ripley, Amanda. “College Is Dead. Long Live College! Can a new breed of online megacourses finally offer a college education to more people for less money?” Time U.S.. Time Inc., 18 Oct. 2012. Web. 18 Dec. 2013. <http://nation.time.com/2012/10/18/college-is-dead-long-live-college/>.

Rogers, Richard A. “Debanalizing Twitter: The Transformation of an Object of Study.” Proceedings of the 5th Annual ACM Web Science Conference (WebSci ’13) (2013): 356–365. Print.

———. Digital Methods. Cambridge, MA: The MIT Press, 2013. Print.

———. The End of the Virtual: Digital Methods. Amsterdam: Vossiuspers UvA, 2009. Print.

Russell, Walter. “Is The MOOC Hype Dying?” The American Interest. The American Interest LLC, 19 Nov. 2013. Web. 18 Dec. 2013. <http://www.the-american-interest.com/blog/2013/11/19/is-the-mooc-hype-dying/>.

Stallings, Dees. “The Virtual University Is Inevitable: But Will the Model Be Non-Profit or Profit? A Speculative Commentary on the Emerging Education Environment.” The Journal of Academic Librarianship 23.4 (1997): 271–280. Print.

Starr, Donna R. “Virtual Education: Current Practices and Future Directions.” The Internet and Higher Education 1.2 (1998): 157–165. Print.

Tait, Alan. “Reflections on Student Support in Open and Distance Learning.” The International Review of Research in Open and Distance Learning (2003): n. pag. Web. 18 Dec. 2013. <http://www.irrodl.org/index.php/irrodl/article/view/134/214>.

Tiernan, Peter. The Influence of Twitter on Lecture Engagement and Discussion. Dublin City University, 2012. Print.

Tough, Allen. The Adult's Learning Projects: A Fresh Approach to Theory and Practice in Adult Learning. Toronto, Canada: Ontario Institute for Studies in Education, 1971. Print.

van Treeck, Timo, and Martin Ebner. “How Useful Is Twitter for Learning in Massive Communities? An Analysis of Two MOOCs.” Twitter and Society. Eds. Katrin Weller, et al. Bern: Peter Lang International Academic Publishers, 2013. 411–424. Print. Digital Formations.

“Twitter Donates Entire Tweet Archive to Library of Congress.” LoC.gov. The Library of Congress, 15 April 2010. Web. 17 Dec. 2013. <http://www.loc.gov/today/pr/2010/10-081.html>.

“Update on the Twitter Archive At the Library of Congress.” Library of Congress, Jan. 2013. Print.

Vaccari, Cristian, and Augusto Valeriani. “Follow the Leader! Direct and Indirect Flows of Political Communication During the 2013 General Election Campaign.” New Media & Society (2013): n. pag. Print.

de Waard, Inge, et al. “Exploring the MOOC Format as a Pedagogical Approach for mLearning.” Proceedings from mLearn (2011): n. pag. Print.

Wallace, J. Faculty and Student Perceptions of Distance Education using Television (TV). Unpublished dissertation, Ball State University, 1991. Print.

van Weigel, B. Deep Learning for a Digital Age: Technology's Untapped Potential to Enrich Higher Education. Jossey-Bass, 2001. Print.

Yang, Dennis. “Are We MOOC'd Out?” HuffPost Business. TheHuffingtonPost.com, Inc., 14 Mar. 2013. Web. 18 Dec. 2013. <http://www.huffingtonpost.com/dennis-yang/post_4496_b_2877799.html>.

Young, Jeffrey R. Beyond the MOOC Hype: A Guide to Higher Education's High-Tech Disruption. Washington, D.C.: The Chronicle of Higher Education, 2013. Print.

Info
Title: Following UvAMOOC on Twitter
Subtitle: A Network Analysis of a Massive Socio-Technical and Cross-Platform Online Learning Environment
Type: Research article; Assignment
Author.name: T. (Tessa) de Keijser; F. N. (Fernando) van der Vlist
Author.affiliation: Graduate School of Humanities, Faculty of Humanities, University of Amsterdam
Instructor.name: Dr. B. (Bernhard) Rieder; E. K. (Erik) Borra; Prof. Dr. R. A. (Richard) Rogers; Dr. A. (Anat) Ben-David
Instructor.affiliation: Dept. of Media Studies, Faculty of Humanities, University of Amsterdam
Abstract: In this article we argue for a conceptualisation of MOOCs as socio-technical and cross-platform learning environments. We explore different competing visions and expectations of a contested space, and observe the kind of actors that are actually attracted by these visions and expectations to grasp what kind of learning environment these actors help constitute. To do this, we trace distance learning and interactivity as two initially separate historical and conceptual lineages of contemporary educational technology, and paid careful attention to the particular visions of learning these technologies carry with them. We argue that MOOCs build on Web 2.0 as a platform (O'Reilly 2005) as well as on online social networks to conjure up a specific socio-technical learning environment. We specifically focus our attention on Twitter, and how this microblogging platform is deployed as part of massive online learning environments. To operationalise this research, we have devised an outline for an analytical framework for “following Twitter”, both as medium and as a network of actors (Latour [2005] 2007). This framework would not just render productive particular issues such as uncertainties and instabilities with which we are presented, but also allow us to “put ahead of us” that which we want to study. Consequently, it enabled us to do a network analysis and describe what is actually happening inside the “data-space” as it has been captured by the DMI-TCAT, a device we deploy to “follow the medium” (Rogers 2009) in our acquisition of Twitter data. Ultimately, we debunk some of the idealised visions and inflated expectations following from Web 2.0 “possibility thinking” in the case of our pilot study of UvAMOOC on Twitter. This pilot study then caused us to reconsider the concept of “context collapse” (Marwick and boyd 2010) and the co-existence of a range of actors involved to different degrees within the same “data-space”.
Keywords: educational technology, learning environment, digital methods, network analysis, Twitter, UvAMOOC
Length.words: 8,026
Length.reading: 45 mins
Element.table: Table 1; Table 2; Table 3; Table 4; Table 5; Table 6; Table 7; Table 8
Element.figure: Fig. 1a; Fig. 1b; Fig. 2a; Fig. 2b; Fig. 3; Fig. 4
Date.submitted: 19 Dec. 2013
Date.evaluated: 24 Jan. 2014
Date.publishedonline: 1 Feb. 2014
Language: English (United Kingdom)
Documentation.style: Modern Language Association (7th ed.)
Export.citation: BibTEX
Export.print: javascript:window.print()

