The Hashtags Rivalry behind the Controversial Bill: A comparative study on the Opposition and Support Movement of Omnibus Law Bill in Indonesia.

(1)

The Hashtags Rivalry behind the Controversial Bill:

A comparative study on the Opposition and Support Movement of Omnibus Law Bill in Indonesia.

Master Thesis – Applied Social Analysis

Author: Imelda Riris Damayanti Supervisor:

Victoria Yantseva

Giangiacomo Bravo

(2)

Abstract

A controversial bill aimed to stimulate investment and boost the economy in Indonesia, called the Omnibus Law Bill, is followed by both protest and support expressed in social media prior to its signatories in October 2020. During that time, the Twittersphere is packed with both the Opposition and Support movement of the bill, who both benefit from the use of hashtags. To distinguish an organic grass- roots movement from a propaganda that fits the agenda of the government and elite, a comparison study is conducted with a framework of top-down and bottom-up- mechanism of information virality (Nahon & Hemsley, 2013). The top-down mechanism combined with participatory propaganda theory is designated to explain the Support movement. Vice versa the bottom-up mechanism is combined with connective action theory designed to explain the Opposition movement as its character in line with a contemporary and digital protest movement (Bennett &

Segerberg, 2012). As existing research only often studies both networks alone, this unique case provides an opportunity to compare both networks. A mixed-method of Social Network Analysis (SNA) and Topic Modelling used to differentiate the characteristics of both groups, based on both network structure and topics discussed.

The finding in regards to the SNA is corresponding to the theoretical framework and previous studies. The loosely organized nature of connective action is reflected in several characteristics of the Opposition Network, in contrast to the element of coordination found in the Support Network. Findings from bi-term topic modeling, however, both contradict and support the hypothesis that suggests more variations in the topics within the Opposition Network as a result of the self-motivated participant and personalized messages (Leong et al., 2019).

Keywords:

Connective Action, Participatory Propaganda, Social Network Analysis, Topic Modelling, Omnibus Law, Indonesia

(3)

Acknowledgment

I would like to express my gratitude for the support that I have received throughout the process of writing this thesis. I am beyond thankful to my supervisor, Victoria Yantseva, for her constructive feedback and supportive guidance that help me to improve my work. Furthermore, I would like to thank lecturers for their dedication and inspiration throughout my journey throughout the Master’s Program on Applied Social Analysis. Particularly to Gergei Farkas, who was in charge of the course Social Network Analysis. I also recognize the technical and administrative support from the Department of Social Studies. Last but not least, I would like to thank my family for their enormous support.

(4)

Two Opposite Movements in Response to the Omnibus Bill

4

1.3

Research Objectives

4

2 Theoretical Framework 5

2.1 Bottom Up and Top Down Mechanism: The Virality Framework 5

2.2 The Logic of Connective Action 6

2.3 The Role Dynamics in Connective Action 6

2.4 Participatory Propaganda 7

3 Methodology 8

3.1

Social Network Analysis (SNA)

8

3.1.1

The Measure of Participation: Average Number of Ties

9 3.1.2

The Measure of Communication Approach and Information

Spread: Diameter 10

3.1.3

The Measure of Coordination and Organizational Approach I:

Density and Clustering Coefficient. 10

3.1.4

The Measure of Coordination and Organizational Approach II:

Centralization 11

3.1.5

The Measure of Most Prominent Nodes based on Two Centrality

Measure 12

3.2

Textual Analysis (Topic Modeling)

12

3.2.1 The Measure of Coordination of Message I: Unique Words 13 3.2.2

The Measure of Coordination of Message II: Variations of Topics

and Provocative Content 13

3.3 Ethics 13

3.4 Limitation 14

4 Result and Analysis 14

4.1 The Measure of Participation: Average Number of Ties 14 4.2

The Measure of Communication Approach and Information

Spread: Diameter 15

4.3

The Measure of Coordination and Organizational Approach I:

Density and Clustering Coefficient. 15

4.4

The Measure of Coordination and Organizational Approach II:

Centralization 16

4.4.1 Indegree Centralization 16

4.4.2 Betweenness Centralization 17

4.5

The Measure of Most Prominent Nodes based on Two Centrality

Measure 19

4.4.1 Most Prominent Nodes in the Opposition Network 19

(5)

4.4.2 Most Prominent Nodes in the Support Network 20 4.6

The Measure of Coordination of Message I: Unique Words

21 4.7

The Measure of Coordination of Message II: Variations of Topics

and Provocative Content 22

5 Theoretical Framework 28

5.1 Discussion on Network Structure and Characteristics 28

5.2 Discussion on Topic and Context 30

References 32

(6)

(7)

1 Introduction and Background

The use of hashtags as a tool for a social movement is nothing entirely new, particularly when it comes to the Twittersphere. The increasing role of Twitter as an alternative platform for activism roots in how this social media outlet enables users to craft and disseminate their message in order to gain support (Majchrzak et al., 2013). One feature, the Twitter hashtag, stands out in the midst of this new approach of activism as it helps in the spread of counter-narrative, mobilization of participation, and build up such diverse networks of support (Wang et al., 2016).

The term hashtag activism, defined as “act of fighting for or supporting a cause with the use of hashtags as the primary channel to raise awareness of an issue and encourage debate via social media” (Tombleson & Wolf, 2017, p. 15), allows an individual to shows their self-association to a certain socio-political issue in a digital sphere (Gleason, 2013; Gruzd et al., 2011). One single hashtag is able to create a social context that benefits in connecting like-minded users (Xiong, Cho &

Boatwright, 2019; Xu, 2020).

For the past years, hashtag activism has been highly linked to wide-ranging political protests and movements all over the world due to its ability for effective information spread that further boosts the coordination of action (Shirky, 2011). Including the events of large-scale movements such as the Arab Spring and Occupy Wall Street (Castells, 2015). Although hashtag activism alone is not considered powerful enough to single-handedly create a political revolution (Fuchs, 2012), it is still significantly crucial in social movements nowadays. Hashtag activism plays a unique role through the culture of autonomy and self-organized participation (Castells, 2015).

