The Bottom-Up Formation and Maintenance of a Twitter Community: Analysis of the #FreeJahar Twitter Community

Purpose – The article explores the formation, maintenance and disintegration of a fringe Twitter community in order to understand if offline community structure applies to online communities Design/methodology/approach – The research adopted Big Data methodological approaches in tracking user-generated contents over a series of months and mapped online Twitter interactions as a multimodal, longitudinal ‘social information landscape’. Centrality measures were employed to gauge the importance of particular user nodes within the complete network and time-series analysis were used to track ego centralities in order to see if this particular online communities were maintained by specific egos. Findings – The case study shows that communities with distinct boundaries and memberships can form and exist within Twitter’s limited user content and sequential policies, which unlike other social media services, do not support formal groups, demonstrating the resilience of desperate online users when their ideology overcome social media limitations. Analysis in this article using social networks approaches also reveals that communities are formed and maintained from the bottom-up. Research limitations/implications – The research data is based on a particular dataset which occurred within a specific time and space. However, due to the rapid, polarising group behaviour, growth, disintegration and decline of the online community, the dataset presents a ‘laboratory’ case from which many other online community can be compared with. It is highly possible that the case can be generalised to a broader range of communities and from which online community theories can be proved/disproved. Practical implications – The article showed that particular group of egos with high activities, if removed, could entirely break the cohesiveness of the community. Conversely, strengthening such egos will reinforce the community strength. The questions mooted within the paper and the methodology outlined can potentially be applied in a variety of social science research areas. The contribution to the understanding of a complex social and political arena, as outlined in the paper, is a key example of such an application within an increasingly strategic research area and this will surely be applied and developed further by the computer science and security community. Originality/value – The majority of researches that cover these domains have not focused on communities that are multimodal and longitudinal. This is mainly due to the challenges associated with the collection and analysis of continuous datasets that have high volume and velocity. Such datasets are therefore unexploited with regards to cyber-community research.


Introduction
Can communities form within Twitter? The social networking and multiplatform micro-blogging service that allows a limited tweet of 140 characters is generally associated with the spread of information.
Users however are able to follow, or subscribe to the posts of other users, which create a network of followers and followees. It must not be argued that the followers-followees are a community as observations of users are mainly inactive with outbursts of retweets around certain viral news. It may be argued that the followers-followees phenomenon is at most a network with equal ties. We must however probe deeper in order to identify and isolate potential community behaviour within which the Twitter environment can provide. This is important partly due to the fact that many recent world events that tipped the balance of powers of governments originated from coordinated activities within Twitter, and partly due to the need to understand collective behaviours in cyber-communities. Twitter as a social media however, is certainly a research environment (Golder & Macy, 2013) where we can explore our questions.
Twitter service interprets keywords prefixed by a hashtag '#' as topical, and users are preceded with a @ in the tweets. Unlike other social networks, the service does not allow formal group creations. The sequential presentation of tweets as viewed within Web browsers or on mobile devices is chronological.
Can communities form under such a limited environment? If communities are able to form, in what way do they maintain and support their existence, especially when Twitter does not allow formal groups?
The interest of this research lies in the controversial online teen #FreeJahar movement calling for the freedom of the Boston bombing suspect Dzhokhar Tsarnaev because teenage girls believe he is "too beautiful to be a terrorist" (Nelson, 2013). Concerns were raised that in on-line forums the younger Boston Bombing suspect appeared to attract a cult teen following expressing affection and concern for him (DailyMail, 2013). Facebook, Tumblr tribute accounts were set up in support of the teen with the #FreeJahar Twitter tag, which were trending when these activities first appeared. Teen activities in Twitter do need to be monitored (Wiederhold, 2012) as observed below: "I can't be the only one who finds the suspected bomber to be sexy, can I?" 19 April 2013. "i don't even care if jahar is a terrorist he's cute i don't want him to die." @******, 20 April 2013. "I'm not gonna lie, the second bombing suspect, Dzhokhar Tsarnaev, is hot. #sorrynotsorry" @******, 20 April 2013. "Yes I like Justin Bieber and I like Jahar but that has nothing to do with why i support him. I know hes innocent, he is far too beautiful" @******, 25 April 2013.
The term community has two major uses. The first being 'territorial' and 'geographical', which refers to the notion of community-neighbourhood, town and city, whereas the latter refers to the relational aspect of a community (Gusfield, 1975). Whilst geography plays an important role in the formation of social ties within online communities (Takhteyev, Gruzd, & Wellman, 2012), it is believed that online social networks such as Twitter can be highly relational. Such communities are concerned with the quality of character of human relationship (spiritual, professional, ideological, etc.) without reference to territories. It is noted early that modern society develops community around interests and skills more than around locality (Durkheim, 1964), the consistency of such a notion of community has been maintained in the Information Age. Community is better defined by the nature of relationships between individuals rather than geographical proximity (Preece & Maloney-Krichmar, 2005).
It was noted a decade ago that "the Internet has altered our sense of boundaries, participation, and identity" (Renninger & Shumar, 2002), and much ink has been filled on the topic of whether online communities are really communities (Bruckman, 2006). Twitter, unlike other services such as Usenet newsgroups, Internet Relay Chat (IRC), Facebook, SecondLife, etc., that allows a formal formation of communities (Wellman & Gulia, 1999) is different as it was originally created as a messaging service. It therefore may not be suitable to study Twitter using a similar approach in the literatures, see (Preece & Maloney-Krichmar, 2005;Wellman & Gulia, 1999).
How then does a community look like in Twitter, if the definition of a community that we know can be formed at all? To answer this question, we must first look at communities familiar to us, i.e., those that have been formulated in the literatures. A community should exhibit a 'Sense of Community' (D. W. take a nominalist approach in defining the concepts, particularly of community boundaries, in reality, the automated computational approach that is employed in the research gathers datasets within the realist strategy (Laumann, Marsden, & Prensky, 1989). Considering the volume of Twitter data that this research collects, it will be extremely tedious to attempt to find a boundary using qualitative approaches, thus, a computational, quantitative approach in social networks analysis is used. Comparisons thus can be made and hypotheses tested when data has been analysed. We may begin to see patterns of community as we explore the datasets in the subsequent sections.