Research Report

Following MOOC on Twitter

A Network Analysis of a Twitter Dataset on Massive Open Online Courses

Tessa de Keijser, Fernando N. van der Vlist [alphabetical]

(Graduate School of Humanities,) University of Amsterdam, the Netherlands

Working paper

Published online: 1 February 2014

Abstract

We report on our investigation of Twitter activity around a particular set of keywords or hashtags by using the DMI-TCAT “mooc” dataset. For our data, we relied on the “mooc” dataset available through the DMI-TCAT, an open-source toolset that captures tweets and allows for multiple types of analysis (hashtags, mentions, users, search, and so forth). For further analysis of this dataset, we relied on Gephi to help us reveal patterns and trends, find the influential hubs, link clusters and giant components, betweenness centrality (bonding and bridging nodes), and content-sharing communities and subcommunities within that dataset. More specifically, we visualise and analyse a social graph by mentions, a social graph by in_reply_to_status_id, and a bipartite hashtag-URL graph. The results of our analysis show that debates around MOOCs are still happening. This debate is spread out over a number of social groups, as the types of actors that tweet on MOOC vary greatly and the URLs and hashtags they use mostly relate to broader themes on new media technologies and online education. Furthermore, we found that the discussion space centreing on MOOCs is multilingual and that this multilingual space is for a big part dominated by English, French and Spanish users.

Keywords

Internet research, digital methods, platform studies, network analysis, graph visualisation, Twitter, MOOC

Internet research
digital methods
platform studies
network analysis
graph visualisation
Twitter
MOOC

1. Introduction

Twitter was founded by Jack Dorsey and associates in 2006 in “a long line of squawk media, dispatch, short messaging as well as citizen communications” (Rogers, “Debanalizing Twitter”, 356). According to its own description, it is now a “social networking and microblogging service utilising instant messaging, SMS or a web interface”. Statistics from Twitter, the Huffington Post and eMarketer indicate that it now has 554,750,000 active registered users, resulting in a continuous stream of as much as 9,100 tweets posted every second (or 1 billion tweets in 5 days). It should come as no surprise, then, that Twitter has also been conceived of a valuable resource and object of research for industry as well as for academics from the beginning.

1.1. Twitter Studies

Rogers (2013) has suggested a periodization based on how Twitter has been conceived as an object of study. Starting from it's initial development in 2006 until roughly 2009, Twitter research could mostly be characterised as an ambient friend-following and messaging utility, or an “urban lifestyle tool” with many tweets on mundane (that which is “of the earth”) activities such as lying in bed or taking a walk in the park. Research on Twitter in this first stage or tradition (“Twitter I”) is often directed at determining the value of tweets and the degree to which messages are “banal”, thus focussing mostly on the content (357). Characterised by a change of the tagline from “What are you doing?” to “What's happening?” in 2009, a second phase or tradition (“Twitter II”) conceives of Twitter as an “event-following tool”, a way to monitor, for instance, elections or disasters (359). Yet, as Rogers argues, this stage of Twitter research is gradually shifting towards a third phase or tradition (“Twitter III”) that is interested in Twitter as it is settling into a massive data set (362), which is of value to itself but is also an object of cultural heritage as their public tweets are being archived and preserved by the Library of Congress (LoC) since April 2010 (“Twitter Donates Entire Tweet Archive to Library of Congress”). In a more recent report from January 2013 (“Update on the Twitter Archive At the Library of Congress”), the LoC states that it is processing and organizing this collection to make it usable and accessible for researchers and that it is “exploring the possibility of developing a research- and scholarship-focused interface to the archive using Gnip's existing historical Twitter product offerings” (360). Although we are not able to access this collection at this point, it is nevertheless this third tradition of Twitter research that most closely relates to our study here.

Furthermore, this ambitious project is part of LoC's larger mission to “preserve and provide access to a universal collection of knowledge and the record of America's creativity for Congress and American people.” (emphasis added). This project thus can be read as an indication of the socio-cultural value of Twitter for many different parties and purposes, but cannot completely account for the multitude of ways in which Twitter is used differently by individual users. How people actually use Twitter can be very different: some use it only for following particular accounts or as information resource, others use it only for posting excessively about their own lives, and probably most are somewhere in between or use in still other ways. Building on this observation, Crawford (2009) has suggested listening as metaphor for paying attention online and has identified “background listening” (528), “reciprocal listening” (529), and “corporations ‘listening in’” (531) as modes of listening on Twitter and has considered how these are experienced and performed by individuals, politicians and corporations each with their own interests. The purpose of using Twitter can therefore also be very different: for example, a politician might want to influence the reception of a story and shift public opinion, or teens may want to share details of their personal lives in a cry for attention.

1.2. Following Twitter

There are some obvious limitations in understanding Twitter as “data provision machine” (Rogers 362): first, Twitter is ephemeral, so how does one study such an unstable object? Second, if the researcher is not employed by the organisation owning the data, he or she may have little or no access to some aspects of the data. Third, regarding LoC's archive of tweets, Twitter retains a six-month delay before tweets may be archived, thus effectively rendering impossible any form of real-time research with that dataset. Back, et al. (2013) and Elmer (2013) have both enquired into how real-time research poses new challenges as a result of the problem of ephemerality. For Elmer, streams of data especially on Twitter pose new challenges to research on political communications campaigns, and the analysis of “vertical tickers” highlights the fleeting character of networked and socially mediated communications (25). Marres and Weltevrede (2013) offer an operational analytical framework as they investigate the device of scraping. In particular, they show how scraping imports analytical presumptions into the research, and also that scrapers import data that is already pre-formatted. As a consequence of this, they raise the question of whether the research helps “to bring into view the dynamics of media platforms or also those of the social phenomena they enable?” To operationalise that question, they suggest a distinction between two kinds of real-time research: “those dedicated to monitoring live content (which terms are current?) and those concerned with analysing the liveliness of issues (which topics are happening?).” (356). Where live social research renders “analytically productive the formatted, dynamic character of digital networked data” (15), which means that it makes possible the study of changing “natively” digital objects, research on liveliness on the other hand aims to capture what is currently happening in a particular social space; which actors are the most active and make things happen in the “eternal now”.

