YouTube’s Algorithms and the Cartography of Controversy: Unboxing the algorithmic framing of the Middle East on YouTube

(1)

YouTube’s Algorithms and the

Cartography of Controversy:

Unboxing the algorithmic framing of

the Middle East on YouTube

Layal Boulos

Student ID: 12156183

MA New Media and Digital Culture

28 June 2019

Supervisor: Dr. Tim Highfield

Second Reader: Dr. Bernhard Rieder

(2)

4.1.2: Islam - Algorithmic Associations 36 4.1.3: Islam – Comment Space 40 4.2: Jamal Khashoggi - Algorithmic Associations 42 4.2.2: Jamal Khashoggi – Comment Space 45 4.3: Syria - Algorithmic Associations 47 4.3.2: Syria – Comment Space 51 4.4: Kuwait – Algorithmic Associations 53 4.4.2: Kuwait – Comment Space 57

4.5: To Sum Up 59

Chapter 5: Discussion

5.1: An Algorithmic Reflection of Reality 60 5.2: An Informative Algorithmic Engine 62

5.3: A News Algorithmic Engine 63

5.4: A Loop of Misinformation and Radicalization 65 5.5: The Ripple Effect of the Algorithms’ Inner Workings 68 Chapter 6: Conclusion and Further Research 70

References 73

(4)

ACKNOWLEDGMENTS

I would first like to thank my thesis supervisor, Dr. Tim Highfield, for constantly being there to answer my flood of questions and for pointing me in the direction I wanted to go but could not seem to grasp.

My immense gratitude goes to my parents for their continuous encouragement and support. I would not be here without them. I am equally thankful for my sister, Laura, who has constantly believed in me and pushed me to be the best version of myself.

I thank my other half for being my rock throughout this past year, for his graciousness and support. For being my source of joy and comfort.

To Victoria Andelsman, thank you for being my partner throughout this journey. For teaching me so much about the world of academia and research and for having faith in my abilities. Most importantly, thank you and Ketaki Chand for being wonderful friends that I will hold dear to my heart no matter the distance.

(5)

Abstract

This thesis examines how YouTube frames controversial social and/or political topics by mapping related videos and examining the comment space. More specifically, by looking into queries pertaining to the Middle East, an understudied issue space on YouTube, this study pursues the cartography of controversy by observing and describing social topics online. The analysis of four queries [Islam], [Jamal Khashoggi], [Syria] and [Kuwait] has led to the conclusion that YouTube frames topics as a reflection of ‘reality’ by presenting itself as an informative, news algorithmic engine. This thesis further adds to studies about YouTube’s misinformation claims and radicalizing potential, arguing that different queries induce diverse associations. Additionally, those that expressed a highly polarized space encouraged user-engagement, sparked political sentiment or comments of hostile nature. This study draws on these findings to reflect on the importance of examining ‘algorithmic associations’ within their social and political contexts, in order to put data into forms of meaning.

Key Words:

(6)

Chapter 1: Introduction

1.1 Setting the scene with YouTube

“Broadcast yourself”, the original slogan and purpose of YouTube as a platform when it launched in 2005. It was the initial stepping-stone that encouraged users to become more that just merely consumers of content but rather sparked user-generated content turning “video consumers into video producers” (Ding et al. 361). YouTube is one of the many web 2.0 technologies that replaced passive consumers with active, discerning, and vocal ones (Langlois 92). This has stimulated user-engagement and public involvement in social and political matters. Thus, the platform may have started as a medium that encouraged users to express themselves to the public, which we still see essence of with microcelebrities and vloggers; it has now become a platform where people share more than just content regarding their personal lives. Due to the various topic categories and ideological heterogeneity that it affords its users, as well as its involvement in amplifying public affairs, YouTube has become the second most visited website in the world making it a worthy platform of analysis (Alexa).

Statistics show that the majority of people turn to YouTube for entertainment and comedic purposes, either “to relax” or “to feel entertained”, thus the top most watched videos fall into the categories of comedy, music, entertainment/pop culture, and “how to” (“The Latest YouTube Stats”). Due to the aforementioned facts, users do not see YouTube as a platform that affords political and/or social debate or content. In fact, scholars have stated that due to YouTube’s commercial aspect of paying creators dividends from advertising videos, it encourages the creation of content that appeals to the majority of audiences, therefore orienting video content towards entertainment for the benefit of viewer-attraction; and consequently advertisers (Guo and Harlow; Guo and Lee; Burgess and Green; van Dijck, “Users like you?”). As a result, common perceptions of YouTube’s early years saw it as space for fun rather than political dialogues (Hess 427). However, as the platform has grown and as its position has changed overtime this has become more complicated.

Scholars have highlighted certain types of controversial-oriented videos that spark user-engagement and debate such as those that perpetuate stereotypes (Nakamura). Even though this could be interpreted negatively, scholars have found that it sparks racial discussions by allowing

(7)

YouTube to be a space to voice opinions on different ideologies (Guo and Harlow; Kopacz and Lawton). In fact, the number of adult users who consume news on YouTube in the United States alone has doubled between 2013 and 2018 (Smith et al.), further emphasizing on the shift of turning to YouTube for more than just entertainment. Others have pointed to YouTube’s “political potential” by referencing its live video streaming feature that allows users to be part of certain events (Arthurs et al. 6). Nonetheless, they have also pointed to the other side of the coin by referencing ISIS propaganda videos that have led the platform to pledge to identifying and removing such extremist content (Arthurs et al. 6; Gibbs). Moreover, one cannot fail to mention the fake news circulation and crisis that has infiltrated YouTube, embodying the essence of a political space (Arthurs et al. 6).

With YouTube as the platform of study, it is imperative to define the term engagement. Scholars have defined it as users “producing, consuming, or disseminating information” (Ksiazek 504; Deuze; Hargittai and Walejko; Jenkins; Napoli, Audience Evolution; Oh et al.). This paper acknowledges all three definitions because YouTube offers its users the capability of engaging in all of the stated forms, but also because my goal is to identify how YouTube frames controversial political and/or social topics via its algorithmic curation of related videos and its encouraged affordances. However, unlike previous common stances on algorithms, this paper aims to examine how these topics are framed by the ‘algorithmic associations’, which are made up of elements that drive YouTube’s narrative. Thus, without user’s engagement on the platform, whether via producing content, consuming it or sharing it, associations would not exist to begin with.

1.2 The Middle East and the Cartography of Controversy

Controversial topics range from political narratives like war photography to what some may say softer social issues like paparazzi and the end of privacy in the 1970’s. These stated controversies might stem from different issues; however, they have one thing in common which is that they are “situations where actors disagree”, as defined by Tommaso Venturini (261). Hence, when actors disagree, sides are taken forming clusters of associations based on a combination of actors’ interactions and the networks they form (Venturini 264). This stems from Latour’s work on tracing associations of the social within a controversy (23).

