“We Learn Through Mistakes”: Perspectives of Social Media Creators on Copyright Moderation in the European Union

Abstract

Artists promoting their cultural production on social media platforms must navigate an algorithmic environment that involves visibility, content distribution, and growing algorithmic copyright moderation. The latter was experienced by EU-based social media creators when Article 17 of the Copyright in the Digital Single Market Directive was enforced in the EU (Directive [EU] 2019/790, 2019). This research focuses on the understanding and experiences of such creators regarding copyright content moderation of their work on social media platforms. Semi-structured interviews were conducted with 14 artists from EU countries, who were surveyed on digitization and digital access to cultural content. A multimodal framework was employed to analyze the copyright governance of creative practices and products, focusing on the regulative (i.e., the adoption of Article 17), normative (assumptions about legitimate and illegitimate behavior in specific communities), and discursive dimensions, as well as the influence of technological affordance on creative work. The findings contribute to the fields of cultural production and platform governance, as well as having policy implications. The anticipation of punishments from platforms due to copyright-related fears and the inability to use appeal processes directly influenced cultural production. Most interviewees employed self-censorship, avoidance, and content adjustment in their creative work before posting on social media platforms.

Introduction

“YouTube and its users face an existential threat from the EU’s new copyright directive” was one of numerous similarly apocalyptic headlines that appeared in the spring of 2019 in European and US media (Feiner, 2019, May 12). Indeed, concerns about the implementation of Article 17 of the Copyright in the Digital Single Market Directive (CDSMD) brought many into the streets in Germany (Martin, 2019, March 2), affected platforms to the extent to which they stated they would put up a fight, and created fears that users would flee elsewhere (Feiner, 2019, May 12). The contested Article 17 introduces a new liability regime for platforms that was thought to endanger free speech. The clauses mandates that in the absence of authorization from the copyright holder, online content-sharing service providers (OCSSP) must meet challenging conditions to be exempt from direct liability for copyright infringement of their users, and has caused widespread concern that it would destroy the diversity of users and content. In consequence, the step taken by EU legislators has drastically changed the regulatory aspects of platform governance, making social media platforms much more liable for users’ content. In previous years, the rhetoric used by platform stakeholders contained claims that they were not media companies, thus allowing them to evade liability for user-generated content (Gorwa, 2019). The copyright reform was the first attempt to do platform regulation on the European level.

However, at the same time, copyright has been an important exception to the stated immunity of intermediaries (Poell et al., 2021, p. 82). Wishing to avoid litigation arising from notice and takedown procedures (Urban et al., 2017), social media platforms had created their own robust copyright moderation systems even before the implementation of Article 17. Since its inception, YouTube alone has invested over US$100 million in its content-filtering system through its content ID, whose purpose is precisely to identify copyrighted content (Sawers, 2018).

Now, EU policymakers have seemingly added a further layer to the surveillance of creative work. An interesting question is how the top-down approach to platform governance and state efforts to improve governance for content creators (Cunningham & Craig, 2019) have been reflected in creators’ experiences.

Previously, social media content creators has been a term used very broadly, describing cultural producers on social media platforms, whose creative labor is directed at building self-presentation for future gains, be it reputational or monetary, in direct and indirect forms (Caplan & Gillespie, 2020; Duffy & Meisner, 2023; Poell et al., 2021).

Today, producers of cultural content on social media platforms must navigate the algorithmic environment, which relates to not only the visibility that content creates (Bishop, 2019; Bucher, 2017) and content distribution (Hallinan & Striphas, 2016; Willson, 2017), but also increasing levels of algorithmic and human moderation, particularly in relation to copyright (Gorwa et al., 2020; Gray & Suzor, 2020).

In this study, we contribute to scholarship on both algorithmic and platform governance on one hand, and social media creators on the other hand, by investigating algorithmic copyright moderation in the post-Article 17 environment through the experiences of producers of cultural content on social media platforms in the EU member states. Through qualitative semi-structured interviews, the authors examined the understanding and experiences of creators regarding copyright moderation in relation to their media production labor on social media platforms.

The current research draws on a multimodal framework to analyze the copyright governance of creative practices and products, focusing on the regulative dimension, the normative dimension (prevalent assumptions about legitimate and illegitimate behavior in a specific community), the discursive dimension (shared [or contested] understandings and framings of issues in certain contexts), and a technological dimension (affordances and rules embodied in infrastructures, devices, and algorithms relevant to creative work) (Katzenbach, 2018, p. 10). Thus, governance is a broader term than regulation, including all patterns and exploring the construction of social order and social coordination (Bevir, 2009). With regard to copyright, a governance perspective seeks to understand “how certain frames and norms constitute, inform, and question the practices of creative people, as well as users of creative products” (Katzenbach, 2018, p. 13).