Nevertheless, the use of the hashtag is also known to be a tool for participatory propaganda, to benefit the powerful and the rich. Evidence shows how social media has been used by the government in China able to detect, monitor, and suppress collective expression (King, Pan, & Roberts, 2013). Such alarming exert of power can be scaled up to the extent of limiting internet access, and planting provocateurs (Qiang, 2011; Zittrain, 2008; Morozov, 2009).

The Omnibus Law Bill on Job Creation in Indonesia, a rather controversial bill aimed to stimulate investment and boost the economy, is followed by both protest and support expressed in social media (Ghaliya, 2020; Patrick, 2020). Labor unions expressed their frustration upon the content of the Law Bill as it puts the working class into a more vulnerable position (Paddock, 2020). Experts have also pointed out the flaws in this law-making process, including the lack of transparency and inclusive public participation (Beech & Suhartono, 2020). At the same time, a contradicting narrative has been spread based on the trust in the current government and the urgency of job creation (Patrick, 2020). Quickly the Twitter platform has turned into a debate space where both parties try to pursue two completely different agendas.

Existing literature in the field of social network and text analysis have provided valuable insights on both digital activism and participatory propaganda,

(8)

individually. In the area of digital activism, some studies managed to explore the characteristics of such networks with the focus on identifying the key players (Lotan et al., 2011), the distinct roles played by activists in the network (Meraz &

Papacharissi, 2013), and how activists across different social movements can form a close community on Twitter (Wang et al., 2019). Some others draw more attention to the content analysis such as co-occurrence patterns and which type of hashtags have a higher tendency to become viral (Blevins et al., 2019; Wang et al., 2016). In regards to participatory propaganda, one study reveals the characteristics in the case of Trump’s Facebook pages based on both social networks and content analysis, including the formation of an echo chamber, and the utilization of provocative content (Wanless & Berk, 2017). However, it is difficult to find a specific study that systematically compares the two networks that often collide against each other in the digital world. With a unique opportunity provided by the Omnibus Law case, this paper seeks to close this gap by comparing the different characteristics between two Twitter hashtag networks; the Opposition network, and the Support network of the law bill.

With a framework of top-down and bottom-up mechanisms for the information dissemination of hashtags, this study is designed to identify how two different networks with two opposing political agendas can be differentiated through the identification of their social network characteristics. Thus, it will also be an attempt to analyze the textual context of tweets that appeared on each network to spot differences in the nature of the discussions by conducting topic modeling. By doing so, this paper will make a contribution based on the mixed method and two-level of analysis; the network and the content. It will provide evidence to separate the two different hashtag networks, one built in a top-down approach, and one as a bottom- up movement.

Furthermore, the comparison between the two networks is crucial in the midst of increasing use of the hashtag and digital platform by both grass-root activists and political propagandists. While answering the two research questions, the effort to separate such networks apart can benefit the discussion regarding the power of information dissemination through digital media and the two sides of the coin from free speech in a democratic country. As political propaganda takes place in the digital platform, it is not only possible for it to be disguised but also to manipulate people to spread it unknowingly (Haigh et al., 2017). Worst, people can be driven to produce, belief, and spread inaccurate content (Mejias & Vokuev, 2017) that can lead to massive misinformation benefiting the status quo (Haigh et al., 2017).

(9)

1.1 The Omnibus bill on job creation

Right at the beginning of his inauguration speech in October 2019, President Jokowi Dodo expressed the great ambition of his second presidential term to lead Indonesia out of the middle-income trap (Tehusijarana, 2019). To achieve it, five priorities of the administration are introduced. The third priority stands for the elimination of obstructive regulations through the establishment of two laws that will revise several previous laws. One of them is the Omnibus Bill on Job Creation.

The Omnibus approach is considered new in the process of law-making in Indonesia (Indrati, 2020). The Black’s Law Dictionary defines an Omnibus Bill as follows:

“(1) A single bill containing various distinct matters, usu. drafted in this way to force the executive either to accept all the unrelated minor provisions or to veto the major provisions.; (2) A bill that deals with all proposals relating to a particular subject, such as an ‘omnibus judgeship bill’ covering all proposals for new judgeships or an ‘omnibus crime bill’ dealing with different subjects such as new crimes and grams to states for crime control” (Garner, 2004, p.175). According to the Center for Constitutional Law Studies from the University of Indonesia, this method is entirely alien where none of it is regulated under Law Number 12 the Year 2011 regarding the Establishment of Legislation (Putri, 2020). The official page of the Omnibus Bill mentioned how such an approach is often used in Anglo- Saxon countries that adhere to the Common Law system (Coordinating Ministry of Economic Affairs, 2020). A rather strange comparison since the system in Indonesia is based on Civil Law.

The Omnibus Law Bill is expected to live up to its name, by simplifying and harmonizing the existing regulations. It is expected to contribute to economic development by creating jobs, stimulating new investments, and empowering small businesses (Coordinating Ministry of Economic Affairs, 2020). However, the controversial content of this law bill is not a reflection of the glorified claim in which it is considered to be in favor of the elite investor and none others (Firdaus, 2020). According to labor unions, Omnibus Law will put the working class in danger, lowering access to welfare and social protection (Paddock, 2020). The labor articles in Omnibus Law Bill will affect Law Number 13/2003 on Manpower, Law Number 4/2004 on National Social Security System, and Law Number 24/2011 on Social Security Organizing Agency by changing the previous provisions on severance pay, the scope of work, foreign workers, remuneration, working hours, labor rights, layoffs, and social security (Samboh, 2020). Such changes are seen to harm the people, including by the reduction of severance pay, subtraction of mandatory leave, and permission to implement longer work hours, (Samboh, 2020).

Apart from the impact on labor rights, the content of this bill is seen to be a major setback in environmental protection in Indonesia since it will abolish environmental permits to simplify the overall business processes and therefore will amend Law No 32/2009 on Environmental Protection and Management (Sembiring, Fatimah &

Widyaningsih, 2020).