(8)

One of the biggest global controversial topics, which this thesis uses as a case study, is the Middle East. Even though several tensions and opinions occurred along the history of the Middle East, one can say that further controversy was sparked post 9/11. The staged terrorist attacks increased a fluctuation of attitudes towards Islam – a religion that stems from Middle Eastern heritage - one of discrimination and racism (Oswald). These types of historical events and political relations generates situations where people disagree, setting the context for a controversy. When people are being exposed to certain perceptions by the media – ethnicities in this case -, it primes them into thinking in a specific way (Gerbner and Gross; Shrum). It has been emphasized that much of what occurs in our lives and our beliefs are related to the fluctuating nature of our environment and the groups that belong in it (Tajfel). Thus, when people are being shown and reported specific images of a group in a certain light, categorization serves to “introduce simplicity and order where there is complexity” (Tajfel 82); and complexities are where tensions arise (Law and Hassard 12).

Practicing the complexities of tension is what Latours Actor-Network Theory (ANT) suggests (Venturini 258), however for the purpose of this thesis’ theme I aim to focus on the cartography of controversy which is “the exercise of crafting devices to observe and describe social debate especially, but not exclusively, around technoscientific issues” (Venturini 258). The reason I pose the Middle Eastern world as a controversial topic of dispute is because it checks many of Venturini’s list of what makes up a controversy. To begin with, controversies involve all kinds of actors, including non-human ones, and a large actor in this controversy is the politics that shapes the Middle East sparking debate and involvement of the ‘outside’ countries (261). Moreover, he states that “controversies are reduction-resistant”, hence human-actors cannot seem to agree on a question to be able to solve an issue (262). An example of this scenario encompasses headlines around the world explaining why the Middle East, “once a civilization that used to lead the world”, has become a “tragedy” in “ruins” in the midst of the Arab Spring (“The Tragedy of the Arabs”). Others downplayed the West’s foreign policy involvement in the unravelling of “chaos” within countries’ internal affairs (Aly). The aforementioned articles exemplify the Middle East’s association with controversy as they spark conflict (Venturini 262). An example that emphasizes conflict in the Middle East is when Donald Trump launched a military attack on Syria in 2018. Trump clearly saw that “social order” and “social hierarchy” were at stake, it was a matter of assertion and “distribution of power” (Venturini 262). Therefore, controversies, as Venturini

(9)

explicitly stated, often “involve force and violence” (262). The Middle Eastern world is where “the largest and most diverse” actors get involved, where friends and foe’s fluctuate “recklessly”, where everyone is “shouting and quarrelling”. Consequently, this makes the Middle East– or issues that stem from it- an object of the cartography of controversy (262).

1.3 Framing political and/or social issue spaces

As Rogers et al. explained, the web has opened multiple channels for “action, communication, and participation for the actors involved in a controversy” (29-30); this emphasizes the significance that these intermediaries (human and non-human) have in the “evolution of a debate, and the unfolding of an issue” (29-30). Thus, understanding and demarcating these political and/or social issue spaces is important in order to stress the fact that the web influences the way topics are studied and understood. With the unfolding of recent worldwide events, it has further highlighted that digital media platforms are indeed framing these narratives. However, before tackling these examples and the role of social media, it is relevant to define the use of ‘framing’ in this thesis. The notion of framing is used to stress the significance of a text by “diagnose[ing], evaluate[ing] and prescribe[ing]…” it (Entman 51-52). Simply put, the main purpose of framing theory is that the same issue may be interpreted in different ways and that “the way it is presented has an effect in how the audience thinks about it” (Andelsman and Mitchelstein 462; Chong and Druckman; Entman and Rojecki; Scheufele). This is important to highlight because one of the main points that this thesis aims to vocalize is the role that digital media platforms hold in framing the discourse on controversial topics.

Research that looks into the way social media frames controversial topics is of significance because these platforms are the ones driving narratives on issues, especially as they have been largely associated with real-life disasters. Mainstream, on the other hand, media appears to be catching up with them by reflecting and examining the position of platforms amidst these political and/or social issues. This is seen after global disasters struck not too long ago this year. One of these tragedies is the Christchurch attack that occurred on 19th March 2019, where an Islamophobic-white-supremacy act took place in a mosque resulting in the death of several Muslims amidst their prayers; it was all live-streamed online emphasizing a gaming culture style.

(10)

The stream was uploaded on 8chan, “a particularly lawless forum that often features racist and extremist posts”; it also linked to a manifesto expressing anti-immigrant and anti-Muslim ideologies and to a Facebook page that also streamed the attack (Marsh and Mulholland). CNN has called this “an internet driven hate speech” as the shooter told his followers to subscribe to a well-known YouTuber, PewDiePie, who has previously expressed right-wing ideas. CNN also questioned whether social media has become “a tool for terrorists” in which they get the chance to publicize their horrific acts. Therefore, as VICE put it, it is necessary for people to believe that “online spaces played a big role in his [the shooter’s] attempt to control the narrative surrounding the attack” (Haskins). On the other hand, VICE also stated that the “internet-centric narrative” of the attack misses the fact that the Christchurch attack is part of a much bigger Islamophobic violence that has manifested deeply within our world for decades (Haskins). The article quotes Professor Whitney Philips, saying that if the problem we had was rooted in the platforms then that would make things a little easier, however the “technologies themselves, they exacerbate existing tensions...” (Haskins). Philips’ quote explains this papers view, in which humans are the source of the issues but it is the non-human actors – as in platforms and their algorithms – that frame, stir or amplify the narrative that is presented on the web. Nonetheless, the human entities behind the decision-making processes cannot be disregarded; this will be addressed in chapter two.

More specifically, and what this thesis argues, is that the ‘algorithmic associations’ on YouTube are what frame the political and/or social issues. Another recent crisis, which highlights the latter statement, is that of Notre Dame’s fire which took place on 15th_{April 2019. YouTube}

provided a live stream of the fire in which it offered what they call “knowledge panels”, a banner that presents related information that linked to Encyclopaedia Britannica articles about the September 11 attacks (Paul). Linking a fire caused by accident with a terrorist attack creates other issues in relation to YouTube and misinformation, which will be further addressed in chapter two. A machine-learning expert explained that as long as we are using automated process to stir a narrative there is always a chance for errors to occur (Paul). The failure of the algorithm makes it even more imperative to be rather transparent with how it works (Paul); this is one of the reasons why studying the YouTube algorithm and its associations in relation to controversial-associated topics is of significance. While Notre Dame was in flames, another beloved and holy place in Jerusalem was also in flames: Al-Aqsa Mosque - considered to be the third holiest place in Islam (Solly). The interesting thing is that the former disaster received much more attention than the

(11)

latter (Solly). When you Google or YouTube search this disaster larger and more prominent news outlets do not appear in the top search in contrast to when you search for the Notre Dame Fire. This brings into question what is considered newsworthy to media outlets, but more importantly how that influences what the platforms present, referring back to the idea that associations frame narratives – in this case via channels, videos and users.