Therefore, this article focuses on two main aspects: first, how a relatively recent legal regulation in the Copyright Directive has affected platform governance in the case of copyright enforcement and, second, how perceptions of the affected creators regarding their interactions with the “algorithm” can shape governance agendas.

The following research questions were addressed in this study:

• RQ1: Do creators have a shared understanding of the new and pre-existing policies on the copyright regulations of platforms?

• RQ2: How does the newly introduced copyright legislation, which increases automatic moderation on streaming platforms, affect cultural creations posted on such platforms?

• RQ3: Have creators experienced any changes to their creative process due to copyright-related content moderation?

Creators’ Perspectives on Content Moderation and Algorithms

This article builds on literature that has studied the effects of algorithms and governing practices on cultural production (Bishop, 2019; Bucher, 2017; Caplan & Gillespie, 2020; Duffy & Meisner, 2023; Poell et al., 2021), and investigates these questions with a focus on copyright content moderation Our primary approach is to investigate this complex setting and questions from the perspective of social media creators themselves. This is important, as prior studies have indicated that discussions on copyright policies frequently fail to include or adequately represent individuals who use social media platforms (Edwards & Moss, 2020; Kaye & Gray, 2021).

We specifically build on an extensive amount of literature that has explored the perceptions of content creators regarding the “algorithm” and its implications for their cultural work. Cultural creators seeking mainly online audiences depend strongly on social media platforms. They must constantly pursue algorithmic visibility, as measured by quantified metrics, such as likes, views, and shares (Bucher, 2017; Duffy & Meisner, 2023). Meanwhile, the ways that platforms curate and govern content and interactions on their sites, as well as the dynamic and untransparent character of such actions, evoke the threat of invisibility to creators (Duffy & Meisner, 2023). Such governance has been described as dangerous for creators (Cunningham & Craig, 2019). Bucher (2012) highlighted how the algorithm had established participatory norms through validation and punishment, with the same author later mentioning the affective dimensions of the algorithm, such as how it makes people feel, and thus analyzing algorithmic imaginaries (Bucher, 2017). Subsequently, Bishop (2019) introduced the concept of “algorithmic gossip” as a way for communities of creators to engage with algorithms, and form a shared understanding of them.

Describing creators’ labor on social media platforms, Abidin (2016) introduced a concept of visibility labor among Instagram creators, while DeVito et al. (2017) discussed algorithmically driven content curation and user resistance. Caplan and Gillespie (2020) examined how YouTube creators develop their own theories about the demonetization of their content. Bishop (2020) studied self-proclaimed YouTube “experts” who claimed to help users “mitigate the risk of algorithmic invisibility” and Cotter (2019) conceptualized “practical knowledge” of algorithms through a case study of a YouTube community’s practices. With a focus on copyright, Kaye and Gray (2021) explored copyright understanding among creators of YouTube content using a qualitative analysis of videos that discussed copyright moderation, and defined a particular branch of Bishop’s concept of algorithmic gossip as copyright gossip.

Interestingly, in our study—when giving their perspectives on copyright moderation—the creators involved in our study described most of the concepts mentioned above. Visibility labor, algorithmic gossip, user resistance, algorithmic folklore, shadow banning (Savolainen, 2022), folk theories of content moderation, and algorithmic invisibility (Duffy & Meisner, 2023) were all discussed by the interviewees.

Creators on Social Media Platforms

Who are social media creators? Caplan and Gillespie (2020) refer to creators as users who rely on the revenue from their media production labor on YouTube. The social media creators discussed in this article sometimes relied on platforms for their income; however, they were not always producing media specifically to gain revenue but using platforms to promote the creative work they had done elsewhere. Despite not making careers out of creating for social media (with one exception), they still engaged in creative labor for these platforms (Duffy & Meisner, 2023) to promote the creative work they were doing outside (or inside) of the platforms.

Creative labor on social media platforms has been conceptualized as the work of self-presentation with the belief that it would bring benefit to one’s professional status and future opportunities (Craig, 2019; Duffy, 2017; Meisner & Ledbetter, 2022). Following this line of thought, we define social media creators in this article as artists whose workload consists of marketing their creations on social media platforms, with the aim of creating self-presentation and self-brand for future opportunities. In this way, social media creators evade the traditional distinctions in the creative industry and copyright policy between authors and producers, performers, and consumers. Even in recent studies in the field of copyright law, the understanding of authors, users, consumers and infringers as separate actors is increasingly questioned (Meese, 2018; Pappalardo & Meese, 2019).

