Let’s unpack the title, just a little bit:

During the time of writing, we are nearing the end of 2021 and the Covid-19 pandemic is nowhere near over. The global health crisis has impacted every aspect of social life and continues to boggle governments and societies around the world. The struggles and consequences are continuously debated, restrictions vary per country and the rolling out of massive vaccination campaigns are an unprecedented phenomenon. The discussions on various social media platforms concerning the pandemic are a compelling point of analysis, because the highly contested strategies of trying to limit the impact of the virus are a moving target. In 2020, Instagram rolled out new features to help combat misinformation about the virus (Clark 2020).

On Instagram, the use of certain hashtags connected to the pandemic return results that are from sources Instagram deems credible – like the World Health Organization and the RIVM in the Netherlands – and posts that either include tags, or certain words (ie. “Vaccine”, “Covid”) get tagged with a link to a local site that provides advice and strategies (Instagram Blog 2021).

They have also added an information center where they in short form provide explanations about vaccines, their side effects and how they work (Instagram 2021). The information centers are location based, so they can provide guidelines in local contexts, as the pandemic and the strategies around it vary per country. Such interface changes and strategy rollouts have considerable consequences for how the discourse around the health crisis is shaped. The use of data, algorithms and machine learning to find, recognize and tag posts which lead to the information center needs to be considered from a critical perspective.

Scholars and theoreticians such as…

Boyd and Crawford (2012, 674) have sketched out several considerations about the use of data; one that rings particularly with Instagram’s updates features concerning Covid-19 information is their mapping of the divide between ‘Big Data Rich and the Big Data poor’. Their paper opens up critical conversations with regards to how data is used to shape conversations. They propose that Big Data and its role in doing research brings about an epistemological and ethical shift, where the process of conducting research and how reality is categorized is completely reframed(ibid).

What is interesting for this text is how Instagram uses its affordances to tag content, and the role it plays in providing quality information about a global health crisis, with the intent of stopping the spread of falsehoods – at least that is the justification for the interface changes.

There are necessary considerations when a platform that has around a billion monthly users implements strategies which accompany posts that users share. It is not immediately clear how exactly Instagram tags posts where people discuss the pandemic – the use of certain hashtags seems logical, however, posts with no hashtags where an image has keywords that could be connected to the phenomenon can get tagged and linked as well. It could be that Instagram does not want to share how they are flagging posts, as that could provide a manual for circumventing the tagging process. Yet, it leaves very little space for critical examination, when their methods and decisions are unavailable for scrutiny.

Data visualizations and journalism exploded with this pandemic, with many outlets and platforms offering infographics with local and global contextualization. While the potential of data to be used for reporting and framing global phenomena cannot be overstated, boyd and Crawford’s text still rings true in their call for questioning how data is used, by whom, and to what ends. According to Petrina (2021, 224) data has even become a subject, which reshapes our understanding of what a subject is.  Usually, it would be understood as a coherent entity, while in the case of data which is fragmented and needs to be put into context to be unified, the philosophical conundrum of data as a subject presents a radical reimagination of its role and identity(ibid). Petrina’s proposal functions as an extension of boyd and Crawford’s call for critical examination; one must consider the role, status and access to data that Instagram has, and how they subsequently employ their strategies, and further, how we can see data as a subject in the provided discourses on the pandemic.

When companies as Instagram roll out new features in their algorithmic and datafied interfaces, the question of who gets insight is omnipresent within media studies. The black-box nature of tech behemoths is nothing new in the conversations on the use of data, algorithms, privacy and machine learning; however, in this specific crisis there needs to be space for critique of access and accountability. If Instagram’s changes to their interface and the tagging of content is indeed so crucial for people to have access to certified information by health authorities, who is in position to examine the platform’s role, their ethical considerations and the teams behind the processes.

In order to roll out such major changes and have an almost infinite stream of content be affected, decisions have to have been made about how this will be implemented. According to Gillespie (2014, 167) algorithms do not only help us find and filter information, but act as a means of participation in socio-political discourse, because they are a crucial component in the flow of information upon which people then assign meanings, importance and truthfulness.

CONSIDER THIS…

From a scholarly perspective it is compelling to question why and how these decisions were made by the platform, are there possible alternatives and who holds Instagram accountable. Upon initial examination it might seem harmless to tag content with the intent of providing scientifically corroborated information in a health crisis. Yet, Instagram is not an official authority on information and takes considerable decisions that can have massive impact with little transparency in their methodology. Again, it is in no way peculiar that a technology company makes decisions that have major implications without much input from the user base, or with very little communicated clarity. Another interesting development is the creativity with which people avoid their content being tagged. For instance, by changing letters in the text or using specific keywords.

It is far beyond the scope of this commentary to consider all the implications of such tactics; however, it points to several compelling phenomena that tie to the use of digital platforms. One of them could be the creation of counterpublics – Milner’s (2016) concept where he proposes that such public challenge dominant discourses. Or perhaps the political nature of content creation that is meant to avoid tagging. At the core of the proposed considerations are the use of data and platform interfaces to shape narratives and discourse during a global pandemic. When a platform is as massive as Instagram is, the role it plays in contributing to such a discourse needs to be acknowledged and scrutinized. Critically examining its use of data, algorithms and interface choices that shape narratives, decide who has access and authority and in turn function as a catalyst for users to employ tactics and mobilize to try and ‘trick’ the system.

BIBLIOGRAPHY

boyd, danah, and Kate Crawford. “CRITICAL QUESTIONS FOR BIG DATA: Provocations for a Cultural, Technological, and Scholarly Phenomenon.” Information, communication & society 15, no. 5 (2012): 662–679.

Clark, Mitchell. 2020. ‘Instagram Is Rolling out New Notifications about COVID-19 Information’. The Verge. 17 December 2020. https://www.theverge.com/2020/12/17/22187298/instagram-coronavirus-covid19-misinformation-notifications.

‘COVID-19 Information Center’. n.d. Accessed 30 November 2021. https://www.instagram.com/coronavirus_info.

Gillespie, Tarleton. “The Relevance of Algorithms”. In:Media Technologies: Essays on Communication, Materiality, and Society, edited by Tarleton Gillespie, P J Boczkowski, and K A Foot, 167-194. Cambridge MA: MIT Press, 2014.

‘Helping to Inform People about COVID-19 Vaccines | Instagram Blog’. n.d. Accessed 30 Novemberr 2021. https://about.instagram.com/blog/announcements/continuing-to-keep-people-safe-and-informed-about-covid-19.

Petrina, Denis. “Affect Trapped: Algorithms, Control, Biopolitical Security”. In: Living and Thinking in the Postdigital World, edited by Wróbel, S.  Skonieczny, 219-231. Krakow, University of Warsaw, 2021.

Project idea

It is no news that algorithms can be sexist when autocompleting a sentence, performing facial recognition or displaying search results (Sheng et al. 2018; Buolamwini and Gebru 2018; Noble 2018). To confront this problem, artist and researcher Caroline Sinders (2020) started a multi-year project Feminist Data Set, to “interrogate every step of the AI process” (p.4) with the aim of designing a feminist chatbot. Currently the project has dealt with data collection and data labeling, the first two steps of the AI process.

Inspired by Sinders’ project, we follow the guideline in her project documentation to create a feminist dataset ourselves. Through this project, we seek to situate Sinders’ project in the current academic debate on algorithmic bias and data feminism. Both our research on existing literature, and the practice and reflection on data collection inform us to provide a critique on Feminist Data Set. 

Current academic debate

Algorithmic bias

Algorithms can be simply understood as “a sequence of instructions telling a computer what to do” (Domingos 2015 cited in Brogan 2016). Unlike rule-based algorithms, in which the instructions are determined, machine learning algorithms write their own instructions based on given input and output data, through a mathematical process of “brute-force approximation” (Joler and Pasquinelli, 2020, p.11). The application of machine learning algorithms is in explosive growth due to the rising volume of data and increased computational power, the continuous innovation in image technology, video technology and human-computer interaction, and the increasing use of algorithm-based autonomous decision systems for data mining, search engines, biometrics and many other areas.

While some scholars take the social determinism view on algorithms and are optimistic about their neutrality (Cahn et.al., 2019), most scholars agree that algorithms are biased. Various examples of algorithms leading to gender and racial discrimination repeatedly demonstrate that the seemingly neutral, automated systems can produce unfair decisions and biased results (Danks, 2017). In other words, the existing biases in gender, race, ability and class in society are amplified by algorithms (Joler and Pasquinelli, 2020).

  Existing research on the cause of algorithmic bias can mainly be summarized in threefold ways: First, the bias existing in the society before technological intervention; second, the bias related to the training dataset; third, the bias related to the algorithm model itself (Friedman and Nissenbaum, 1996; Baeza-Yates, 2016; Danks, 2017; Joler and Pasquinelli, 2020). This research project focuses on the process of building a training dataset, so we will review relevant research on the bias related to the training dataset below. 

As the data used for training an algorithm is the cornerstone of the algorithmic program, its degree of objectivity and neutrality directly affects the decision outcome of the algorithm. An important feature of machine learning is that they require a large amount of input data to ‘learn’. If the input data itself is biased, then the results produced are biased as well. Particularly in machine learning algorithms, the absence of data samples from marginalised groups will infinitely circulate and reinforce society’s structural biases. Lerman (2013) argues that, due to poverty or geographic location, people living on the edge of big data are invariably left out in a non-random, systematic way, and their lives are less ‘datafied’ than the general population. Buolamwini and Gebru’s study “Gender shades” also illustrates this problem. Due to the lack of photos of many black women in data used to train facial recognition AI, the detection of black women is less accurate than for other groups of people. Some famous black women such as Oprah Winfrey and Michelle Obama are incorrectly identified as men by AI (Buolamwini and Gebru, 2018). 

Understanding the current data collection practice for natural language processing AI is pertinent, because this research project is about collecting training data for a text-based AI. Due to the data-hungry nature of machine learning algorithms, a big amount of text-based data is mined from Reddit and used to train the latest natural language processing model GPT-3 (Piper, 2020). As Reddit is known for its misogynistic culture, it is not hard to imagine the algorithm trained by data from Reddit will also become misogynistic. Hence, a feminist intervention on AI training dataset is necessary.

Data feminism & Intersectional feminism 

The concept of “intersectionality” was first coined by scholar Crenshaw (1989) to refer to multiple identities, which include gender, race, class, ethnicity, nationality, sexual orientation, religion, age, etc, and she proposes that “the experience of intersectionality is far from a superposition of racism and sexism”. D’Ignazio and Klein (2020) link “intersectionality” to “feminism” and argue that when discussing intersectional feminism, gender should not be used as a single framework for analysis, but needs to be examined alongside issues of race, migration status, history, social class, and especially about the individual female experience. As the intersectional approach suggests that people’s social identities can overlap, resulting in complex experiences of discrimination, intersectional feminism focuses on the voices of those who experience overlapping and simultaneous forms of oppression, in order to understand the depth of inequalities in any particular context (Dastagir, 2017).

The concept of intersectionality has been popular as a critical concept among academics since its emergence as an important dimension of race/class/gender studies, and has even been referred to as the ‘academic kernel of difference studies’ (Dill, 2009). Furthermore, the idea of intersectional feminism inspires some researchers, such as Sinders, D’Ignazio and Klein, to critique the status quo in the development of data and algorithms by examining the unequal power in data and how it leads to algorithmic bias. D’Ignazio and Klein (2020) present a new way of thinking about data science and data ethics, “data feminism”. Building on the notion of “data as power”, they point out that data collection and analysis is biased against anyone who does not fit the upper-class white male mould, and provide suggestions on how to deconstruct this power. Furthermore, as the data is envisioned and produced in a society significantly influenced by a history of white supremacy and patriarchy (Garbee, 2020), no matter how data is collected, it still remains confined within that system and still cannot be seen as truly equal.

Thoughtful data collection in practice

Methodology

Informed by an overview of existing scholarly work on algorithmic bias and the intersectional feminist approach to data, we attempt at building a dataset from scratch following the guidelines in Caroline Sinder’s research and art project “Feminist Data Set,” in which she notes that the data collection process is designed to be slow and thoughtful, countering the opaque and extractive nature of data collection in the current machine learning practices. We seek to critically analyze Feminist Data Set by going through the process of data collection ourselves. Although we will create a dataset as a result of the project, the main goal is not to produce a usable training dataset, but to be able to reflect on the process and to deepen our understanding on data collection, algorithmic bias and intersectional feminist interventions.

As Sinders notes in her documentation, there are two main criteria for building the dataset: 1) Data collected has to be feminist and intersectional; 2) For the purpose of building a chatbot, data should be in the format of written word, and part of the data has to be in colloquial language. In order to operate the project on a feasible scale, we focus on the topic of childcare as the theme of data collection.

Search engine bias

We start our data collection on Google fully aware of the search engine’s bias. Although this project’s focus is not a critique on Google’s algorithmic search per se, Google search engine is part of the problem that Feminist Data Set sets out to solve, i.e. algorithmic bias. Therefore, our data collection starts with critiquing the search engine we employed, and designing our query to confront the flaws in the tool we use in order to find intersectional feminist results effectively and efficiently.

Researchers, scholars and journalists have written about the search engine’s in-built bias, specifically towards women and people of color (Pariser, 2011; Noble, 2018; Tenner, 2018). According to Hindman (2008, p.43), “sites are ranked in a popularity contest in which each link is a vote, but the votes of popular sites carry more weight” in the ranking of Google search results. Through black feminism critique, Noble (2018) points out that Google “[biases] information toward the interests of the neoliberal capital and social elites in the United States.” Sinders (2020) also warns in her documentation that written text from marginalized groups can be hard to find. It is with this awareness that we collect data on the topic of childcare.

Query Design and data collection

We situate our data collection topic, childcare, in the context of intersectional feminism, which we have explored in literature review. It informs us to expand our query keywords from the stereotypes of “mother” and “female” to “unpaid work”, “migrant labor”, “race”, and policies welcomed by intersectional feminists such as “universal childcare”. 

After searching with different keywords and browsing, we realize that it is not efficient to take the entire web as the source for query, as intersectional feminist contents are not the mainstream on the internet, and some of the results we’re looking for are not privileged by the search engine algorithm. Inspired by the ideas behind digital methods such as cross-spherical analysis (Rogers, 2013) and source distance (Rogers, 2019), we decide to use [site:] queries to look for childcare-related data purposefully within the intersectional feminist part of the internet. Out of those query results, we go beyond the first few pages of the results, and select ones that are truly intersectional and feminist. A list of the collected data can be seen here.

The same search approach applies when we look for conversational text for the purpose of training a chatbot AI. We first identify online forums where discussions on feminism may happen, and then check if the text is intersectional. We take into account the space where these conversations take place, and only collect data from spaces that are deemed fully public and open, such as Reddit and other online forums. We did not search on semi-open social platforms such as Facebook. We believe that taking extra care assessing what data to and not to collect is part of the practice that Feminist Data Set advocates for.

Discussion

Feminist Data Set’s contribution

Perhaps the most valuable contribution of Feminist Data Set is that it reclaimed data collection as a practice of resistance. The development of AI has been shaped by U.S. military and intelligence agencies since its early days, dating back to the 1950s; and data are collected to support the priority of military and state control (Crawford, 2021). Even though now AI technology has been largely commercialized, it doesn’t change the extractive nature of data collection––now serving the purpose of capitalist profit generation. Within the realm of academic research on AI, data collection is done under the guidance of specific ethical codes. Sinders’ Feminist Data Set is intentionally designed to start a public dialogue by subverting those power dynamics in data collection. In her toolkit documentation, she wrote, “What is a data set about a community that is made by that community? It can be a self-portrait, it can be protest, it can be a demand to be seen, it can be intervention or confrontation, or all of the above” (Sinders, 2020, p.7). 

Feminist Data Set is a radical intervention in that it recognizes problems in existing datasets and seeks to solve the problem by creating new and better datasets. There are other types of intervention examining discriminations in algorithms, but most of them stop at identifying discrimination within existing algorithmic systems without offering solutions. In their research, Sandvig et al. (2014) borrowed the concept of audit study from social science and designed an analogous approach, algorithm audit, “to investigate normatively significant instances of discrimination involving computer algorithms operated by Internet platforms” (p.8). In the five audit study designs proposed, they mostly focused on identifying algorithmic discriminations from platforms, users and their interactions (Sandvig et al., 2014). Other interventions attempt at understanding how algorithms work without necessarily looking into the “black-box,” such as through what Bishop (2019) coined as “algorithm gossip” and what Bucher (2017) called “algorithmic imagination,” both examine how people perceive algorithms at work, to make the invisible visible.

Feminist Data Set’s limitation

There are some limitations with data collection as intervention to algorithmic bias. The most obvious one is that addressing the problems in datasets is only the first step of such intervention, as bias exists in each step of the process of designing an AI, which is outlined in the literature review. This section will discuss other less obvious limitations of data collection.

Data collection has its own paradox. When Taylor (2017) proposes the framework of data justice, she points out that data visibility can result in both access to representation and concerns over informational privacy. She emphasizes the importance of reconciling these seemingly contradicting perspectives by situating them in specific social and political conditions, without offering a one-fit-for-all answer (Taylor, 2017). 

Informed by the data justice framework, careful assessment on the nature of data collected and the collection methods is necessary when applying Feminist Data Set in practice. Our project shedding light on the intersectional feminist perspective on childcare aims at collecting public available articles on the topic, hence can be evaluated as more of an effort to improve the visibility and representation of the topic from the perspective of the marginalized. However, we acknowledge that on some other issues, collecting sensitive and private data amounts to surveillance or violation of privacy. Raji et al. (2020) point out that efforts to collect images from a particular population to improve facial processing technology raise ethical concerns over consents and exploitation, even though the intention was to improve the technology to prevent it from discriminating against the population.

Moreover, on the practical level, we have to confront the mismatch between the intentionally slow process of data collection guided by Feminist Data Set and the huge amount of data necessary for machine learning purposes. Although we’re clear that the purpose of this project is not to create a usable dataset, the experience of creating a dataset from scratch helps us realize the demanding workload if this method were to be implemented in reality. A successful model that comes to mind is the Wikipedia Edit-a-thon organized by Art + Feminism, dedicated to adding information about women and people of color on the community-run online encyclopedia website Wikipedia. Moving forward, Feminist Data Set can explore the possibility of mobilizing a network of people to collectively build intersectional feminist datasets, so that the goal of having a dataset to train a machine learning algorithm can become achievable.

Conclusion

This research project situates Sinder’s Feminist Data Set in the academic debates on algorithmic bias and data feminism, and seek to critique it through the practice of building a dataset from scratch. The prevalent bias against women and marginalized groups in algorithms makes intersectional feminist interventions all the more necessary for data collection. 

During the data collection process, we first examine the bias present in search engines, and design our queries accordingly. Based on the practice, we found that there are both limitations and contributions with data collection as an intervention to algorithmic bias: One of the shortcomings is that depending on the type of data collected, data collection can improve the representation of marginalized groups while raising ethical concerns. Another shortcoming of the data collection approach is its feasibility. We recognize Sinders’ project as a practice of resistance, questioning the current power imbalance of data collection mostly serving the interests of the state and for-profit companies. Feminist Data Set represents a force to reclaim data collection for the interest of the public, especially the marginalized group, a creative intervention on algorithmic bias, and a willingness to work towards its elimination. 

Reference

Baeza-Yates, R., 2016. Data and algorithmic bias in the web. In Proceedings of the 8th ACM Conference on Web Science (pp. 1-1).

Bishop, S., 2019. Managing visibility on YouTube through algorithmic gossip. New Media & Society 21, 2589–2606. https://doi.org/10.1177/1461444819854731

Brogan, J., 2016. What’s the Deal With Algorithms? Slate. Available at https://slate.com/technology/2016/02/whats-the-deal-with-algorithms.html (Accessed 27 October 2021).

Bucher, T., 2017. The algorithmic imaginary: exploring the ordinary affects of Facebook algorithms. Information, Communication & Society 20, 30–44. https://doi.org/10.1080/1369118X.2016.1154086

Buolamwini, J., & Gebru, T. 2018. Gender shades: Intersectional accuracy disparities in commercial gender classification. In Conference on fairness, accountability and transparency (pp. 77-91). PMLR.