1.4 Research question and relevance

Platforms are important because they have the capability of intervening in public discourse as some “content is chosen, some is given prominence, some is discarded, and some is expelled”, ultimately platforms pick and choose (Gillespie, “Platforms Intervene” 1-2). Therefore, understanding how platforms “shape the social dynamics” via their design (Gillespie, “Platforms Intervene” 2), which this thesis argues is visible via the ‘algorithmic associations’ of actors, allows us to assemble and trace the way they frame discourse that takes part in public culture. This thesis is largely influenced by Latour’s work on associations, and Neyland and Möllers’ work that points to the influence that stems from ‘algorithmic associations.’Furthermore, it is inspired by Rieder et al.’s digital methods approach and conclusions on YouTube’s algorithms - all of which will be explained further in chapter two. While scholars have examined the consequences of algorithms and the power or impact they hold in shaping society, there is a dearth of research that seeks to understand algorithms alternatively. I seek to fill this gap by focusing less on how YouTube’s algorithm works and more on what the ‘algorithmic associations’ formed between actors - channels and videos – entail. I am not suggesting that consequences do not arise because of algorithms, instead I am proposing that the assemblages that are made visible by these associations is what shapes discourse. Thus, by combining a background in framing theory within the context of new media, the aim is to accentuate on what is outlined by platforms and how people respond to certain topics.

Moreover, based on statistics discussed earlier, YouTube is largely considered an entertainment-oriented platform, hence it is interesting to examine how these actors are framing the discourse on controversial topics, in order to further illustrate the shift to political processes. Broadly speaking, my paper seeks to answer what is the role of digital media platforms in framing

(12)

controversial political and/or social topics. More specifically, what is the discourse framed because of the ‘algorithmic associations’ of YouTube’s related videos? In addition, what is the discourse framed because of YouTube’s affordances via the comment space?

Another aspect of this thesis’ relevance to the development of the scientific field stems from the examination of an under-studied group on YouTube, which is the Middle East. The aim is to take existing research further by adding and expanding on previous studies that drew from Latour or Venturini’s work on associations and the cartography of controversy to understand the impacts of algorithmic curation (Rieder et al.; Bucher; Neyland and Möllers). To tackle the aforementioned questions, the structure of this thesis is presented in the following ways: studying YouTube as a platform by exploring its algorithm, examining YouTube’s affordances, specifically the comment space and how it encourages users to interact, and finally looking at YouTube’s culture by examining the spread and role of creator culture and misinformation within its space. In chapter three, an empirical methodology that uses digital methods’ tools to trace the related videos of four Middle-Eastern-specific queries along with a content analysis of the videos, channels and comment space is presented. Following the methodological approach, the results are demonstrated, ending with a discussion and conclusion of the findings, as well as possible points for further research.

Ultimately, this thesis’ most significant conclusions is, firstly, that the framing of each issue space varied according to each queries’ current political or social state. This emphasizes on the fact that YouTube’s space acts as a reflection of world events by presenting itself as an informative, news algorithmic engine. Secondly, more polarized issue spaces are more likely to spark political sentiment or hostility in the comment section. Finally, although not prominent, essence of misinformation and radicalization are present.

(13)

Chapter 2: Literature Review and Theoretical Framework

2.1 YouTube as a platform

Algorithms are elements within a platform; therefore, it is best to take a step back to understand how platforms work and what function they serve. Platforms host algorithmic processes in order to control and manage the flow of information, “contents, ideas, and sociability” (Rieder et al. 51; Gillespie, “the politics of platforms”). They are made by the distribution of users and interfaces; it is those interactions and exchanges that are capitalized (Bratton 42; Bucher 6). In regards to YouTube, it is what the platform affords, via likes and view counts that is capitalized. Based on that, advertisers come into play; making the platform a multi-sided market that operates by bringing together at least two types of end-users (Rieder and Sire 199). This is important to consider especially when taking into account the policy and business model that YouTube works under as it affects the way they may intervene in debates (Gillespie, “Platforms Intervene” 2). As stated by Gillespie, this is largely due to the idea that they “pick and choose” what and how they present their content, uphold their responsibilities and respond to criticism. This is based on both clear and ambiguous cultural standards “at the behest of offended users or concerned lawmakers, and in ways that best suit their economic aims”; usually by putting advertisers’ interests first (“Platforms Intervene” 2). Arguably, this led to what has become known as “Adpocalypses”, which refers to the demonetization of creator content; not being able to convert their views into revenue (Lewis, “YouTube's Bungled PR Announcements”). Nonetheless, YouTube continues to adopt regular changes within their algorithms to combat their continuous ruptured reputation, which will be further discussed below (Roose).

2.1.2: Algorithms

Algorithms have been labelled as “black box[es]”; we are aware of what they are but have no idea how they work or what exact purpose they serve, they lack transparency (Beer , “Management in Cybernetics Terms” 113; Hargittai 770). As a result, many regular internet users do not realize that we are entwined with algorithms on a daily basis as they have infiltrated the internet space influencing our choices (Rieder et al.). Most of these choices occur based on what the platform presents to us in terms of content, setting the framing of issue space discourse. However, in order

(14)

to understand how this discourse is framed on platforms, especially one like YouTube which is formed on the basis of participatory culture, one must not only study how users navigate through it according to the affordances of the platform but also pay attention to the “networked conditions that enable it” (Langlois 91). Those networked conditions involve the way the platform navigates its users through content, such as via recommended, related or suggested videos; allowing certain associations to form. This makes it possible to understand the information converted into cultural signs that outline users’ opinions and activities (Langlois 91).

Because YouTube is a complex and networked system, it functions based on a collection of algorithms (Bucher 47). Therefore, YouTube makes a distinction between three types of algorithmic videos. Firstly, the suggested videos are the list on the right side of the YouTube interface when a user is watching a video, there are about 19 videos in that column. Secondly, related videos are a subdivision of the suggested videos and are usually the same for everyone viewing that video and in the same order. They make up 16 out of the 19 suggested videos, however if the person is not logged in they can make up to 18 out of the 19 videos. Finally, the recommended videos are based on personalized options; they have a “recommended for you” label on the bottom. Hence, if the viewer is logged in it correlates with a higher percentage of the videos on the right side of the column to be a part of this category (Golden).