The debate surrounding this bill also covers the law-making process that is considered to be “the worst legislative process in Indonesia history” (Beech &

Suhartono, 2020). Critics argue that the current regime poses a threat to democracy

(10)

due to the lack of inclusive public participation, and weak transparency hence conflicts against Law No 12/2011. From the beginning of the drafting process, public criticism was quickly arise following the formation of a task force for public participation in December 2019 because of the underrepresentation of crucial civil- society elements whereas the particular task force only consists of the government and business stakeholders (Sembiring, Fatimah & Widyaningsih, 2020). Another crucial point is the poor access to the draft bill and its development. The public faces major difficulties in obtaining the content of the bill even up to the day into a law that evidently is different from the latest accessible version and has expanded from 812 to 1,187 pages (Beech & Suhartono, 2020).

1.2 Two Opposite Movements in Response to the Omnibus Bill

The criticism over the Omnibus Law followed with a phenomenon in social media where one hashtag, #TolakOmnibusLaw (#RejectOmnibusBill) is widely used to express resistance against this law bill, particularly on the Twitter platform. The hashtag appeared for the first time on Labor Day, 1st of May 2020, as a form of protest by workers and labor unions (Ghaliya, 2020). Due to the pandemic situation, the use of this hashtag provides an alternative way for the larger public to join the protest (Patrick, 2020). In this paper, the social network formed based on this hashtag movement will be referred to as the Opposition Network.

On the other side, the Indonesian Twittersphere has also become the arena for groups who support the Omnibus Bill to express their opinion. A growing number of public concerns regarding who will benefit from this narrative appear following the spread of several supporting hashtags, including #DukungOmnibusLaw (#SupportOmnibusLaw), #kitabutuhciptakerja (#WeNeedJobCreationLaw), and

#rakyatbutuhkerja (#PeopleNeedJobs) (Patrick, 2020). The suspicion in regards to the nature of this group, whether or not it is an organic movement by the people or a propaganda movement, was shortly answered. Starting in August 2020, several well-known social media influencers and public figures with a high number of followers were exposed (Pangestika, 2020). Later on, they acknowledged and apologized for their involvement in promoting the Omnibus Law, admitting that they indeed receive payment to spread such narrative by using hashtags across different social media platforms (Pangestika, 2020). This second social network will be called the Support Network.

1.3 Research Objectives

This research aims to see the differences between the Opposition Network and Support Network, with a concern given to both network characteristics and content analysis. The research questions are:

1. What are the key differences in the structure and characteristics of the support and opposition networks of the Omnibus Law Bill in Indonesia?

2. How do the narrative and discussion differ between the support and opposition networks of the Omnibus Bill in Indonesia?

(11)

To answer the above questions, the paper will use Twitter data, based on several prominent hashtags that appeared in both support and opposition narratives, and build two distinguished social networks. Social Network Analysis (SNA) will be used to answer the first question where the main characteristic and structure of both networks will be identified and compared to one another. The second question will be answered with the method of topic modeling where the author will further qualitatively analyze the discussion that arises in both networks.

2 Theoretical Framework

2.1 Bottom Up and Top Down Mechanism: The Virality Framework

Nahon and Hemsley (2013) deconstruct the occurrence of virality in this era of social media by two distinguished processes, a top-down process, and a bottom-up process. Top-down forces drive viral information through the role of promotional effort and network gatekeeping. The essence of the top-down process lies in how the fast spread of information is designed by people and institutions in power, including politicians and mass media. On the other hand, the bottom-up process rather represents the ideal or more common perception about virality where it is driven organically by anyone who agrees that the information is worth spreading. It is about a massive amount of people that share similar interests to disseminate the message.

The bottom-up process is often found in networked social movements as self- motivated participants are the key to spreading the message (Castells, 2015). Within this process, two major factors affect the decision of an individual to share certain content to their social networks (Nahon & Hemsley, 2013). First, the message itself.

Among a few information characteristics that influence information sharing behavior, salience and relevance of the message are particularly important (Nahon

& Hemsley, 2013). Second, the network structure shows that some viral cases can circumvent gatekeepers (Nahon & Hemsley, 2013). The dynamics in such a structure also represent the combined influence of weak ties and strong ties in a social network that allows some to be more influential than others. Therefore, it is still possible for opinion leaders to emerge in the bottom-up process as people tend to be easily influenced by not only those with personal connections to them but also people that they knew (Allsop et al. 2007; Watts & Dodds; 2007). The bottom-up process is therefore relevant to explain the Opposition Network, as this network was started by the working-class people to protest against the ruling administration.

The top-down process indicates that virality can be controlled to some extent (Nahon & Hemsley, 2013). Promotion of content is one of the top-down forces that can also be a form of interaction between social media and mass media (Nahon &

Hemsley, 2013). This force is accompanied by gatekeepers that are influential in the process of information flow (Nahon & Hemsley, 2013). In the case of the Support Network of Omnibus Law, the promotional aspect and gatekeeping aspect are

(12)

merged, proven by the paid involvement of social media influencers and public figures to spread such supporting narrative (Pangestika, 2020). With this definitive difference in comparison to the Opposition Network, the Supporting Network can be explained by the top-down process.

2.2 The Logic of Connective Action

Bennet and Segerberg (2012) introduced the Logic of Connective Action as a framework to analyze contemporary protest, in comparison to the Logic of Collective Action for the conventional and traditional protest. There are few distinctions to identify contemporary protest in regards to the organizational and communication style. Connective action refers to loosely organized organizations where those who play significant roles in this area are not necessarily seasoned in conducting similar roles before (Bennett & Segerberg, 2012). It also means that protest is no longer organized by conventional organizations such as political parties and unions (Bennett & Segerberg, 2012, p. 744). Within a connective action, the spread of the message occurs with the lack of central organization. ‘Personalized communication’ is the way to reach as many individuals as possible, through both personal and digital networks, hence pointing out a crucial utilization of social media (Bennett & Segerberg, 2012, pp. 747-748). This use of digital media is one argument of how connective action should be studied separately compared to collective action (Bennett & Segerberg, 2012). By doing so, it doesn’t imply that one is better than the other but rather points out how participation differs because of the shift in organizational and communication style.