Cahn, N., Carbone, J., & Levit, N. 2019. Discrimination by Design. Ariz. St. LJ, 51, 1.

Collins, Patricia Hill, and Sirma Bilge. Intersectionality. Malden: Polity, 2016.

Crawford, K., 2021. Atlas of AI: power, politics, and the planetary costs of artificial intelligence. Yale University Press, New Haven.

Crenshaw, K. 1989. Demarginalizing the intersection of race and sex: A black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics. u. Chi. Legal f., 139.

Danks, D. and London, A.J., 2017, August. Algorithmic Bias in Autonomous Systems. In IJCAI (Vol. 17, pp. 4691-4697).

Dastagir, A., 2017. Bodies: Alia E. Dastagir, “What is Intersectional Feminism? A Look at a Term You May be Hearing A Lot”. [online] Bodies: A Digital Companion. Available at: https://scalar.usc.edu/works/bodies/alia-e-dastagir-what-is-intersectional-feminism-a-look-at-a-term-you-may-be-hearing-a-lot (Accessed 21 October 2021).

D’ignazio, C., & Klein, L. F. (2020). Data feminism. MIT press.

Dill, B. T. 2009. 10. Intersections, Identities, and Inequalities in Higher Education. In Emerging intersections (pp. 229-252). Rutgers University Press.

Friedman, B. and Nissenbaum, H., 1996. Bias in computer systems. ACM Transactions on Information Systems (TOIS), 14(3), pp.330-347.

Garbee, E., 2020. Review of DATA FEMINISM by Catherine D’Ignazio and Lauren F. Klein. [online] Issues in Science and Technology. Available at: https://issues.org/doing-the-work-data-feminism-review/ (Accessed 21 October 2021).

Lerman J. 2013. Big Data and Its Exclusions.Social Science Electronic Publishing，66:P.3，P.5.

Noble, S.U., 2018. Algorithms of oppression: how search engines reinforce racism. New York University Press, New York.

Pariser, E., 2011. The filter bubble: how the new personalized web is changing what we read and how we think. Penguin Books, New York.

Pasquinelli, M., & Joler, V. 2020. The Nooscope manifested: AI as instrument of knowledge extractivism. Ai & Society, 1-18.

Piper, K., 2020. GPT-3, explained: This new language AI is uncanny, funny — and a big deal. Vox. Available at https://www.vox.com/future-perfect/21355768/gpt-3-ai-openai-turing-test-language (Accessed 19 October 2021).

Raji, I.D., Gebru, T., Mitchell, M., Buolamwini, J., Lee, J., Denton, E., 2020. Saving Face: Investigating the Ethical Concerns of Facial Recognition Auditing, in: Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, AIES ’20. Association for Computing Machinery, New York, NY, USA, pp. 145–151. https://doi.org/10.1145/3375627.3375820

Sandvig, C., Hamilton, K., Karahalios, K. and Langbort, C., 2014. Auditing algorithms: Research methods for detecting discrimination on internet platforms. Data and discrimination: converting critical concerns into productive inquiry, 22, pp.4349-4357.

Sheng, E., Chang, K.-W., Natarajan, P., Peng, N., 2019. The Woman Worked as a Babysitter: On Biases in Language Generation. arXiv:1909.01326 [cs].

Sinders, C., 2020. Feminist Data Set. Available at https://carolinesinders.com/wp-content/uploads/2020/05/Feminist-Data-Set-Final-Draft-2020-0526.pdf (Accessed 19 October 2021).

Taylor, L., 2017. What is data justice? The case for connecting digital rights and freedoms globally. Big Data & Society 4, 2053951717736335. https://doi.org/10.1177/2053951717736335

Tenner, E., 2018. How to Get Better, Less Biased Search Results. Time. Available at https://time.com/5318918/search-results-engine-google-bias-trusted-sources/ (Accessed 24 October 2021).

Who doesn’t love an introduction?

The notion that machines can mimic features of human cognition has intrigued the new media art world in recent years. During the last half century, machine learning algorithms have progressed far enough that artificial neural networks can produce images that artists have written in computer programs in order to generate algorithmic art. According to Broeckmann (2016, 108) algorithms have found many different uses in art – they can generate images, shape human and machine interactions, and design objects.

Peters (2015, 28) polemicized that there is a technicality in mind and body, hence the humanities function as disciplines where culture is stored, interpreted and transmitted. According to him, in music, dance and poetry there is counting, and there is writing. From there one could posit that all art forms carry a technological quality which combines mind, body, space and technology http://mirziamov.ru/zaym-bez-otkaza. This research follows the line of thought where artistic creation is an amalgamation of the material, the technical, conceptual, physical and ideological. Therein lies the broad potential of algorithmic and AI creation, because it questions authenticity, agency, and our own beliefs of what creativity is, and how we can understand and/or develop it further. Miller (2020, chap.VI, para.1) proposes that when we program computers according to our understanding of creativity, that we cannot yet imagine a creativity that would belong to a computer that develops it without us. Herein lies our main question: How to treat the potentials that algorithms and AI provide in New Media art production? 

To answer the research question and respond to the emerging debates from the concepts proposed, a literature review is conducted together with a close reading of the attached artworks. An experiment is conducted at the end of the research project with the free tool Artbreeder, which allows for AI image manipulation across several categories.

An inquiry into AI and algorithmic art is interesting for media and other scholars alike. The interdisciplinary nature of the research and the many possible angles of examination make the field compelling, as well as challenging for inspection. In new media studies it opens debates on the use of data, the agency of technology, the potentials and pitfalls of artificial intelligence and algorithm use.

Some historical Context is even better…

Since the late 1950s, when a group of engineers at Max Bense’s laboratory at the University of Stuttgart began experimenting with computer graphics, artists have been employing computers to make artworks. Frieder Nake, Georg Nees, Manfred Mohr, Vera Molnár, and a slew of other artists experimented with the use of mainframe computers, plotters, and algorithms to create visually appealing artifacts (Garcia 2016). It is interesting to see how humans and machines are working together with art, which many consider to be a cultural practice that is linked to the human mind and skill rather than a machine. One of the earlier practitioners of this form is artist Harold Cohen, who wrote the program AARON in 1973 to produce drawings that followed a set of rules he had created.  In describing the creation of AARON, Cohen wrote that the versions prior to 1980 worked with aspects of human cognition, making it possible for the program to understand similarities, positions, repetitive patters; he described it as a system for an expert, because it served as research tool to expand the researcher’s knowledge, instead of encapsulating the knowledge for others to use (ibid). Cohen continued to develop and refine AARON for the rest of his career, but the program maintained its core design of performing tasks as directed by the artist. 

A theoretical Framework is especially lovable…

Big Data has been a compelling research topic across scholarly disciplines and its exacerbation as a crucial component of digital infrastructures has only furthered the polarizing debates concerning the concept. For the purpose of this paper, we look at Big Data as one of the necessary components in the creation of algorithmic and AI art. The working definition that is used as the lens for this paper follows Boyd and Crawford’s (2012, 663) division into three dimensions: (1)the technological – which looks at computation power and algorithmic accuracy to tackle large datasets for analysis; (2) the analysis dimension – which is relevant for identifying patterns in order to make overarching claims; (3) the mythology surrounding data – which proposes that through the use of big data insights and possibilities occur that offer new intelligence previously unattainable.

The strict focus of the research is not an analysis of Big Data and how it is used, but the interplay of these dimensions is vital to consider because they are intertwined with the creative processes that the paper’s examples treat. By considering the technological, analytical and mythological dimension, it is possible to further the understanding into algorithmic creativity and the particular ‘intelligence’ that the research questions.

Crawford (2021, 215) in her book directly addresses this ‘magical’ illusion of AI that is so commonly taken as fact; she criticizes the Cartesian dualism that is often employed in discourse on AI where they are seen as independent brains that produce knowledge completely independent from the structures in which they are deeply embedded. Every technology is embedded into its social, political and institutional context. That is why it is necessary to consider how in the process of creating new media artworks that these technologies have a deep connection to the artist, the data and environment in which they are contextualized. 

There are sets of rules and functions that are put into the designs which are not alien or otherworldly, but very much material and not as intangible as they might come across. The interconnectedness of the processes involved in employing these technologies and the media that they produce is the space that the paper questions. As a site of experimentation, it can be argued that what such technologies do for artistic production is generate a different arena and expand possibilities. However, they do not do so completely on their own as they rely on infrastructures and datasets. 

Peters (2015, 23) proposes that old media do not die, moreover, they are absorbed into what we call new, and are then reimagined. He further claims that all media raise enduring questions of life; the digital affordances of media have enabled services of collection and management for prayer, profit and power – which is both ancient, and modern(ibid). 

These parallels are a guideline through the research process. It is far beyond the scope of this research to argue within art what is indeed new, and what is reimagined, as many art and history scholars have written extensively on such debates. However, what is relevant for this paper is situating the case studies, reflecting upon what the new-ness of algorithms and AI in art can be indicative of, when considering new media artistic production. 

Media scholar, theorist and philosopher Marshall McLuhan in his book ‘Understanding Media’ has elegantly combined the how media, technology and art are intertwined: 

No society has ever known enough about its actions to have developed immunity to its new extensions or technologies. Today we have begun to sense that art may be able to provide such immunity. In the history of human culture there is no example of a conscious adjustment of the various factors of personal and social life to new exceptions except in the puny and peripheral efforts of artists. The artist picks up the message of cultural and technological challenge decades before its transitioning impact occurs. He, then, builds models or Noah’s arks for facing the change that is at hand.(McLuhan 1994, 71)

Pasquinelli and Joler (2020) in their mapping of AI attempt to secularize it into an understanding of how it functions as an instrument of knowledge, and not as an ‘intelligence’. Such treatment directly tackles the mythological tendencies around algorithms and AI because it gives it a more concrete theoretical grounding, where one can actively avoid ascribing legend-alike cognitive capabilities. Rather, it gives materiality and technological affordances a place where they are an extension of creation, not a mind of its own, free of any human intervention. 

Where they are especially critical is in their claim that AI – like many forms of automation – functions as a map of approximations and potential outcomes; however, still unavailable for community consent and access(ibid). In the area of new media algorithmic and AI art, one can see that access is indeed a crucial component into what can and cannot be appropriated for creative production. Yet it can then also be posited that exactly new media art can offer potential interventions into access to AI for artistic production. Digital media art co-created by algorithms and AI fills this gap, because it relates the producer, tool, data and consumer together in an almost circular relation to one another.

Examples are the cherry on top…

PATTERNS

Early noticeable examples of AI and art were made by engineer A. Michael Noll in 1962, who experimented with an IBM machine’s creativity by having it make random patterns. Later, this would be called ‘computer art’ by fellow programmers and cultural historians (Pepi, 2020). Noll however, decided the results should simply be called ‘patterns’. Arthur I Miller (2020) wrote about these ‘patterns’, comparing the AI used to the human brain, arguing that the production of new knowledge from already existing knowledge is what happens to both. This is an example of minimal human interference, where the result is still deemed artistic. However, this notion could also be the result of the time it was discovered, with it being new and relatively aesthetically pleasing. It also falls neatly in line with Pasquinelli and Joler, because in spite of minimal human intervention, the machine is an extension, a tool to produce these patterns of material creation.

Figure 1 Michael Noll’s patterns ‘A. Michael NOLL at the Digital Art Museum’. n.d. Accessed 25 October 2021. https://dam.org/archive/noll/artworks_04.htm.

Figure 2 Michael Noll’s ‘Gaussian-Quadratic’ ‘History of Information’. n.d. Accessed 27 October 2021. https://historyofinformation.com/image.php?id=4737.

AICAN

Art made by AICAN differs greatly from Noll’s work in the way that this system was specifically designed to autonomously make art. It also differs in that it generates art based on other artworks and existing art genres, opposed to the ‘new’ art style of ‘patterns’. It builds off of something that already exists. The controversial issue is here, that if the system generates something novel, engaging or moving, it is hard to decide who is credited for it. What is also interesting, is that an experiment proved many people could not tell the difference between AICAN and human art. This proposes questions about the importance of authenticity and puts a lot of weight on the human perception of art. Crawford’s framework neatly complements the questions surrounding these works: there are no magical dimensions, but an interconnection of human created art and AICAN.

 Figure 3 AICAN artworks Elgammal, Ahmed. n.d. ‘Meet AICAN, a Machine That Operates as an Autonomous Artist’. The Conversation. Accessed 17 October 2021. http://theconversation.com/meet-aican-a-machine-that-operates-as-an-autonomous-artist-104381.

GAN 

Another example is the work of artist Casey Reas. The main difference between this and AICAN’s artwork, is the many steps involved in the production of art made by the GAN system. Artists who work with GAN need to select and upload a large set of images. Then, the artist must interfere with the system to get a desired outcome, by deleting and adding certain pictures. However, the artist never has full control as the result is always a matter of testing and learning, as nobody is entirely sure how the systems’ layers learn. Whilst the process is therefore machine influenced, Reas argues that each of these steps can be compared to the steps any ‘normal’ artist takes when working. 

Choosing equipment and materials, making changes to get a desired result, and leaving the last bit up to the tool. With the use of AICAN, many artworks have been made that imitate specific genres or even artists (Pepi 2020). What is interesting about this type of art, is that it once again leaves us to think about the way we not only view, but critique art. When critiquing the artist, the main point must be in consideration to the selection process of the artist. This puts more weight on the process of the art, rather than the artwork itself. Peters’ treatment of the old subsuming into the ‘new’ and questioning what about such an artwork is in fact novel are demonstrated in the example of the Mona Lisa reimagined in different styles.

Figure 4 Variations of ‘Mona Lisa’ Pepi, Mike. 2020. ‘How Does a Human Critique Art Made by AI?’ ARTnews.Com(blog). 6 May 2020. https://www.artnews.com/art-in-america/features/creative-ai-art-criticism-1202686003/

THE NEXT REMBRANDT

The Next Rembrandt is a project that was initiated by ING Bank and company Walter Thompson, where they worked together with Microsoft to bring the work of the artist Rembrandt van Rijn back to life, almost four hundred years after his death. This project is very different from the previously discussed AICAN artworks, because there is a large team involved. Data scientists, AI developers AI and 3D printing experts worked on the project for 18 months. In this time, more than 160 thousand fragments were scanned from 346 paintings and processed with a deep learning algorithm. Even his brushstrokes were replicated by using depth scanning and 3D printing brushstrokes (Schlackman 2020). In terms of the paper’s questions on AI art, this project opens a lot of debates. For example, the final product is not an individually desired result from an artist, but the result of technical experts and machine learning. Many people were involved in the process, yet it seems as if the result is mostly machine generated, with every final ‘decision’ being made on the basis of existing work and carefully acquired data – which again proposes questions of authorship, and new-ness.

 Peters’ argument would then propose that AI and algorithmic art stand at a similar frontier of new and old, as does the much-contested debate on what media is indeed ‘new’ and what makes it such.  Boyd and Crawford’s three dimensions of data are also particularly relevant here as they are deeply embedded in this work: an extensive technological and analytical part took place, to create an almost mythological ‘new’ or ’next’ Rembrandt. 

 Figure 5 ‘The Next Rembrandt’ Schlackman, Steve. “Who Holds The Copyright in AI Created Art.” Art Business Journal. September 29th, 2020. https://abj.artrepreneur.com/the-next-rembrandt-who-holds-the-copyright-in-computer-generated-art/

MODERN DAY USE OF AI 

The Next Rembrandt is a fascinating example of using highly capable technology by experts to re-work a text. Yet other, more accessible forms of AI creation are showing up in online spaces. An example of this is seen on TikTok, where multiple users entertain large quantities of viewers by generating requested AI images. For example, the creator makes an AI system generate visual images based on certain words or questions. The results have evoked all kinds of reactions from viewers. This is partially due to the themes of the AI visuals, like mental illness and many controversial topics like death, life, emotions, etcetera. 

This shows how integrated AI can be in everyday life, but also how thin the line is when discussing AI and art. Following McLuhan’s train of thought, here the artist re-creates messages of the cultural and technological. The impact is difficult to discuss or map, which also falls in line with his argument that the models for changing perception might be ahead of their time. If art is defined by being authentic, moving, or even valued by perception, simple generated visuals like this can in spite of their simplicity when compared to large-scale projects like The Next Rembrandt, are indeed new media artworks. It can be argued that in the Tik Tok example, Pasquinelli and Joler’s call for intervention is precisely what’s happening. Community interaction with technology to create algorithmic art, at a very low barrier of access. 

Figure 6 Screenshots of User @acrxhd. Posts with AI generated viuals. TikTok. Accessed October 27th, 2021.https://vm.tiktok.com/ZM8antDqJ/

We all experiment a little…

ArtBreeder 

An example of a Generative Adversarial Network, is the website Artbreeder. It uses BigGAN and StyleGAN models, and strives to be a new type of creative tool that encourages people to cooperate and explore their ideas. Initially known as Ganbreeder, it began as a test to breed and collaborate techniques of exploration in high-complexity areas of AI. This experimental website allows the use of machine learning with image combination, and permits users to create different types of images to their own personal customization. The range of image types varies from portraits, landscapes, buildings, painting, sci-bio art, characters, albums, furries and anime portraits. With a free account one can pick from the options mentioned and, slightly change the settings that the tool offers and create an image based on the inputs the user decides upon. 

Artbreeder is becoming a popular tool for people to incorporate AI to make art and visuals. It can act as a site of intervention for AI creation as it has a low barrier of entry and offers various options. The possibilities of Artbreeder are endless and the results are diverse, creative and often times aesthetically pleasing and even emotion-evoking. This proposes a lot of questions within the proposed debates. A relatively new and obscure artform like AI art, is suddenly becoming ‘common’. This questions how we define art and how art and how we critique it. It makes us think about how we can define art and the importance of the process versus the result. Other projects included in the research were funded differently, did not offer much open access in the process and are connected to institutions. Tools such as this one has the potential to make AI and new media art even more participatory as its digital aesthetic has a certain kind of ‘known’ feel to it. Lastly, it questions the reimagination of art as a specific style and how we position it within the more classical and longer existing forms of art.

Figure 7 Screenshot from a created character that was built with the tool by Jeanelle Grech

Figure 8 Screenshot from an exploration of a portrait creation by Goran Kusić

And for the grand finale…

The chosen examples and experimented artworks have shown the sheer versatility and potential of using algorithms, datasets and artificial intelligence for artistic production. The debates presented in the paper concerning new-ness, authorship and technological agency do not have any one absolute response. Moreover, the deeper one delves into exploration of AI and algorithms in new media art, the more questions arise and additional layers provide additional challenges. However, there is a certainty: the potential of creative expansion is practically limitless. What needs to be taken into consideration are the dimensions of technological materiality and their embeddedness into their contexts. To see AI and algorithms as possible extensions of creativity as co-producers, is a starting point for further examination of the creative potentials and theoretical and practical conundrums concerning their use. With special attention to where and how the AI and algorithms are employed in new media art production; their practically limitless potentials can be treated as the next frontier in digital creativity. 

Nobody ever reads this, but you really should, the sources are great!

boyd, danah, and Kate Crawford. 2012. ‘Critical Questions for Big Data’. Information, Communication & Society 15 (5): 662–79. https://doi.org/10.1080/1369118X.2012.678878.

Broeckmann, Andreas. 2016. Machine Art in the Twentieth Century. Cambridge, Massachusetts London, England: The MIT Press.

Crawford, Kate. 2021. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven: Yale University Press.

Elgammal, Ahmed. “Meet AICAN, a Machine That Operates as an Autonomous Artist.” The Conversation. October 17th, 2018. https://theconversation.com/meet-aican-a-machine-that-operates-as-an-autonomous-artist-104381.

Garcia, Chris ‘Harold Cohen and AARON—A 40-Year Collaboration’. 2016. CHM. 23 August 2016. https://computerhistory.org/blog/harold-cohen-and-aaron-a-40-year-collaboration/.