The function of algorithms has been a continuous area of examination. For example, Machill et al. claimed that the returned results of a search engine such as the way they are ordered or how genuinely complete they are, remains unclear to the users (1). Other scholars such as Rieder et al. have built on those claims by examining the YouTube search algorithm in order to understand how it works but more importantly its possible impact. They conducted a study focused on YouTube’s search results by querying controversial sociocultural issues, with specific concentration on how they are ranked over a period of time. Two of their queries, [Syria] and [Islam], were adopted into this thesis considering their relevance in terms of pertaining to the polarization of the Middle East. They found that subcultures, such as war photography, within YouTube were very likely to dominate a query as broad as [Syria] and that the [Islam]’s issue space contained antagonistic voices but there were also a domination of Islamic religious channels during stable news times (60). Ultimately, their study found that the YouTube algorithm gives actors that are native to the platform dominance over others and the more “newsy” or prominent the video is, the more likely it is to appear at the top of the search (64). Such type of research points

(15)

to the possibility of ‘power’ clusters that arise by the algorithm and seeing what it gives prominence to emphasizes that it prioritizes certain channels over others. This means that there is a higher chance for exposure of certain types of videos, based on what the algorithm deems ‘worthy’ forming its own narrative. A similar line of research further emphasized the latter point by concluding that the YouTube algorithm favours feminized hegemonic content in order to create focused markets (Bishop).

Algorithms are sculpted on visions of the social world, their outcomes are influenced by commercial or other schemas (Beer, “The social power of algorithms” 4; Williamson 5). This is why scholars have stated that algorithms should not be studied outside of their social contexts, as their “existence and design are a product of social forces” (Beer, “The social power of algorithms” 4; Neyland and Möllers). Bucher also emphasized on being aware of those who manipulate the outcomes of algorithms according to their interests (25; van Dijk, “The culture of connectivity” 12). Thus, it becomes important to question and examine the narrative that is being pushed and presented to users, especially when it is capable of causing “real-world consequences” (Rieder et al. 53). An example of these consequences, as termed by Matamoros-Fernández, could be ‘platformed racism’ where a platform’s technicalities, affordances and algorithms displays racist discourse and encourages the spread of hate speech (933). Similarly, a study found that Google’s search algorithm can produce racist, sexist or homophobic results, perpetuating negative stereotypes in our society (Baker & Potts; Noble). Another example is the conspiracy theory known as “Pizzagate”, a right wing fake news story claiming that Hilary Clinton was involved in turning a restaurant into a child sex ring. This is an example of how a fake story presented on YouTube is capable of “motivat[ing] a gunman to fire a weapon inside the restaurant” (Wong and Levin), further stressing its ability to cause dangerous real-world consequences especially when not taking into account passive or undiscerning consumers who take what is presented at face value.

According to Napoli recommendation systems, such as YouTube’s, serve as a “digital best-selling desk”; meaning that the most popular videos watched - like those that fall into the entertainment category - might conceal the “niche content” that emphasize alternative options (Audience Evolution; Guo and Harlow 285). Similarly, Guo and Harlow stress that the “bandwagon effect” makes popular content even more popular (285). Hence, while YouTube is a place that encourages the expression of diverse opinions, some of them could be diffused by the

(16)

louder and more prominent ones. Though they may exist, they are harder to find, allowing the more popular ones to monopolize the platform (Gillespie, “The politics of ‘platforms’” 359). This highlights the possibility of YouTube making decisions on what information to include or exclude (Guo and Harlow 285; Bucher 34; Bozdag; Helberger et al.; Napoli “Social media and the public interest”). This is relevant to the purposes of this thesis because I aim to uncover the framing of political controversial queries, which are potentially considered ‘less popular’ content. Thus, it is interesting to examine what type of framing is encouraged and what is considered ‘niche’ content for each issue space by examining the associations formed via the top related videos.

2.1.3: YouTube’s radicalization

Several studies have examined YouTube’s radicalism in regards to its recommendation algorithm (Madrigal; Friedersdorf; Lewis; Nicas; Kaiser & Rauchfleisch) - sometimes used interchangeably with the related algorithm (Golden). A New York Times article called YouTube “the great radicalizer.” The researcher started watching Donald Trump videos and noticed the YouTube algorithm recommending and auto playing right wing, supremacist videos (Tufekci). When starting with videos on Hilary Clinton and Berny Sanders the algorithm recommended conspiracy theories. This pattern of extremism was also prevalent with non-political content such as when starting with a video about vegetarianism she was directed to veganism. Tufekci explains that this is largely because Google owns YouTube, so their system is related to Google’s artificial intelligence; which functions by selling our attention to companies that pay for it. The longer they keep us on the platform, the more they benefit as it leads to more ads (Tufekci; Roose).

Therefore, YouTube’s algorithm is found to be systematically strengthening the voices of videos that are “divisive, sensational and conspiratorial” (Lewis, “Fiction is Outperforming Reality”). YouTube’s response to these findings was that “its algorithm was a neutral mirror of the desires of the people who use it – if we don’t like what it does, we have ourselves to blame” (Lewis, “Fiction is Outperforming Reality”). The Guardian challenged YouTube’s response saying that we might click on disturbing content “instinctively” and questioned if “those in-the-moment impulses” are really a reflection of the content we want to consume (Lewis, “Fiction is Outperforming Reality”). Albeit that may be true, the algorithm does not know if the videos we are clicking on are a result of rapid instinct or active content consumption. Thus, it could certainly be said that YouTube’s algorithm is not creating something that does not already exist via our own

(17)

decisions; whether passive or active. As explained by Kaiser and Rauchfleisch, these channels, actors and videos not only exist but also overlap and interact; YouTube’s algorithm simply makes these networks visible to us (“Unite the Right?”). On the other hand, there are studies that contradict these statements believing that such Big Data drive our digital decision-making processes by hyper-nudging us to click on preferred content created by the choice architect, the platform itself (Yeung 121). Although that might be true in the case of YouTube pushing commercial content for revenue, we should still not conclude that a social force is what is characterizing interaction; particularly without “associations that explain the social” (Latour 238). Similarly, Bucher explained that the way algorithms work is “what we get is what we did and that is what we see” (2). If we frame algorithms on the basis of having some form of power it jeopardies ignoring the human decision-making processes that are involved (Bucher 34), whether that was the curator or the user.

Another impact of algorithms that is seen as part of this wider discourse is the formation of “filter bubbles” and “echo chambers.” Filter bubbles are the personalization of content, and echo chambers is the exposure to content of compatible opinions (Pariser; Sunstein). These are consequences of user choices; although “filter bubbles” and “echo chambers” are not the subject of this thesis it is relevant to emphasize on to highlight on the attention surrounding algorithms and their possible influences, but also to stress the role of users in algorithmic content curation.