Social creators’ relationship to platforms is ambivalent. While they generally use many platforms to promote their products, many creators remain relatively platform-independent (Poell et al., 2021). But platforms make a difference. As described by Duffy et al. (2019), platform operators not only intervene in an established cultural production site but also re-establish their authority over distribution and marketing channels. Nevertheless, it has long as been noted that immense change has been occurring in creative labor markets, and that this change is shaped by platforms as the new environment where cultural production occurs and is being governed and shaped by social media platforms (Arriagada & Ibáñez, 2020; Cunningham & Craig, 2019; Duffy et al., 2019).

Copyright Content Moderation by Platforms

Platforms exercise their governance of cultural production, particularly through automated content moderation (Gorwa et al., 2020). With a view to copyright, we understand with Quintais et al. (2022) copyright content moderation as the activities of platforms aimed at detecting, identifying and addressing content or information that is illegal under EU copyright law and is incompatible with their Terms of Service. These activities can include a broad repertoire of measures: Gorwa et al. (2020, p. 4) distinguish with regard to consequences between “hard” measures such as blocking and removal, and “soft” measures, such as downranking and flagging content, and with regard to technological measures between “matching” approaches (seeking additional copies of known content) and “prediction” approaches (seeking to extrapolate features from known to unknown content). The most prominent example is probably YouTube’s Content ID system that matches each upload on the platform against a database of content registered by rightholders.

Critical assessments of the decision-making mechanisms of platforms have demonstrated that such processes are not transparent or easily understood by creators, if they are understood at all (Eslami et al., 2015; Gillespie, 2018; Poell et al., 2021). Copyright moderation by platforms is not a new experience for creators. YouTube imposes various monetization and content rules (Caplan & Gillespie, 2020), while YouTube and Meta had both established robust systems of automated copyright detection long before 2019 (Poell et al., 2021). With the adoption of Article 17 of the CDMSD, EU policymakers have attempted to protect the rights of creators and added further incentives for social media platforms to utilize automated copyright moderation.

As the European Commission (2021, June 4) explained in a communication to the European Parliament:

Article 17 provides for a specific regime of authorisation and liability for copyright and rights related to copyright (“copyright”) which applies to certain information society service providers defined as online content-sharing service providers. Under the previously applicable legal framework 3, the copyright liability of these service providers for the acts of their users was unclear. Article 17 provides legal certainty as to whether online content-sharing service providers engage in copyright-relevant acts in relation to the acts of their users, as well as legal certainty for users.

In the research by Cunningham and Craig (2019), the main concern was the lack of recognition for creators as stakeholders in contemporary academic or policy debates on platform governance. The current study investigates how cultural production is changing in the views of artists who are simultaneously social media creators, given the impacts of more robust copyright moderation due to the enforcement of Article 17 in the EU.

Cultural Industry and Social Media Platforms

In the context of cultural industries, the term “platform practices” refers to the various methods, procedures, experiences, and manifestations of creativity, work, and citizenship that influence cultural production via online platforms (Duffy et al., 2019). Social media platforms have now been recognized as primary infrastructural gatekeepers and intermediaries in cultural production (Gillespie, 2018; Langley & Leyshon, 2017; Siciliano, 2022). Similar to the adjustments that traditional cultural producers made to their practices to suit conventional gatekeepers, platformized cultural producers must also adapt to the rapidly changing platform algorithms, interfaces, and governance. However, the infrastructural elements of these platforms remain unclear to both users and downstream intermediaries, such as platformized organizations, according to research by Duffy et al. (2019), Nieborg and Poell (2018), and Poell et al. (2021).

The current research attempts to bridge the gap concerning the correspondence between regulatory and governance regimes and the empirical realities of understanding and responding to online copyright moderation.

Data and Methods

The present article focuses on analysis of 14 semi-structured interviews with content creators conducted between May 2022 and July 2022. The design of this empirical study was such that the only creators to participate were those who agreed to be interviewed after a large-scale EU survey of artists’ experiences with the digital sphere (Poort & Pervaiz, 2022).