Jochim, Beth ‘A Short Overview on AI Art’. 2021. Libre AI. 24 May 2021. https://www.libreai.com/a-short-overview-on-ai-art/.

McLuhan, Marshall. 1994. Understanding Media: The Extensions of Man. Cambridge, MA, USA: MIT Press.

Miller, Arthur I. 2020. The Artist in the Machine: The World of AI-Powered Creativity. Cambridge, Massachusetts: The MIT Press.Apple Books. 

Pepi, Mike. 2020. ‘How Does a Human Critique Art Made by AI?’ ARTnews.Com (blog). 6 May 2020. https://www.artnews.com/art-in-america/features/creative-ai-art-criticism-1202686003/.

Peters, John Durham. 2015. The Marvelous Clouds: Toward a Philosophy of Elemental Media. 1st edition. Chicago ; London: University of Chicago Press.

Schlackman, Steve. “Who Holds The Copyright in AI Created Art.” Art Business Journal. September 29th, 2020. https://abj.artrepreneur.com/the-next-rembrandt-who-holds-the-copyright-in-computer-generated-art/.

‘Scrying Pen by Andy Matuschak – Experiments with Google’. n.d. Accessed 28 October 2021. https://experiments.withgoogle.com/scrying-pen.

Simon, Joel ‘Artbreeder’. n.d. Accessed 27 October 2021. https://artbreeder.com.

‘The Nooscope Manifested: AI as Instrument of Knowledge Extractivism’. n.d. The Nooscope Manifested: AI as Instrument of Knowledge Extractivism. Accessed 24 October 2021. http://nooscope.ai/.

User @acrxhd. Posts with AI generated viuals. TikTok. Accessed October 27th, 2021. https://vm.tiktok.com/ZM8antDqJ/

Facebook, now a common name in communication all over the world, faced a complete shutdown of services on October 4, 2021. This outage reflected on the entire “Facebook Family” of applications, including Instagram, Messenger, Whatsapp, Mapillary and Oculus. The shutdown lasted for about six to seven hours and effectively cut down the use of what we now consider essential modes of communication. This ordeal brought about a wave of domino effects, whether it was a hit on small businesses, Facebook’s stock price, or just a complete stop to virtual communication for many. Moreover, the infrastructural fallout affected not only the end-product consumers of the media outlet and its branches, but also the employees and producers at the company. It was reported that even Facebook employees had trouble accessing the mainframe systems at the main offices of Facebook, and weren’t able to enter the premises or even make calls on company issued phones. 

News outlets and users were quick to address the issue through their respective channels with the latter taking to Twitter to express their frustration or opinion on the matter. While the reaction to the event was, for the most part, lighthearted with users posting memes and poking fun at the whole event, it seems that some impactful consequences arose as a result of the outage (Jiménez and Patel). In particular, users were made more aware of the underlying systems that kept Facebook functioning. This infrastructure is invisible, as its designers purposely obfuscate its presence from the overall user experience (Bowker and Star 230). In the particular case of the Facebook outage, the material infrastructure that facilitates data flow and keeps the platform working was compromised. In such cases, the infrastructure renders itself visible and vulnerable to analysis and criticism (Bowker and Star, 231). 

As such, the case study of the Facebook outage is an interesting one, because the infrastructure was made materially visible to users on a massive scale. We take from the definition for infrastructure as proposed by Bowker and Star, taking into account its flexibility to respond to outside forces, its embeddedness within socio political and economic spheres, and the implementation of the aforementioned forces into its technical design and execution (231, 232). This paper therefore takes this incident as its case study by 1) gauging the general sentiment towards Facebook and its platforms and 2) determining the potential for such infrastructural malfunctions to make way for critical introspection. 

With Facebook’s own discourse made so vulnerable during this event, individual users were handed an opportunity to position themselves in a way that allowed them to criticize the platform as well as their own habits. Such an opportunity is rare and we believe that the potential this carries is wide and far-reaching. A collective awareness of the infrastructural makeup of Facebook is powerful insofar as it allows for critical intervention on a global scale, challenging the hegemonic discourse and allowing for alternative opinions on Facebook’s management and accepted cultural practices to be heard. Casual users and social media professionals were united in their shared incapacity to access their desired data. This paper will explore these effects by amalgamating a general impression of the public opinion through news media coverage and tweet content. 

Methodologies 

As mentioned, we have decided to use Twitter, because apart from Facebook and the platforms they own, it is one of the largest social networking sites. Also, Twitter was “flooded with users”, as a result of the Facebook outage (Timsit and Mateus). Through looking at news outlet’s response to the outage, we try to build an argument about the implication that the media makes about Facebook following the outage, as well as the current disposition concerning Facebook in general, and how this is reflected through this infrastructure disruption

After, based on this movement from Facebook’s platforms to Twitter, we hypothesise that an audience of reflective users that mention the effect of the Facebook outage in their tweets as well as engagement behavior (specifically retweets). Consequently, we will look at the top tweets that are most engaged with (retweeted) mentioning the #facebookdown tag during the outage, since this most accurately represents the disposition of the users concerned with the Facebook outage during that time. Though it would be interesting to compare the tweets before, during and after the outage, our main goal is to look into the direct social response to the disruption of a global online infrastructure.

Additionally, we decided to use discourse analysis for the tweets that mention the #facebookdown, a method that has been used in similar studies (Cravens et al. 374; Dambo et al. 6). Even though hashtags based on breaking events might not have settled (Bruns and Burgess, 5), the Twitter community was quick to develop an ad hoc public surrounding the event, going beyond the “follower” networks. We are considering the fact that not all tweets about the outage contain the hashtag, but we are confident that all posts with the hashtag concern the outage, meaning our corpus is focused, efficient and to an extent representative.

Using Twitter’s built-in advanced search method, we found our corpus of tweets by looking for items containing the #facebookdown tag with at least 500 retweets, that were posted between Sunday to Tuesday, October 3 and October 5, 2021 since the outage took place on Monday October 4. Unfortunately, Twitter does not allow it’s users to sort the tweets they looked for by date or popularity, meaning it is sorted by their algorithms, causing a potential lack in representability in the order of the tweets that are shown. However since the amount of tweets is relatively small, this should not matter for our research. Through trial and error, we found that a minimum of 500 retweets strike a good balance between publicity and volume of tweets that were shown. 

In order to ensure the privacy of the users posting the tweets we are analyzing, we will not be disclosing any information that could identify the users. Though all tweets analyzed are public, we try our best to ensure the privacy of all non-public figures posting with this hashtag. Following Boyd and Crawford‘s principle (672), it is not ethical to collect just because it is accessible.

News Coverage 

Much of the general sentiment towards the event can be felt by looking at the media coverage by news platforms during this time. Major outlets capitalized on the opportunity to not only cover the most trending story of the day but also to make use of the situation’s potential to spark political conversation. In a New York Times op-ed, attention was taken towards Facebook’s infrastructural design, criticizing its ethical validity and overall direction (Bokat-Lindell). For instance, the article mentions how the platform’s algorithmic infrastructure prioritizes user action that is conducive to benefiting the company’s profit with callous disregard to the user’s wellbeing. In an interview with CBS, former Facebook employee and whistleblower Frances Haugen stated: “The thing I saw at Facebook over and over again was there were conflicts of interest between what was good for the public and what was good for Facebook. And Facebook, over and over again, chose to optimize for its own interests, like making more money” (Pelley). While this statement was given independent of the outage that took place, it seems that the outage proved to be the perfect opportunity to highlight and make these types of opinions heard.  

This is in reference to the company’s approach to their platforms’ infrastructure. It would seem that despite the flexibility that infrastructures afford, Facebook chooses to prioritize functionalities which increase user engagement so as to maximize the profitability and value of the platforms’ ad space. In Facebook’s quarter-by-quarter breakdown of their revenue from the 2020 calendar year, it was estimated that a grand total of 84-billion dollars was made from the platforms’ advertising (Facebook). Naturally, with advertising the largest source of income, much of the company’s effort targets user engagement and attention. 

As such, much of the design that is incorporated into the platform is targeted towards this objective. The algorithmic infrastructure which filters content based on personal interest does much for this enterprise. An expansive body of research has been dedicated to the negative impacts of such an infrastructure on user well-being (e.g. Chatzopoulou et al. 1289). Moreover, infrastructure standardizes interactivity and by extension user experience and habit (Bowker and Star 234). Instagram’s and Facebook’s algorithm depend heavily on “Likes” as a quantitative tool of gauging interest and keeping track of information flow. Research done on this infrastructural element shows the harm it has on user wellbeing, in particular to affect and loneliness by quantifying popularity and social acceptance (Wallace and Buil, 4). As a result, users might feel alienated or not accepted by the response that they receive. 

Such issues clearly derive from the platform’s infrastructural backbone. The Facebook outage undermined the invisibility of these infrastructures and helped foster a critical discourse. Not only were casual users’ habits interrupted, but social media managers, ‘influencers’ and professionals who depended on the platform for their livelihood had to put their projects on halt. During this time, much debate arose surrounding Facebook’s infrastructure and questions were brought up as to current practices associated with social media and their psychological effect on its users. Due to the infrastructure’s malfunction, it rendered itself visible, thereby making Facebook vulnerable to criticism.  While the major news outlets used this opportunity to make existing concerns heard, individual users took to Twitter to voice their own opinions and personal experiences. Using discourse analysis, we decided to look at whether the opinions shared by the news outlets were also reflected in the general discourse of the event. 

Discourse analysis

Due of the significant number of people that Facebook and its owned services reach, many people quickly noted its outage and proceeded to discuss it on other platforms, which represents a kind of collective awareness of the infrastructural disruption. Serving as one of the major means of communication, Twitter was then under the spotlight, where they said “hello” to “literally everyone” (Figure 1). The current situation brought users to the same platform, which potentially caused more convenience in communication, but also revealed their problematic heavy reliance on so many apps from the world’s largest social media company at the same time. 

Hence, we selected some of the most retweeted tweets and search with the hashtag #facebookdown between October 3 to October 5 to figure out the most related and time-effective posts. Through this, we try to understand how the users behind those tweets roughly think about and joke about the outage. We did this through discourse analysis of those tweets including the hashtag #facebookdown.

Each individual has very different angles and concerns about the breakdown. We simply analyze posts from two categories: users and public figures or companies to better understand their attitudes and how they view the problem from different perspectives.

Users across the platform were generally playful with memes about Facebook, Instagram, and WhatsApps icons (Figure 2). The theme of most posts is funny and joyful, engaging in banter and generally lightening the mood of the public. Jokes and venting ensue during the event. Some even teased those who work in social media with GIFs of busy electricians, who seem to have no ideas to deal with technical problems. Moreover, one played jokes on Netflix’s new hit show Squid Game (Figure 3), showcasing that Facebook and its family of apps (contestants) are hit by Twitter (the machine doll). Furthermore, in the same series, the broken pieces represent the losers, and only Twitter’s logo is intact (the winner) (Figure 4).

Facebook CEO Mark Zuckerberg appeared to be in the center of the event and from the public view. His personal wealth dropped by nearly $7 billion in a matter of hours since shutdown and became the 5th richest man in the world, said by social entrepreneur Chief Erican @EricanSA on Twitter (Figure 5). Many social media users also posted pictures of ‘Zuckerberg’ getting tangled in the wires (Figure 6).

Facebook’s outage soon makes everyone online and be more concentrated without some apps’ distraction. In this way, some of the big companies, organizations, and brands thought it was a great opportunity to market their products and services via Twitter. Several company’s official accounts quickly tweeted to check in with their customers and even run Twitter’s giveaways to increase the engagement, such as entering a prize draw or posting a new release. Chasing after the hot trend, WHO also posted that, “Never let your masks down” to catch attention (Figure 7). Meanwhile, politicians like Josh Shapiro (@JoshShapiroPA), the attorney general of Pennsylvania, used for political purposes, called for actions by saying in a tweet that, “Now that you’re all here, I want to take this opportunity to say: Raise the minimum wage. Defend the right to vote. Protect abortion access.”

The shutdown had sparked heated discussions and divisive conversations on specific causes of this issue among professionals and social media users. The company did not disclose the cause of the outage. Six hours later, they attributed the downtime to a “server configuration change.” 

Computer experts and professionals made thoughtful assumptions and offered possible solutions. Before the outage, Facebook has already experienced heavy criticism. One of its main and most famous critics, Edward Snowden (@Snowden), the president at Freedom of Press and a former computer intelligence consultant, tweeted that: “Facebook and Instagram go mysteriously offline and, for one shining day, the world becomes a healthier place.” “You and your friends should probably be using a more private, non-profit alternative like @Signalapp anyway (or another open-source app of your choice). It’s just as free, and takes like 30 seconds to switch,” he added. Also, the Domain Name System (DNS) and DDoS attacks on providers or servers are most likely to be interpreted as the symptom of hours-long inaccessibility of Facebook and its owned platforms (Figure 10). For instance, Minds (@minds) retweeted 60 Minutes’ (@60Mintes) post to relate the issue to the company whistleblower (Figure 11).

Discussion & Conclusion

Following our analysis of the Facebook outage, it is clear that while this event lasted only for about 6 hours, it spread havoc over a variety of fields. An event of this magnitude shook the functioning of the tools of new media and communication to such an extent that it barred users from even logging on. As mentioned earlier, it was rather interesting to speculate over the concerned infrastructure and upon analysing it closely in this essay, we can safely say that there were significant effects reflective in the functioning of the platform.

Firstly, we delved into coverage by news outlets such as the New York Times and CBS News to analyse the reflection of an event this significant on journalism and general reportage. The news coverage showed a critical reflection the Facebook algorithm, both through items that discussed the outage itself, as well as coverage critical of our dependence on the concerned infrastructure itself.

Given the nature of our research, we were able to find user-generated data from Twitter, and were able to finally hypothesize not only that users visibly switched to an alternate platform but also their reception to the shift in media sites. It also stands relevant to Bowker and Star’s theory on Infrastructures as it breaks down the function of user experience on a media platform and habit. To conclude, we can safely state that while the Facebook outage struck out as a haptic break for several users and businesses (including Facebook itself), it also proved to be an outlet for users to post their opinions on an alternate media site and uncovered the infrastructural makeup of our usage habits. 

Though we got a general impression of the response to the Facebook outage, our tools of collecting the data were limited, given that we were not granted access to the Academic Twitter API. This meant that we could not collect actual data from Twitter, and we had to use the ‘advanced search’ provided in the general user interface of Twitter. This caused the tweets to be sorted algorithmically by twitter, jeopardizing the representability of the content we analyzed. Also, because we had to use this method, we could not sort by time, which would have helped in analyzing the disposition of users as the outage progressed. These are all issues that could be resolved in future work. 

Word count: 2744

Bibliography

Bokat-Lindell, Spencer. “Opinion | Facebook Was down for a Few Hours. Should It Go Away Forever?” The New York Times, 5 Oct. 2021, www.nytimes.com/2021/10/05/opinion/facebook-whatsapp-instagram.html.

Bruns, Axel, and Jean Burgess. “The Use of Twitter Hashtags in the Formation of Ad Hoc Publics.” Proceedings of the 6th European Consortium for Political Research (ECPR) General Conference 2011, The European Consortium for Political Research (ECPR), 2011, pp. 1–9.

Chatzopoulou, Elena, et al. “Instagram and Body Image: Motivation to Conform to the ‘Instabod’ and Consequences on Young Male Wellbeing.” Journal of Consumer Affairs, vol. 54, no. 4, Sept. 2020, https://doi.org/10.1111/joca.12329.

Cravens, Jaclyn D., et al. “Why I Stayed/Left: An Analysis of Voices of Intimate Partner Violence on Social Media.” Contemporary Family Therapy, vol. 37, no. 4, Springer, 2015, pp. 372–85.

Dambo, Tamar Haruna, et al. “Office of the Citizen: A Qualitative Analysis of Twitter Activity during the Lekki Shooting in Nigeria’s# EndSARS Protests.” Information, Communication & Society, Taylor & Francis, 2021, pp. 1–18.

Facebook. “Facebook Reports Fourth Quarter and Full Year 2020 Results.” Investor.fb.com, 27 Jan. 2021, https://investor.fb.com/investor-news/press-release-details/2021/Facebook-Reports-Fourth-Quarter-and-Full-Year-2020-Results/default.aspx 

Jiménez, Jesus, and Vimal Patel. “Users Turn to Twitter after Facebook Outage. Jokes and Venting Ensue.” The New York Times, 4 Oct. 2021, www.nytimes.com/live/2021/10/04/business/news-business-stock-market. 

Pelley, Scott. “Whistleblower: Facebook Is Misleading the Public on Progress against Hate Speech, Violence, Misinformation.” Www.cbsnews.com, 4 Oct. 2021, www.cbsnews.com/news/facebook-whistleblower-frances-haugen-misinformation-public-60-minutes-2021-10-03/. 

Star, Susan Leigh, and Geoffrey C. Bowker. “How to Infrastructure.” Handbook of New Media: Social Shaping and Social Consequences of ICTs, edited by Leah A. Lievrouw and Sonia Livingstone, SAGE Publications Ltd., 2002, pp. 230–45.

Wallace, Elaine, and Isabel Buil. “Hiding Instagram Likes: Effects on Negative Affect and Loneliness.” Personality and Individual Differences, Nov. 2020, p. 110509, https://doi.org/10.1016/j.paid.2020.110509.

Timsit, Annabelle, and Sofia Diogo Mateus. “‘Hello Literally Everyone’: Twitter Flooded with Users during Facebook, Instagram Outage.” Washington Post, 2021. www.washingtonpost.com, https://www.washingtonpost.com/technology/2021/10/05/twitter-users-facebook-outage-instagram-whatsapp/.

Twitter Posts and Figures. Retrieved by Oct. 29, 2021. https://twitter.com/search?q=(%23facebookdown)%20min_retweets%3A500%20until%3A2021-10-05%20since%3A2021-10-03&src=typed_query&f=top

Games becoming a staple of daily life, gamers have come to attach much greater importance and meaning to the virtual worlds they inhabit with their avatars, as they identify with them to a certain extent. So much so that, they are willing to purchase “virtual items” for their in-game characters, i.e. things in games that are of purely cosmetic nature -such as skins, clothes or accessories- and have no functional value; that “do not grant any advantages to their owners in the gameplay.”

In order to begin to get a sense as to why people would spend real life money on items that are merely decorative, we shall first try to understand what games and gaming means to people. Sheth et al. in their 1991 article have theorized that the choices of individuals are driven not only by functional value; but also social, emotional, epistemic and conditional values . It is therefore apparent that our behavior as consumers is also triggered by motivations other than pure function. Park and Lee have built upon this premise, and examined in-game purchases “in order to investigate online game users’ perceived value of purchasable game items”

People’s main motivation to purchase virtual items is to further their enjoyment of the game in question. Virtual items such as skins can enhance the player’s immersion in the virtual world the game takes place in, as they generally offer higher production quality possibly with better special effects. Another major reason is that separately sold virtual items provide a perceived increase in prestige and respect commanded in others; it serves as a status symbol, signaling that the bearer of the item is an important member of the community. They might also signal skill level, implying that the time or money invested to get the special item must indicate how good the player is in that game. Players may also choose to acquire exclusive cosmetic items to differentiate themselves from the crowd and express their individuality.

Fortnite is a prime example of today’s video game prominence. With a total registered player base of over 350 million and a record of 15.3 million concurrent players, it stands today as a pinnacle of online multiplayer gaming. Not only by virtue of its player base, but also through ground-breaking collaborations with celebrities, such as musicians giving virtual concerts within the game to record audiences. Fortnite manages to, more than merely staying relevant, push boundaries and maintain its integral place in pop culture.

The popularity of Fortnite skins seems like an indicative factor of Fortnite’s success as a free-to-play video game. For this reason, we aim to make a critical intervention on the phenomenon of Fortnite skins by performing a visual analysis of several Fortnite skins. In turn, our goal is to relate our findings with current debates and topics within the field of academic research on video game consumption and its implications.

To explore the questions, we attempt to look at the skins on Fortnite through direct visualization. Direct visualization, as introduced by Lev Manovich, is data “reorganized into a new visual representation that preserves its original form.” Instead of abstracting the skins into graphic signs or reduced charts, we aim to employ direct visualization as a tool for the intervention.