2.1.4: Mapping ‘algorithmic associations’

Despite the fact that the majority of previous research focuses on how algorithms are the culprits that perpetuate powerful acts on our society, I argue that the impact is caused via their associations, which are making connections visible. Neyland and Möllers’ paper on the conditions and consequences of algorithmic power argues that power stems from ‘algorithmic associations’, which is the assemblage of “people, things, resources and other entities held together by practice and process” (46). They cite the work of Latour on associations, which are methods of interactions in which things take a “social shape” (47). According to Latour we should not conclude that a social force - power derived by algorithms in this case - characterizes interaction (238). Instead, power has to be “produced, made up and composed” (Latour 64), hence to understand the impact that algorithms have one has to explore their associations. Which is why viewing algorithms as black boxes suggests that they are obscuring networks and assemblages that they are composed of

(18)

(Bucher 50; Latour 183), exploring their associations thus allows for the unboxing of the black box. Neyland and Möllers’ paper examines two algorithmic surveillance systems to inspect how they are developed and what consequences that has. By examining the latter, they traced associations to understand the conditions and consequences of algorithms (46). Even though I chose to align myself with the general idea of their paper – ‘power’ stems from algorithmic associations - they focus on tracing associations to see what makes up algorithms; questioning dynamics of power in the process. My thesis, on the other hand, focuses on associations formed by algorithms, which frames discourse. This is encouraged by Bucher as she explained that “the notion of algorithmic power may not even be about the algorithm”, her book adopts a different stance on algorithms by examining the power they hold through the kinds of “encounters and orientations” these algorithmic systems are composed of (3). Therefore, instead of pinpointing where the power comes from, what should be examined are “the algorithms in practice, the places and situations through which algorithms are made present…” (Bucher 34), which is the premise that this thesis concerns itself with as they are made present via associations. This is stressed by the term “programmed society”, which emphasizes Latour’s work on actors’ interaction that symbolise the social world (Bucher 4). This point is further highlighted by “ontological politics”, which explains that “realities are never given”, instead they are shaped and made visible via interaction (Bucher 56). I draw upon Bucher's argument that these associations, which are construed, create the circumstances through which data are put into forms of meaningfulness, which is shaping what the user sees (82). However, while in her work Bucher examines what causes visibility on Facebook, in this thesis I am concerned with the narrative of visibility that is produced by YouTube’s ‘algorithmic associations.’

Therefore, the previously mentioned real-world consequences that algorithms have would be made evident if mapped and traced. This thesis adopts Latour’s concepts about mapping associations by taking a form of social cartography approach. As a participatory platform, YouTube is a space that does not simply encourage actors to contribute but their contribution is what brands the platform. Consequently, YouTube can be seen as a lens through which society is represented; where actors and information connect leaving traces of associations and assemblages (Rogers et al. 16), which are made visible via the related algorithm.

(19)

2.2 YouTube’s affordances

In October 2006, Google purchased YouTube, gaining a foothold in the world of video and user-generated online content instigating a form of dominance due to the combination of “search engine, blogging tools and interactive online spaces” (Gillespie, “The Politics of ‘Platforms” 347). These forms of engagement are now the key to “cultural discussion” that navigates the internet (Gillespie, “The Politics of ‘Platforms” 348). Which is why we should be questioning how discussions are framed, not just by the users but also by what the platform affords to the users and how that leads to the formation of ‘algorithmic associations’; pointing to the framing of discourse. YouTube ‘encourages’ user-interaction and engagement via the like/dislike or subscription button. Similar to Davis and Chouinard’s interpretation of Facebook and Instagram’s like button, the buttons on YouTube’s platform also create network interaction, giving regular feedback to content creators (243). For the purpose of this study, I define affordances as understood and expanded by Evans et al., which is the variable that connects and relates technology to specific interactive outcomes (Evans et al. 36; Faraj & Azad). This is because to understand how YouTube frames issues it is necessary to explore the relationship that is formed between those using the platforms’ technology and how it is being used based on what the platform presents.

Subsequently, where interaction takes place, associational assemblages are formed. Interaction is initiated through what the platform affords in terms of user-engagement. For example, affordances such as the like/dislike button, the comment section, and the view counts are all measures of video engagement that emphasize popularity (Burgess and Green); which is why other scholars have referred to them as “popularity metrics” (Chatzopoulou et al.). Moreover, Neff et al. highlighted the significance of such metrics as they are used to “track attention and visibility” (93). As a result, the videos that have such responsiveness, in terms of audience engagement, attract and gain the highest prominence. This study chose to focus on the comment space, as it is interesting to study more than just how platforms frame issues but also how users respond. Specifically, via the ability to give context to their opinions through the comments; thus allowing the examination of the general framing of the comment space alongside the ‘algorithmic associations.’

(20)

2.2.2 The comment space

Scholars have studied user interaction via YouTube’s affordances, specifically by examining how users respond to certain videos in the comment space (Kopacz and Lawton; Thelwall et al.; Ksiazek et al.; Murthy and Sharma). Kopacz and Lawton, for instance, have analysed YouTube video comments in response to the presence of Native Americans (NAs) by selecting queries based on labels commonly used to describe NAs. They quantitatively coded comments based on the tone toward videos, tone toward NAs, and comments’ affirmation of racial discrimination (22-25). They found that the comment space initiates debate by allowing users to discuss racial discrimination and they point to YouTube as a medium that has the capability of changing established social perceptions via user-engagement (34). Similar studies were adopted that examined the comment space of other group related videos, such as Asians and African Americans (Guo and Harlow; Guo and Lee).

Moreover, others found that a controversial music band that posts videos with “subtle” racist songs on YouTube creates a network of racialized hostility; nevertheless, there were instances where people actually engaged in political and/or social debate by having “meaningful” discussions about racial issues (Murthy and Sharma 209). They based their corpus on the five most viewed videos and then used grounded theory to develop a scheme that they quantitatively analysed to create a visualization of the comment space (196-198). Ksiazek et al., on the other hand, examined user-engagement with online news, by focusing specifically on the News labelled category ranked by views on the basis of seven days.They quantitatively studied the correlation between popularity and interactive engagement metrics, as well as between these metrics and different types of news content (508-510), finding a strong correlation between popularity and commenting (514).

As highlighted by previous research, being able to assess user responses via the comment space has led to conclusions such as controversial topics receiving high user-engagement and popularity; this makes unexplored controversial topics on YouTube a worthy subject of study. However, scholars have emphasized the harm that can be caused because of these affordances, specifically because they split the online persona from the physical offline body making it highly more possible to cause misunderstanding and a lack of ethical communication and action (Milner and Philips).