The Logic of Connective Action has been referred to as a framework to study digital activism, including hashtag activism in particular. It is used to analyze one particular phenomenon, for example in the research on #MeToo (Xiong, Cho &

Boatwright, 2019), #HongKongPoliceBrutality (Wang & Zhou, 2021),

#BoycottNFL (Chung et al., 2021). In this paper, the Logic of Connective Action will be a framework for the Opposition Network due to the nature of the network as a digital protest based on Twitter. As the framework highlights two essential attributes of connective action being loosely organized and depending on personalized communication, the paper will use SNA and topic modeling to measure the centralization, key players, and the message being discussed. It is expected that the Opposition Network will have a rather small centralization measure with a less conventional organization or individual as the center. Thus, the messages are expected to be less organized or identical, giving enough room for various topics.

2.3 The Role Dynamics in Connective Action

Scholars have divided social movement actors into three categories, the victims, the elites, and the non-elites. Vulnerable groups who are directly victimized by a certain policy or action tend to be the initiator of traditional social movements and play the

(13)

leadership role throughout the whole movement (Bauermeister, 2016; Melucci, 2003; Palmer et al., 2014).

The elites, such as political leaders, government officials, social activists, intellectuals, human rights groups, and journalists tend to join the movement later on due to the influence of the first group or the popularity of the movement (Cress

& Snow, 2000; Soule & Olzak, 2004). It is argued that the first group will put efforts to influence the elite to gain several strategic benefits, including political influence and media attention (Brym et al., 2014; Giugni, 1998). In the context of Twitter networks, elite groups, with the privilege of large follower numbers, can potentially have a widespread effect on their online participation in a social movement (Tremayne, 2014).

The last group, categorized as non-elites, consists of people who do not directly receive negative impacts from a particular policy or action, nor are part of the social elites. This group is often seen as the least prominent actor in the study of the traditional social movement, yet surprisingly become the key actor in several cases of digital social movement (Ansari, 2012; Brym et al., 2014; Hodges & Stocking, 2015; Lim, 2012; Ranney, 2014). This group of ordinary people manages to show their influence in the era of social media in regards to spreading information and mobilization of people in the digital social network. The way that social media change social movement to become more informal is believed to be one of the main reasons why the non-elites can play a more active role online (Coretti, 2012;

Gerbaudo, 2012).

Some scholars believe in the notion of a leaderless network, defined as “networks of individuals, groups, and masses, whose actions are generally uncoordinated but either they are able to coordinate themselves using solely the social media platform(s) or the platform is considered having the metaphysical ability to rid the masses of such organizational and coordination tasks.” (Spier, 2017, p. 79). The identification of prominent actors behind the Opposition Network became interesting to find out the extent of elite and non-elite participation.

2.4 Participatory Propaganda

A new form of modern propaganda based on persuasive communication appeared to be quite different compared to traditional propaganda due to the existence of digital platforms and the many new tools offered. The term participatory propaganda builds upon the definition of propaganda by Jowett and O’Donnell as “the deliberate, and systematic attempt to shape perceptions, manipulate cognitions and direct behavior of a target audience while seeking to co-opt its members to actively engage in the spread of persuasive communications, to achieve a response that furthers the desired intent of the propagandist.” (2015, p.7).

The participatory aspect points out that although one propaganda project is reflecting top-down communication, there is an additional vehicle that helps the massive information dissemination which none other than ordinary users of a digital platform. It marks the main character of participatory propaganda as the spread of

(14)

information is no longer about “one-to-many”, but rather “one-to-many-to-many more” (Wanless & Berk, 2017). This aspect enables propaganda to disguise as users can receive the message from people they trust in their social network (Haigh et al., 2017). However, not every involvement in the digital platform is genuine (Wanless

& Berk, 2017). Technological advances make it possible for propagandists to create and use social bots (Shao et al. 2017), and troll armies (Aro, 2016). Not to mention paid participation, as proven in the case of the Opposition Network (Pangestika, 2020).

The exploration of the dynamics and structure of the Support Network, and comparing it to the Opposition Network can help in distinguishing which network weighs more on control, and which one weighs more on participation. There are a handful of tactics that scholars have identified within participatory propaganda.

Among the six tactics elaborated by Wanless and Berk (2017), one of them is in line with the top-down mechanism explained in the previous section, which is the use of traditional media. The most interesting one that later on will be useful for the textual analysis, is how propagandists are relying on provocative content. Content such as fake news, memes, and leaks are used to trigger an emotional response from the readers that is followed by a call for action (Wanless & Bark, 2017). This nature of the content, once again, shows the remaining control aspect even in a participatory form of propaganda.

3 Methodology

The research is designed based upon a comparative approach on SNA and topic modeling where two networks are studied to find differences to distinguish both networks apart. With the purpose to earn a deeper understanding regarding the bottom-up and top-down networks, such utilization of mixed methods will allow to identify and visualize characteristics of both networks from two different aspects, the structure of the network and the context of the discussion.

3.1 Social Network Analysis (SNA)

A social network is formed by a set of actors, called nodes, who are connected by a certain type of social relationship, called ties. Such network systems in the society can be found in the most simple way as a dyad, made by a pair of people, but can be more complex, such as Facebook friends (Christakis & Fowler, 2011; Spier, 2017).

The two opposite hashtag movements in response to the controversial Omnibus Law are representative of two different social networks and can be used to show the individual complexity of each network. To study and compare both networks, SNA will be used to visualize the networks, analyze several key characteristics of the networks, and find out about the network structures and dynamics (Luke, 2015).

The first network will be called the Opposition Network representing the hashtag against the Omnibus Law. Twitter data is collected using the hashtag

#TolakOmnibusLaw, the Indonesian of “refuse Omnibus Law”. The second dataset is collected to represent the group in favor of the law bill and therefore will be

(15)

called the Support Network. Due to the different amounts of data available between the two group, this dataset is built upon seven hashtags that are all representing a similar narrative in order to minimize the difference in the size of the two networks.