A digital methods approach may offer one solution to the problems posed by live social research: by following the medium, and it's natively digital objects it is better prepared to handle such changing objects. Our study is thus situated in the broader frames of Internet research and platform studies. Specifically, the use and deployment of medium-specific objects, methods, devices and techniques that characterise this research situate it in the line of work on “digital methods” (Rogers 2013). We will be studying Twitter with medium-specific approaches that “follow the medium” by repurposing the devices and techniques of the platform, in order to see whether we may develop claims about culture and societal change more broadly. According to Rogers, “Twitter is particularly attractive for research owing to the relative ease with which tweets are gathered and collections are made, as well as the in-built means of analysis, including RT (retweets) for significant tweets, #hashtags for subject matter categorization, @replies as well as following/followers for network analysis and shortened URLs for reference analysis. Given its character limit and the fact that each tweet in a collection is relatively the same length, it also lends itself well to textual analysis, including co-word analysis” (362). Each of these elements are “native” or device-specific, and can to a specified degree be accessed through application programming interfaces (APIs), this is part of what Puschmann and Burgess (2013) call the “politics of Twitter data”: the negotiations of the platform provider with the end users through terms of service and application programming interfaces. This device-specificity of the data has implications for our research because it is pre-formatted (which is at the same time what makes it specific) in so-called JavaScript Object Notation (JSON), an open data-interchange standard. Consequently, Borra and Rieder are right to stay as close to follow the “native” format of the data as much as possible to introduce as little bias as possible in developing the DMI-TCAT tool we will be using here (“Programmed Method” see below).

In regards to this pre-formatting of data, it is worth mentioning or referring to research on of some implications that in a way follow from this. Social networking sites (SNSs) such as Twitter mark an important site to study what Rogers (2009) has dubbed “postdemographics.” This is a useful concept for the study of social networks that not simply looks at established and conventional demographics such as age, race or gender that retain an indexical relation to the body. Instead, he argues that this biopolitical concept of demographics should be traded for an info-political one that takes into account the “native” Web as it is becoming increasingly “social” and allows for the study of “databodies” which are aggregates of “tastes, interests, favorites, groups, accepted invitations, installed apps, and other information that comprises an online profile and its accompanying baggage” (Digital Methods 153–154). With postdemographics, this metadata does not refer to data about a particular body – individual or mass – anymore, but rather renders them “indifferent” (Amoore 28) as samples, data, markets or “banks”, resulting in what Deleuze has famously called the “dividual” ([1990] 1992, 5). This implies that demographics and postdemographics also make use of different kinds of sources. Where demographics typically come from official records, “profilers”, as Rogers also calls those who do postdemographics, analyse data that users created or generated themselves on “platforms that create and maintain social relations” (Digital Methods 154). As such, these social networking sites present data that can be treated and analysed in two different ways: either for personalised or depersonalised profiling practices (taste and taste relationships) (156–157). SNSs like Twitter thus constitute a breeding ground or milieu for the production of sociality and connectivity (van Dijck 2011, 2013) where these “native” types of data may emerge, and which has resulted in the “data market” and a new kind of “data scientist” that reads society not up close, but from a distance (cf. Moretti 2000).

1.3. Massive Open Online Courses: Literature Review of Studies and Critiques

Massive open online courses (MOOCs) aim at large-scale interactive participation and open access via the web or other network technologies. As such, they rely on the Web as a platform for interactive forms of community building, engaging with open knowledge, and e-learning. Attending such an MOOC does not only entail listening to a recorded video lecture, but may also entail searching for answers and helping other students at the designated forums, or peer-reviewing each others assignments for actual grades. For these purposes, MOOCs rely on the Web and on social media platforms, which is why an examination of the different uses of these platforms may provide an interesting entry point to study MOOC as it appears on or flows through Twitter.

There is a long-standing debate on how computer technologies can fundamentally alter our present teaching and learning models. These ideas about innovation in education used to be focused on distance education, but also on “interactivity” as an ideal for individual empowerment. In Computer Lib/Dream Machines (1974), Ted Nelson already wrote about the potential of the computer for educational purposes. He compared and criticised ordinary teaching and the then-emerging idea of computer-assisted instruction (CAI) for preventing the student to directly interact and engage with subject matter. In the former case, the teacher would be the obstacle between the student and the subject matter, and in the latter case it becomes the computer. He therefore proposed to create “hyper-media” and “responding resources” to give free play to the student's initiative, permitting the student to control the system and find his own way through the resources available to him. Situating the present discourse around MOOCs in this tradition, it seems that the discourse on “interactivity”, which has since been criticised as simply relocating control (Barry 2001), has shifted to a rhetoric of “inversion” or “flipped classrooms”, “openness”, “access”, “connectivism” and “sharing” (Cheverie 2013; Carr 2013; Haggard 2013).

In the middle of the hype associated with MOOCs, the year of 2012 was exclaimed “the year of the MOOC” (Pappano 2012), but this subsequent year has seen quite a few MOOC critiques. Will Oremus (2013) wrote a critique arguing that free online classes should not be focusing on replacing teachers and classrooms, but rather on improving them in the first place. In a similar vein, Bogost (2013) critiqued the radical innovation MOOCs seem to promise by arguing that the traditional learning system has not been “inverted” or “flipped”, so much as abstracted. Dan Colman (2013), founder of Open Culture (a “high-quality cultural & educational media for the worldwide lifelong learning community”) identified a number of reasons why readers had not finished their MOOCs, and found as problems that it takes too much time, either assumes too much or too little knowledge, lecture fatigue (relying on formal video lectures), bad communication tools, bad peer review and trolls, hidden costs, and the idea that you are not there for the credential at the end. The “massive” scale in itself also introduces logistical problems related to evaluation and assessment (Head 2013). Brent Chesley, who teaches 17th- and 18th-century British literature at Aquinas College, has evaluated MOOCs based on his own experience of teaching such a course. Mainly, he does not think its a good idea to equate MOOCs with other courses. Rather, he argues that the combination of conventional structures with online additional resources to enhance the class (Chesley 2013). Finally there are also critiques regarding centralisation. Google and edX created spin off Web sites for “the rest of us”, thus critiquing MOOC providers for only recruiting elite, high-profile institutions and their professors (Kolewich 2013). Similarly, distributed open collaborative courses (DOCCs), developed by feminist professors under the name FemTechNet, are a response to centralisation problems by challenging the role of the instructor, money, hierarchy, and of the value of the “massive”. For instance, the consequence of some “best” professors becoming worldwide superstars educating the rest of us, and the discouragement of group-learning in MOOCs (Jaschik 2013).