Up till October 2021, there are altogether 1,191 skins that are released or to be released on Fortnite. The skins consist of all specified and non-specified genders, and are inspired by various categories such as sports, animals, daily objects, mystic creatures, holidays, robots, celebrities, aliens and more. While a major part of the skins depict fantastical human figures in an array of imaginary outfits, many are essentially non-human or anthropomorphized characters. Mythical, eccentric, and utterly impossible in real life, these characters therefore highlight the virtual aspect of video games and serve as a reminder that the cosmetic items are ultimately infeasible in the players’ regular daily lives. Hence, out of 1,191 Fortnite skins, we selected every food-inspired skin and generated a direct visualization as an exemplary category that represents all of the designs. Through a thorough intervention of the direct visualization, we venture into the patterns of these designs, and analyze how the skins could continue to encourage and motivate virtual cosmetic purchases.

Figure 1. Direct Visualization of All Food Inspired Skins on Fortnite. Generated on October 26, 2021.

In the direct visualization above, the food inspired Fortnite skins are shown in their original forms with all details preserved. In the meantime, they are completely taken out of the contexts of the gameplay and thereof displayed as individual characters. When laid out in the direct visualization, it becomes immediately apparent that all the skins are still designed upon the base of a human form, while certain physical parts are either realistic representations of edible items, or completely covered in distinct patterns that are archetypal of the particular item. For instance, “Crustina” [Fig. 3] is a female character with the head of a grinning tomato. While her arms, torso and legs remain human, she is dressed in a red-and-white-striped vest and green trousers that are stained with tomato sauce. The image on her vest indicates that “Crustina” is in fact inspired by pizza. On the other hand, “Dappermint” [Fig. 4] is a candy cane that conveniently takes the peppermint candy’s original shape as head and torso, with human arms and legs attached to the character. In Buying the Unreal: Drivers of Virtual Item Purchase in Video Games, Syahrizal et al. discuss several factors in the playability of a game, one of which is attractiveness, that is “defined as the pleasure and satisfaction of players with aspects of the game, which are of interest and gain the confidence of the player. It increases the feeling of presence and enjoyment and creates higher levels of engagement.”

Figure 3, Crustina. Accessed on 26 October 2021.
Figure 4, Dappermint. Accessed on 26 October 2021.

As confidence and the feeling of presence comes to discussion, the food inspired skins become an intriguing subject – a tomato head or a candy cane body will not necessarily make a person more attractive or provide a self esteem boost in everyday life. Moreover, they do not grant any direct advantages in the overall gameplay. Nevertheless, the skins become desirable items sold for real money in a virtual world. To understand some aspects behind the motivation of cosmetic purchases, direct visualization can offer some clues. When taking a closer look, it is perceptible that 93.9% of the food inspired skins are depicted without a racial or ethnic representation, with only two of them depicted with a customizable human face (“Gingerbread Raider” and “Guernsey”). In Getting into the Game, Casey Hart examines the personalities of gamers and their projection in the video game avatars. In his findings, Hart discusses that “Subjects appear to use video games in order to experience alternate personalities, instead of projecting actual or ideal self into their avatars” and that “players may use video games more for escapism than projection of self.” In correlation to the direct visualization of Fortnite’s food inspired skins, it is arguable that these particular skins are a drastic and unusual step in representation of oneself in a game. While the color of the skins are already tunable in Fortnite, most of the food inspired skins completely omit the concept of race and ethnicity. The players therefore have the option to express themselves creatively, and freely choose their identities in ways that are not affected by stereotypes or cultural backgrounds – a nearly impossible mission in the real world. With the increasing awareness and global discussions of the racial conflicts and non-binary popularities, these sexually ambiguous and racially obscure skins might also be a correspondence that reflects these issues.

Upon closer inspection, it is also noticeable that one skin can develop into multiple enhanced versions and alternatives. “Peely” is a popular skin in Fortnite that was first released in February 2019. The skin depicts a slightly peeled banana with a wide set of beady eyes and a smile. The original “Peely” is not dressed or accessorized in any special costume. Yet through the direct visualization [Fig.1], “Peely”’s character forms a recurrent pattern, as the skin evolves into “P-1000, “Agent Peely”, “Peely Bone”, “Potassius Peels”, and “Unpeely”. [Fig. 4]

Figure 4, Peely and variants of Peely. Generated on October 26, 2021.

The “Peely” character in the game narrative goes through various phases and encounters challenging scenarios, as revealed by a special Loading Screen called The Ripening Ritual in Fortnite’s gameplay. In Season 9, “Peely” was blended into a smoothie, and was able to develop into “P-1000”, a namesake of T-1000 in Terminators, after dressed in a robot suit designed by another character in the game. “Peely Bone” was also introduced in Chapter 2, followed by Agent Peely who could dress in a secret agent black suit. Afterwards, in Season 3 of the Chapter, “Peely” peels himself off, and enjoys the summer dressed in beach shorts and a straw hat as “Unpeely”. In An Odyssey into Virtual Worlds, Animesh et al. argue that “participants’ interaction with a virtual world is enacted through the actions of their avatars.” Thus, the pattern of “Peely” in the direct visualization illustrates the interactivity of the gameplay behind the skins – as the basic skin experiences a myriad of events, the players who identify with the skin immerse themselves in the virtual adventure as well and follow the narrative of Peely himself. The purchases of the “Peely tribe” then transform into an extension of a self development journey, which will bring more pleasure as the collection completes.

Before we can navigate towards any conclusion drawn from the visual analysis it is important to shed light on how the game mechanics of Fortnite are designed. Fortnite is a free-to-play game with no performance-enhancing items available in its in-game shop. This implies that everything that can be bought in the shop is purely cosmetic and only offers personalization and customization of one’s in-game character, including all the skins. As shown in the direct visualization [Fig.1], skins are categorized in different colors. These colors indicate the rarity of the skins ranging from grey (common), green (uncommon), blue (rare), purple (epic) to orange (legendary) with the occasional exception of a different color used for skins in special events.

A basic understanding of how skins are attained demonstrates a key element in Fortnite’s success in selling cosmetic items; through scarcity marketing. First, skins can be bought directly using V-bucks, Fortnite’s in-game currency which can be bought in exchange for real-world currencies. 1000 V-bucks cost around eight euros and the average skin price varies between 800 and 2500 V-bucks. Secondly, players can purchase a Battle Pass that enables them to participate in daily challenges and special events through which V-bucks or skins are rewarded. The latter approach of collecting skins exemplifies the way in which Fortnite’s developers, Epic Games, create the demand for skins. Epic Games employ a strategy of artificial scarcity where particular skins are only available in limited quantities, indicated with the earlier-mentioned color code system, or within a limited time frame in which one must complete a challenge to unlock or purchase them before the possibility passes. The fear of missing out on skins increases the interactivity and engagement of players. Moreover, the achievement of unlocking skins adds an extra layer of enjoyment to the gaming experience that plays a role in the meaning and significance of Fortnite skins.

Our main objective for this intervention was to relate our findings in the visual analysis of Fortnite skins to existing debates and potential topics of research in academia regarding the phenomenon of cosmetic items. Through direct visualization, we were able to categorize, depict and highlight aspects that relate to various topics of research that allow for further exploration. 

While this article does not aim to focus or elaborate on marketing tools, we found it relevant to mention this factor relating to the experience of skins. As explained above, the way in which Fortnite is designed contributes a lot to the value and experience of engaging with skins. The act of earning skins is directly connected to a player’s engagement with the game as it requires the player to be proactively involved in order to participate in challenges and events. This adds to the game’s experience as skins are to some extent indicators of achievement and dedication. 

An important finding we like to emphasize in this article is the notion of identity forming through digital representations of the self. In 1988 R. Belk coined the term of the “extended self” for the way in which we use technological devices to disembody ourselves and engage in communication and interacting through different representations. With technological developments, the concept of the extended self naturally expanded into the digital realm where we find ourselves with limitless possibilities of self representations through social media and avatars. 

“In the digital world, the self is now extended into avatars, broadly construed, with which we identify strongly and which can affect our offline behavior and sense of self.”

This idea strongly relates to the ways in which Fortnite allows a player to choose from a wide range of skins with the possibility to change into anything. One’s digital representation is not limited to gender, race, or even to living beings as showcased in the various food-related skins. The limitlessness of virtual characters embodies a certain liberation one can experience from engaging in a digital environment that is as immersive as the one of video games. Skins enhance the immersion of players into the virtual world of video games, therefore, it can be interesting to examine the psychological or social implications this can bring to light.

Bibliography

Andy Syahrizal et al., “Buying the Unreal: Drivers of Virtual Item Purchase in Video Games,” in Proceedings of the 3rd International Conference on Software Engineering and Information Management (ICSIM ’20: The 3rd International Conference on Software Engineering and Information Management, Sydney NSW Australia: ACM, 2020), 205.

Animesh et al., “An Odyssey into Virtual Worlds: Exploring the Impacts of Technological and Spatial Environments on Intention to Purchase Virtual Products,” MIS Quarterly 35, no. 3 (2011): 792.

Belk, Russell W. “Possessions and the Extended Self.” Journal of Consumer Research 15, no. 2 

(September 1988): 139. https://doi.org/10.1086/209154.

Belk, Russell W. “Extended Self in a Digital World: Table 1.” Journal of Consumer Research 40, 

no. 3 (October 1, 2013): 477–500. https://doi.org/10.1086/671052.

Casey Hart, “Getting into the Game: An Examination of Player Personality Projection in Videogame Avatars,” n.d., 1.

“Fortnite Skins List – All Characters & Outfits!,” Pro Game Guides (blog), accessed October 24, 2021, https://progameguides.com/fortnite-skins-list/.

Gilbert, Ben. n.d. “Video-Game Industry Revenues Grew so Much during the Pandemic That They Reportedly Exceeded Sports and Film Combined.” Business Insider. Accessed 28 October 2021. https://www.businessinsider.com/video-game-industry-revenues-exceed-sports-and-film-combined-idc-2020-12.

Hernandez, Patricia. 2019. “Fortnite Is Free, but Kids Are Getting Bullied into Spending Money.” Polygon (blog). May 7, 2019. Accessed 25 October 2021. https://www.polygon.com/2019/5/7/18534431/fortnite-rare-default-skins-bullying-harassment.

Hurley, Leon. 2021. “Here’s How Many People Play Fortnite.” Gamesradar. October 13, 2021. Accessed 27 October 2021. https://www.gamesradar.com/how-many-people-play-fortnite/.

“Infographic: Gaming: The Most Lucrative Entertainment Industry By Far.” n.d. Statista Infographics. Accessed October 28, 2021. https://www.statista.com/chart/22392/global-revenue-of-selected-entertainment-industry-sectors/.

Lev Manovich, “What Is Visualisation?,” Visual Studies 26, no. 1 (March 15, 2011): 41.

Mäntymäki, Matti, and Jari Salo. “Why Do Teens Spend Real Money in Virtual Worlds? A
Consumption Values and Developmental Psychology Perspective on Virtual
Consumption.” International Journal of Information Management 35, no. 1 (February
2015): 124–34. https://doi.org/10.1016/j.ijinfomgt.2014.10.004.

“Most Played Games in 2021, Ranked by Peak Concurrent Players.” 2021. Twinfinite. March 18, 

2021. Accessed 27 October 2021. https://twinfinite.net/2021/03/most-played-games-in-2020-ranked-by-peak-concurrent-players/.

Park, Bong-Won, and Kun Chang Lee. 2011. “Exploring the Value of Purchasing Online Game Items.” Computers in Human Behavior 27 (6): 2178–85. https://doi.org/10.1016/j.chb.2011.06.013.

“Peely (Outfit),” Fortnite Wiki, accessed October 27, 2021, https://fortnite-archive.fandom.com/wiki/Peely_(outfit).

Perrotta, Matthew. 2020. “Business Models of Video Games: Past, Present, and Future.” Medium (blog). April 6, 2020. Accessed 28 October 2021. https://medium.com/@mjperrotta46/business-models-of-video-games-past-present-and-future-2b2aafe8ade1.

Pra, Vitor Dal. 2020. “Why People Buy In-Game Items With Real Money?” Medium (blog). July 1, 2020. Accessed 26 October 2021. https://vitordalpra.medium.com/why-people-buy-virtual-items-in-games-with-real-money-and-why-that-matters-7e0cd2199762.

“Ripening Ritual (Loading Screen),” Fortnite Wiki, accessed October 27, 2021, https://fortnite-archive.fandom.com/wiki/Ripening_Ritual_(loading_screen).

Sheth, Jagdish N., Bruce I. Newman, and Barbara L. Gross. 1991. “Why We Buy What We Buy: A Theory of Consumption Values.” Journal of Business Research 22 (2): 159–70. https://doi.org/10.1016/0148-2963(91)90050-8.

Syahrizal, Andy, Betty Purwandari, Muhammad Mishbah, and Muhammad Fadhil Dzulfikar. 2020. “Buying the Unreal: Drivers of Virtual Item Purchase in Video Games.” In Proceedings of the 3rd International Conference on Software Engineering and Information Management, 203–9. Sydney NSW Australia: ACM. https://doi.org/10.1145/3378936.3378948.

Webster, Andrew. 2020. “Travis Scott’s First Fortnite Concert Was Surreal and Spectacular.” The Verge. April 23, 2020. Accessed 27 October 2021. https://www.theverge.com/2020/4/23/21233637/travis-scott-fortnite-concert-astronomical-live-report.

Webster, Andrew. 2021. “Ariana Grande’s Fortnite Tour Was a Moment Years in the Making.” The Verge. August 9, 2021. Accessed 27 October 2021. https://www.theverge.com/2021/8/9/22616664/ariana-grande-fortnite-rift-tour-worldbuilding-storytelling.Wohn, Donghee Yvette. 2014. “Spending Real Money: Purchasing Patterns of Virtual Goods in an Online Social Game.” In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 3359–68. Toronto Ontario Canada: ACM. https://doi.org/10.1145/2556288.2557074.

Introduction

With the arrival of video game live streaming platforms, gaming has become a spectator form of entertainment. Live streaming lets players build an audience, brand, and income while streaming their game practices – often straight from their bedroom, turning the traditional consumer into content creators (Digital Surgeons 2015). Twitch is one of the biggest live-streaming platforms that host massive amounts of live video game content (Taylor 2018). The American video streaming service launched in 2011, blending in two distinct mediums: broadcast media and games. On Twitch, users can construct channels to stream their gameplay and competitions to the world. Thanks to the live comment section, interaction from viewers can take place in real-time. As of February 2020, Twitch has millions of users and a total of 3.8 million unique broadcasters, and it is the fourth-largest source of peak Internet traffic in the US (Delfino 2020).

Twitch offers millions of broadcasters all kinds of opportunities to build online communities and transform their private play into public entertainment. However, underneath it all, technology companies are building and sustaining economic models that allow them to survive (Taylor 2018). On Wednesday, the 6th of October 2021, Twitch became the victim of a data leak, divulging confidential company information. More than 125GB of data leaked on 4Chan, including streamer pay-outs and moderation strike guidelines. (Tidy and Molloy 2021). News9 (2021) described the document as follows: “Twitch’s confidential guidelines for moderation includes pictures and examples of what content should be ‘denied,’ ‘allowed,’ or ‘escalated,’ giving distinctions and rules for what to leave and takedown.”. Another file in the leaked data is the “do_not_ban_list,” which caused a considerable controversy (Tidy and Molloy 2021). This list, filled with popular streamers, gives insight into the inner workings of Twitch, where it seems like certain streamers are given more leeway than others (Nightingale 2021). 

Twitch is an exciting research topic because it is a relatively new yet unexplored medium (Sjöblom and Hamari 2017). Twitch has regularly been under fire from creators and users who feel like the site does not do enough to stop inappropriate content from being shared. The “do not ban list” leak resulted in even more controversy, with users feeling like high earning streamers are treated differently, implying that Twitch’s moderation practices are ruled by money and power. The debate related to Twitch’s moderation practices is not an isolated issue. Many social media platforms are facing problems related to moderation practices. Social media platforms are designed for engagement and, therefore, revenue. Nevertheless, unfortunately, the most engaging content is usually the most controversial (The Economist 2020).

Moderation is a widely studied topic, and it is essential to consider how moderation strategies shape the image of tech companies. Moderation guidelines are usually non-public documents, meaning users have minimal insight into platform content moderation operations (West 2018). This makes it hard for users to understand why certain users are banned or ‘de-platformed’ for breaking the rules and why others are not. This paper will analyze the Twitch data leak to explore if it constitutes a data point of great interest for research regarding the platform economy and governing practices, thus providing future researchers with enough information to warrant further investigations and highlighting possibilities, limitations, and requirements. We will conduct a content analysis of the leaked documents, including: (1) the internal moderation strike guidelines, (2) lists of the highest-earning streamers on the website, and (3) parts of the do_not_ban_list.

Theoretical Framework

Over the past years, online platforms’ moderation practices have garnered growing scholarly attention. From studies ranging from exploring how social media platforms police our online behaviour (Gillespie 2018) to the impact of moderation practices on marginalized groups (Sablosky 2021), the body of work on this subject is still growing. 

As defined by Grimmelman (2015), moderation refers to “the governance mechanisms that structure participation in a community to facilitate cooperation and prevent abuse.” Today, most of us access the Internet through online platforms; being subjected to these mechanisms is no surprise. This moderation is an “essential, constitutional, definitional” practice for platforms (Gillespie 2018). Underlining the prominent role of moderation in the practices of online platforms, Gillespie (2018) even argues that platforms not only need moderation to survive, “they are not platforms without it.” Today, the platforms we use increasingly “determine what users can distribute and to whom, how they will connect users and broker their interactions, and what they will refuse.” (Gillespie 2018). 

However, determining what is and is not allowed has proven to be a difficult task, as platforms have to consider a variety of stakeholders when formulating their rules and norms. Firstly, as most platforms operate globally, one has to consider that what is considered unacceptable in one country might be accepted in another. Besides that, one common characteristic of online platforms is that they serve multi-sided markets. Facilitating end-users and advertisers and sometimes other stakeholders, companies must keep all these different groups in mind when shaping their moderation protocols. As stated by Myers West (2018) when it came to the moderation practices of social media platforms: “although a social media company may have an interest in free expression that enables users to post as much content as possible, it may not desire the kinds of expression that scare away advertisers. Alternatively, it may seek to balance the need to maintain the perception of being an open platform with demands by governments to police certain kinds of content.” 

The external influences described above play a part in shaping a platform’s moderation policies. As expressed by Gillespie: “Nearly all social media platforms are commercial enterprises, and must find a way to make a profit, reassure advertisers, and honor an international spectrum of laws.” This makes online platforms tread a fine line as they design their moderation mechanism, which becomes apparent when looking at the community guidelines of most of these platforms. The purpose of a platform’s community guideline is to provide users with an overview of the types of behavior and content that are allowed and disallowed. As found by Myers West (2018), however, most of these guidelines “tend to be fairly generic, to allow companies to navigate a core tension in their moderation of content.” This strategic lack of precision in the community guidelines leaves users with “limited insight into the evolving scope of platform content moderations” (Myers West 2018).

These unclear protocols have frequently been the cause of contention between platforms and end-users. The main reason for this is that users who are punished for acting against a platform’s policies are sometimes unaware of what they did wrong. Because of this, users often feel that they are being wrongfully or unfairly punished. With punitive measures ranging from content takedowns to temporary suspensions to being completely banned from a platform, these measures can substantially impact users (Myers West 2018, 4376). In her study of users’ experiences with content moderation on social media platforms, for instance, Myers West (2018) found that respondents acknowledged how these platforms provided them with the public good, granting them access to “support systems of communication that are deeply interwoven with social, political, and economic life.” (4376). The latter is especially the case for Twitch. As many users can garner income from their Twitch streams, being banned from the platform exceeds mere annoyance as it can have a tangible financial impact on them.