The hashtags are #DukungOmnibusLaw (support Omnibus Law),

#DukungRUUCiptaKerja (support Bill on Job Creation) #kitabutuhciptakerja (we need Job Creation Law), #rakyatbutuhkerja (people need jobs), #omnibuslawhalal (Omnibus Law is halal), #omnibuslawjadisolusi (Omnibus Law is the solution),

#manfaatciptakerja (benefit of Bill on Job Creation). There are a few more hashtags used that were excluded from this network due to the relevance of the tweets under the hashtag. The timeline is limited to two months, September and October of 2020 of Western Indonesian Time (GMT+7). The selection of this timeline is designed to capture the period where the discussion gained high attention from the public, right before and after the signing of the bill on the 5th of October 2020 (Beech & Suhartono, 2020).

Both datasets were downloaded through Rstudio (R Core Team, 2020), a certified and open statistical computing program with the additional use of a Twitter Developer Account that allows the writer to access the data legally (Luke, 2015).

The number of tweets downloaded for the Opposition Network is 25.000, and the number of tweets downloaded for the Support Network is 15.350. The difference in this number is due to the availability of the data where the Support Network represents a maximum number of tweets that exist using the selected set of hashtags.

Both networks are built on identified ties of mention, retweet (RT), reply, and quote tweet. The data is also filtered to remove the isolated tweet and self-loop tweets which are considered unnecessary to the study. In addition, ties that are connected into non-active nodes, the ones who only act as recipients, are excluded from both networks to specify the focus on networks based on active nodes. The result is a total of 24,632 data units (edges) from 9,396 users (nodes) for the Opposition Network, and 11,824 data units (edges) from 2,073 users (nodes) for the Support Network. Even though both networks can be categorized as large social networks, this different size will be considered during the whole analysis process.

Nodes in both networks are designed to possess 2 vertex attributes, the number of followers, and the status of verification to differentiate the verified accounts and non-verified accounts). Both attributes are crucial to be examined further concerning the characteristics of prominent actors of both networks.

3.1.1 The Measure of Participation: Average Number of Ties

One concept that highlights the difference between the bottom-up process and the top-down process is the type of participation in both processes; in which the first one is self-motivated and the second one is a result of a designed attempt (Nahon &

Hemsley, 2013). Using a static network, it is impossible to analyze the motivation of user participation. Therefore, the concern on participation will focus on seeking how

(16)

effective and active are users in the networks are in creating interactions with each other.

Although identification of size, nodes, and ties, of networks is the first step in describing basic characteristics of a network (Luke, 2015), the comparison will start with the measurement of average ties on both networks because of the different sampling sizes. This calculation can show how participants (nodes) in both networks participated. The higher the value of average ties means that nodes in that respected network create more relations, whether in a form of mention, reply, retweet, or quote tweet. Based on the theoretical framework used, it is expected that the average number of ties will appear higher in the Support Network as a result of the more control, and more organized element of the propaganda network that can be seen through the “one-to-many-to-many more” concept of information spread (Wanless

& Berk, 2017).

H1: Higher average number of ties in the Support Network.

3.1.2 The Measure of Communication Approach and Information Spread: Diameter

The diameters of the largest network component are identified to see the number of steps separating nodes that further apart. According to Newman (2000) in “Models of the Small World”, this aspect of network structure is highly linked with the communication process on how faster the spread of information is obtained in a social network. The smaller the diameter number, the faster information will spread.

As the bottom-up mechanism and connective action point out how organic digital activism arises with the lack of gatekeeper participation and personal communication approach, the diameter value of the Opposition Network is therefore expected to be higher (Nahon & Hemsley, 2013; Bennett & Segerberg, 2012). In comparison, top-down mechanism and participatory propaganda benefitted with the additional role of promotional content spread, gatekeeper, and even bots and trolls (Nahon & Hemsley, 2013; Wanless & Berk, 2017). Hence, the Support Network is expected to have a smaller diameter value.

H2: Higher diameter value in the Opposition Network.

3.1.3 The Measure of Coordination and Organizational Approach I:

Density and Clustering Coefficient.

One key element to separate the Bottom Up mechanism from the top-down mechanism; similar to differentiate organic connective action from participatory propaganda that will be highlighted in this paper, is the organizational approach.

Suggested by the theoretical framework, the Opposition Network is expected to show characteristics of a loosely organized network led by those who traditionally are not the prominent players in social movement (Nahon & Hemsley, 2013;

Bennett & Segerberg, 2012). Meanwhile, the Support Network is supposed to show

(17)

the characteristics of a more robust and organized network as it is designed by propagandists (Nahon & Hemsley, 2013; Wanless & Berk, 2017). In this case, this distinction will be examined with density and clustering coefficient. This paper expects to see such distinction reflected in both measurement values to be higher in the Support Network and vice versa.

With the existing SNA tools, density (d) is calculated to show the proportion of ties to the maximum number of possible ties, reflecting the interconnection of each network (Luke, 2015). The clustering coefficient will be examined with the value of transitivity to show the presence of clustering in a social network. Tranisitivy value is based on the proportion of closed triplets to the entire number of existing triplets in the network. Closed triplet occurs when three nodes form a triangle that is fully connected, in other words, when “two people who share a common friend also become friends themselves” (Luke, 2015, p.16).

H3: Higher density and clustering coefficient value in the Support Network.

3.1.4 The Measure of Coordination and Organizational Approach II:

Centralization

To continue the analysis on dissimilarity between both networks in the organizational approach, the paper will look further into the networks’

centralization. In this case, SNA measurements based on degree centrality and betweenness centrality are chosen. The centralization measurement in SNA can be used as one tool to seek how far the network is controlled where the Support Network is expected to be more centralized due to the certain control that remains in the participatory propaganda network. Similar to how the theoretical framework supports the argument to the measure for density and clustering coefficient, the Opposition Network is expected to be less centralized compared to the Support Network due to the reason that participatory propaganda with top-up mechanism identic with more coordinated, and systemic movement (Nahon & Hemsley, 2013;

Wanless & Berk, 2017).