(21)

2.3: YouTube’s culture

2.3.1: Creator culture

One of the major actors that take part in the formation of associations on YouTube are microcelebrities, “those who regularly post videos on their own YouTube channel” (Rogers, “DMI” 1). This new form of celebrity that was born via the shift towards user-generated technologies embody an essence of unique authenticity that scholars have studied (Whitaker). That authenticity is built on microcelebrities being their ‘genuine self’, which leads to a formation of a bond of friendship between the YouTuber and the viewer; one where the viewer is affected by certain points of views (Westenberg; Nouri; Whitaker; Perez-Torres et al.). Although creators upload diverse content, a substantial number of videos on YouTube can be found in the entertainment section where, for example, humour often gets used as means for expression on discussions surrounding political and/or social issues; as it has been used for decades on different traditional media (Hall et al.). YouTube also became a space where native political content is distributed, whether it was news, or YouTubers commenting and debunking certain topics. For instance, they host a “dominant native political community: the YouTube right” who have become commentators that challenge and mock the “mainstream media” (Herrman; Boulos et al.).

Previous scholars have pointed to elements that contribute to the persuasiveness of media content, such as the formation of parasocial interaction, specifically via the distribution of news by traditional radio commentators (Moyer-Gusé et al.). This is largely due to some commentators being perceived as “friends” who are “witty, informed, engaged...” (Polletta and Callahan 63; Norton). Thus, YouTubers have augmented these traditional commentators, as they have become commentators on political content themselves, just like social media platforms have augmented traditional media reporting. Parasocial interaction raises concerns about the extent of audience influence, as it leads to higher persuasion levels (Moyer-Gusé et al.), increasing the potentiality of the diffusion of subjective news according to certain opinions. Consequently, this forms the aforementioned ‘algorithmic associations’ because assemblages are being formed and clustered based on the formation of collective identities. The spread of ideologies is used as a motivating factor to upload political content on YouTube, along with the general monetization of videos (Zhang and Zubcsek). These two aspects are important to consider, specifically in the misinformation crisis that digital media finds itself in today, as the “two main motivations [that]

(22)

underlie the production of fake news” are “financial and ideological” (Edson et al. 138; Boulos et al.).

2.3.1: YouTube and fake news

Despite the fact that Facebook and Twitter have been the centre of most of the academic debates on fake news, following the 2016 elections (Shepard), YouTube is a relevant platform that offers a space convulsed with spider webs of conspiracy theories and polarized information, creating its own storyline in regards to prominent issues (Debczak). This makes seeing ‘algorithmic associations’ that are manifested surrounding controversial debates ever-more interesting. One may say that the misinformation crisis amplified largely because of social media platforms. This is mainly because the online world affords user-generated content and thus “allows for citizen participation in news reporting”, giving users’ the ability to disseminate news whether fake or real (Haigh et al. 2063; Boulos et al.).

A typology of fake news, which is important to consider due to the supposed manifestation of entertainment on YouTube, is the use of humour to distribute information or news updates (Edson et al. 141). The promotion of entertainment becomes the main element rather than the dissemination of information; such components are seen when news is distributed satirically or via parodies (Edson et al. 141-143). This is illustrated on shows such as the Daily Show on Comedy Central, which encompasses a relatively high subscriber count on their YouTube channel of 4.5 million. An example of a misinformation scandal on YouTube is the increase of ‘Flat Earthers’, which has been found to be a result of videos watched because of the recommendation algorithm (Sample).

With the spread of the misinformation crisis across platforms, YouTube aims to use its algorithms to recommend reliable sources (Robertson). One of YouTube’s solutions to the fake news crisis involves including links to articles in order to support credible video journalism with the aim of making it easier for its audiences to detect the real news from fictitious ones (Debczak). The news articles that they aim to link are described as “authoritative”, which are channels that YouTube has authenticated as part of their algorithms (Castillo). However, this article was published in July 2018, and since then one could argue about the extent of fake news that remains infiltrated within the platform especially when taking into account recent scandals. An example is

(23)

the previously mentioned Notre Dame Fire videos on YouTube as they linked to 9/11 articles, which is also seen as part of the misinformation crisis because it links a fire that allegedly occurred by accident to a terrorist attack (Paul). However, according to YouTube the reason they have created the idea of involving third parties like Encyclopaedia Britannica and Wikipedia is to help contain misinformation (Paul).

Consequently, questions arise regarding their machine learning detection systems in which they relate to unrelated problematic content, further adding to the misinformation crisis. Other questions include which channels does the YouTube algorithm consider “authoritative” and according to whom? What sort of narrative does this present to the users? What kind of associations are formed as a result? How does this frame issues on such controversial political and/or social topics? Such questions are aimed to be explored by this thesis. The previous chapters have explained positioning the Middle East amidst the cartography of controversy and stressed on examining how YouTube works as a platform and what it affords to its users. This was done to distinguish the relationship between user-action and platform presentation. Additionally, including YouTube’s creator culture and its misinformation crisis was imperative to situate my study within its real-time context, especially as social media platforms have been associated with recent tragedies. This pointed to the importance of examining the framing of issues produced by YouTube’s ‘algorithmic associations’. Hence, in the next chapter, the methodological approach will be outlined based on mapping the aforementioned ‘algorithmic associations’ along with an analysis of the comment space.

(24)

Chapter 3: Methodology

3.1 Data collection

In order to analyse what is the discourse framed because of the ‘algorithmic associations’ of

YouTube’s related videos, my methodology revolved around mapping and analysing algorithmic

networks of related videos in regards to four manually selected Arab-related queries – [Syria], [Kuwait], [Jamal Khashoggi] and [Islam]. As exhibited in previous sections, many of the former studies focused their methods on quantitative approaches but this thesis adopted a qualitative approach to add to the diversification of the field.

To retrieve the data, I used two different modules from the YouTube Data Tools: the ‘Video List’ and ‘Video Network’ module available via the Digital Method Initiative (DMI) (Rieder). The parameters for both modules were set to ‘relevance’ as it is the default setting for users on the interface (Rieder et al. 54). However, as Rieder et al. stated, the tool neglects elements of personalization and localization; thus reproducing the exact same results for end-users is a skewed process (54). They further emphasized on the hardships for researchers to create methodological stability (54), but in the case of this paper as in Rieder et al.’s, I acknowledge that personalization and localization are imperative to YouTube’s inner-workings “but take the API data to be as close to a ‘baseline’ perspective as one can get” (54). The stated process was repeated three times within the span of three weeks: 8 April 2019, 15 April and 21 April. Although this paper does not concern itself with ranking, which is why I only used the first extracted dataset for the purpose of the analysis, it is important to show acknowledgement to the way the algorithm functions as mentioned in previous sections (Rieder et al.).

3.2 Query assemblage

The four manually selected queries used for the purposes of this thesis are [Islam], [Jamal Khashoggi], [Syria], and [Kuwait]. This selection was made to focus on controversial sociocultural issues that stem from the Middle East. As emphasized by Rieder et al., search results are connected to such controversial and polarized debates as they take part in defining how issues are shaped and understood (54). Even though some of the chosen queries do not pertain to particularly active

(25)

controversies, the main theme is that they pertain to the controversy of the Middle East as a whole, which is certainly an active issue where actors disagree as stated in chapter one.