Considering the effects of platforms’ moderation policies on users – ranging from social to economic implications – it becomes clear why insight into the hidden moderation mechanisms is crucial. Users deserve to know the exact reason behind their punishment, and this allows reflecting and changing their behaviour to avoid being (repeatedly) penalized. As mentioned by Myers West (2018), besides the community guidelines designed for users, it has been found that platforms also have “non-public documentation that operationalizes the community guidelines at a much more granular level of detail” (4370). Through a study of these more detailed guidelines, one should be able to figure out what platforms expect from their users and draw conclusions as to why some things are allowed and some are not.

Methodology

The research conducted in this paper aims to explore if the twitch leaks constitute a data point of great interest for research regarding the platform economy and governing practices, thus providing future researchers with enough information to warrant further investigations in highlighting possibilities, limitations, and requirements. It does so by doing a content analysis of published documents, including the internal moderation strike guidelines for twitch operators determining how to handle behaviour and content produced by streamers on the website, lists of the highest-earning streamers on the website based on the payouts by Twitch regarding the number of subscribers and ad revenue, and parts of a do_not_ban_list. The content analysis angle is guided by contemporary literature from researchers on the platformization of the web, such as Gillespie and West, giving the investigation a firm stance in the scientific discourse and venturing beyond. In order to understand how the paper approaches the research concerning the twitch leaks, it is essential to understand the circumstances by which the data in this paper was obtained. All information, including screenshots and files used in the research, were made public across the web by journalists, news websites, and Twitch and remain visible as of publishing this paper. Non-journalistic sources or sources not directly related to Twitch in an official capacity posting illegally obtained data from the leak, for example, but not exclusively on social media platforms such as Reddit or the imageboard 4Chan were not used in the creation of this paper. This is due to a few apparent reasons:

1. Legal Uncertainty: It is unclear to which extent the data made public in the wake of the breach is confidential and if engagement with the data through non-official Twitch or journalistic sources may be illegal on the national or international stage. Platforms other than Twitch quickly took down any data traces or sources included in the leak, warranting concern of possible repercussions.
2. Technical Limitations:

2.1 The twitch leak is roughly 125 Gigabyte large, requiring a stable and high-speed internet connection to download all of the information without interruption. Since administrators of various sites are constantly taking down the download links to protect their operations from legal inquiries by Twitch. The remaining links are mostly magnet links allowing one to download the files via torrent; since the peer-to-peer torrent network relies on the internet speed of its contributors instead of large commercial servers, download speed and stability are further reduced.

2.2 The 125 Gigabyte of files are most likely compressed, thus requiring a machine powerful enough and extensive memory to assist in timely processing and saving of 125 Gigabyte that will at the very least double in size. Furthermore, once the data has been uncompressed, the files have to be indexed by the computer in a time-intensive process to be made searchable.

2.3 The data being searchable by the machine still requires the researcher to be computational literate and capable of understanding file formats, techniques of extraction, and processing to make good use of the search function to produce any valuable results. Despite the technical limitations and legal concerns, the data used in the following content analysis of the strike operator guidelines, the do_no_ban_list, and the streamer revenue chart can be regarded as genuine given the reputable sources they have been obtained from. Moreover, the relentless vigor by which hegemonic players of the platform economy are issuing takedown notices and banning digital traces of leaks reminds one of a whistleblower spelling state secrets, giving further credibility to the scope of possible future implications regarding the leaked data; therefore, a first glance at the surface may be obligatory.

Analysis

The Do Not Ban List

The first leaked document to be analyzed is the ‘Do Not Ban’ list. In 2016, this list first surfaced on social media platform Reddit. It includes the usernames of certain Twitch streamers. One of the names on the list, ‘Sarbandia,’ is followed by the statement: ‘(do not ban for literally any reason).’ It could be concluded that different rules apply to different users when it comes to banning. If the user Sarbandia breaches the guidelines established by Twitch, the account would not be banned from the platform, while other accounts would be banned in such cases. A double standard arises when multiple people are treated differently even though they should be treated the same way, which would be the case if Twitch does not apply the guidelines they have set to all their users. 

The ‘Moderation Strike Guide’

The leaked document starts with information on ‘what to deny,’ ‘what to allow,’ ‘what to escalate,’ and ‘other hateful conduct violations.’ This is followed by examples and pictures to clarify the information and a list of hate symbols that are not allowed on the platform. An example of one of these hate symbols is the ‘White Power’ hand sign. As stated by Myers West (2018), platforms could not desire content that scares advertisers away. The symbol of White Power could be recognized as undesirable content, harming Twitch’s advertising revenue. In the leaked document, one could find more information on what is allowed on even more subjects such as adult nudity, porn or sexually explicit material, gambling referral and gore, and other obscene, obscene content. This corresponds with the statements of Gillespie (2018) about how platforms increasingly determine what users can distribute and what content the platform will refuse. This is shown in the Strike Guide by the lists of information on, for example, what to deny and allow.

The Payout List

The last Twitch document that has been leaked is the list of the highest-paid Twitch streamers. The list shows the usernames of the streamers in a column on the left, and in the right column, one can see what they have earned according to the data from the data breach. These are the earnings since September 2019. First place on this list is held by a streamer called Critical Role, which has earned more than 9.6 million dollars. Following up on the second place is Canadian Twitch gamer xQc with 8.4 million dollars. It is important to note that the earnings on the list only include income directly paid by Twitch. Most popular streamers generate even more income through, for example, sponsorship deals or merchandise sales.  An expression of Gillespie (2018) that has been included in the theoretical framework is how nearly all social media platforms are commercial enterprises wanting to profit. If the highest-paid Twitch streamers are banned from the platform because they breach the community guidelines, Twitch likely loses many of their users and, thus, income. This perspective could be a reason for Twitch’s double standard when it comes to banning users. 

Conclusion 

The leaked Twitch documents provide insight into the moderation practices of Twitch concerning their economic interests, exposing the dominant role of the platform. After analyzing the leaked documents, it could be stated that these documents and how Twitch handles their guidelines should be studied in more detail. First of all, the confidential nature of the ‘Strike Guide” raises questions about why exactly this guideline is hidden from users. As pointed out in our theoretical framework, making users aware of the exact reasoning behind their punishment can prove beneficial as it allows them to reflect on and learn from their mistakes, enabling them to prevent further repercussions or an eventual banning. The importance of this has also been underlined by Myers West, who stated that “if the overall objective of a content moderation system is to encourage better behavior on the part of users,” the system which obscures the reason behind the punitive measures fails, as it does not educate users and gives them “no opportunity for engagement with the platform to learn.” (Myers West 2018, 4379). Therefore, future research could investigate the exact motives behind Twitch’s confidential “Strike Guide.” Why do they feel that the users’ are better off without having access to these guidelines? Moreover, what are the contents that the platform is trying to hide from outsiders?

Secondly, the ‘Do Not Ban’ list raises the assumption that some users are treated differently than others. Further research in this subject could investigate whether certain sets of users are being treated more favorably than others and answer why this is the case.

Lastly, the relation between the ‘Do Not Ban’ list and the leaked payout list could also be questioned since there could be economic reasons not to ban specific users from the platform. One could test the statement that Twitch makes its moderation decisions based on commercial values.

Overall, possibilities for future research imply overcoming legal and technical limitations related to the Twitch data as described in the methodology, as well as commanding a certain extent of computational literacy as part of the (research-) effort to be able to conduct an informed and focused investigation of the leaks. 

Bibliography 

Delfino, Devon. 2020. ‘What Is Twitch? All You Need to Know About the Livestream Platform’. Business Insider, 11 June 2020. https://www.businessinsider.com/what-is-twitch?international=true&r=US&IR=T.

Digital Surgeons. 2015. ‘Twitch and YouTube Are Blurring the Lines between Consumers and Content Creators’’. Digital Surgeons, 18 August 2015. https://www.digitalsurgeons.com/thoughts/strategy/twitch-and-youtube-are-blurring-the-lines-between-consumers-and-content-creators/.

Gillespie, T. (2018). ​Custodians of the Internet: Platforms, content moderation, andthe hidden decisions that shape social media.​ Yale University Press. (1-23)

Grimmelmann J (2015) The virtues of moderation. Yale Journal of Law and Technology 17(42): 42–109.

Myers West, Sarah. “Censored, Suspended, Shadowbanned: User Interpretations of Content Moderation on Social Media Platforms.” New Media & Society 20, no. 11 (November 2018): 4366–83. https://doi.org/10.1177/1461444818773059. 

News9. 2021. ‘Twitch’s Confidential Internal Strike Guide for Moderating Content Leaked’. NEWS9LIVE. 11 October 2021. https://www.news9live.com/technology/twitchs-confidential-internal-strike-guide-for-moderating-content-on-the-platform-leaks-125526.

Nightingale, Ed. 2021. ‘’Twitch “do not ban” list used to protect prominent streamers’’. Eurogamer (blog), 19 October 2021. https://www.eurogamer.net/articles/2021-10-19-twitch-do-not-ban-list-used-to-protect-prominent-streamers.

Roberts, Sarah T.. 2014. ‘’Behind the screen: the hidden digital labor of commercial content moderation’’. PhD Thesis, University of Illinois, Chicago, IL. https://www.ideals.illinois.edu/bitstream/handle/2142/50401/Sarah_Roberts.pdf?sequence=1&isAllowed=y

Sablosky, Jeffrey. 2021. ‘“Dangerous Organizations: Facebook’s Content Moderation Decisions and Ethnic Visibility in Myanmar”’. Media, Culture & Society 43 (6): 1017–42. https://doi.org/10.1177/0163443720987751. 

Sjöblom, Max & Hamari, Juho. 2017. ‘’Why do people watch others play video games? An empirical study on the motivations of Twitch users’’. Computers in Human Behavior. 75. 985-996. 10.1016/j.chb.2016.10.019.

Taylor, T. L. 2018. Watch Me Play: Twitch and the Rise of Game Live Streaming. Princeton University Press.

The Economist. 2020. ‘Social Media’s Struggle with Self-Censorship’’. The Economist, 24 October 2020. https://www.economist.com/briefing/2020/10/22/social-medias-struggle-with-self-censorship.

Tidy, Joe and Molloy, David. 2021. ‘Twitch Confirms Massive Data Breach’. BBC News, 6 October 2021, sec. Technology. https://www.bbc.com/news/technology-58817658.

Amina Mohamed- aminanasree@gmail.com

Geneviève Lemaire – lemaire.gen@gmail.com

Miazia Schüler – miazia.schueler@student.uva.nl

New Media Research Practices – WG4- 5

October 29, 2021

How Online Slacktivism and White Saviorism Undermine Social Movements: KONY 2012 and Black Lives Matter

In the age of ubiquitous connectivity and social media activism, it makes sense that movements can begin, thrive, or die online. Older millennials and Gen Zers have come of age in a world where online engagement is synonymous with real life action, and where political change can start with a hashtag. However, the relationship between meaningful online engagement and slacktivism requires a critical eye, as measuring the impact of a movement can depend on its intentions, organizing methods and outcomes. In 2020, the world watched as Black Lives Matter, an organization which had first come into existence in 2013 after the acquittal of George Zimmerman, morphed into a global movement. However, while it is decidedly more robust in its activity, it is not the first group to experience that transition. In March 2012, Invisible Children launched the KONY2012 campaign that prompted a widespread-if short lived- campaign against an unknown African warlord. 

While both instances demonstrate a form of social media activism that successfully captured global attention, one must ask whether public engagement with activist related hashtags signify engagement with a cause, or is it an example of what Leonard describes as “apolitical consumption”, whereby “participation in such campaigns rarely brings whiteness and the role of white people in propagating violence and inequality into the spotlight” (Leonard 3)? Further, do the patterns illustrated around Black Lives Matter’s #blackouttuesday campaign echo the same trends that surrounded the KONY2012 campaign, namely widespread uninformed engagement and participation? In order to engage with these ideas we must acknowledge the differing popularity of platforms through the years, and as a result, we will examine each phenomenon as it performs on the most popular social media platform of its time. Therefore, we will examine this divide by engaging with the KONY2012 campaign on Facebook – the most popular social media app of the mid 2010’s, and the more current Black Lives Matter’s #blackouttuesday hashtag on Instagram – the most popular engagement app of the moment – as a means to interrogate the ways apoliticism is visually represented on the platform. 

The Rise of Slacktivism

In March 2013, Invisible Children, an organization whose mission is to “end violent conflict and foster thriving ecosystems in solidarity with [the] world’s most at-risk communities” (Invisible Children), published a short documentary to YouTube, profiling the alleged war crimes of  Ugandan war criminal Joseph Kony. The video garnered over half a million views in the first three hours, and within six days became the fastest video to reach one hundred million views on the platform. The premise was to “make [Joseph Kony] visible” (“KONY 2012” 22:56-22:58) which later translated to making him famous (see fig 1) by identifying April 20, 2012 as a global night of action, “cover the night” (26:27), where people would put up posters and other KONY paraphernalia across their respective cities to pressure their governments to address Kony.

Figure 1: Still from KONY2012 Youtube video               

KONY2012 demonstrated one of the earliest and most widespread instances of what has come to be known as “slacktivism…[a] low‐risk, low‐cost activity via social media, whose purpose is to raise awareness, produce change, or grant satisfaction to the person engaged in the activity…[and] eviden[ced] where our digital effort makes us feel very useful and important but ha[s] zero social impact” (Chazal, Pocrnic 8). As KONY 2012 grew in popularity, and as more users shared the video and bought KONY Action Kits – which sold out “almost immediately” (Wilkerson) (see fig 2)- , the video’s virality and humanitarian tone prompted a global movement to #stopKony. 

However, as we will demonstrate later in the paper, the unprecedented amount of online engagement translated into very little, if any, real world impact facilitated by the “low risk, low cost” model of social media participation through Facebook, which, through its like and share features, exacerbated a prominent white saviour-ist mentality that “allowed for the white man’s burden to manifest without having to confront the ideological and real-life tensions that manifest within face-to-face contact between whiteness and its others” (Leonard 6). KONY 2012 highlighted slacktivist tendencies that were – and continue to be- prominent online, and was further facilitated by the saviour-ist framing that both favour a – in this case digitally- participatory public over a nuanced, factually informed public.

Figure 2: Reddit thread discussing donations to KONY2012 

From KONY to Black Lives Matter

While the KONY 2012 campaign was the first of its kind at the time, it illustrated many troubling characteristics that are present in contemporary social media movements. By creating an online first campaign, one that relied on likes, shares, and online donations, KONY2012  “lowered involvement thresholds, preventing participants from deviating too far from everyday life practices or investing too much time, energy or resources to participate”(Chazal, Pocrnic 8 ). Invisible Children’s video claimed that “in order for the people to care, they have to know. And they will only know if Kony’s name is everywhere” (22:11- 22:19). The associative language tying fame and notoriety with care demonstrates a distinct understanding of internet culture, where recognition and attention are a form of currency, but does not set the stage for long term engagement. However, KONY 2012 was also one of the first campaigns to exemplify what would eventually become influencer culture, calling on “diverse and influential culture makers”  (23:49)  to use their platforms to amplify the movement’s message. The campaign called out “celebrities, athletes, and billionaires” as they “have a loud voice and what they talk about spreads instantly” (23:20-23:27). This initial use of celebrity endorsements is directly representative of our current discourse on influencer responsibility. 

This phenomenon is also evident in the #blackouttuesday movement, where influencers and celebrities blindly participated in a popular hashtag to remain relevant without taking the time to understand its purpose. From the Instagram infographic industrial complex to the study of influencer culture and what has come to be known as performance activism and performative allyship, contemporary academic and popular dialogue continues to engage with these intersecting patterns of saviourism and slacktivism as it concerns the evolution of online culture and its real world impact. 

Methodology 

For the comparative study of the Kony 2012 campaign and #blackouttuesday, we will examine and critically analyze various sets of qualitative data, that provide contemporarily appropriate and cross-sectional insights for an efficient approach to systemically answer the research question. Despite small sample sizes and the underlying challenge of measuring success or failures of (ongoing) decentralized movements, we nonetheless manage to accurately capture recurring patterns: White saviours engaging in slacktivism. 

First, we examine Google Trends graphs of movement-specific keywords for stability and consistency. Although Google was not the main platform to host the majority either of the movements online activistic engagement, Google is nonetheless the most popular search  engine which according to van Dijck is a “co-producer” of knowledge (575). Furthermore, we analyze a handful of contemporarily appropriate memes, that were popular and highly circulated following the days after Kony2012 and #blackouttuesday. They were selected from the top rows of intentionally brief Google searches which are ranked according to popularity (van Dijck 577) or from related critical articles. We also address the ratio of on and offline protesters for Kony 2012 rally attendances, referring to local news outlets and a study conducted by the Pew Research Center. Lastly, we identified key Twitter debates and analyzed the individual feed aesthetics of popular Instagram pages both engaging with #blackouttuesday which paint a clear picture of online performativity and physical activistic engagement, or the lack thereof.

Analysis

Google trends

The Google Trends graphs for the keywords “Kony 2012”, “Invisible Children”, “Blackout Tuesday”, and “Black Lives Matter” (Figures 3-6) show the worldwide popularity of the respective google searches in a timeline roughly ranging from of a month before and one year after the respective online protests. The steep spikes and fast declines can be observed across both campaigns and clearly demonstrate a situational interest that rapidly dropped after a matter of days.  

Additionally, the Google Trends reports show that the interest per region largely came from western, colonial, and/or predominantly White countries in North America, Australasia, and Skandinavia. Interestingly, Uganda ranked highest in google searching “Invisible Children”. While the movements did have different goals and mainstream activistic attention is typically situational, a lack of consistency and therefore a lack of stability (especially from the global north), are nonetheless evident and evidently essential for sustainable success of a movement. Fisher criticized Kony 2012, claiming that “Central African violence is the kind of issue that could benefit from a small but passionate and knowledgeable group of people, not from a 50 million-person mob with a 30-minute attention span” (Fisher).

Figure 3: Google Trends Screenshot for “Kony 2012”. Source: Google Trends, accessed 22.10.2021

Figure 4: Google Trends Screenshot for “Invisible Children”. Source: Google Trends, accessed 22.10.2021

Figure 5: Google Trends Screenshot for “Blackout Tuesday”. Source: Google Trends, accessed 22.10.2021

Figure 6: Google Trends Screenshot for “Black Lives Matter” (post-Floyd). Source: Google Trends, accessed 22.10.2021

Offline Engagement

Kony 2012

Following Kony 2012’s rise in March that year, attention quickly declined (as seen in Figures 3-6) and the movement eventually “failed in its attempt to go offline” (Hager). As previously mentioned, the original video reached 100 million views on YouTube. 58% of 18-29 year olds had heard of the campaign and engaged with it through social media (Rainie et al).  An estimated 10 million views came from the UK alone, of which 300,000 signed up to support the campaign and 10,000 users publicly confirmed their rally attendance on Facebook. In Birmingham however, the UK’s second largest city after London, only 35 people showed up (Walker). Similar patterns occurred globally. In Vancouver, Canada 21,000 protesters registered to attend on Facebook, but only 17 people came (Hager). This simple top-down approach hence shows the utter lack of physical engagement and speaks for a high level of performative online engagement. On the other hand, by the time rallies were scheduled, skepticism about the campaign had started circulating the web, also reaching Facebook events’ comment sections (Walker). This was quickly followed by Invisible Children’s founder Jason Russel’s very public psychotic episode which did not contribute to the non-profit’s credibility, either (Hager).

Twitter commentary

Kony 2012

Celebrities ranging from Oprah Winfery, Rihanna to Justin Bieber, came out with endorsements for the campaign, urging their followers to do the same (Flock). By March 7th, 2012, the hashtags #makekonyfamous, #kony2012, and #stopkony were trending worldwide on Twitter (Know Your Meme). By the end of the week, there were an estimated 5 million tweets about the Kony video (Walker). A whooping  66% of the conversation taking place on Twitter was in favour of the anti-Kony campaign (Walker). These statistics show the magnitude and popularity of the movement. However, as journalists began sounding the alarm and busting some of the myths surrounding Kony and Uganda’s situation, the narrative shifted and within a few days, Kony’s 15 minutes of fame were up. People had gone back to their regular programming, forgetting the movement all together. Just like that, a social movement had come and gone within a matter of days. Arguably, the design affordance of platforms such as Twitter and Facebook made it simple for people to like and share the video and other content related to the movement with their social networks. As easy as it was to engage with the movement, as easy as it was to disengage. 