Interviews were conducted with 14 creators from the following European Union countries: the Netherlands, Bulgaria, France, Romania, Croatia, the Czech Republic, Portugal, Estonia, and France. Most participants were visual artists (such as painters, animators, photographers, or illustrators), while one was a full-time professional vlogger who presented educational art topics on YouTube. For the other interviewees, the use of social media platforms, although associated with their creative work, functioned more as a marketing and visibility tool.

The artists used a wide range of social media platforms, including Instagram, Facebook, TikTok, YouTube, Twitter, Behance, Etsy, LinkedIn, Vimeo, Pinterest, and Dailymotion.

Conducted on Zoom, the interviews followed a semi-structured interview protocol and lasted between 30 and 90 min. Participants received a gift card worth US$50 in exchange for their time and insights. All the interviews were recorded after acquiring the participants’ consent. After the interviews, the audio recordings were transcribed and edited to address any discrepancies. To analyze the copyright governance of creative practices and products (Katzenbach, 2018), a multimodal framework was adopted for the thematic coding of the content. The data set was coded in an instrumental sense, namely, to summarize, identify, and organize themes in the interview transcript corpus according to the adopted framework.

Results and Discussion

Regulative Dimension: Understanding Legislation, Terms, and Conditions

In this section, the authors analyze how creators have assessed and understood the regulative dimension of copyright moderation. This refers to not only the recently adapted Article 17 of the CDSMD but also the contractual terms and conditions accepted by every social media user when they use social media platforms.

The creative content producers displayed a lack of or minimal understanding of the regulatory mechanisms that apply to social media platforms. Assumptions about “right” and “wrong” practices differed, and they usually did not correspond to legal realities, which calls into question the “regulative dimension” of the provision and enforcement of formal rules.

The European CDSMD (2019/790) was adopted and came into force in June 2019. Countries had 2 years to implement it within their national legislation but almost 1 year after the deadline, on May 19, 2022, the EU Commission issued a press release saying that Belgium, Bulgaria, Cyprus, Denmark, Greece, France, Latvia, Poland, Portugal, Slovenia, Slovakia, Finland, and Sweden had not yet notified the Commission on changes to their national laws.

However, even the cultural producers who used platforms extensively in their marketing strategies did not read the terms and conditions of these platforms, and they displayed only very basic awareness of copyright regulations, such as policies and legislation (including Article 17 of the CDSMD). An artist from Portugal said: “I couldn’t tell you if there [are] any differences between Instagram regulations in Portugal, . . . Spain, and Italy, no.”

Even interviewees from countries that had adopted the Directive were either unaware of the law or had not seen it enacted to protect their own work. An artist from Romania said:

My opinion on platform copyright regulation is that it is hectic and illogical for things that I would like to be copyright-regulated and should be obviously regulated, since it’s art theft or something like that. Then it’s not regulated at all, because the content was changed, so the platform doesn’t care. I find it just confusing and illogical, and I don’t find that it works. Like when it should work, it doesn’t and when it shouldn’t work, it does.

A Greek painter thought that his creations were not protected via copyright on social media platforms:

Let’s say, I drew a picture, and somebody copied my image and posted it on another social media [platform]. I can’t track that, you know. That’s the issue, I think, and I don’t want to track every social media [platform] to see what’s going on. I’m gonna waste a lot of time doing that.

A tattoo artist from Croatia, who actively used Instagram for her creative work, said:

I think every time an artist wants to show their work, if they put it online, it’s not only ours anymore. If we want the world to see our work, then we have to accept that someone else might like it enough to steal it. [That’s] just the way it is. I don’t think it can be fixed.

These findings about perceptions of creativity in our study with a focus on EU-based creators correlate with studies in other copyright regimes. For example, a study on creators’ self-perception across different copyright regulation regimes in the United States and Australia, showed that despite of what the law actually said, creators have ideas about the process of creation which the scholars call a “Romantic trope” (Pappalardo & Aufderheide, 2020); similarly for the United States, Jessica Silbey (2014) identified stark mismatches between creatives’ positions and informal norms, on one hand, and formal copyright regulations on the other hand. As we have not studied the copyright provisions of platforms in this study, future research could explore how platforms policies (e.g., based on data sets, such as the Platform Governance Archive, Katzenbach et al., 2023) are exactly positioned between copyright law on one hand, and users’ imaginations and experiences on the other hand.

Another interviewee, a visual artist from Portugal, was unsure when the “laws” had changed. Like many other interviewees, they was also confused about whether the rules and regulations set by platforms for their users were legal obligations (i.e., laws):

Because the laws regarding music in YouTube videos and posts have changed in the last couple of years, or, like four years [ago], people were used to having no copyrights associated with videos. [Then] suddenly, four years ago, everything changed.