Besides these critiques, research has been done to solve particular problems of MOOCs with networked social media platforms. Purser, Towndrow and Aranguiz (2013) explored the realisation of peer-to-peer learning by embracing networked social media to “tame” MOOCs and completely bypass traditional notions and tools of online learning support. Twitter hashtags (#edcmooc), Google Plus and blogs were suggested as useful for connecting with one another, and from there on students could create further subgroupings (potentially also in other social media). Mak, Williams and Mackness (2010) explored specifically the role of blogs and forums as facilitators for communication and learning tools in an MOOC, and found that the use of these “platforms” to a large extent correlated with specific individual learning styles and media affordances. Use of blogs was associated with the ability to create personal space for personal learning, quiet reflection and personal relationships, and forums were associated with fast paced challenging interaction, building learning communities, open discussion, and more links to broader themes and the bigger picture (275). Van Treeck and Ebner (2013) analysed two particular MOOCs to study the usefulness of Twitter as a micro-blogging platform for learning in massive communities. Using hashtag analysis, they were able to show that 70% of tweets were directly related to the course, 39% of those refer to specific topics, and 31% to organisation of the course. Finally, Bruff, et al. studied student perceptions of attempts to blend MOOCs with conventional models into “blended learning”, and advocate for drawing from multiple MOOCs as well as other online sources. As also shines through in the reflection of Chesley (2013), the success is dependent on the course design, and so naturally, MOOCs and the platforms they require to support them will not work the same or with the same success for every discipline or study.

2. Research Questions

We want to investigate Twitter activity around a particular set of keywords or hashtags by using the DMI-TCAT “mooc” dataset. Based on our current understanding of the subject matter, we expect both actual MOOC courses and critical debate centreing on MOOCs to be represented in the same dataset. Consequently, we wish to investigate just how the semantic space on Twitter, opened up by tweets is actually or empirically organised.

Firstly, based on our selected dataset, we would like to examine whether we could make claims about online learning communities, MOOC providers, and their specific use of Twitter. Can we locate different communities, subcommunities and conversations in this Twitter data space? And if so, then how are these communities and conversations geographically dispersed, and what can we find out about the use of language? To what extent do the critical claims and the results of the studies introduced above resonate with the dataset? What actors are positively invested in MOOC debate, and which are critical?

Secondly, besides questions more specific to our topic, the study poses questions that relate to the broader debates introduced above. Can we do Twitter studies? And if we do, are we then doing social research, or are we rather studying the medium, that is Twitter dynamics and trends? Can we indeed retain this distinction between live social research and social research on liveliness as Marres and Weltevrede (2013) have proposed? Must one work at Twitter to do research with Twitter data?

3. Methods

3.1. DMI-TCAT

To operationalise our research questions, we rely on the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT), an open-source toolset that captures tweets and allows for multiple types of analysis (hashtags, mentions, users, search, and so forth). As Borra and Rieder note, The toolset enables epistemic plurality and provides “robust and reproducible data capture and analysis, and interlinks with existing analytical software.” (“Programmed Method”, 1). Moreover, it calls attention to methodological, epistemological, logistical, legal, ethical, political issues arising from the use of software in the study of (proprietary) data extracted from social media platforms. For data acquisition, the DMI-TCAT is built on top of Twitter's streaming API and the REST API and as such is bound to their possibilities and limitations (7–8). As mentioned before, Borra and Rieder remain as close as possible to the “native” formats of this data, so that they introduce as little additional bias as possible. They do, however, enrich the data in two directions: URL expansion and the addition of Klout scores. Thus, the DMI-TCAT offers us a way around some of the limitations and restraints of similar tools, but also keeps in mind the crucial methodological repercussions.

The toolset allows us to select the “mooc” dataset, which is a set that has been capturing tweets with the keywords “mooc” or “uvamooc” since 7 March 2013. Due to the method of capturing of tweets over time we were limited to work with this dataset as it has been captured. As of 5 December 2013, this dataset contains 406,464 tweets. In this set, we queried [2013-06-03] as the starting data, and [2013-12-03] as the end date to get a slightly smaller subset spanning over the past six months. The subset contains 106,316 distinct users, and 278,685 tweets, of which 72.6% (202,302) contain links, and 27.4% (76,383) do not. Plotting this data on the built-in timeline presents us with quite a regular pattern of peaks and valleys, with a smaller overall number of tweets during the summer holidays roughly from June up until the end of August (Figure 1). The built-in analytics tools, however, are not as useful when we actually want to study the semantic space delineated by this dataset, and the various actors, relations, communities that make up that space.

Fig. 1. Overview of our dataset selection generated by the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT). Screenshot created on 5 Dec. 2013, 10:49:30.

3.2. Network Analysis

For further analysis of the dataset, we relied on Gephi, which is an open-source software for network visualisation and analysis. Importing the output of the DMI-TCAT, it helps us to reveal patterns and trends, find the influential hubs, the outliers, link clusters and giant components, betweenness centrality (bonding and bridging nodes), and so forth within the dataset we have downloaded. Network analysis thus allows us to look at the “Twitter data space” from the ground up in an empirical fashion, and to see what emerges as significant based on nodes and their relations. Consequently, we used it for a kind of interactive probing, the iterative process of playing with the dataset and seeing what happens when applying various structuring algorithms that alter the relations between the nodes; thus exploring different ways to “make hills” in a given “flatland” (Rieder 2012).

Besides the “wiggling around” with properties of the nodes and edges that is possible in Gephi, the spatialisation of these nodes and edges follows a layout algorithm. The layout algorithm that we used most in our analysis is the built-in ForceAtlas2, which is a continuous force-vector graph layout algorithm based on a linear-linear model, and developed by Mathieu Jacomy in 2011 (Gephi.org, “ForceAtlas2, the new version of our home-brew Layout”). Because force-directed algorithms are actually physical simulations, it is able to run homogeneously while displaying the “live” result of the spatialisation (Jacomy et al. 2012, 3) and attraction and repulsion between nodes are proportional to distance between these nodes. For our visualisations and analysis, this means that the edges between the nodes are assigned forces relative to their positions, thus allowing us to distinguish and compare subcommunities and the strength of ties. It also means that no specialised knowledge about graph theory is required to draw or read the graphs.

4. Results

Before continuing with presenting our results, we want to point out that this study is complicated and suffers by the data acquisition of our dataset. Not just because of the limitations of the DMI-TCAT outlined by Borra and Rieder themselves (“Programmed Method”), but also as a consequence of the including only tweets containing the keyword “mooc” or “uvamooc” into the collection. As a result, our dataset is simply not complete, and is presumably even skewed by including only “uvamooc” as a keyword referring specifically to an MOOC on “Introduction to Grid Computing” at the University of Amsterdam (UvA). This might mean that particular discourses that uses the buzzword “MOOC” might be overrepresented in our dataset. Furthermore, as an inherent consequence of collecting tweets that contain the same keyword, it will also include tweets that are completely irrelevant but simply happen to use that keyword. This is significant to the extent that each node does influence the form and structure of the network visualisation and thus of the analysis. Finally, besides methodological transparency, algorithms could always still contain bugs or might just be “wrong”’ (Gillespie 2012).