#blackouttuesday 

Celebrities like Lil Nas X, Lizzo, Sadé, and Kehlani, among others, came out on Twitter, criticizing #blackouttuesday (Castillo; Welk). Their critique stemmed from the doubt on the impact this small act would have on the movement. They suggested that people should be engaged and learning, rather than logging off for the day after making their post. 

Aesthetics 

#blackouttuesday Black square analysis 

By looking at Hannah Godwin’s feed on Instagram after #blackouttuesday (see figure 5), we see that she returns to her regular programming as an influencer, posting about her lifestyle and all things aesthetically pleasing. Her feed is not unique, rather it is a symptom of a bigger phenomenon we see across several instagram accounts. Several influencers, celebrities, and everyday citizens alike only posted about BLM during the peak times when everyone else was as well. 

Figure 5. Hannah Godwin’s Instagram feed around #blackouttuesday 

Another trend that took place during this moment was people posting a more aesthetically pleasing black box on their feed (See figures 7,8,9). Influencers and celebrities, including Emma Watson, were criticised by this choice in the comment sections of their posts (See figure 10). Her decision to maintain the ‘white border’ and posting the black square three times attracted thousands of comments, suggesting “that she had focused on maintaining the aesthetic of her Instagram feed instead of communicating a message of solidarity” (Blum). 

Figure 7, 8, 9: Different types of black squares / aesthetics posted during #blackouttuesday  Source: Instagram

Figure 10. Comments under Emma Watson #blackouttuesday post. Source: Instagram

While influencers, celebrities and everyday citizens may be doing work to advance the BLM movement outside of their social media, it would appear as though their online space is reserved for aesthetics and performance activism, prompting just enough engagement to appear active during peak social movement moments.

Meme analysis

Kony 2012

After the ‘truth’ came out about Kony, so did a series of memes mocking those who had failed to comprehend the reality and complexity of the situation, and what ‘real’ activism it would require from participants to change things in Uganda. (See figures 11-16). 

The memes we selected capture the essence of slacktivism. We also posit these memes as a mechanism through which to understand the conception of white saviourism as many of the characters in the memes either showcase or inference white people. These memes capture a tale as old as time, that White people like to be the hero and ‘save Africa’, without over exerting themselves in the process.

Figure 11. “Share Kony Video, I fixed Africa” Source: Google images

Figure 12. “One does note simply destabilize a Ugandan warlord by liking a status” Source: Google images

Figure 13. “Supports Kony2012, can’t locate Uganda on a world map.” Source: knowyourmeme.com  

Figure 14. “Kony 2012 made you cry? You’d better donate some money to the cause without doing any further research on the matter” Source: knowyourmeme.com

Figure 15: “Shared the Kony video…I ended the war in Uganda!” Source: knowyourmeme.com

Figure 16. “Omg soooo inspiring if you don’t watch then you’re heartless!!!” Doesn’t actually do anything to help the campaign. Source: knowyourmeme.com 

#blackouttuesday 

#blackouttuesday also had similar memes to that of Kony, which mocked the concept that something as simple as posting a black square could help progress the movement and end police brutality (see Figures 17,18, 19).

Figure 17. “Posts blackout Tuesday and proud of being in part of BLM”. Source: Google images

Figure 18. “Girls on Instagram. Police brutality”. Source: Google images

Figure 19. “Is this activism?”. Source: Google images

Discussion

The main contentious debate surrounding slacktivism stems from the fact that many have opted to use social media as a replacement for traditional activism as opposed to a hybrid or heightened communicative tool. While it would be disingenuous to disregard the organizing power of social media activism, “at some point one must convert awareness into action, and this is where tools like Twitter and Facebook proved much less successful” (Morozov 190-191). Further, over-communication, discussing events on any platform can result in an information vacuum where “engaging in social causes from behind a screen can devolve into regurgitating platitudes that don’t get enough done […] social media should be regarded as a powerful tool to aid activism, not a substitute for it” (Talbot). 

The issue with “movements [that are] modeled after fast food delivery” (Morozov 196) or the “low risk, low cost” (Leonard 6) model of social media participation is that awareness and expressing solidarity are often “simply too shallow” (Garza 157) and for performative and/or saviorist reasons. Several celebrities, influencers, and everyday citizens are guilty of embodying traits of the White saviour when they engage with social movements. Approaching such matters with conditional or momentary solidarity and “empty slogans”, as opposed to practicing more “authentic solidarity”, undermines and distracts from the work it will take to build a strong and sustainable support system and foundation “in the face of oppression, dysfunction, and marginalization” (157): 

“Solidarity can never be expressed by hearing someone’s pain and then turning the conversation back to yourself. Solidarity means trying to understand the ways our communities experienced unique forms of oppression and marginalization. It means showing up for one another to bear witness and then expanding our fight to include the challenges faced by other communities besides our own.” (Garza 157)

As seen in both Kony 2012 and Black Lives Matter, public engagement does not necessarily translate into action and thereby conforms to  Leonard’s notion of “apolitical consumption”. Performative online actions and short-lived attention spans for crucial matters were largely coming from the Global North. However, much like the Central African violence, Black Lives Matter too will benefit most from long-term engagement, that is not simply practiced for narcissistic reasons. 

Bibliography: 

“About.” Invisible Children, 14 Feb. 2017, https://invisiblechildren.com/

Blum, Jeremy. “Emma Watson’s Trio of Bordered Blackout Tuesday Squares Prompts Backlash.” HuffPost, HuffPost, 3 June 2020, https://www.huffpost.com/archive/au/entry/emma-watsons-trio-of-blackout-tuesday-squares-brings-backlash_au_5ed6f7d6c5b67913aa26a188/amp . 

Castillo, Jessica. “Lizzo, Kehlani, Lil Nas X Question Black Square Activism.” Teen Vogue, 3 June 2020, https://www.teenvogue.com/story/blackout-tuesday-lizzo-kehlani-lil-nas-x-black-squares-instagram . 

Chazal, Nerida, and Adam Pocrnic. “Kony 2012: Intervention Narratives and the Saviour Subject.” International journal for crime, justice and social democracy 5.1 (2016): 98–112. Web.

Fisher, Max. “The International Obsession with Joseph Kony Is Already Ending.” The Atlantic, Atlantic Media Company, 15 Mar. 2012, https://www.theatlantic.com/international/archive/2012/03/the-international-obsession-with-joseph-kony-is-already-ending/254510/ . 

Flock, Elizabeth. “Kony 2012 Campaign Gets Support of Obama, Others.” The Washington Post, WP Company, 8 Mar. 2012, https://www.washingtonpost.com/blogs/blogpost/post/kony-2012-campaign-gets-support-of-obama-others/2012/03/08/gIQArnHkzR_blog.html .  

Garza, Alicia. “The Purpose of Power: How to Build Movements for the 21^st century.” New York: Transworld, 2020

Hager, Mike. “Kony 2012 campaign fails to go offline in Vancouver.” Global News, Vancouver Sun, 21 April 2012, https://globalnews.ca/news/236629/kony-2012-campaign-fails-to-go-offline-in-vancouver/ 

“Kony 2012.” Know Your Meme, 10 June 2021, https://knowyourmeme.com/memes/events/kony-2012 . 

Leonard, David J. “Remixing the Burden: Kony 2012 and the wages of whiteness.” Critical Race & Whiteness Studies 11.1 (2015).

Morozov, Evgeny. “The Net Delusion: The Dark Side of Internet Freedom.” New York: Public Affairs, 2011

Rainie, Lee, et al. “The Viral Kony 2012 Video.” Pew Research Center, 15 March 2012 https://www.pewresearch.org/internet/2012/03/15/the-viral-kony-2012-video/ 

“R/Askreddit – What Scam Did You Get Sucked into?” Reddit, https://www.reddit.com/r/AskReddit/comments/9wdhrq/what_scam_did_you_get_sucked_into/. 

Talbot, Shelby. “The Pandemic Is Exposing the Weaknesses of Celebrity Activism.” Queens Journal, 1 June 2020, https://www.queensjournal.ca/story/2020-06-01/pop-culture/the-pandemic-is-exposing-the-weaknesses-of-celebrity-activism/ . 

Walker, Peter. “Kony 2012 Charity’s Cover the Night Protest Draws Less Visible Support.” The Guardian, Guardian News and Media, 20 Apr. 2012, https://www.theguardian.com/world/blog/2012/apr/20/kony-2012-cover-the-night 

Welk, Brian. “Lil Nas X, Kehlani, Sade and More Criticize Blackout Tuesday: ‘People Need to See What’s Going on’.” Yahoo!, Yahoo!, 2 June 2020, https://www.yahoo.com/entertainment/lil-nas-x-kehlani-sade-164836393.html?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAADNGIyIsWwIb5TYehc2qgITIQZdfzdPts-aP8YK0cEB0ReWWOO7HrQ_LslbDDqvCWKCfyoj8qz2TRtYEUTuSmEQz3aaeLB3vLrsO7tuiGZJ_5v0UUBizILNQpaUMwfiTYDOlEcSfqeDt-moOt7PdO9wTxrVEB6p7V2o2ZSmZuJj_.  

Wilkerson, Michael. “No Longer Invisible.” Foreign Policy, Foreign Policy, 24 Mar. 2012, https://foreignpolicy.com/2012/03/23/no-longer-invisible/.

Keywords: Network Graph Analysis, Gephi, YouTube, Red Pill, Social Network

Abstract

During the last few years, the Red Pill movement has established strongholds on Youtube through a network of diverse channels promoting content ranging from anti-liberalism, anti-feminism to extreme right-wing political ideologies. Although the channels vary in quantities, they seem to share close thematic relationships. Using Youtube Data Crawl Netviz and Gephi data visualisation, we explored the dynamics of channels and videos from the search queries “Red Pill” and “Blue Pill”.

Introduction

A democratic society where widespread information is the norm is neither utopian nor progressive but potentially a breeding ground for conspiracy theories, toxicity, and extremism. Services such as YouTube, which allows the publication of audiovisual products with loosely enforced regulations, is an ideal platform for diverse communities, enabling culture production and the circulation of ideologies (Röchert 2, Herrman). Red Pill stands out among these ideologies as a loosely defined phenomenon equated to a codeword indicating neo-reactionary enlightenment associated with a wake from the false consciousness of liberal political correctness (Aikin 422, Benjamin 3). The term has roots in The Matrix (1999), taking inspiration from the scene when the character Neo chooses to take the Red Pill over the Blue Pill to awake from a simulated reality; however, when appearing in the context of Youtube, it can reference several ideas. Regarding these Red Pill vernaculars, the essay will first explain Red Pill as a phenomenon before using Youtube as a gateway to explore content associated with the Red Pill by visualizing it using Gephi.

Theoretical Background / Literature Review

Youtube is a space with conditions enabling users to create audiovisual content and circulate data around it. Röchert (2021) argued that Youtube afforded diverse communities in the forms of different individual channels but loosely and thematically related (2). Within this domain, the algorithm also plays a vital role as a tool to circulate similar content to consumers according to their viewing history (Gillespie 173). Through this mean, Red Pill managed to find a firm breeding ground within YouTube as a vernacular. De Zeeuw and others defined vernaculars as specific to a community but remain obscure to others outside the group (216). Under different contexts, these vernaculars are also meant to express wills and emotions towards particular issues. Similarly, in the case of Red Pills on Youtube, Tuters identified Red Pill memes specifically and Red Pill in general as a means of expression, meaning waking up from the false world, oriented towards a violent reaction against governing system of power (3).

Terminology

Having established one of the environments from which Red Pill stems, the question returns to what exactly is Red Pill. Cunha traced the term’s meaning back to The Matrix, taking the Pill to wake up from a false conscious society. From the term Red Pill, other vernacular terminologies are closely associated with the original phrase, which appears during the discussion sector of this text. In a general sense, the manosphere refers to the network of websites, blogs, forums that promote masculinity and strong opposition to modern feminism. A vital part of the manosphere is its members called incels, meaning males who fail to get a romantic or sexual partner. Within this, 1 percent is a metaphorical representation of the ideal goal for masculinity to achieve in terms of being the top ultra-rich in income and wealth. The Pick-Up Artist (PUA) is another term that falls under toxic masculinity, meaning the movement from man to seduce and achieve sexual success with women. Finally, in association with this term, coaching refers to training men in the “arts’ of seduction and manipulating women. All of these terms are specific vernaculars within the Red Pill toxic masculine environment.

Methodology

In the following section, we first describe why and how we selected the datasets and search queries from YouTube to analyse, including limitations and specifications on the tool used to scrape data from Youtube without including personal biases. Following this, we outline the method of translating the datasets into visual network graphs using Gephi. We also explain how the data was extrapolated from the YouTube Data Tools. Lastly, we look at different ways of using these values to help layout the network graphs to identify clusters of similar content or other notable patterns.

Video And Channel Search Module Parameters

 In order to extract data from YouTube and build a network graph of Red Pill-related content, we used the Netvizz YouTube Data Tool Set to scrape results from different search queries and channel recommendations. In order to maintain a workable sample of data – and for ease of replication, we scraped both YouTube Channel and Video Modules with the Search Queries of ‘Red Pill’ and ‘Blue Pill’. Our findings focused on the Video Channel Module, which uses search results to allow for more easily accessible content than the networks of ‘featured channels’ found when scraping data via the Channel Network Module. The Video Network Module creates a “network of relations between videos, starting from a search or a list of video ids [to] generate a network of channels based on the same relations”. (YouTube Data Tools) More specifically, the scraping tool’s code targets the “related videos” dataset from the “search/list#relatedToVideoId” API endpoint and transforms this into a. PDF file format (YouTube Data Tools).

For this research project, we used the parameters of 5 iterations with a crawl depth of 1 (i.e. we scraped five iterations of 50 results each – with a depth of 1 which indicates how far from the seeded search query the script goes (YouTube Data Tools).

Network Analysis: Gephi Layout and Degrees

In order to visualise the video and channel search networks, we used a combination of modularity ranking and betweenness centralities – as these can help to visually demarcate specific clusters of similarity and influence potential, respectively. As Newman explains, the value of modularity is useful for identifying communities “a closely connected social community will imply a faster rate of transmission of information or rumour among them than a loosely connected community… [And] leads to the appearance of communities in a given network” (Newman 2007). 
Node size was ranked according to the betweenness centrality, which detects the amount of influence a node has over the flow of information by looking at how many nodes are intercepted via a connection. Modularity was used to help partition the clusters into different colour groups for easier identification. Networks with high modularity have dense connections between nodes within modules but fewer connections with the nodes in other modules – i.e. they are closed communities or clusters of related actors. The algorithm used to work out the communities and modularities was developed by Blondel et al. l in the Journal of Statistical Mechanics: Theory and Experiment (2008).

Research Limitations

Overall, there are many limitations when dealing with network analysis, considering the quantity of data, multiple cross-platform contaminations, and the dynamic nature of fast-changing online platforms like YouTube. While more research and focus needs to be put into how networking analysis methodologies can be improved considering these challenges – we identified three primary limitations that influenced the scope and focus of our research.

a) Language limitations: Several videos, channels, and clusters identified in our research data are produced in multiple languages and dialects. Without extensive language resource tools, it is impossible to objectively deduce inferences about a large amount of the content – however, we can still infer geopolitical conclusions regarding which languages or dialects appear and how they are represented.

b) Limitations of ambiguity within large data sets. One of the biggest challenges when conducting any data research is the quantity and scope of the sample data used. With more resources, time, and computing power, it would be possible to deduce more correlations between the networked graphs and the specific video content relating to themes of Red Pill toxicity. Considering this, when working with large sets of social network metadata, we have to find alternative ways to infer possible meanings from the structures, clusters, and patterns of connections evident when visually translating the data into network graphs.

c) Lastly, there are also specific platform limitations present. These exist both within YouTube’s ecosystem – where some protocols regarding how videos are searched and sorted within their algorithms affect our overall graphs. While we cannot find a way to subvert this limitation to our research – it is essential to consider YouTube’s platform affects how the data is manifested and represented – while still trying to conclude both the producers, consumers, and ecosystem of YouTube. Another platform limitation that affects the sample size used is the current computer processing power available to our research. Because of the limited processing power and time for scraping data – sample sizes need to be kept smaller enough to analyse on accessible computers – yet still large enough to avoid non-uniform samples.

Findings

Red Pill Search Query

The modularity values as indicated by the different colors on the graph (refer to Figure 7) indicated 22 primary clusters. Not all 22 clusters will be discussed as it is visually apparent that some clusters are significantly larger than others. Clusters with a modularity class percentage of > 3% containing nodes with a betweenness centrality measure > 0 are for the purpose of this paper deemed influential.

The Main Cluster

The most influential cluster is represented in pink, which will be labeled the main cluster. This represents the largest community of nodes making up 21 percent of the modularity class as well as containing the node with the highest betweenness centrality measure of 362336.66 (rounded to the nearest decimal point). These nodes within this cluster can be viewed as the most relevant and influential video channels when searching the term Red Pill on Youtube. This cluster represents the most influential nodes of both this cluster as well as the entire data set.

Red Pill Sub-Clusters

The remaining sub-clusters are represented by different colours (refer back to Figure 7). There are a total of 8 sub-clusters with a modularity class of over 3%. It is important to note that this does not mean the smaller clusters are unimportant to the entire network but are less influential. Upon closer analysis of the clusters that formed, it became evident that common themes were apparent among channels within a cluster. For example, five out of the eight sub-clusters had videos from the same region and generally were also from the specific language of that region. The tables below (refer to Figure 9) indicate the top 5 highest betweenness centrality channels within sub-clusters. The betweenness centrality measure is used in order to determine the relevancy of the channel to both the sub-cluster as well as to the overall dataset.

The last three clusters also had consistent themes running through the nodes. One cluster was dedicated to financial coaching videos and predominantly consisted of cryptocurrency and trading advice videos (refer to Figure 10). Another cluster comprised videos based on consciousness, truth, and spirituality. The last cluster is composed of film channels that were directly related to the Matrix content.

Discussion

The channel clusters mentioned previously are some of the most influential channels in the Red Pill search network. However small, these clusters represent an important portion of the Red Pill Youtube search query. From this information, it is possible to derive what the Youtube Red Pill vernaculars look like as well as what some channels individuals would likely stumble across when searching the term ‘Red Pill’. 

A general overview of the video content within the clusters concludes that the most influential channels are American-based. All 14 channels pulled from the data set (refer to Figure 8: Main Cluster) are American channels targeted towards English-speaking audiences. That being said, the data reveals that the Red Pill influence extends beyond American communities as sub-red pill communities from other regions have formed. Although analysis of thumbnails can infer that there are prevalent themes. Figure 11 to 13 illustrates this as all three thumbnails all feature a picture of a man as wealthy with click-baiting, coaching titles.

Identity Formation based on Toxic Masculinity and Heteronormativity

Taking a closer look at the top 5 channels of the most relevant clusters gave us a sense of what the red-pill community on Youtube looks like. Within the videos, there is an overwhelming presence of videos dedicated to the improvement of the body and mind. This became evident with recurring keywords such as mindset, coaching, diet, and exercise. The theme of transcending and finding the ‘ultimate truth’ is evident in the prevalence of videos based on epistemology and religion. That is not to say that they are closely related but rather that they both fall under the theme of ‘awakening’. There seems to be the assumption that taking the Red Pill would result in some kind of higher power, strength, and success. This was revealed by looking closer at some of the sub-clusters. For example, the investment cluster revealed the link between financial success and being the “ideal” male. The ideal male in this case is referred to either as the 1 or 5 percent (refer to Figure 13), a term used by influential Red Pill Youtube Channels such as Fresh and Fit and Kevin Samuels. The 1 percent is a metaphorical representation of the ideal goal for masculinity to achieve in terms of being the top ultra-rich in income and wealth, in which a man becomes the ‘alpha’ male. The Red Pill claims that there is an ‘ultimate’ man commonly represented by the term ‘alpha male’. Somehow through metaphorically ingesting the Red Pill, you will reach a higher power by knowing how to speak to women, becoming physically fit and financially successful.