Creators often expressed a desire to know more about the rules and regulations concerning copyright moderation. A photographer from Greece said: “I never read them, to be honest. I never read any things about what I [have consented to] already . . . but, you know, I wish that I did [at] the beginning. I wish I knew more.” Some artists stated that they would even like to receive formal education on the issue, such as “seminars for young creators.”

A jewelry maker from Estonia said:

That’s something I want to make clear for myself: How can I have music as background for my stories, or how [can] people use a lot of different music? And I always think: Is this allowed? Or how does it work?

Because the creators understand platform governance policies as “laws,” it is possible to suggest that there is a disconnect not only with regard to their understanding of the copyright legislation regime of the EU and nation states, but also with regard to platform policies that govern copyright and what creators actually think about it. The large-scale qualitative study of creators in Australia by Pappalardo and Meese (2019) has identified a similar disconnect. While creators think that attributing something is fair and logical, this does not align well with copyright law and platform policies, especially in the context of automated copyright moderation as incentivized by Article 17. As Pappalardo and Meese (2019) suggest, formal copyright policies are still not well equipped to handle content engagement and re-use.

Another issue that may have led to ignorance of formal regulations is the existing technical affordances, as it had been raised in our interviews. That is, the very designs used by platforms to obtain user consent. For instance, as an Estonian artist mentioned: “I never really read them. Like, when there’s . . . pops-ups [that say] “Do you agree with the terms and services?,” then I just click: “Yes, of course.”

Normative Dimension: Problematic Assumptions of Copyright Content Moderation

Consistent with previous studies on understanding the algorithms, creators on social media platforms often acquired their knowledge of copyright issues through “algorithmic gossip” (Bishop, 2019). “I don’t know much actually. What I know is from personal stories,” said one interviewee. Learning through mistakes was also common practice: the creators knew of no formal methods or mechanisms by which they could learn about copyright moderation. As one interviewee stated:

I think it’s not explained for us. We just learn by [our] mistakes, for example, [by] putting a song [online] and then understanding that we can or can not [use it in this way]. Or we learn by what fellow colleagues tell us. I never came across an actual workshop or seminar, anything like that.

A creator from Spain explained that they knew about only two aspects of automated platform moderation:

I don’t think I’m aware of it. The only thing that I know that it’s that they don’t allow very explicit images, for instance, very sexual images . . . And also [copying] . . . someone’s work [is not allowed]. But I’m not sure because I see a lot of the time someone copying someone’s work without permission before anyone actually notices and reports it.

A Bulgarian visual artist agreed: “For example, do not use other people’s content in terms of music, and everything about pictures, without crediting them. [That’s] pretty much it.”

The creators often felt that their lack of knowledge of automated algorithmic copyright moderation prevented them from undertaking other forms of creative work. “That’s something I want to make more clear for myself because I want to have music for background for my stories, or [because] people use a lot of different music. And I always think, how is this allowed? Or how does it work?” said a visual artist from Estonia. A visual artist from Greece added:

I don’t have reels, for example, because of the copyright laws, because reels need music to be able to play. But it means a lot of work for me to figure out which one will go with which platform and which algorithm will suppress it, because it’s copyright violated. So even if you have read the rules, it feels a bit like a guessing game.

Analyzing these themes enabled research question 1 to be answered: the visual artists were found to have no shared understanding of what copyright moderation on platforms entailed for them. Instead, they continued to rely on algorithmic gossip (Bishop, 2019) and folk theories (DeVito et al., 2017; Myers West, 2018) while developing their own theories (Caplan & Gillespie, 2020), learning “by their mistakes,” and having their content removed by algorithmic moderation.

This observation led to another theme that emerged in the interviews: that the regulatory dimension of the platforms’ copyright legal obligations and the technological dimension of copyright algorithmic moderation change, and occasionally even disable creative processes. Our findings indicate that this is indeed happening in practice. The regimes of discipline and punishment that structure the social media economy (Duffy & Meisner, 2023), with copyright moderation being an integral part of it much like algorithms that give or take visibility, have influenced experiential practices of creators. These range from self-censorship to relying on “word of mouth” from the community.

Discoursive Dimension: Platforms and Creative Content with Regard to Copyright

In this study, several themes were identified that could be assessed as the discoursive dimension. First, the interviews revealed both highly negative and highly positive attitudes toward the role of social media in cultural production processes. On one hand, creators described how platforms were promoting “cheap superficial content” (artist, the Netherlands), generating “hype” (photographer, Greece), and being “algorithmic treadmill[s]” (artist, Bulgaria). A graphic designer from Romania stated that in their opinion, Instagram was “ruining the art world” because it forced artists to focus more on visibility labor (Abidin, 2016) than the creation process.