Centrality measure is essential in SNA as it can provide the prominence of certain actors through their visibility in the network (Knoke & Burt, 1983). First, the overall centralization measure will be calculated. Thus, to analyze the hashtag networks, the degree centrality measure and betweenness centrality measure will also be examined. Degree centrality will put the nodes with the highest number of direct ties as the most prominent actor in each network. In this case, the in-degree measure is selected to show the incoming interaction instead of the outgoing interaction.

Meanwhile, the betweenness centrality is based on how strategic the position of each node is, where prominent nodes will be difficult to be avoided and therefore become a path between other nodes. Both measures of centrality will be communicated through the visualization of each network, drawn using the Fruchterman-Reingold algorithm (Csardi and Nepusz 2006).

H4: Higher centralization value in the Support Network.

(18)

3.1.5 Measure of Most Prominent Nodes based on Two Centrality Measure

In addition, the writer highlights the 10 most prominent nodes based on the two centrality measures. Identifying the most important users within several SNA tools and can help to map out the most influential Twitter users within this network. The paper will further break the list down to seek the attributes of these key players on both networks based on their number of followers and verification status. The key players under this Support Network are expected to be more similar to another in regards to their attributes to show the higher degree of organization in such a network where not every participant is genuine (Wanless & Berk, 2017). On the contrary, the key players in the Opposition Network are expected to be more diverse in terms of their vertex attributes due to the more organic nature of such a connective action network where individuals with less to no experience, or the

“non-elites” can also arise as essential players (Bennett & Segerberg, 2012. On analyzing the prominent nodes, both of the node’s attributes will also be taken into account, especially for the follower number. In the context of Twitter networks, the privilege of having large follower numbers can potentially have a widespread effect on their online participation in a social movement (Tremayne, 2014).

H4: More similarities of node’s attributes in the Support Network.

3.2 Textual Analysis (Topic Modeling)

In addition to SNA, textual analysis is seen to be another method that can bring more depth to the analysis process of comparing the two hashtag networks. Tweets from each network will be processed using the Biterm Top Model (BTM) in the Rstudio. BTM is based on word co-occurrence in which “a bi-term consists of two words co-occurring in the same context” (Li et al., 2013). This process allows the writer to map out the topics being discussed in each network. BTM is chosen as the method is suitable for corpus arrived from short text, such as tweets.

Before conducting BTM, there are few steps of text cleaning and pre-processing of the corpus to obtain meaningful results. First, data cleaning by removing unnecessary components of text, including hashtags, user mention, punctuation, emoji, newline character, and web addresses. To standardize the corpus and prevent wrong interpretation of the text, slang words are replaced with their standard form by using the Colloquial Indonesian Lexicon (Salsabila, et al. 2018). Thus, the Stemming and Lemmatization processes took place to remove the prefix and suffix of a word and replace it with its basic form. Tokenization is the step to breaking the text sample into tokens, in which each token means one word. Last, stop words are removed to prevent bias in the BTM process.

Following several trials ahead, it is chosen that BTM on each network will generate 8 topics to have a more specified result as the trials with 10 and 12 topics ended up with too much repetition on the terms shown in each topic. The result of BTM will

(19)

be visualized using Worldcloud. The BTM is conducted to further analyze the level of coordination that is expected to separate both network and context of the topic with finding expectations stated in the next section.

3.2.1 The Measure of Coordination of Message I: Unique Words

One significant difference between the bottom-up and top-down mechanism of virality that also appears if one compares connective action with participatory propaganda is to what extent the movements are being coordinated. The bottom-up mechanism and connective action are dominated by self-motivated participants with personalized messages (Nahon & Hemsley, 2013; Bennett & Segerberg, 2012). This is something that is not found in participatory propaganda. To see how far the messages are being coordinated, the unique words found in both networks in ratio to the total number of words. It is expected that the ratio will be higher in the Opposition Network due to less coordination.

H6: Higher unique word ratio in the Opposition Network.

3.2.2 The Measure of Coordination of Message II: Variations of Topics and Provocative Content

The last measurement is taken into account to understand better the context of topics being discussed in each network and to follow up the measurement based on vocabulary and unique words. Following the identification of 8 topics per network, the terms formulating each topic will show the main discussion in each topic. This will be shown by visualization of the 20 most discussed terms on each topic.

First, these topics will be examined in which the Opposition Network is expected to have more variations of issues being discussed. This is because self-motivated participation in connective action is suggested to have a more personalized approach and more personalized content (Leong et al., 2019, p. 174). Meaning that the element of coordination is not as expected to be found in the Support Network.

Second, a closer look to identify provocative content will also be done. As participatory propaganda is often linked with provocative content to attract emotional response from the target (Wanless & Bark, 2017). Hence, provocative issues are expected to be found more in the Support Network, concerning the context of the current political situation in Indonesia.

H6: Higher unique topics in the Opposition Network and more provocative topics will be found in the Support Network.

3.3 Ethics

In regards to ethical consideration, there is always a concern over consent, particularly with samples taken from digital records such as social media data. The sampling method of extracting the designed number of tweets for the study is not breaching the Twitter API nor the legal consent given by Twitter users that agree

(20)

upon the accessibility of public posts for third parties (William, Burnap, & Sloan, 2017). To protect private information the writer has to avoid exposing this type of information in any section of the paper (Boyd & Crawford, 2012). The solution is by avoiding the use of highly identifiable information (Bechmann & Kim, 2020).

For example, when profiling the most prominent nodes based on centrality, the writer will not explicitly mention the user name and the individual or organization of the Twitter account, rather by only categorizing these accounts based on their nodes attributes such as follower number and verification status.