The ‘transformative’ theme of becoming the ‘alpha male’ is not only determined by a man’s financial success but also his physical strength and success with women. This is evident by the abundance of Pickup Artists (PUA’s)  that are present within the clusters in the Graph.

Some examples are Austin Dunham Vlogs, Mundo Alfa, and Casey Zander (refer to Figure 14) which all combine fitness content, relationship advice, and most prevalently cold approaching women in public spaces. This all falls under the illusion that financially successful, fit men are a ‘prize’. Subsequently, the transformation of the ‘beta’ male into the ‘alpha’ relies on heteronormative ideals of masculinity.

Subverting Red Pill Discussions

There are also groups of actors within the network that may not be promoting Red Pill content, but rather subverting, disproving, or debating their ideas. These types of Blue Pill, ‘pro-woke’ or centralist views are identifiable within clusters of the Red Pill video graph and can create difficulty making any absolute claims regarding clusters and potential toxic influence. This can also work vice versa – with red pill content being masked or disguised under other names – which creates research challenges when trying to identify specific clusters of influence.

As Röchert et all (2021) state: “exposure to conspiracy beliefs can occur even without being polarized or deliberately seeking information”. Furthermore, their research goes on to explain how it is important to be cognizant about the ambiguity of intention within YouTube datasets – as often titles, comments or metadata can be misleading or in reaction to ideas. As they explain:

“counter-messages are directly associated with extremist videos. .. [Users] who view counter-videos are likely to receive a recommendation for extremist videos” (Röchert et all 2021). Or in a similar way – users searching for Red Pill content may be served anti-Red Pill content and vice versa. A good example of how different actors subvert the toxic aspects of the Red Pill communities by using their same ideas can be seen in the ‘Red Pill Black’ movement by Candace Owens, which is a YouTube Channel that promotes black conservatism by using the term metaphorically for “the process of rejection of previously believed leftist narratives” (Ames 2007).

Conclusion

The research concludes that while the most influential Red Pill channels are mostly concentrated in the United States, its influence has reached other regions. Within these channels, the most prominent themes are toxic masculinity and heteronormativity where the acts of coaching men in romantic and sexual tactics with females with the goal to achieve a metaphorical “alpha-male” image are highly motivated.

Works Cited:

Aikin, Scott F. ‘Deep Disagreement, the Dark Enlightenment, and the Rhetoric of the Red Pill’. Journal of Applied Philosophy, vol. 36, no. 3, July 2019, pp. 420–35. DOI.org (Crossref), https://doi.org/10.1111/japp.12331.

Ames, Elizabeth. ‘Liberals Sick of the Alt-Left Are Taking “the Red Pill”’. Fox News, Fox News, 12 Sept. 2017, https://www.foxnews.com/opinion/liberals-sick-of-the-alt-left-are-taking-the-red-pill.

Ball, Siobhan. Lilly Wachowski Tells Musk and Invanka To ‘Fuck Off’ On Twitter. https://www.dailydot.com/unclick/lilly-wachowski-elon-musk-red-pill-fuck-off/. 

Blondel Jean-Loup Guillaume, Renaud Lambiotte, and Etienne Lefebvre. ”Fast unfolding of communities in large networks.” Journal of Statistical Mechanics: Theory and Experiment, (10), 2008, pp. 1000

Cunha, Darlena. Red Pills and Dog Whistles: It Is More than ‘Just the Internet’. p. 2.

de Zeeuw, Daniël, and Marc Tuters. ‘Teh Internet Is Serious Business’. Cultural Politics, vol. 16, no. 2, July 2020, pp. 214–32. DOI.org (Crossref), https://doi.org/10.1215/17432197-8233406.

Gillespie, Tarleton. ‘The Relevance of Algorithms’. Media Technologies, edited by Tarleton Gillespie et al., The MIT Press, 2014, pp. 167–94. DOI.org (Crossref), https://doi.org/10.7551/mitpress/9780262525374.003.0009.

Herrman, John. ‘For the New Far Right, YouTube Has Become the New Talk Radio’. The New York Times, 3 Aug. 2017. NYTimes.com, https://www.nytimes.com/2017/08/03/magazine/for-the-new-far-right-youtube-has-become-the-new-talk-radio.html.

Lovink, Geert, et al. ‘Rude Awakening: Memes as dialectical images’. None (EN), Non.copyriot.com, 3 Apr. 2018. www.narcis.nl, https://research.hva.nl/en/publications/8df871b1-fedc-49dc-8159-a2bb23732a2f.

Newman, M. E. J. “Mathematics of networks”. The New Palgrave Encyclopedia of Economics (2   ed.). Palgrave Macmillan, Basingstoke (ed.), 2007. 

Rieder, Bernhard (2015). YouTube Data Tools (Version 1.22) [Software]. Available from https://tools.digitalmethods.n et/netvizz/youtube/.

Röchert, Daniel, et al. ‘Caught in a Networked Collusion? Homogeneity in Conspiracy-Related Discussion Networks on YouTube’. Information Systems, vol. 103, Jan. 2022, p. 101866. ScienceDirect, https://doi.org/10.1016/j.is.2021.101866.

‘Search: List | YouTube Data API’. Google Developers, https://developers.google.com/youtube/v3/docs/search/list..

YouTube Data Tools. https://tools.digitalmethods.net/netvizz/youtube/mod_videos_net.php. 

A project by: Dilara Akdemir, Betsy Brossman, Gaurika Chaturvedi, and Noyan Er

Introduction

On YouTube, amateurs and professionals are free to upload or consume content ranging from a 30-second video of a boy at the zoo to full-length news reports. A user’s first interaction with the site is the homepage, but this homepage, however, is not an identical place for all users, for an opaque algorithm is used to personalize user experience. Our guiding question is – “Does a YouTube user’s age, gender, language, and country of usage affect what they are recommended on their homepage”. To answer this question, we create eight different sock puppet accounts that differ in these categories. We compare their associated homepages before and after subscribing to the same YouTube channels to investigate the differences that might appear based on the socket puppet’s demographic factors.

We are also creating YourTube so that users can see these differences for themselves. YourTube is a place where users can view the homepages of other YouTube users and compare their own homepage experiences. There is not much research completed specifically on the homepage of YouTube and our project aims to fill that gap. Greater emphasis has been put on the search results of YouTube which have a similar algorithmic issue, for “The entanglement of algorithmic work in the attribution of relevance is clearly visible: the arrangement in the form of ranked lists reinforces the idea that some contents deserve more prominence than others’’ (Rieder et al., 2018). With the creation of YourTube, we aim to spread this awareness beyond solely media professionals/students and make the platform easily accessible for those who may have no former realization of how curated their content truly is, for simply being aware of the filter in which you live is the first step in decreasing communication gaps.  

Theory

Recommender algorithms are agents that define a crucial aspect of the user experience on social media platforms. The effects of algorithms on individualization through the process of curating content have been a focal point of academic debate. User-generated content is being created at such a pace that the technicalities tasked with the curation of these (i.e. algorithms) might not be able to perform the task. (Bozdag, 2013).

Bozdag (2013) proposes a “model of filtering for online web services including personalization’’ within which location, user preferences, and user actions are included as factors determining the processing of information by personalization algorithms (p. 214). These algorithms used in creating a personalized experience, however, are also being held responsible for the “filter bubble” phenomena. Pariser (2011) conceptualizes the filter bubble with an emphasis on user actions: “The new generation of Internet filters looks at the things you seem to like—the actual things you’ve done, or the things people like you like—and tries to extrapolate” (p. 14). Pariser argues that although publishers were always interested in the preferences of their audiences, the “filter bubble” is a different phenomenon in that the individual is alone in it through specific personalization, the bubble is invisible since it is to be observed in platforms that target the general public as a whole whereas political rhetoric in traditional media publishers are easily detectable by the audience and finally that the individual doesn’t actively choose to enter the filter bubble whereas, in conventional broadcast media, individual curation is a more active process (Pariser, 2011, p.14).

The invisibility of the filter bubble calls for an approach that would not only discuss the problems of algorithmic bias but would rather put emphasis on the visibility of this phenomenon. It would also be an innovative approach to implement a collective approach to tackle the loneliness aspect of the filter bubble. Visualization of this problem through a collective call to action would also have an emphasis on user demands: “It is evident that an individual embracing tool for encrypting communications represents a rather different phenomenon than a global advocacy network established to demand greater algorithmic accountability’’ (Beraldo & Milan, 2019, p. 7).

YourTube Website 

We have chosen to achieve our purpose through a website dedicated to data visualisation. Reliance on another platform requires subscribing to the values inscribed in it. However, a website is customisable and scalable. This allows us to structure the user experience in a way that better guides visitors to be critical. Furthermore, we can add more capabilities of data visualisation and opportunities for profile comparison. Although just looking at one’s homepage might be interesting, the opportunity to compare that data with a much larger aggregate is what is appealing about our project.  

Since we aim to be able to compare many diverse homepages, data visualisation can be a useful way to communicate this information (Tufte, 2001). Thus, the visualisations will be the landing page for the website. The user will have the option to read the instructions on how to submit their homepage data or jump directly to the submission and visualisation section. Before a submission is made, a pre-submitted random profile is selected, and the website will show the visualisations associated with it. 

To submit their own data, users must first install the ytTREX (YouTube Tracking Exposed, n.d.) extension to their browser. Then they must go to their personal page and click on the “Homepage Content” button next to “download csv”. They can submit this csv file, along with self-reported age, gender, primary language of usage, and country of usage, through a form on the website.  

One of the limitations of using the ytTREX extension is that it is primarily used on a desktop. On mobile devices, YouTube is also accessed through the YouTube application rather than a browser. However, we argue that this limitation is overcome by the fact that the homepage on Desktop is informed by the account’s YouTube usage on other devices. 

For the purposes of comparison, we intend to cluster the homepages submitted using a Jaccard indicator which measures “the similarity of two sets” (Xu & Tian, 2015, p. 170). This will be done on the level of videos and the level of channels. One of the drawbacks of using Machine Learning for classification is that there may be clusters which are difficult to interpret. Thus, the human intervention of reviewing clusters for coherency before being present for comparison on the website will exist. This is not to say incoherent clusters will not be accessible, however, coherent clusters will be named to make them easier to find and review. These clusters can be viewed in the form of a network visualisation employing a “force vector” algorithm (Venturini, 2014). 

Informed by the principles of reduction and spatial variables (Manovich, 2011), we will provide some macro level visualisations regarding the most popular channels and videos in that profile cluster. Furthermore, visitors can interact with the visualisations by controlling the aforementioned demographic factors and the level of analysis. Additionally, to allow users to explore the data, we will be taking a “direct visualisation” approach (Manovich, 2011). “Direct visualisation” refers to the visualisation of visual objects without the reduction through representation (Manovich, 2011). Thus, we will have a “feed competition” where users can compare their homepage with that of other clusters to see the differences in the subjective experience of the platform (Rogers, 2021). 

We think it is important to communicate our role as curators and how the actual data curation is occurring. Thus, we shall have a subpage communicating our values, role, skillset, and knowledge base to be reflexive about how our experience has shaped the form of the project. Additionally, two pages communicating how the visualisations are being made based on the data submitted — one on a surface level and one on a technical level — will also be present. This is intended to communicate what knowledge can be drawn from our visualisations. For instance, the network visualisations of profile clusters will be accompanied by Venturini’s reading principles (2021). 

To avoid violating the trade dress copyright of YouTube’s User Interface, our comparative profile visualisations will not use YouTube’s UI. Videos will be displayed in the same order as present on the YouTube homepage without any characteristic features such as the sidebar, search bar, and YouTube logo. Furthermore, the videos will be embedded into the visualisation to avoid using copyrighted thumbnails. 

We expect that some users may have privacy concerns regarding the data collection of this project. Firstly, we ask that each user consents to our “Terms and Conditions” agreement that gives us permission to access their one-month YouTube data and share only their gender, age, language, and country of usage with others using the platform. This statement will declare in detail the data privacy policies and data security policies we must legally abide by. Unlike sites like Instagram, this personal data will be shared with no other companies, and the Terms and Conditions will legally explain “the ways in which users can control how and what information about them is being collected and stored” (Steinfeld, 2016). The social security numbers of our users will not be collected; each user will be administered a unique reference number to safeguard their privacy.We expect to rent a web/database server from Microsoft Azure which will then be stored in their European region. With this European-based storage, we understand that our site will be subject to the GDPR, the General Data Protection Regulation. This agreement will ensure that we responsibly store this data to comply with European law. Our main demographic is likely YouTube users, but we might have researchers who will be interested in the comparisons we make. To demonstrate how our project might be useful in this regard we are also running our own research in parallel as discussed below. 

Methodology: Sock-Puppet Research 

The selection of the method adopted for this study was made based on the emphasis on intervening and highlighting the biases of the YouTube recommender algorithm affecting the “home” page recommendations and visualizing these biases. Therefore, we concluded that a “sock puppet” method paired with data collection using the YouTube Tracking Exposed tool would be a fitting and direct approach in collecting and analyzing data. 

Sock puppets are usually associated with fraud, spamming, cyberbullying, and discrimination (Sweney, 2008). In digital media, a sock puppet is defined as “A user account that is controlled by an individual (or puppetmaster) who controls at least one other user account” (Wang, et al., 2018). Sock puppets in this study, however, are used differently since the actions taken by them would be limited to visiting the homepage and subscribing to channels only.  

For this study, language, country, age and gender were controlled variables as we wanted to see the differences in recommendations related to these factors. Another point of inquiry is the effect of subscriptions to YouTube channels on the different “home” pages. To observe these, eight different sock puppets were created of which four selected “Germany” and “German” as location and language preference. For the other four accounts “Turkish” and “Turkey” were selected. Half of the accounts were female, and the other half were male and half of them were twenty and the other half fifty years old. These accounts subscribed to thirty channels which were selected from the fifty most influential YouTube channels on the social media analytics platform Socialblade (Socialblade, n.d.).

Another issue when constructing the methodology were the policies and regulations that would concern our method. The terms of the YouTube community state that users should not “access, reproduce, download, distribute, transmit, broadcast, display, sell, license, alter, modify or otherwise use any part of the Service or any Content except: (a) as specifically permitted by the Service;  (b) with prior written permission from YouTube and, if applicable, the respective rights holders; or (c) as permitted by applicable law…’’ (YouTube, 2021). In the light of these statements, it would seem problematic to use copyrighted material (in our case, the video thumbnails) in the research. However, drawing on the statement emphasizing fair use under the applicable law; the EU directive The harmonisation of certain aspects of copyright and related rights in the information society states that exceptions regarding the limitations in the use of copyrighted material may apply for “use for the sole purpose of illustration for teaching or scientific research, as long as the source, including the author’s name, is indicated, unless this turns out to be impossible and to the extent justified by the non-commercial purpose to be achieved, without prejudice to the exceptions and limitations provided for in Directive (EU) 2019/79’’ (Directive 2001/29/EC). It is therefore concluded that the method does not infringe with the community guidelines.

Analysis: Sock-Puppet Research and Discussion  

Our research shows that by only following 30 channels, the YouTube algorithm suggested different videos stereotypically more related to the younger, 20- years old, female sock puppets for both Turkish and German users. Comparing the homepages of both younger females before and after subscribing to other channels with the male sock puppets, it is noticeable how the younger female accounts are more pop music-oriented. However, both the German and Turkish male sock puppets, on the other hand, got football-related videos, before and after subscribing to the channels. Consequently, we noticed a slight categorization where the female sock puppets were recommended more music and food-related videos, whereby the male users were led towards watching sports.

Apart from that, we noticed how only the first four suggested video rows seem to take the interest of the user based on the subscribed channels and their identity criteria, such as age, gender, location, and language, into account since the suggested video of the 5^th and last row have similar content compared to from and after subscribing to channels.  

Another observation was that on the homepage from the German, 50-year-old female, the suggested videos related to Covid-19 remained after subscribing to other channels, different from the 20-year-old female who got different more interest-related videos suggested. Speculating, it could be due to the age criteria. On the contrary, all the sock puppet users seem to have a variety of the same videos recommended, such as Squid Game content, DIYs, or food challenges, after subscribing to the same channels despite their age, gender, location, and language.  

The result of our research is solely based on viewing and visually presenting the homepage of our different sock puppets before and after subscribing to the same 30 YouTube channels. But even though we did not actively use the subscribed channels, nor our sock puppet accounts by watching videos, liking, or commenting, we still noticed slight differences in the recommended videos based on gender, age, location, and language. Due to this approach, we are aware of the limitation of our research and assume based on our findings that the results of the homepages would be more severe if there was more individual action of the users with their YouTube accounts. However, since our results did find stereotypical video recommendations, we argue for the importance of the intervention of our app where users can compare their homepages to see how their homepage might differ from others based on their intersections.  

Another limitation is that we got Dutch videos recommended despite our location and language since Google knows the current location of our laptops. Thus, Turkish or German users with the same intersections as our sock puppets in Germany or Turkey might get different results displayed than we did.  

Conclusion 

Our findings show that YouTube makes some assumptions about the user based on the information entered during account creation, that is, before the user has used the platform. After completing our research, the answer to our research question – Does a YouTube user’s age, gender, language, and country of usage affect what they are recommended on their homepage? – is yes. In this way YouTube is implicated in the reification of hegemonies associated with various identities, for instance, gender roles.  

A methodological challenge we encountered was associated with the lack of academic understanding of how YouTube’s recommendation system concretely works. Since we don’t know exactly what metrics the YouTube recommender uses in which way, we relied on the “subscribe’’ action to provide each sock puppet with a different “home” page whereas it can be argued that other metrics such as average time spent on each page or how a sock puppet is landing on a page can affect the recommendations. A subsequent project can aim to consider these factors and account for the influence of user behavior on recommendations instead of being limited to demographic factors. 

References

Beraldo, D., & Milan, S. (2019). From data politics to the contentious politics of data. Big Data & Society, 6(2), 1-11. https://doi.org/10.1177/2053951719885967 

Bozdag, E. (2013). Bias in algorithmic filtering and personalization. Ethics and Information Technology, 15(3), 209-227

Directive 2001/29/EC. The harmonisation of certain aspects of copyright and related rights in the information society. European Parliament, Council of the European Union. https://eur-lex.europa.eu/eli/dir/2001/29/2019-06-06 

Manovich, L. (2011). What is visualisation?, Visual Studies, 26:1, 36-49, DOI: 10.1080/1472586X.2011.548488 

Pariser, E. (2011). Bias in Algorithmic Filtering and Personalization. Penguin Books. 

Rieder, B., Matamoros-Fernández, A., & Coromina, S. (2018). From ranking algorithms to ‘ranking cultures.’ Convergence: The International Journal of Research into New Media Technologies, 24(1), 50–68. https://doi.org/10.1177/1354856517736982

Rogers, R. (2021). Visual media analysis for Instagram and other online platforms. Big Data & Society, 8(1), 20539517211022370

Steinfeld, N. (2016). “I agree to the terms and conditions”: (How) do users read privacy policies online? An eye-tracking experiment. Computers in Human Behavior, 55(), 992–1000. doi:10.1016/j.chb.2015.09.038 

Sweney, M. (2008). Should stealth marketing be regulated? The Guardian. https://www.theguardian.com/media/organgrinder/2008/may/21/shouldcommercialbloggingbe 

YouTube Tracking Exposed. (n.d.). Retrieved 26 October 2021, from https://youtube.tracking.exposed/ 

Top 50 Influential YouTube Channels (Sorted By SB Rank). (n.d.) Social Blade. Retrieved October 23, 2021, from https://socialblade.com/youtube/top/50 

Terms of Service. (2021, June 1). YouTube. Retrieved October 23, 2021, from https://www.youtube.com/static?gl=TR&template=terms 

Tufte, E. R. (2001). The Visual Display of Quantitative Information. 2nd ed. Cheshire, Conn: Graphics Press. 