4.1. Social Graph by Mentions

We used a first graph to analyse interaction or communication patterns, to find the “hubs” and the communities and subcommunities in the dataset. First, we generated a .GDF network file from the DMI-TCAT containing data on users and their mentions. Because the complete dataset would simply be unmanageable, we have filtered the set to only include the top 1% (1,063) of users. Second, we could produce a directed graph based on those interactions between users. Whenever one users mentions another user, a directed relation is created. The more often this happens, the stronger the link becomes (link weight). Furthermore, the number of tweets is also included in the file, on the basis of which we determined who where the most active users. Third, nodes were scaled according to their number of mentions (“no_mentions”), and colours follow a heat map scheme (red means most, blue means least). In this network, the number of tweets sent out by a user roughly correlates with their number of mentions. We then ran the PageRank statistics algorithm to do a recursive status or authority ranking on top of this (cf. Rieder 2012). As can be seen in the visualisation, the biggest nodes do not simply have the highest PageRank (Figure 2a). Rather, there is quite a big difference between the inDegree (Figure 2b) and outDegree values (Figure 2c), which means that authority (according to this measure) is very hierarchical. The nodes with the highest PageRank (“coursera”, “futurelearn”, “scoopit”) are course and content directory websites, followed by “udacity”, “chronicle”, “gsiemens”, “audreywatters”, “universite_num”. After this we ran the network diameter statistics algorithm to generate betweenness centrality. This calculates which users are the most important for bridging different subcommunities. Accordingly, “thegoodmooc”, “moocnewsreviews” (which are both sites for reviews, interviews, analyses, and so on), “eraser”, “iversity”, and “gsiemens” have the highest betweenness centrality (Figure 3).

Fig. 2a. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours are set according to a heat map schema representing their PageRank. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 2b. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours are set according to a heat map schema representing their PageRank-outRank. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 2c. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours are set according to a heat map schema representing their PageRank-inRank. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 3. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours are set according to a heat map schema representing their betweenness centrality. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Visualising interactions between users based on their mentions, it appears that there are a lot of different actors that form clusters, with larger clusters of interconnected users in the middle and on the right (Figure 4). On the bottom left, a cluster of users is visible with the user “entrepenez” as a dominant node in the middle. When we organise the graph table by the number of tweets, we get a list of the top Twitter users in terms of the number of tweets containing the keywords “mooc” or “uvamooc” (Table 1). Upon examination of the top users Twitter profiles and the websites and organisations they are related to (listed in Table 1 as “organisation” and “type of organisation”) we can see that these top users represent different actors in the MOOC debate (Table 2). In total, there are five MOOC providers in the top 25, with three of them being commercial parties and two of them academic. A second group of organisations consists of two accounts related to a website that function as MOOC directories. What is also noteworthy is that four Twitter accounts are related to organisations that do not provide MOOCs themselves, but instead offer technology to develop MOOCs. A large segment of the top 25 users consists of Twitter users that are related to a blog or website that either gives an overview or discusses developments in the world of online learning (Table 1). Note here that only two users are connected to an organisation that comments on or evaluates the form of the MOOC specifically (Table 1). Therefore, it would seem that the MOOC is seen as part of the broader field of online learning, or of using new media technologies in education. The largest group is made up by exactly this type of websites, one that evaluate, discuss or report on developments in online learning as a whole (11 users). The David Webb Show is the only account not directly related to MOOCs, which is a United States based radio show with no apparent connection to MOOCs or education types. A second point that presents itself from this analysis is that there are three dominant language sets in the top 25: English, French, and Spanish. 9 Users Twitter exclusively in English, 6 in French, and 6 in Spanish. There is also one user that tweets in Basque and two users that are multilingual. The language used by their related organisation, on the websites for example, is almost always the same as the language in which a user tweets. Furthermore, the top French users cluster together (on the far left), as do the top Spanish users (in the top cluster in the middle) (Figure 4).

Fig. 4. Social graph by mentions. Nodes are users scaled to their number of mentions, edges are mentions, and node colours are set according to a heat map schema representing their relative degrees. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

4.2. Social Graph by in_reply_to_status_id

We used a second graph for the same reason as the social graph by mentions: to analyse interaction or communication patterns, to find the “hubs” and the communities and subcommunities in the dataset, but we will focus more strongly on language now. First, we generated a Gephi-readable network file from the DMI-TCAT containing data on interactions between users (1,914 nodes; 1,503 edges). To reduce the size of the graph, we set the minimum frequency for data to be included in the export to 4. Whenever a tweet was written in reply to another, a directed relation is created. Second, because those interactions are directed, a directed graph could be visualised based on those specific interactions between users. Third, within this set of Twitter conversations, we scaled nodes according to the number of tweets each user has sent in total (“from_user_tweetcount”) so that the weight of these users is qualified. This has given us an overview of the significant reply-interactions in the set, and as we can see in the graph, there are many small chain and star networks, but they are all relatively small (Figure 5). Furthermore, while some of these users tweet a lot, most of them do not. We then introduced an additional parameter (colour) to highlights language distribution of these tweet conversations. We found that for this subset, 59.61% has English set as its default account language (turquoise), 13.64% has French (brown), 9.61% has Spanish (blue), the other languages are selected by less than 2% of the users, and 10.97% has no set value (“null” green) (Figure 6a–6c). Looking at the actual tweets in this set, we found that even when a particular language is set as the account language, quite some users are in fact posting tweets in different languages as well (Table 3). In another version of the visualisation, we also scaled nodes according to the number of followers (“from_user_followers”), and coloured according to a heat map scheme by tweetcount (red means most, blue means least) (Figure 7). This gives us a few influential nodes and their tweet counts, which we investigated further by analyzing the user accounts and their related organisations of the top 10 users with the most followers (Table 4). The most followed user in this dataset is “QueenRania,” the Twitter account of Queen Rania of Jordan, who is an advocate for quality education. She is the only user in the top 10, the rest of the user accounts being mostly accounts of large (online) magazines in the fields of business, education, news, and culture.

Fig. 5. Social graph by in_reply_to_status_id. Nodes are users scaled to their number of tweets, edges are tweet replies, and node colours are set according to a heat map schema representing their number of followers. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 6a. Social graph by in_reply_to_status_id. Nodes are users scaled to their number of tweets, edges are tweet replies, and node colours are set according to language-based graph partitioning. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

Fig. 6b. Overview of node partitions for the social graph by in_reply_to_status_id (list view). List generated with Gephi, data from DMI-TCAT (4a05476c87).