I guess I kind of miss the social media from five years ago or 10 years ago, when you could actually build a more serious community and have a bigger reach for your audience, because for me, personally, I’ve kind of started my online career from social media. But that was 10 years ago, when Facebook and Instagram basically didn’t have any kind of engagement algorithm, and you would have a really big reach.

On the other hand, social media was viewed as a useful tool for marketing and sometimes, although to a lesser extent, sales. Although many artists preferred not to do this kind of work, many felt that it was a prerequisite for modern creative industries. As a graphic designer from Portugal said: “In my day-to-day life, I’d rather just focus on creating art and then not having to worry about how it will be posted . . . But in contemporary society, it’s kind of a requirement for artists.”

On a related note, some copyright discourses could be traced to the seemingly unavoidable platform labor undertaken by these creators to promote their work. A strong emerging theme was the sense of “unfairness” among the artists of the need to surrender copyright rights to social media platforms:

They’re probably . . . the pictures are probably copyrighted by them [i.e., the platforms]. They are owned by them, which is, I think, terrible. [It] shouldn’t be like that, [they] should be open source. I think that’s really bad. Actually, I think it should be prohibited

said an artist from Portugal.

Such opinions on governing copyright were connected to the issue of platforms being private entities but not regulated by the respective national governments.

I think the platform should be more regulated by, I think, governments or, you know, something more legal. Not by a company, you know, it shouldn’t be commercially driven. It should actually be a world that is governed by governments.

The prevalent opinion among the cultural producers was that the moment they posted a creation on social media, they surrendered their rights to copyright to that platform. “One . . . reason [for that] is that Instagram and Facebook . . . consider that they are private networks, so . . . copyrights don’t apply to them from the moment you put something on [these platforms],” said an artist from the Netherlands.

Creators felt this was unfair: “I think that even Instagram shouldn’t be owning your images. It’s my art, it’s my personal page, it’s my business page. They should be happy that we share it on their platform,” said another artist.

The trade-off between the labor that cultural producers must expend in chasing their social media visibility (Bishop, 2019; Bucher, 2017) and the benefits they obtain from the platforms was generally regarded as unequal. An artist from Paris mentioned that social media is now considered a “good [form of] promotion for your work” but some artists felt they could not protect their copyright when they posted something.

In the end, it’s . . . mostly [the case that] all the creators . . . work for free and don’t get much back out of it. . . . Obviously, we don’t have with every picture you post on Instagram or Facebook a message that you . . . own the copyrights. . . . None of those platforms have developed a way to address that topic to the creators themselves.

Technically, the terms of use of most social media platforms say that the platforms do not own posted images but only have a license to use them (see, e.g., Instagram’s, 2023 Terms of use). However, while these rules are stated on each platform, creators did not understand the differences between these contractual provisions.

Technical Dimension and How Creators Understand It

Nine of the 14 interview participants either had had their creations taken down for copyright infringement reasons or knew someone who had experienced this. One creator from the Netherlands recounted:

I remember once or twice using pieces from movies to post on Instagram, once, something from a generally good movie from the sixties. And strangely, it was blocked . . . because within that movie there was a soundtrack by the Rolling Stones, and I guess the algorithm works really well although it [the content] really did not feature anything related to the Rolling Stones. I tried several times to upload it and it was immediately removed.

An illustrator from Portugal described a similar situation with a colleague:

I had a colleague . . . who does posters for rock bands. [He] uploaded a video from the drop of his most current poster, and on the video, music from the band was playing in the background, not in the background of the music, like, it wasn’t edited in. It was from the video itself. It was playing live where we shot it. [It] got taken down, even though he works with the band.

As Suzor et al. (2019) had already highlighted, there is a massive lack of transparency from platforms in how content moderation works. Our finding indicates that this is still the case, despite transparency initiatives by platforms; creators and users still experience high levels of confusion and distrust about how content moderation is functioning.

How Creators Understand the Technical Dimension of Complaints and Appeals

In general, the participants found that the appeal, report, and complaint processes on platforms were difficult, opaque and often remained unanswered. They occasionally had to use networks of friends and followers to solve an issue; in other words, they had not complained directly to a platform but used informal mechanisms. As one participant stated:

It was very difficult of a process to solve. I had an artist friend who was catfished. So someone created an account, sharing her work as if it was by them on Instagram. [Then], she asked everyone to report it [so] we [her friends and followers] all reported [it] and the account was taken out.