[Islam] was chosen as one of the four queries because it is an ongoing polarized issue that is largely associated with the Arab world, despite the fact that it is a misconception, when one thinks of Islam they think of Arabs and vice versa. According to the Pew Research Center, in 2015 and estimated numbers in 2060 show that the top five countries with the highest percentage of Muslims do not reside in the Middle East but rather in Indonesia, India, Pakistan, Bangladesh and Nigeria (Diamant). Thus, this misconception binding Islam and the Arab world makes it an even more relevant controversial issue space to examine. Furthermore, people of Arab nationalities are commonly found to be involved in ideologies such as Islamic fundamentalism, political Islam or terrorist groups such as Al-Qaeda or ISIS and their specific relation in tragic events such as 9/11. Edward Said, one of the first academic writers to emphasize on the Wests misconceptions of the Arab/Muslim world (Khawaja), highlighted on the notion that Islam means different things to many different people - including to those that practice the faith (Said). Therefore, the way it is defined, demarcated and understood by people is likely to lead to situations where actors disagree. [Jamal Khashoggi], was chosen as one of the queries because a lot of debate stemmed from this controversy, causing a “diplomatic crisis” between several countries. Khashoggi was a journalist at the Washington Post with Saudi Arabian heritage; nonetheless, he criticized the ways of Crown Prince Mohammed bin Salman of Saudi, also known as MBS. On 2 October 2018, he visited the Saudi consulate in Turkey and was later declared murdered. The controversy began with Saudi Arabia denying their involvement, while Turkey saying it was a planned assassination that involved torture. Eventually Saudi Arabia stated that it is was a rogue operation. Even though Trump has criticized the cover-up, he still stands firm and defends the alleged crown princes’ involvement in the murder. While other European countries such as Germany, Finland and Denmark, were amongst the countries that have cancelled arms deals with Saudi Arabia since the killing (“The Jamal Khashoggi”) - further emphasizing on international differences.

In addition to choosing two sociocultural issues as queries, one active and the other dormant, I also included two countries in order to analyse diverse issue spaces. One of the countries, [Syria], was chosen due to the extended conflict beginning in 2011. Inspired by the Arab Spring, the conflict in Syria initially began with a series of uprisings against President Bashar Al

(26)

Assad. However, the situation became far more complex as several groups, countries and actors involved themselves. Religious conflicts arose between the Sunni Muslim majority against the Shia Alawite’s sect, bringing into play terrorists and other groups such as ISIS and the Kurds who have long been fighting for the right to self-govern themselves. With several countries taking different sides, the world split into pro-Assad’s Syria and against-Assad’s Syria; having two of the usual big-power players - Russia and the United States - on different sides of the spectrum, thus causing further polarization (“Why is there war in Syria”).

[Kuwait], on the other hand, was chosen as a query that does not particularly fit into the controversial elements of this study making it a perfect point of comparison between the other internationally polarized terms. Despite the crisis of 1990-1991 where Saddam Hussein invaded the country (Tisdall), Kuwait has had relatively stable politics, with minor parliamentary conflicts. The country is known to have a high number of expatriates making up to 70% of its population with the largest group leading are Indians (“Kuwait Population”). The most recent international conflict was one that involved the murder of a Filipina domestic worker causing tensions between the two countries (Heydarian). Despite the aforementioned diplomatic crisis, Kuwait stands out in contrast to several other regions in the Arab world where international involvement has occurred - or is occurring - making it a relatively low area of coverage in the Wests news (“Kuwait’s Profile”).

3.3 Video List module procedure

This module creates a list of video information and statistics from the platform’s API endpoint (Riedeer). I used the ‘search query’ option with iteration one – which retrieves 50 videos in order to make a list. The list acquired in CSV file consists of 20 different statistics related to each video. This includes its ranking within the YouTube interface, the channel and video ID, the channel and video title, the date published, description, category label and other affordance-related emphasis like view count, like/dislike and comment count. The corpus consisted of the first 20 video IDs that were inserted into the ‘video network’ module. Delimiting my corpus to 20 videos was a choice made to mirror the same amount of videos within a standard YouTube page (Rieder et al. 54).

(27)

3.3.2 Video network module procedure

This module creates a network of relations between related videos, “starting from a search or a list of video ids” (Rieder). It is retrieved from the “API endpoint and creates a graph file in GDF format” (Rieder). I used the ‘seed’ option as a starting point and crawl depth one in order to retrieve more than just the relations between the inputted seeds. Each data sets’ first 20 video IDs were inserted in order to retrieve the related video network for each query. The video network for [Islam] retrieved 1,588 related videos, whereas [Syria]’s data set was composed of 1,384 videos, [Jamal Khashoggi] had 1,149 videos and [Kuwait] had 1,306 videos. Nonetheless, the corpus was further delimited for each query as will be explained below.

A network analysis would not be complete without the examination of actors that leave a trail of connections (Rogers et al. 16). Therefore, the channel titles were also analysed in order to conduct a group formation analysis, studying not only related video networks but also the actors involved. It is because of the visual communication and associations between actors that a network is manifested within the social environment of YouTube. As stated by Latour and emphasized by Rogers et al., the social is composed of the reconfiguration of actors continuously re-associating and reassembling (Rogers et al. 42; Latour 7). Consequently, looking at the channel titles highlights the dominant actors that are amplifying and broadcasting certain “issue spaces”, which are defined as groups of actors engaged in the same issue area (Rogers, Digital Methods 39).

3.3.3 Gephi visualizations

Rogers et al. have stated the effectiveness in combining mapping practices with digital methods as they permit researchers to visualize digital data by positioning issue networks in techniques capable of telling stories (29). This allows for making conclusions on a medium and societal research level (Rogers, Digital Methods). Hence, each dataset has been visualized by gephi, which is a tool that allows the “study [of] social networks…” (van der Vlist & Helmond 23). The node sizes were arranged based on their indegree values, which are the total of inbound edges of a node (van der Vlist & Helmond 51). This was chosen to represent the amount of times a video is connected to another, thus demonstrating related videos and operationalizing ‘algorithmic associations’.

(28)

The layout used for each of the four visualizations was ForceAtlas 2, a “continuous force-directed layout algorithm” which is used to make communities appear as groups of nodes. It allowed for clear visualizations as ForceAtlas 2 is suitable for the qualitative analysis of graphs (van der Vlist & Helmond 48). To further enhance their aesthetics and readability, I tuned the settings for ‘Scaling’ to five, as it is responsible for the size of the network, and changed the ‘Gravity’ to 0.01 to adjust the strength of the attraction among the nodes (van der Vlist & Helmond 48). Moreover, I continued to refine the graphs by using the ‘Prevent Overlap’ option, which adds an insulating layer to the nodes ensuring the lack of overlapping (van der Vlist & Helmond 48). The ‘video themes’ were set as the attribute for the ‘partition’, which emphasizes the colour of the nodes. The indegree values were set as the attribute for the size of the node with a minimum size of ten and a maximum size of 100 to ensure visible differences within the graphs. The text on each node represents the channel theme categorizations.