Fig. 6c. Overview of node partitions for the social graph by in_reply_to_status_id (pie chart view). Pie chart generated with Gephi, data from DMI-TCAT (4a05476c87).

Fig. 7. Social graph by in_reply_to_status_id. Nodes are users scaled to their number of followers, edges are tweet replies, and node colours are set according to a heat map schema representing their number of tweets. graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

4.3. Bipartite Hashtag-URL Graph

We used a third graph to get a rough grasp of how URLs and hashtags are qualified, that is to better understand what circulates between these actors. It gives an account of which hashtags are used in combination with which links the most. Again, we first generated a Gephi-readable network file from the DMI-TCAT containing URLs and the number of times they have co-occurred with a particular hashtag (50,976 nodes; 87,394 edges). Second, those co-occurrences could then be visualised in an bipartite graph, which is an undirected type of graph whose vertices can be divided into two disjoint sets in such a way that every edge connects across these sets. If an URL co-occurs with a particular hashtag, a relation will be created, and the more often they appear together, the stronger this link will be. Third, we partitioned the graph for these two types, and found that 75.74% of these were URLs (red), and 24.26% are hashtags (light blue). Perhaps not too surprisingly, most bigger nodes at first sight simply seem to be more generic hashtags such as #mooc, #education, #elearning, #onlinelearning, #onlineeducation, #elearn, #globaled, #edtech and so on (Figure 8). Besides these, there are also hashtags referring to platforms such as #fb or #ted. Looking closer to the URLs, a similar situation occurs. Again, more generic URLs such as “http://mooc.org/” and “https://www.coursera.org/” are the biggest, but in general there are less large nodes here, suggesting a high level of diversity among these.

Fig. 8. Bipartite Hashtag-URL Graph. Turquoise nodes are hashtags and red nodes are URLs both scaled to their numbers of occurrence, edges indicate their co-occurrence in a tweet. Network graph visualised with Gephi (0.8.2 beta), data from DMI-TCAT (4a05476c87).

5. Discussion

The results of our analysis show that debates around MOOCs are still happening. This debate is spread out over a number of social groups, as the types of actors that tweet on MOOC vary greatly and the URLs and hashtags they use mostly relate to broader themes on new media technologies and online education. Furthermore, we found that the discussion space centreing on “MOOC” is multilingual and that this multilingual space is for a big part dominated by English, French and Spanish users. Still, the graphs, or the empirical reality they aim to visualise, are not in and of themselves meaningful. They are only empirical data represented in a structured form in need of interpretation.

5.1. Platforms are Non-Homogeneous Spaces

In giving an overview of the literature and debates around MOOCs and Twitter, we spoke of MOOCs as relying heavily on the sociality of the web and of Twitter being one of those platforms on which the MOOC community is established. Yet the particular term “platform”, rather than just a description, can also be seen as a concept. As suggested by Gillespie (2010) the term has taken up four different kind of meanings and over the years it has been used to denote architectural (a raised surface), political (standpoint or plank), computational (to launch software programmes), and figurative meanings (something that is empowering) (349–350). Platform thus constitutes a discursive space, and the discourse surrounding MOOCs (as captured in our dataset) strongly emphasises that they are platforms for positive notions such as “openness”, “free” education (as in gratis), “access”, “sharing”, and “connectedness”.

Regarding our initial question of how the Twitter platform is used to support the limitations of MOOCs, we can first of all say that we did not find that much original content on Twitter, in the sense that most tweets are just propelling sources from outside of Twitter (72% contain URLs). Moreover, all top Twitter users in our set are affiliated with particular organisations as well, which they often explicitly stated on their profile pages. Second, we raised the questions whether different communities, subcommunities and conversations in the Twitter data space could be located, and how these communities and conversations would be geographically dispersed. We found that language connects clusters of Twitter users and debates. Most notable, English, French and Spanish subcommunities and sub-debates appeared in the dataset. On the one hand, this language clustering shows that the acronym “MOOC” is used not only in countries that have English as an official language and that a Twitter hashtag in fact constitutes a multilingual space of global communication (cf. Poell and Darmoni 2012). On the other hand, we also wonder whether there are other terms used that substitute MOOCs in other countries, and whether our dataset indeed included tweets about MOOCs. Further research could pay more attention to building a robust list of hashtags that are relevant to this particular problem space. Third, based on our current understanding of the subject matter and the MOOC as a multi-platform type of organisation, we expected both actual MOOC courses and critical debate centreing on MOOCs to be represented in the same dataset. In our analysis, we have not found convincing evidence that MOOC courses themselves were present in this Twitter dataset. But on the other hand, the critical debate centreing on MOOCs is indeed quite strongly represented, as we have seen in the analysis of the top 25 users. Many tweets and users were linking to rather general sources on new online, open and free learning technologies.

5.2. Suggestions for Further Research

In order to develop a better understanding of the results one would need to bring in further information from the outside. Further investigations into the content of the URLs could be useful to further enquire into this point, as would the careful building of a dataset that contains more specific keywords. As mentioned above, with the exception of “uvamooc”, this study has not been able to take into account other similar forms of online distance learning and e-learning, specific MOOC platforms such as Coursera (US), EdX (US), Udacity (US), FutureLearn (UK) (Haggard 2013, 12), or courses like this one at the UvA. A more advanced hashtag analysis could also help research what kind of subject matter is actually discussed in this data space, and which courses lend themselves better for MOOCs. Because there are many specific topics in this set, one could not only explore which hashtags are linked to which types of information, but also see which universities or programmes (e.g. #uvamooc, #pennstate, #stanford, or #harvard) are associated with particular sources and topics. Thus, it allows for an analysis of content-sharing communities and subcommunities. Including this might lead to a more inclusive picture of MOOCs within this Twitter data space.

6. References

Andrew, Barry. “On Interactivity.” Political Machines: Governing a Technological Society. London: Athlone Press, 2001. 127–152. Print.

Amoore, Louise. “Data Derivatives: On the Emergence of a Security Risk Calculus for Our Times.” Theory, Culture & Society 28.6 (2011): 24–43. Print.

Back, Les, Celia Lury, and Robert Zimmer. Doing Real Time Research: Opportunities and Challenges. National Centre for Research Methods Methodological Review paper. Goldsmiths College, University of London, 2013. Print.