Others had not appealed at all: “I didn’t appeal because I thought that it was in there, right, the song wasn’t ours,” said a Bulgarian artist. “[We] were only lip-syncing so we weren’t singing over it, or we were only singing over, like some parts. So I thought that it’s, you know, stupid, but fair, I guess.”

Another artist from Croatia remembered how she had no idea why her video had been taken down, and she had to use Google to discover what was happening since she received no answers from the platform. Neither had she appealed afterward.

I was struggling with it for a few days before I realized what happened. I just saw my video was taken down. And there was a note saying, like a little yellow sign, that something was wrong. It said that the video cannot be played right now. [There was] no other explanation. I tried to write and I tried to upload it again, but nothing happened. Then I Googled why.

Another creator gave up on appealing after reading an article (not on the platform) explaining that it would not “make sense” for platforms to hire someone (a human) to respond to all the questions. Therefore, she said “[I] just stopped expecting the answer every time I have a problem.” Artists often think there is nothing they can do to influence the decision of the platform: “When there’s a problem, there’s no way of solving it,” concluded a Croatian tattoo artist. In consequence, there is still ample room for improvement on the side of platforms to educate users about the guidelines and exact processes of appeals and decisions (Myers West, 2018; Suzor et al., 2019).

Regulative and Technical Dimensions Influencing the Normative: Self-Censorship and Cultural Content Adjustment Due to Algorithmic Copyright Moderation

Even if creators themselves had not experienced copyright moderation, they tended to anticipate it and thus adjust the content beforehand. A Bulgarian visual artist said:

The music I’ve had [to adjust] too, especially when we’re shooting from shows and we have music playing at the shows. I have to . . . either delete or distort the whole sound of the videos that I’m posting because I’ve heard people [have problems because of that]. It’s not what I would wish to do since it is influencing the work. I don’t know how to explain it, it changes the content when you change the music. . . . I think I’ve done it once or twice with paintings, because they have been replicas and were very close to somebody else’s work. So I would change them a bit for social media.

Several other interviewees talked about music issues they needed to “solve” before posting anything on social media. “I wasn’t used to it at first and then I had to adjust a few things,” said a creator from Romania. “For example, [I asked] my friends that have bands or my boyfriend who is a composer to do his own thing, to send me his own music so that I don’t have any copyright issues.” She added that she did not know how others were coping with this strain, as not many people have “boyfriends who write music.”

A creator from France added: “When I need to have cover music, I post my own or I post one [track] without copyright claims. So I’m very careful about not having any [issues] with copyright.” Some interviewees confessed that they had quit posting certain products to platforms altogether due to their uncertainty about copyright moderation. As one participant stated: “I stopped uploading videos on Facebook because I don’t know how to make them interesting without all the filters and music and everything, so I when I post a video I always post it only on Instagram.” Here, the technological dimension (the Instagram platform design at the time of writing), had a significant influence on cultural production. Instagram had a special list of music that one could use without incurring a copyright claim. Despite making it easier for creators, this governed the creative process by assigning pre-approved music.

“I think we had a Lip Sync Battle with gallery workers,” said a participant from Bulgaria. “We sang a very popular song and we uploaded it because it was like a promo with gallery workers. [It] got taken down in a matter of an hour, I think. It was [on] Instagram.” After this, she only used “approved” music from the list.

“I don’t upload my own music,” said another interviewee.

I just use the music that is there because it would take too much time. I think if I upload my own music, they will recognize [it], and you will get an email that it’s the copyright and a scam or, you know, they will delete it. . . Unless you compose the music yourself, obviously, then you’re good, but I’m not a musician.

An artist from the Netherlands who also used pre-approved music for reels on Instagram said that this directly influenced his creative process:

I think it’s a collaboration, more than stealing someone’s copyrights or something. For example, my art is many many times inspired by vintage or classical stuff. I use music from the 1920s or 1930s. For example, last time, I drew Marilyn Monroe. Obviously, many artists do that. But you will put in that reel . . . you will put the song of Marilyn Monroe [in], right? To enchant it, to highlight it. And that’s it.

The empirical findings suggest that self-censorship and content adjustment are prevalent in artists’ experiences with cultural production on platforms in relation to copyright moderation. Duffy and Meisner (2023) have identified similar forms of “suppression” in their study on creators.