3.4 Video and channel content analysis

To examine what is the discourse of the ‘algorithmic associations’, I looked at the framing of the subject of English and Arabic videos by content analysing them. This was done by selecting the top 200-indegree values within my corpus from each query as created by gephi, since the indegree values represent the number of related videos. Even though 200 is not the total number of videos within each network, since I focused on the most connected or related videos, the corpus chosen should fulfil the goal of understanding the discourse created because of the ‘algorithmic associations’. Specifically, as the aim of this study is to analyse the general discourse framed in each issue space.

I proceeded to adopt a deductive thematic approach because the study is driven by my interest in the area; therefore, I created categorizations based on the stated research questions and the content of each set of videos per query (Braun and Clarke 84). Additionally, the ‘level’ at which themes are identified were adopted via a semantic approach as patterns are “identified within the explicit or surface meanings” (84). The categories were established by highlighting the general idea of the videos within each issue space separately, as the distribution of the narratives differed per query. The same process was implemented for the analysis of actors by identifying the general

(29)

commitment of each channel. This was derived by looking at the channel descriptions, as well as by looking over the general videos posted to confirm that the description matches the uploads.

In regards to [Syria], the video variables were ISIS-Related, Reporting on the Syrian Crisis, Unrelated Conflict, Unrelated and News. The channel variables were News, Native News, Informative and Infotainment (Table 1). [Kuwait]’s variables included Footage, Facts, Life in Kuwait, Celebrating, Advertising, News and Unrelated. The channel categorizations were Company, Informative, Entertainment, Infotainment, News, Vlogger and Travel & Tourism (Table 2). [Jamal Khashoggi]’s variables on the other hand were News, Related to Mohammed Bin Salman and Saudi Arabian relations, Pre-crime, Reporting on murder and Unrelated. The channels were composed of News, Native News, Informative, Entertainment, and Comedic News (Table 3). The video variables for the final query [Islam] were Anti-Islam, Debating Islam, Islamic Education, Preaching Islam, Informative, Human Interest, Public Views, News and Unrelated. The channel variables were Religious, Informative, Entertainment, Infotainment, Atheist Content, Anti-Islam and News (Table 4).

Some videos and channels presented an overlap between the categories causing a mix of narratives; however, they were categorized according to the majority of the main purpose of each content. Several videos were not in English nor Arabic, causing a methodological limitation; nonetheless, they were either categorized using the subtitle captions available on the video or by understanding the purpose of the video via the title and description offered by the uploader. To further understand and dissect the findings I created an Alluvial Diagram using RAWGraphs, a visualization software that allows researchers to work with their data both visually and analytically (van der Vlist & Helmond 65). The Alluvial Diagrams serve the purpose of cross-referencing and seeing correlations between the two “categorical dimensions” of the video and the channel themes (van der Vlist & Helmond 24-25).

Table 1: [Syria] Video and Channel Categorizations’ variables and details

Video Categorizations Variable Details

(30)

Table 2: [Kuwait]’s Video and Channel Categorizations’ variables and details

ISIS-Related Mentioning any association with terrorist organizations/jihadists/jihadi or ISIS brides

ISIS footage on battlefield in Syria or other countries Reporting on

Syrian Crisis

Videos that explain the Syrian crisis: how it started, updates on the situation, news on the president Bashar Al Assad and his predecessor, Russia’s

involvement, soldiers talking about the war etc. Combat footage on different Syrian land

Unrelated Conflict

Reporting/news on a current political, humanitarian or war conflict unrelated to the Syrian crisis

(War) footage of conflicts

Unrelated Videos that are meant to entertain, unrelated to political conflict News Live streaming news videos

Channel Categorizations Variable Details

News News outlets, magazines, or newspapers around the world

Native News Channels that are native to YouTube and report or comment on news Includes YouTubers that specifically comment or report on news Specialized in showing war footage

Informative Documentaries

Provides facts or accounts on historical events

Infotainment Channels that include both informative and entertaining facts or content Includes TV programs/channels

(31)

Video Categorizations Variable Details

Footage Footage around Kuwait

People Vlogging/driving/landing in Kuwait Specific touristic location footage in Kuwait

Facts Videos that give information about Kuwait: places to visit, its history, its geography, its upcoming projects etc.

Facts for tourists/expats to visit or live: visa information, job processes Updates for foreigners/expats in Kuwait in regards to visas/jobs

Life in Kuwait

People talking about what it is like living in Kuwait (negative/positive) What it is like working in Kuwait

Experiences in comparison to their hometowns or what people should know before coming to live in Kuwait (excluding bureaucratic processes)

Advertising Videos that are made by companies (banks, telecommunications companies, shops etc.) with the aim to advertise – appeals to culture and celebrations Celebrating Videos with the sole purpose of celebrating Kuwait’s youth, National &

Liberation day, Eid

Shows people singing/dancing for Kuwait and dressed in Traditional Kuwaiti clothes

News Live news streaming videos

Videos that report on Kuwaiti news or celebrities Unrelated Videos that have nothing to do with Kuwait Channel Categorizations

Variable Details

Company Channels that belong to banks, telecommunications companies, shops and Production companies

(32)

Table 3: [Jamal Khashoggi]’s Video and Channel Categorizations’ variables and detail

Informative Channels that give facts, places to visit, historical/geographical information about countries

Entertainment Music/celebrity news channels

Channels that include random funny/weird videos

Infotainment Channels that include both facts about countries and random entertaining facts about the world or celebrities

News News outlets, magazines, or newspapers around the world

Vlogger Channels where individuals’ majority of videos include them recording themselves talking to the camera about their lives/lives in Kuwait Travel &

Tourism

Channels that are dedicated to photography and filming countries/cities around the world

Video Categorizations

Variable Details

News Live news streaming videos SA & MBS (Related to

MBS & Saudi Arabian relations)

Discourse that shifts from reporting on the murder to majorly discussing Saudi Arabia as a country or relations with it & MBS

Criticizing MBS as a ruler or Saudi Arabia as a country

Explanations on how MBS came to power or what he has done since then

Pre-crime Any footage of an Interview with Khashoggi Khashoggi talking about MBS and Saudi Arabia Reporting on murder Videos that report on the murder

Any news/footage that further adds to the investigation & puts the pieces together

YouTube’s Algorithms and the Cartography of Controversy: Unboxing the algorithmic framing of the Middle East on YouTube