Bogost, Ian. “The Condensed Classroom.” The Atlantic. The Atlantic Monthly Group, 27 Aug. 2013. Web. 3 Dec. 2013. <http://www.theatlantic.com/technology/archive/2013/08/the-condensed-classroom/279013/>.

Borra, Erik K., and Bernhard Rieder. “Programmed Method: Developing a Toolset for Capturing and Analyzing Tweets.” Unpublished Manuscript.

Bruns, Axel, and Jean Burgess. “Researching News Discussion on Twitter: New Methodologies.” Journalism Studies 13.5–6 (2012): 801–814. Print.

Bruff, Derek O., et al. “Wrapping a MOOC: Student Perceptions of an Experiment in Blended Learning.” Journal of Online Learning and Teaching (JOLT) 9.2 (2013): n. pag. Web. 4 Dec. 2013. <http://jolt.merlot.org/vol9no2/bruff_0613.htm>.

Carr, David F. “Udacity Hedges On Open Licensing For MOOCs.” InformationWeek. UBM Tech, 20 Aug. 2013. Web. 4 Dec. 2013. <http://www.informationweek.com/software/udacity-hedges-on-open-licensing-for-moocs/d/d-id/1111226>.

Chesley, Brent. “My Problem With MOOCs.” Inside Higher Ed. Inside Higher Ed, 6 Aug. 2013. Web. 4 Dec. 2013. <http://www.insidehighered.com/views/2013/08/06/essay-mooc-debate-and-what-really-matters-about-teaching>.

Cheverie, Joan. “MOOCs and Intellectual Property: Ownership and Use Rights.” EDUCAUSE Blog. EDUCAUSE, 8 April 2013. Web. 4 Dec. 2013. <http://www.educause.edu/blogs/cheverij/moocs-and-intellectual-property-ownership-and-use-rights>.

Colman, Dan. “MOOC Interrupted: Top 10 Reasons Our Readers Didn't Finish a Massive Open Online Course.” Open Culture. Open Culture, LLC, 5 Apr. 2013. Web. 3 Dec. 2013. <http://www.openculture.com/2013/04/10_reasons_you_didnt_complete_a_mooc.html>.

Crawford, Kate. “Following You: Disciplines of Listening in Social Media.” Continuum 23.4 (2009): 525–535. Print.

Delbanco, Andrew. “MOOCs of Hazard.” New Republic. 31 Mar. 2013. Web. 3 Dec. 2013. <http://www.newrepublic.com/article/112731/moocs-will-online-education-ruin-university-experience>.

Deleuze, Gilles. “Postscript on the Societies of Control.” October 59 (1992): 73–77. Print.

van Dijck, José F. T. M. “Facebook as a Tool for Producing Sociality and Connectivity.” Television & New Media 13.2 (2011): 160–176. Print.

———. The Culture of Connectivity: A Critical History of Social Media. Oxford: Oxford University Press, 2013. Print.

Elmer, Greg. “Live Research: Twittering an Election Debate.” New Media & Society 15.1 (2013): 18–30. Print.

“ForceAtlas2, the new version of our home-brew Layout.” Gephi.org. The Gephi Consortium, 6 June 2011. Web. 5 Dec. 2013. <https://gephi.org/2011/forceatlas2-the-new-version-of-our-home-brew-layout/>.

Gerlitz, Carolin, and Bernhard Rieder. “Mining One Percent of Twitter: Collections, Baselines, Sampling.” M/C Journal 16.2 (2013): n. pag. Web. 4 Dec. 2013. <http://journal.media-culture.org.au/index.php/mcjournal/article/viewArticle/620>.

Gillespie, Tarleton. “Can an Algorithm Be Wrong?” Limn 2 (2012): n. pag. Web. 4 Dec. 2013. <http://limn.it/can-an-algorithm-be-wrong/>.

———. “The Politics of ‘Platforms’.” New Media & Society 12.3 (2010): 347–364. Print.

Haggard, Stephen. “The Maturing of the MOOC: Literature Review of Massive Open Online Courses and Other Forms of Online Distance Learning.” Department for Business Innovation and Skills (BIS), 2013. Print.

Head, Karen. “Inside a MOOC in Progress.” The Chronicle of Higher Education. The Chronicle of Higher Education, 21 June 2013. Web. 3 Dec. 2013. <http://chronicle.com/blogs/wiredcampus/inside-a-mooc-in-progress/44397>.

Jacomy, Mathieu, Sebastien Heymann, Tommaso Venturini, and Mathieu Bastian. “ForceAtlas2, A Graph Layout Algorithm for Handy Network Visualization.” Sciences Po Médialab, 2012. Print.

Jaschik, Scott. “Feminist Professors Create an Alternative to MOOCs.” Inside Higher Ed. Inside Higher Ed, 3 Dec. 2013. Web. 20 Aug. 2013. <http://www.insidehighered.com/news/2013/08/19/feminist-professors-create-alternative-moocs>.

Kolowich, Steve. “Google and edX Create a MOOC Site for the Rest of Us.” The Chronicle of Higher Education. The Chronicle of Higher Education, 10 Sept. 2013. Web. 3 Dec. 2013. <http://chronicle.com/blogs/wiredcampus/google-and-edx-create-a-mooc-site-for-the-rest-of-us/46413>.

Kosinski, Michal, David Stillwell, and Thore Graepel. “Private Traits and Attributes Are Predictable From Digital Records of Human Behavior.” Proceedings of the National Academy of Sciences (PNAS) (2013): 1–4. Print.

Moretti, Franco. “Conjectures on World Literature.” New Left Review 1 (2000): n.pag. Web. 4 Dec. 2013. <http://newleftreview.org/II/1/franco-moretti-conjectures-on-world-literature>.

Nelson, Ted. “From Computer Lib/Dream Machines.” The New Media Reader. Eds. Noah Wardrip-Fruin and Nick Montfort. Cambridge, MA: The MIT Press, 2003. 301–338. Print.

“News from the Library of Congress.” LoC.gov. The Library of Congress, 15 April 2010. Web. 4 Dec. 2013. <http://www.loc.gov/today/pr/2010/10-081.html>.

Mackness, Jenny, Sui Fai John Mak, and Roy Williams. “The Ideals and Reality of Participating in a MOOC.” Proceedings of the 7th International Conference on Networked Learning (2010): 266–275. Print.

Mak, Sui Fai John, Roy Williams, and Jenny Mackness. “Blogs and Forums as Communication and Learning Tools in a MOOC.” Proceedings of the 7th International Conference on Networked Learning (2010): 275–285. Print.