3.4 Limitation

A limitation lies in the availability and access to sampling data. The opposition movement provides an extensively larger size of data available compared to the support movement and therefore it is decided that the sampling size will be different so the sampling in the opposition movement can still be representative enough from the population. Lastly, the characteristics of the data also provide some challenges in the pre-processing and text cleaning. Tweets in Bahasa Indonesia can be very formal and very informal, depending on the users. Several pre-processing steps need to be done to standardize the corpus using this kind of sample.

4 Result and Analysis

4.1 The Measure of Participation: Average Number of Ties

Table 1. Size and Average Number of Ties

Opposition Network Support Network

Number of Ties 24632 11824

Number of Nodes 9396 2073

Average Number of Ties 2.622 5.704

It is rather interesting to point out that the different ratios found in the number of ties are not reflected in their respective number of nodes. This fact is shown by the gap on average ties values presented in Table 1 above. The directed interactions in both networks are formed by a highly different ratio of Twitter users. The Support Network, as part of a governmental effort to increase public favor of the law bill (Pangestika, 2020), indicates a much lower number of actors who are highly interacting with each other. In other words, participants of the hashtag movement in the Support Network can create more relations with each other. As it is stated in the method section, it was hypothesized that the average number of ties would be higher in the support network. In fact, the value is more than two times higher.

(21)

4.2 The Measure of Communication Approach and Information Spread: Diameter

Table 2. Diameter

Diameter 8 7

Comparing the Opposition and Support Network in terms of the diameter of their largest component, one can see that value is higher in the Support Network. This finding is supporting the hypothesis even with only a slight difference. As the diameter of the largest component of each network suggests how fast the information spreads within the network (Newman, 2020), this distinction means that the Support Network shows a more effective process of information dissemination.

However, both values are considered high diameter values of a network as 6 Degree of Separation emphasizes that everyone around the world can be connected by only 6 steps (Barabasi, 2014).

4.3 The Measure of Coordination and Organizational Approach I: Density and Clustering Coefficient.

Table 3. Density & Clustering Coefficient

Density 0.00028 0.00275

Tranistivity 0.105 0.258

In terms of density, both networks possess rather small values. The low-density value is indeed something uncommon to be found in a social network (Faust, 2006).

One important point is how huge the difference between both values is. Even though the size of the networks is different, to begin with, the Support Network appears to possess a significantly higher density value and therefore is more interconnected.

Similar to density, the transitivity measurement or clustering coefficient is also larger in the Support Network. This means that there are more proportions of closed triplets found in the Support Network.

As the two measurements are used to separate the type of coordination and organizational approach in both networks, one can see that this result is consistent with the hypothesis (H2). A more robust network as found in the organized participatory propaganda with a bottom-up mechanism (Nahon & Hemsley, 2013;

Wanless & Berk, 2017), results in a more interconnected network proven by both values of density and clustering coefficient. Meanwhile, the connective action with top-down mechanisms is less organized and allows those who do not necessarily have experience in a social movement to play a bigger role (Nahon & Hemsley,

(22)

2013; Bennett & Segerberg, 2012), the Opposition network is proven to be less interconnected.

4.4 The Measure of Coordination and Organizational Approach II: Centralization

Table 4. Centralization

Degree Centralization 0.384 0.445

Betweenness Centralization 0.001 0.027

Apart from density and clustering coefficient, another attempt to look into the difference in coordination and organizational approach of both networks is through centralization. First, the measurement of the overall centralization value is presented in Table 3. Centrality calculation allows us to see the position of nodes in these networks based on the number of ties (degree centrality) and locations between others (betweenness centrality). The result indicates how the Support Network is more centralized compared to the Opposition Network under both centrality measurements taken. The gap found in the betweenness centralization, in particular, is extremely different between both networks. This finding is in line with the hypothesis and the first look into the coordination and organizational aspect of both networks where the loosely organized connective action with bottom-up mechanism (Nahon & Hemsley, 2013; Wanless & Berk, 2017) corresponds to the lower centralization measurement and vice versa. The visualization on each network taken based on these centrality measures will be drawn in the next section.

4.4.1 In-Degree Centralization

The first measure used for this paper to visualize and communicate the centralization is degree centrality, particularly in-degree centrality, to map the positions of actors within the network based on the number of incoming ties.

(23)

Figure 1.A Opposition Network based on degree centrality Figure 1.B Support Network based on degree centrality

Looking at Figure 1.A on the Opposition Network in comparison to Figure 1.B on the Support Network, one can see that the different dispersion where the first network is more spread and the second network is more centered. Examining further to the most influential users, presented with bigger sizes in the graph, we can find that those users are not necessarily the ones with a high number of followers in both networks (shown with color; the bolder the color the higher the follower number).

However, this finding is more prominent in the Opposition Network. As seen in the graph, those nodes with bigger sizes in the Opposition Network appeared with a less prominent purple color. Meanwhile, the more bold red color is seen more often with bigger size nodes in the Support Network. Thus, in both graphs, the verified users, presented by square objects, are also not dominating the list of influential nodes.

These node attributes will be discussed further in the later sections of this paper to find the composition of the 10 most influential actors in both networks, and whether or not the finding is linked with previous studies about key players in the digital social movement.

4.4.2 Betweenness Centralization

The second centrality measure is betweenness centrality, based on the position on nodes where nodes who are located between a lot of other nodes will have a high betweenness value (Luke, 2005). The higher the betweenness values, the higher the influence of a node in the network in terms of how they can control the flow of information.

(24)

Figure 2.A Opposition Network based on betweenness centrality Figure 2.B Support Network based on betweenness centrality

Based on the visualization above, and the value of betweenness centralization found earlier in Table 4, Support Network appears to be a more centralized network. In terms of the nodes attributes, the Opposition Network points out more nodes with prominent centrality values (shown with bigger size) that correspond with higher followers number (shown with more bold purple color), compared to what was found based on in-degree centralization. Similar to what occurred in the Support Network where nodes with more follower numbers occupied the central part of the graph, compared to what was found based on in-degree centralization. However, the Twitter users at the center of the Opposition Network (figure 2.A) indicate more variability concerning their follower number, shown by the variability in the color shades.