Methods
As opposed to the conventional method of mapping the follower-followee network, the approach defined here practically maps the actual evolution of instantaneous activities occurring within a timescale, over a series of days encompassing both large and small events. This is more useful as activities define the true interactions of active members.
A Big Data Twitter streaming software was used (Ch'ng, 2014) for mapping Twitter users (egos) and tweets as nodes, and edges representing links between egos and their tweets. Tweets are represented as nodes so that the flow of information is made obvious. Only tweets containing the keywords #Dzhokhar, #FreeJihad, #FreeJahar, #Tsarnaev are recorded. The reason for using only four keywords was that these were the keywords that were consistently used in the Tweets. In fact, #FreeJahar would have been sufficient as it appeared in all of the tweets. This resulted in 60 longitudinal datasets, each containing 5 hours of continuous data from 17 May 3.00pm and to 31 May 12.05pm (15 days). The data were recorded from 17 May onwards as news of the 'movement' were not reported until then. An additional 30 days of data records the decline of the #FreeJahar activities.
The file sizes of the series are shown in Figure 1. Peaks and valleys are consistent with the time of activities during the hours spanning both days and nights, except when the keywords were trending. The relative importance of nodes uses Betweenness, Closeness (Freeman, 1979;Newman, 2005;Sabidussi, 1966) and Eigenvector (Bonacich, 1987) centralities measures. Betweenness is a measure of information brokerage between parties, Closeness measures the spread of information from a node to all other nodes, where lower closeness has shorter distance to other nodes. High Eigenvector demonstrates increased numbers of egos who were connected to important egos in the network, an indication of heightened activities.

Results
An analysis of the datasets shows that news events were all related to the Boston Bombers with the topics: 'triple murder', 'FBI kills man', 'Ibrahim Todashev', 'Al Qaeda Mag praises Tsarnaev brothers', 'Dzhokar Recovers', 'Mother and Father', 'Russians provides info about brothers'. Does the #FreeJahar community exists within the graph? Visualising the mapped networks will reveal this information.
Figure 2 visualises 16 graphs, 5 hours each (ranked by file size from the smallest 5708 to the largest 803, left to right, top to bottom). The graphs were reconfigured using Gephi's ForceAtlas algorithms so that connected nodes due to higher interactions, appear closer together (clusters in Figure 2). Each graph shows a different signature as they carry varying explosions of information when news went viral. Egos

The Bottom-Up Organisation of the Community
Communities do not form and then disintegrate. Efforts are needed to maintain the boundary and reinforce membership bonds so that the community becomes stronger over time. The limited and sequential nature of the Twitter environment makes it difficult to maintain an active community boundary, there is a higher probability of disintegration unless members assume some form of leadership.
Conversely, as the #FreeJahar group is formed from disparate actors with strong similar ideology, the community may be organised from the bottom-up where all members have equal importance. This reinforces the theory that positivity and success in the interactions create cohesion (Cook, 1969), external conflict increase internal cohesion (Stein, 1976).