Marres, Noortje S., and Esther Weltevrede. “Scraping the Social?” Journal of Cultural Economy 6.3 (2013): 313–335. Print.

McAuley, Alexander, et al. “The MOOC Model for Digital Practice.” Massive Open Online Courses: Digital Ways of Knowing and Learning. 2010. Print.

Oremus, Will. “Free Online Classes Are an Unsustainable Gimmick. Here's a Better Idea.” Slate Magazine. The Slate Group, 18 Sept. 2013. Web. 3 Dec. 2013. <http://www.slate.com/articles/technology/technology/2013/09/spocs_small_private_online_classes_may_be_better_than_moocs.html>.

Pappano, Laura. “The Year of the MOOC.” International New York Times. The New York Times Company, 2 Nov. 2012. Web. 4 Dec. 2013. <http://www.nytimes.com/2012/11/04/education/edlife/massive-open-online-courses-are-multiplying-at-a-rapid-pace.html>.

Poell, Thomas, and Erik K. Borra. “Twitter, YouTube, and Flickr as Platforms of Alternative Journalism: The Social Media Account of the 2010 Toronto G20 Protests.” Journalism 13.6 (2012): 695–713. Print.

Purser, Emily, Angela Towndrow, and Ary Aranguiz. “Realising the Potential of Peer-to-Peer Learning: Taming a MOOC with Social Media.” eLearning Papers 33 (2013): 1–5. Print.

Puschmann, Cornelius B. A., and Jean Burgess. The Politics of Twitter Data. HIIG Discussion Paper Series. Alexander von Humboldt, Institut Für Internet und Gesellschaft, 2013. Print.

Rieder, Bernhard. “The Refraction Chamber: Twitter as Sphere and Network.” First Monday 17.11 (2012): n. pag. Web. 4 Dec. 2013. <http://firstmonday.org/ojs/index.php/fm/article/view/4199/3359>.

———. “What Is in PageRank? A Historical and Conceptual Investigation of a Recursive Status Index.” Computational Culture 2 (2012): n. pag. Web. 4 Dec. 2013. <http://computationalculture.net/article/what_is_in_pagerank>.

Rogers, Richard A. “Debanalizing Twitter: The Transformation of an Object of Study.” Proceedings of the 5th Annual ACM Web Science Conference (WebSci ’13) (2013): 356–365. Print.

———. Digital Methods. Cambridge, MA: The MIT Press, 2013. Print.

———. “Post-Demographic Machines.” Walled Garden. Eds. Annet Dekker and Annette Wolfsberger. Amsterdam: Virtueel Platform, 2009. 29–39. Print.

Sunstein, Cass. “Fragmentation and Cybercascades.” Republic.com. Princeton: Princeton University Press, 2001. 51–88. Print.

van Treeck, Timo, and Martin Ebner. “How Useful Is Twitter for Learning in Massive Communities? An Analysis of Two MOOCs.” Twitter and Society. Eds. Katrin Weller, et al. Bern: Peter Lang International Academic Publishers, 2013. 411–424. Print. Digital Formations.

“Twitter Statistics” Statistic Brain. Statistic Brain Research Institute, 5 July 2013. Web. 4 Dec. 2013. <http://www.statisticbrain.com/twitter-statistics/>.

“Update on the Twitter Archive At the Library of Congress.” Library of Congress, Jan. 2013. Print.

de Waard, Inge, et al. “Exploring the MOOC Format as a Pedagogical Approach for mLearning.” Proceedings from mLearn (2011): n. pag. Print.

Zimmer, Michael. “‘But the Data Is Already Public’: On the Ethics of Research in Facebook.” Ethics and Information Technology 12.4 (2010): 313–325. Print.

Info
Title: Following MOOC on Twitter
Subtitle: A Network Analysis of a Twitter Dataset on Massive Open Online Courses
Type: Research report; Assignment
Author.name: T. (Tessa) de Keijser; F. N. (Fernando) van der Vlist
Author.affiliation: Graduate School of Humanities, Faculty of Humanities, University of Amsterdam
Instructor.name: Dr. B. (Bernhard) Rieder; E. K. (Erik) Borra; Prof. Dr. R. A. (Richard) Rogers; Dr. A. (Anat) Ben-David
Instructor.affiliation: Dept. of Media Studies, Faculty of Humanities, University of Amsterdam
Abstract: We report on our investigation of Twitter activity around a particular set of keywords or hashtags by using the DMI-TCAT “mooc” dataset. For our data, we relied on the “mooc” dataset available through the DMI-TCAT, an open-source toolset that captures tweets and allows for multiple types of analysis (hashtags, mentions, users, search, and so forth). For further analysis of this dataset, we relied on Gephi to help us reveal patterns and trends, find the influential hubs, link clusters and giant components, betweenness centrality (bonding and bridging nodes), and content-sharing communities and subcommunities within that dataset. More specifically, we visualise and analyse a social graph by mentions, a social graph by in_reply_to_status_id, and a bipartite hashtag-URL graph. The results of our analysis show that debates around MOOCs are still happening. This debate is spread out over a number of social groups, as the types of actors that tweet on MOOC vary greatly and the URLs and hashtags they use mostly relate to broader themes on new media technologies and online education. Furthermore, we found that the discussion space centreing on MOOCs is multilingual and that this multilingual space is for a big part dominated by English, French and Spanish users.
Keywords: Internet research, digital methods, platform studies, network analysis, graph visualisation, Twitter, MOOC
Length.words: 5,971
Length.reading: 34 mins
Sections: Summary; Keywords; 1. Introduction; 1.1. Twitter Studies; 1.2. Following Twitter; 1.3. Massive Open Online Courses: Literature Review of Studies and Critiques; 2. Research Questions; 3. Methods; 3.1. DMI-TCAT; 3.2. Network Analysis; 4. Results; 4.1. Social Graph by Mentions; 4.2. Social Graph by in_reply_to_status_id; 4.3. Bipartite Hashtag-URL Graph; 5. Discussion; 6. References
Element.table: Table 1; Table 2; Table 3; Table 4
Element.figure: Fig. 1; Fig. 2a; Fig. 2b; Fig. 2c; Fig. 3; Fig. 4; Fig. 5; Fig. 6a; Fig. 6b; Fig. 7; Fig. 8
Date.submitted: 6 Dec. 2013
Date.evaluated: 9 Dec. 2013
Date.publishedonline: 1 Feb. 2014
Language: English (United Kingdom)
Documentation.style: Modern Language Association (7th ed.)
Export.citation: BibTEX
2012– fernandovandervlist.nl
v1.2.28