(25)

4.5 Measure of Most Prominent Nodes based on Two Centrality Measure

4.5.1 Most Prominent Nodes in the Opposition Network

Table 5. Top 10 Nodes based on Indegree in the Opposition Network

Rank Indegree

Centrality Followers Verification

Status Top 10 based on Betweness Centrality

Betweenness Centrality

1 7223 1,001-10,000 Non-verified No 4,718.57

2 2868 1,001-10,000 Non-verified No 4,508.00

3 2626 10,001-

100,000

Non-verified Yes 11,0078.10

4 1113 100,001-

1,000,000 Non-verified Yes 83,095.00

5 1096 10,001-

100,000

Non-verified No 0.00

6 792 10,001-

100,000

Verified Yes 37,167.73

7 427 1,001-10,000 Non-verified No 0.00

8 411 101-1,000 Non-verified No 10,221.12

9 342 100,001-

1,000,000

Non-verified No 305.00

10 308 1,001-10,000 Non-verified Yes 40,813.19

Table 6. Top 10 Nodes based on Betweeness in the Opposition Network

Rank Betweness

Status Top 10 based on Indegree Centrality

Indegree Centrality

1 110078.10 10,001-

100,000 Non-verified Yes 2626

2 83095.00 100,001-

1,000,000

Non-verified Yes 1113

3 66250.92 1,001-10,000 Non-verified No 31

4 59519.35 1,001-10,000 Non-verified No 41

(26)

5 47542.92 101-1,000 Non-verified No 31

6 40813.19 1,001-10,000 Non-verified Yes 308

7 37167.73 10,001-

100,000 Verified Yes 792

8 29610.71 1,001-10,000 Non-verified No 271

9 22161.88 10,001-

100,000 Non-verified No 25

10 17282.82 1,001-10,000 Non-verified No 71

By looking at Table 5 and Table 6 on the most prominent users in the Opposition Network based upon the in-degree centrality measure, one can see how diverse these most influential nodes are when it comes to the attribute of follower number.

The same pattern is also portrayed in Table 6 in which the Top 10 nodes according to the betweenness centrality measure have various categories of follower number.

The presence of one Twitter user with a low number of followers in both tables is another thing to point out. But, at the same time, there are a total of three out of twenty users above who possess more than 100,000 followers. In terms of the second attribute, verification status, only two users on each table are stated as verified accounts. Last, only 4 users appear as key players in both measurements.

This total number of sixteen unique users out of twenty users means that nodes with a high number of ties are not necessarily the ones with strategic positions nor the ones that can not be avoided.

4.5.2 Most Prominent Nodes in the Support Network

Table 7. Top 10 Nodes based on Betweenness in the Support Network

Rank Betweenness

Status Top 10 based on Indegree Centrality

Indegree Centrality

1 115861.96 1,001-10,000 Non-verified Yes 1437

2 68730.93 10,001-

3 55880.71 10,001-

4 53620.92 10,001- 100,000

Verified Yes 402

(27)

5 51820.03 10,001-

6 44510.64 1,001-10,000 Non-verified No 288

7 27743.13 1,001-10,000 Non-verified No 35

8 26442.98 10,001-

100,000 Non-verified No 116

9 26228.38 1,001-10,000 Non-verified Yes 427

10 21894.82 10,001-

100,000

Non-verified Yes 357

Different from the pattern found in the previous network, the variation of the most influential actors in the Support Network appeared to be less. First, these key players under the Support Network only fall into the middle follower group, either between 1,001-10,000 or 10,001-100,000. None of them represent the small follower number account and high follower number account. In regards to the verification status, exact numbers of verified accounts identified, similar to the Opposition Network. However, another distinction occurred in which only thirteen unique accounts exist. In other words, seven users are found as the most prominent player in the Support Network both based on the in-degree and betweenness centralization measure. This finding indicates that seven users in the Support Network are those with a high number of ties, strategic positions; all at the same time.

The pattern found in this Supporting Network is a contrast to what was found in the Opposition Network and therefore in favor of the hypothesis that the key players in this network will show less variation or possess more similarity to one another.

Another crucial point is the fact that the prominent nodes only consist of thirteen unique accounts and therefore show the network is centered around a smaller number of people that are powerful both according to their number of ties and their position between other nodes; something that is not found in the Opposition Network

4.6 The Measure of Coordination of Message: Unique Words

Table 8. Unique Words Ratio

Total Unique Words 7116 6881

Total Observed Words 261,721 132,446

Ratio 0.02718926 0.0519155

(28)

As seen in Table 8, following various text processing and BTM, the total observed words are proportionate to network sizes. These total words are considered as meaningful corpus that has been through the series of data cleaning and pre- processing described in the method section. A similar proportion is not found in the case in regards to the unique world. As a result, the ratio found in the Opposition Network is smaller compared to the Support Network, almost by half. This finding is in contradiction to the theory and hypothesis that expect the opposite result due to the nature of participatory propaganda as being more controlled and organized.

4.7 The Measure of Context of Message: Variation of Topic Clusters and Topics Discussed

Table 9. Topics in the Opposition Network

Topic Theta

Value

Visualization of Top 20 Words Topic

Category Omnibus Law

is seen as a neoliberalist movement and protest against it.

0.08 The

problematic factor of omnibus law.

Police brutality during a protest in Jakarta Pusat (Center Jakarta)

0.141 Police

brutality and repressive regime.

(29)

Omnibus law is a

problematic law instrument with the problematic legislative process.

0.109 The

problematic factor of omnibus law.

Disagreement/

refusal on ratification of Omnibus Law on Job Creation.

0.186 Refusal on

ratification.

Repressive apparatus in Indonesia.

0.11 Police

(30)

Allyship for the

demonstration.

0.098 Demonstration

mobilization.

Consequences of the bill on workers’

rights.

0.159 Workers right

Authoritarian regime and police against civil society organizations.

0.116 Police

10. Topics in the Support Network

Topic Theta

Value

Visualization Topic

The Hashtags Rivalry behind the Controversial Bill: A comparative study on the Opposition and Support Movement of Omnibus Law Bill in Indonesia.