Figure 6.
The centrality measures of the cluster corresponds to Leavitt's observation (Leavitt, 1951), that "where high centrality, and hence independence are evenly distributed, there will be no leader, many errors, high activity, slow organisation, and high satisfaction". The edges that play a central role in connecting the small-world network can be traced within 2 steps of egos with high Betweenness centrality within the social movement, the same agent with high Betweenness centrality makes the community cohesive. Removing these egos will invariably disrupt the entire community. Moody and White (Moody & White, 2003) observed that "a group is structurally cohesive to the extent that multiple independent relational paths among all pairs of members hold it together". In this context, removing the egos that keep the cluster alive will disrupt the community. Twitter's removal of highly active members confirmed Moody and White's concept of 'structural cohesion', defined as "the minimum number of actors who, if removed from a group, would disconnect the group". The removal of #FreeJahar members was due to the infringement of Twitter policies. As a result of the infringements, these accounts have since been suspended or deactivated. There was a Twitter post on the 30 July 2013 by one of the active members -"Aint nobody wanna #freejahar no more?" and 31 July -"Why is everyone deactivating their accounts?
This battle is just beginning! #freejahar". This member's account has also been suspended. Figure 5 presents samples from the final decline of the #FreeJahar community.

Discussion
In this article, a longitudinal Twitter dataset associated with the #FreeJahar group calling for the freedom of the Boston bombing suspect were explored. The datasets consist of 5 hourly tweets over 45 days mapped as a network of activities present opportunities for discovering global behaviours from instantaneous contents produced by collective social actors as they interacted desperately at the local level within the confines of a digital display.
The tracking of Twitter activities apart from the follower-followee network reveals distinct spatial expressions between tweets, retweets and conversations. Using this approach, tweet nodes and edges constituting a conversational nature could be identified and isolated from characteristic retweets. The tracking of multimodal connections will give us a more accurate measure of information than a followerfollowee network.
A number of questions were presented at the beginning, probing the possibility of communities forming and maintained within the limited Twitter environment. Data analysis shows that communities do form within Twitter, and as a consequence, raise specific issues on coordinated behaviour and information dissemination within the social media. Twitter community differs from offline community in many ways due to the limits of the Twitter environment, the most apparent is reinforcement, and the support needed amongst members. It is not clear if online Twitter community facilitates offline gatherings, or if Twitter social ties led to other online groups (FaceBook friendships, email and phone exchanges, and etc.) as data could not be obtained. However, to this end, we are at least able to describe the nature of Twitter communities and how they are formed and maintained.
We have learned that Twitter communities are relational, formed via a common ideology and justified by validation of the ideology and the commonality of symbols. These worked together to segregate the in-groups from the out-groups. Members fulfil their needs via discussions and defended their cause against conflicts from another community, which creates internal cohesion. MacMillan and Chavis stated that, "people possess an inherent need to know that the things they see, feel, and understand are experienced in the same way by others" Such a group norm validates their experience. Influence therefore is unidirectional -members influence the group. The community is organised from the bottomup, with equal distribution of leading roles and activities over time. The eventual decline of the #FreeJahar community was due to the suspension of important egos from Twitter, resulting in the destruction of the community structure.
The #FreeJahar event is an exemplar case study that could be generalised to much broader scopes for this sort of work as it demonstrates rapid, polarising grouping, behaviour, growth, disintegration and decline of an online community. The study shows that communities with distinct boundaries and memberships can form and exist within Twitter's limited user content and sequential policies, which unlike other social media services, does not support formal groups, demonstrating the resilience of desperate online users when their ideology overcome social media limitations.
Social networks can increase our range of human connectedness beyond the boundary of users' geographical location. Communications sent now may be retrieved and responded to, much later in time.
This invariably opens up a broad range of opportunities as space and time, in the eye of a user are 'compressed' to within a digital display. The fact that communities can form where services that facilitate group formation are not supported is an interesting phenomenon to look at. It will be beneficial to collate extremely large datasets from ad-hoc communities within Twitter in the future, particularly where revolutions and socially mediated civil uprisings are concerned.