Venturini, T., Jacomy, M., & Jensen, P. (2019). What Do We See When We Look at Networks. An Introduction to Visual Network Analysis and Force-Directed Layouts. SSRN Electronic Journal. Published. https://doi.org/10.2139/ssrn.3378438

Wang, J., Zhou, W., Li, Ji., Yan, Z., Han, J., Hu, S. (2018). An Online Sockpuppet Detection Method Based on Subgraph Similarity Matching. IEEE Intl Conf on Parallel & Distributed Processing with Applications, Ubiquitous Computing & Communications, Big Data & Cloud Computing, Social Computing & Networking, Sustainable Computing & Communications, pp. 391-398, doi: 10.1109/BDCloud.2018.00067 

Xu, D., & Tian, Y. (2015). A comprehensive survey of clustering algorithms. Annals of Data Science, 2(2), 165-193. 

Copyright infringement and fair use are critical issues for YouTube creators, especially ones who primarily work with content they do not own. This research investigates how copyright is understood and navigated by YouTube creators from the genre of commentary YouTube. The investigation sought to uncover discourses by creators through a genre-specific, iterative, and inductive content analysis of 18 YouTube videos. We have found that creators mainly address the topics of monetization, appeal of a claim, YouTube absence, and claimed content. We have found these approaches and overall discourse to relate to broader issues of platformization of cultural content, specifically how entanglements of different actors, such as users, platforms, and legislative institutions, produce a system that is difficult to navigate. With our findings, we hope to shed light on a process of cultural production which, while resulting in extremely popular and relevant content, is increasingly complicated and limiting.

Key Words

YouTube, Platform Governance, Copyright Laws, Fair Use Policies, Content Moderation, Popular Culture

By Marta Ceccarelli, Paola Gardino, Maria Vittoria Ravaioli

Introduction 

As new technologies continue to emerge, governments and authorities must rapidly create, modify, and enforce regulations to ensure conformity and safety for users navigating online platforms (Eggers, Turley & Kishnani 2018). A challenge faced by authorities today is the notion that web spaces inherently transcend regional laws as they cross-national and international borders (Tanneeru 2009), thus making legislation challenging to create and enforce. This study contemplates the implementation of copyright law on the contemporary digital landscape and the introduction of fair use policies as a direct consequence. In its most basic definition, fair use refers to any copying or repurposing of copyrighted material done for a transformative purpose; this includes commenting, criticizing, and parodying any original content (Stim 2019). 

The notion of fair use on YouTube has been subject to scrutiny in the past decade as the platform has not provided adequate clarity on how to practice it. The battle over online copyright infringement was expected to be resolved decades ago. The United States Congress recognized the necessity for updating and improving copyright laws in the late 1990s, consequently enacting the Digital Millennium Copyright Act or “DMCA” (Solomon 2016, 237). The DMCA established a carefully calibrated compromise between the rights of copyright owners and those of online companies and content creators, ensuring that both parties were moderated (Seidenberg 2009, 47). Although unassailable at the time that it was instated, the law has not been altered since, while internet platforms such as YouTube have grown exponentially. The DMCA’s failure to modernize to fit the twenty-first-century media landscape bears heavy repercussions for users and creators operating in the new media environment. In the context of the copyright and fair use regulations on YouTube, new legislation is crucially needed to cater to the user and content creators’ needs as the platform continues to evolve. Due to the outdated nature of the DMCA, the legislation fails to enforce clear and precise guidelines, often resulting in users’ malpractice of the law when navigating digital platforms. 

Copyright infringement has established itself as one of the most pressing constraints to video production for YouTube creators. The danger of having a video demonetized or permanently removed from the platform due to a copyright strike is high especially for ones who employ external content for their own videos, and “copyright remains at the centre of industry struggles for power and control” (Burgess and Green 2018, 44). YouTube has historically struggled to develop precise and homogeneous automated copyright moderators; therefore, manual fraudulent or weaponized copyright claiming has become an issue for the platform’s content creators (Sands 2018). Excessive copyright striking can cause severe problems for YouTube creators. It can lead to suspension of revenues and, in extreme cases, result in a permanent ban from the platform.  

In April 2016, YouTube personality Matt Hoss filed a civil action lawsuit against husband-and-wife team Ethan and Hila Klein, also known as H3H3 (BBC News 2017). H3H3 Production is a commentary YouTube channel, a genre that is based on a critical replication of other individual’s content. Plaintiff Hoss filed the lawsuit under three significant claims: misrepresentation, copyright infringement and defamation. The court concluded that the reaction video H3H3 Productions uploaded on YouTube constituted under fair use, despite it replicating a large portion of Hoss’s copyrighted content (Carrington 2018). Furthermore, Judge Katherine Forrest ruled that “the defendants’ comments regarding the lawsuit are either non-actionable opinions or substantially true as a matter of law”, and for this reason, the plaintiff’s defamation claims failed. Ethan Klein subsequently posted a video celebrating their monumental win, proclaiming that “this is a landmark case, not just for us, but the wording the judge put in is going to strengthen fair use across YouTube”. This lawsuit had profound repercussions on the future of copyright on YouTube as it established a lawful precedent for future legal disputes. Although YouTube copyright laws continue to be ambiguous, this lawsuit has shed light on the situation. 

This study investigates the discourse of copyright infringement and fair use policies on YouTube. Through a qualitative content analysis of 18 videos, we explore how YouTube content creators understand copyright enforcement laws on the platform and how they navigate this exceedingly technical operation system. Our study offers valuable insights into how platform moderation systems such as copyright bylaws influence cultural producers such as YouTube content creators. 

Research Context 

In recent years, the concept of platformization has been developed to explain how platforms have become central organizers of different spheres of life. Most relevantly, Poell, Nieborg, and van Dijck defined it as “the penetration of the infrastructures, economic processes, and governmental frameworks of platforms in different economic sectors and spheres of life […] the reorganization of cultural practices and imaginations around platforms.” (2019, 5). Rather than being independent entities, platforms construct a sort of ecosystem which, at least in the European/North American context, is currently dominated by the so-called Big Five: Google, Amazon, Facebook, Apple, Microsoft (GAFAM) (van Dijck, Poell, and de Waal 2018, 15). YouTube is a connective platform owned by Alphabet. This umbrella corporation encapsulates all of Google’s products and services, making it part of one of the most influential platform corporations worldwide. Through the frame of platformization, platforms have been identified as complex, multi-sided markets in which the interests of several actors intersect (Poell, Nieborg, and van Dijck 2019). Connective platforms such as YouTube depend on creators to provide products or services, in this case, videos, rendering video producers platform complementors. But they also depend on actors such as advertisers to make monetization possible. YouTube is then a multi-sided platform orchestrating the relationships between different actors, from end users to advertisers. This includes relations to the judiciary system, in particular the one of the United States. While operating on a global scale, YouTube must then respond to the legal limitations set and enforced by the US legal system.

Theories of platformization have also addressed how cultural products are inherently contingent on the fast-changing platform ecosystem (Nieborg and Poell 2018, 4289). Because of their role as platform complementors, digital creators are incentivized to optimize their production in the interests of other actors of the multi-sided market, making platform distribution and monetization significant constraints in content production (ibid., 4287). YouTube creators, often referred to simply as YouTubers, have been defined as social media entrepreneurs who professionalize video production of original content while creating and closely interacting with a community (Kaye and Gray 2021). Professional YouTube creators depend on monetization systems dictated by the platform, making them particularly susceptible to changing governance. 

Copyright infringement, then, is a central issue for YouTube creators. According to the platform, when rights holders submit legal and complete takedown requests, the account posting the contested video will receive a ‘strike’. After three strikes, the account may be terminated and content removed from the website (Youtube 2021). While this policy is in place to safeguard content, YouTube’s system of copyright enforcement is “notoriously prone to error” and “largely insensitive to exceptions to copyright such as fair use or fair dealings.”(Kaye and Gray 2021, 2). 

Methodology

The approach by Kay and Gray (2021) was taken as a point of departure for this study, however, an adjustment of methods to the scale and scope of our investigation was in order. Rather than conducting manual sentiment analysis, we opted for an iterative and inductive content analysis of videos. Furthermore, the focus of this investigation was specifically the genre of YouTube creators belonging to the commentary genre. 

Genres are challenging to pinpoint and isolate, posing some methodological challenges to the conceptualization and operationalization of commentary YouTube and hence to the creation of a representative corpus. We define commentary YouTube as an umbrella term for various videos in which the creator actively engages with and proposes a critical perspective on content they did not create for the purpose of entertainment. The videos in question must include the creator watching and actively commenting on such external content, which must be shown in the video. We selected this genre as relevant to the copyright question because of its continuous engagement with material to which they have no rights beyond fair use. The inclusion in this category is also contingent on the understanding and self-categorization of creators. For example, film commentary is also a sub-category under this general definition, but creators engaging with this sort of content tend to categorize their videos as a separate genre. 

For our corpus, we selected 18 videos from a range of YouTubers belonging to the commentary genre, mainly focusing on the ones with a more established audience, viewership, and community. We did this to obtain a representative sample of popular commentary creators who confronted themselves with copyright on YouTube for a longer time throughout different iterations of copyright regulations.

There are a few limitations to this methodology. The selection of videos is somewhat affected by a few biases, including possibly personalized YouTube search results based on our previous use of the platform. Furthermore, since tackling a genre that is somewhat challenging to circumscribe, we partially had to rely on our situated knowledge of such a genre to identify creators whose content was to be considered part of commentary YouTube. Nonetheless, this reflexivity and reliance on personal experiences to create the corpus could be regarded as a valid and positive contribution to the method. This is because our subcultural, community-based knowledge could contribute to selecting relevant individuals within the genre. 

Findings and Discussion

Two problems immediately arose: at the start of our research, we wanted to see how the conversation had changed before and after the H3 lawsuit. Right away, we noticed that the conversation around copyright infringement and copyright claims mainly started after the h3h3production and Matt Hosseinzadeh lawsuit. This finding made us consider the relevance of the H3 lawsuit, which is, in fact, often used by other YouTubers as an example of how fair use was treated in court. As a matter of fact, the H3 lawsuit is the only example (together with the Ray William Johnson and Jukin Media lawsuit in 2014 and 2015) of how fair use was disputed and fought in court. 

The videos where YouTube creators first brought up the issue primarily dated back to 2016, and then resumed in 2018 and 2019. So whereas our initial research wanted to focus on how the conversation about copyright infringement had changed pre and post H3 lawsuit, in the study and analysis of the data, we realized how the conversation was almost wholly absent pre-2016.

Another crucial aspect to indicate is how from 2016 until the beginning of 2018, there were many aspects of confusion among the analyzed creators, questioning YouTube transparency and what fair use truly stands for. Only from the end of 2018 until now creators seem to have a complete understanding of the issue behind fair use. We also noticed how the conversation around YouTube copyright issues revolved around four main themes: monetization, appeal of a claim, YouTube absence, and claimed content.

The main argument of most, if not all creators, was regarding the financial aspect. The idea of creating content for pleasure and passion was often mentioned, but the notion that YouTube is where their primary revenue comes from is persistently reminded to the audience. So that is where the matter truly comes in: the fact that videos are usually claimed in the first three days from the release (which is the time frame in which most of the monetization comes in) and that the revenue from that video will go directly to the complainant constitute a big issue. Furthermore, despite the monetization of the video itself, sometimes the video has to stay up and cannot be removed for contractual reasons. For instance, it is set out in a partnership between a creator and a brand.

The second most debated issue was regarding the appeal of a claim. Most creators explained that once a copyright claim is issued and they, as creators, receive the notice, it is almost impossible, if not ineffectual, to appeal the claim as fair use. Since YouTube decides not to take any part in the issue but instead leaves the resolution of the problem between the creator and the rightsholder. Most of the time, creators described the rights holders as third party companies with little to no understanding of how the platform actually works. So if the creator decides to appeal the claim, the request would directly go to the rights holder, who will still refuse the request in most cases. That would result in the creator receiving a strike on their channel, or if still convinced to be correct, they will have to fight the claim in court. That is a risk nobody is willing to take as YouTube has a 3-strike policy on copyright infringement: the first strike is a warning, the second strike is also a warning. Still, it might also cause demonetization, whereas the third strike will permanently delete the creator’s account and ban him from creating any other account. 

YouTube, in this regard, is entirely absent from supporting its creators. The only cases where YouTube actually intervened in any claim complaint were the ones where creators had connections with YouTube headquarters or in situations where there was public outrage, and so YouTube felt pressured to look into the matter, most analyzed YouTubers point out. This brings us to another point: the only backlash of filing a false claim comes through public outcry towards people who abuse the system. Although failing a false copyright claim is illegal, there are no consequences for a potential system abuser. 

The last theme that YouTube creators significantly talked about was regarding who files the claim and why copyright claims are filed. Rights Holders do not always complain when their content moves from their platform to one of the commentary channels. Every creator that we analyzed noticed that the claimed content were videos where the rights holder was made fun of or criticized. That brought us to the conclusion that rights holders are aware of how commentary channels that re-share and react to their videos bring new subscribers and views to their channels. That is, for example, the case with The Nintendo Creators Program, which shut down in November 2018 when they realized that commentary channels were giving free publicity to their games (Kokatu.com). But, somehow when creators are mocked and made fun of, it is as if the free marketing is not accepted anymore, given that the traffic the rightsholder receives will mainly come from critical viewers. 

Conclusion 

This research project sought to uncover the discourse surrounding copyright infringement and fair use policies on commentary YouTube. In the iterative and inductive content analysis, 18 videos were analyzed to find four main themes that creators deal with, namely monetization, appeal of a claim, YouTube absence, and claimed content. Our study proposes that the discourse around copyright infringement laws on YouTube has become increasingly more prevailing throughout the last 5 years. Furthermore, it found that amongst the analyzed creators, YouTube’s lack of transparency and involvement in the issue is what concerns them the most. 

As cultural content becomes contingent, it is essential to bring a critical perspective towards the objects we can find on platforms, as they are inherently informed by sometimes opaque and restrictive platform policies. To complicate this, such forms of platform governance are deeply entangled with the judiciary system. And even if such institutions remain external to platforms, their position and reach are profoundly influential on the possible field of action for cultural producers, such as in the case of YouTube. Hence, researchers must take critical evaluations and investigations of the pitfalls of copyright law to uncover the constraints to cultural production online. The replication and remix of external content seem to be an integral and endemic aspect of digital culture. Still, as platforms extend their power over more and more parts of the web, such practices become increasingly more challenging to sustain.

Bibliography

BBC News. 2017. YouTube stars H3H3 win ‘landmark’ court case against Matt Hoss. August 24. Accessed October 26, 2021. https://www.bbc.com/news/newsbeat-41037631.

Bridy, Annemarie. 2010. “Graduated response and the turn to private ordering in online copyright enforcement.” Oregon Law Review 81–132.

Burgess, Jean, and Joshua Green. 2018. YouTube: Online video and participatory culture. Cambridge: Polity Press.

Carrington, Terrica. 2018. Exploring the Bounds of Fair Use: Hosseinzadeh v. Klein. February 27. Accessed October 26, 2021. https://copyrightalliance.org/fair-use-hosseinzadeh-klein/.

Dijck, José van, Thomas Poell, and Martijn de Waal. 2018. The Platform Society. The Platform Society. Oxford University Press. https://oxford.universitypressscholarship.com/view/10.1093/oso/9780190889760.001.0001/oso-9780190889760.

Eggers, William, Mike Turley, and Pankaj Kamleshkumar Kishnani. 2018. Principles for regulating emerging technologies. June 19. Accessed 25 2021, October.

Kaye, D. Bondy Valdovinos, and Joanne E. Gray. 2021. “Copyright Gossip: Exploring Copyright Opinions, Theories, and Strategies on YouTube.” Social Media + Society 1–12.

Nieborg, David B, and Thomas Poell. 2018. “The Platformization of Cultural Production: Theorizing the Contingent Cultural Commodity.” New Media & Society 20 (11): 4275–92. https://doi.org/10.1177/1461444818769694.

Plunkett, Luke, Nintendo’s Controversial Creators Program Is Shutting Down. 2018. 

https://kotaku.com/nintendos-controversial-creators-program-is-shutting-do-1830728813

Poell, Thomas, David Nieborg, and José van Dijck. 2019. “Platformisation.” Internet Policy Review 8 (4). https://doi.org/10.14763/2019.4.1425.

Sands, Mason. 2018. Why Copyright Will Be The Biggest Issue For Youtube In 2019 (Updated). December 30. Accessed October 26, 2021. https://www.forbes.com/sites/masonsands/2018/12/30/why-copyright-will-be-the-biggest-issue-for-youtube-in-2019/.

Seidenberg, Steven. 2009. “Copyright in the Age of YouTube.” A.B.A. 46-51.

Solomon, Leron. 2016. “Fair Users or Content Abusers: The Automatic Flagging of Non-Infringing Videos by Content ID on Youtube.” Hofstra Law Review 237-268.

Stim, Rich. 2019. Stanford Copyright and Fair Use Center. October 19. Accessed October 26, 2021. https://fairuse.stanford.edu/overview/fair-use/what-is-fair-use/.

Appendix

H3H3production, We’re Being Sued,  May 25, 2016, https://www.youtube.com/watch?v=fEGVOysbC8w&t=335s 

H3H3Production, WE WON THE LAWSUIT!, August, 24 2017, https://www.youtube.com/watch?v=9eN0CIyF2ok&t=179s 

CodyKo, bad news :( , December 29, 2018, https://www.youtube.com/watch?v=_x2pHhbZheA&ab_channel=CodyKo   

Pewdiepie, About A Copyright Strike, September 14, 2017, https://www.youtube.com/watch?v=ah7LYxysuJ8&ab_channel=PewDiePie 

Pewdiepie, STOP DOING THIS! – Copyright Striking Criticism etc, January 11, 2019, https://www.youtube.com/watch?v=yMuEeUyMfUo 

Pewdiepie, DON’T START YOUTUBE BEFORE WATCHING THIS!, July 24, 2018, https://www.youtube.com/watch?v=yMuEeUyMfUo 

Pewdiepie, “Can we copystrike pewdiepie?” // Twitch Drama #1 ( Deleted PewDiePie Video ), July 12 2020, https://www.youtube.com/watch?v=yMuEeUyMfUo 

Pokiman, My Response To The Copystrike Allegations / PewDiePie / Dark Side Of Pokimane, January 13, 2019, https://www.youtube.com/watch?v=r5AU0JJ2wlA&ab_channel=Pokimane 

Andrei Terbea, The Lamentable Tale of POKIMANE, August 15, 2020, https://www.youtube.com/watch?v=7iRILbewgos&ab_channel=AndreiTerbea 

Nick Nimmin, These 4 Things Will Get YOUR YouTube Channel DELETED, May 21, 2021, https://www.youtube.com/watch?v=Qh4-hTA-Xgg&ab_channel=NickNimmin 

Mumbo Jumbo, YouTube’s copyright system is broken , May 19, 2019, https://www.youtube.com/watch?v=LZplh8rd-I4&ab_channel=MumboJumbo 

Danny Gonzalez, I Have Been Having Some ISSUES, February 23, 2019, https://www.youtube.com/watch?v=Fn8gITxmaSU 

Danny Gonzalez, They Took My Video Down So I Fought Back In the Most Petty Way, November  2, 2018, https://www.youtube.com/watch?v=tgl2TtrD6D8 

Jarvis Johnson, An Animated Story Channel Is Beefing With Me, July 14, 2020,  https://www.youtube.com/watch?v=KnS0Q06E17w 

Angelika Oles, YouTubers STEALING MONEY?! (false copyright claims on PewDiePie, Shookbang and Cody Ko), December 30, 2018,

Angelika Oles, Suzy Lu EXPOSES a HUGE copyright problem…, May 2, 2020, https://www.youtube.com/watch?v=l8mteve_VHI&t=468s 

Smokey Glow, Let’s Talk About Marlena Stell, Makeup Geek and FALSE Copyright Strikes…, January 31, 2020, https://www.youtube.com/watch?v=rDjurev_-HE 

D’Angelo Wallace, youtube’s biggest sponsor broke the law – rAiD: sHadOW LegEnDs, March 10, 2020, https://www.youtube.com/watch?v=OZyM9kfOotk