Regulative Dimension and Timing: Had Copyright Moderation Increased During the Previous Year, in the Views of Creators?

When asked directly, most participants did not think that moderation due to copyright issues had increased during the previous year. However, when they remembered instances of such moderation, these were often from the relatively recent past. The interviewees used phrases, such as “Last Christmas” (referring to a lip sync video by a creator from Bulgaria), “last year” (a reel by a creator from Croatia), or “during the past three or four years” (a video by a creator from Romania). While the adoption of Article 17 of the CDSMD cannot be identified as the only cause of increased perceived copyright moderation, it indicates that according to the interviewees, this form of moderation has recently expanded.

The study was conducted when Article 17 was just in the process of adaptation to national legislative regimes in the EU; that might explain why the results of the temporal dimension were not as conclusive as other results. Future research should investigate these questions systematically from a temporal dimension, as the changes in copyright content moderation might have strong implications for the creative process and the distribution of creative work.

***

As this analysis of the discursive, normative, and technical dimensions shows, anticipating or experiencing copyright moderation does indeed potentially change the cultural process and the cultural products that appear on social media platforms. Creators engaged in self-censorship because they anticipated moderation, which applied to both those who had and had not experienced it themselves. While it was unclear whether the latest regulation in Article 17 of the CDSMD had influenced the increased copyright moderation, the timing of the copyright-related moderation experiences of the creators indicated that this might have been so. Thus, research questions 2 and 3 have been addressed: the increasingly restricted copyright regime and automated copyright moderation were found to influence the creation process and the cultural products posted on social media platforms, according to the creators themselves.

Conclusion

For several years, researchers have expressed concerns that algorithms exercise excessive power in influencing social realities (Gillespie, 2018; Kitchin & Dodge, 2011). In addition, algorithmic content moderation on private platforms has been compared to a “black box” (Pasquale, 2016). One important observation was that “. . . there are multiple sources of opacity—institutional, legal and technological—that make it difficult to evaluate automated private regulatory systems” (Gray & Suzor, 2020, p. 7). Meanwhile, questions about when and whether automated content moderation requires a “human in the loop” have also been asked (Gorwa et al., 2020, p. 15).

This study contributes to these questions of platform governance and algorithmic governance by revealing that both the creative process and the cultural products on social media platforms are strongly influenced by algorithmic copyright content moderation. In extending the understanding of how algorithmic copyright content moderation influences creative work on platforms, perhaps the most important finding is that creators engage in self-censorship by not posting certain content or adjusting it in advance. For many artists, their anticipation of platform punishments directly influenced the cultural products that they produced. In addition, we built on and contribute to the literature on social media creators and how interact with platforms and the regime they afford: Because the creators found the regulative dimension of algorithmic copyright moderation to be opaque, they engaged in algorithmic gossip (Bishop, 2019) and employed user folk theories (DeVito et al., 2017) to attempt to guess which practices were accepted and which were not. They also developed their own theories (Caplan & Gillespie, 2020) and learned through their own “mistakes” and those of the community. Thus, the normative and discoursive dimensions of copyright governance on social media platforms did not align with the regulatory dimension and appeared to lead to frequent self-censorship of cultural production on social media platforms.

Furthermore, the technological dimension of automated copyright moderation algorithms was unclear, and the appeal and complaint processes on social media platforms seemed non-functional. As these were unhelpful, they were not used by the creators. In terms of timing, social media platforms seem to have started implementing more algorithmic copyright moderation after the EU member states approved the European CDSMD (2019/790). Some interviewees had had old and new videos or posts taken down during the previous 2 years due to copyright moderation.

We therefore can conclude that for artists engaging in creative work on social media platforms, the multiple uncertainties about automated copyright content moderation often lead to self-censorship of their creative process and products.

Several important policy implications emerge from this research, such as the discrepancy between regulatory and governance regimes, between discourses and norms of the cultural producers and policies, regulations, and technical affordances of the platforms. A change in this bleak situation would require far more transparency in platform governance from both policymakers and the tech giants themselves as part of their relationships with creators of cultural products on social media platforms. EU’s Digital Services Act (DSA) is clearly a promising regulatory framework for pushing platforms for more transparency and accountability, yet the implementation still yields many challenges.

In conclusion, and to answer the question posed about the need for a human presence in the content-moderation loop, the words of a Croatian artist are especially appropriate:

I would have more actual people look at copyright moderation because when it’s just a computer or an algorithm scanning the internet for repeating sounds and repeating images, it doesn’t really work and content gets taken down when it has the right to be used.