More than fifty shades of grey: Copyright on social network sites

3.1 YouTube videos and comments

In the first case study, I studied the development of antagonistic interaction (called ‘drama’) among a group of users across more than twenty YouTube videos. The drama I observed and subsequently analysed and wrote about occurred among a group of Christian and atheist users from 2007 until 2010 (Pihlaja 2011, 2014). This is now, in many ways, an historical look at YouTube, which has changed significantly since those early days, particularly in terms of how the site controls content. In 2009, when the key videos in my study were published, YouTube was still relatively unmoderated. While work was being done to eradicate the publishing of copyrighted material produced by large media companies, norms for appropriate social activity on the site were still developing and users would often engage in bullying and aggressive interaction. The ‘drama’ I studied involved the insulting of one user by another, the religious and social questions that grew out of his subsequent attempts to justify his actions, and the response of others in the community to this display of ‘hate’ towards one person by another. Linguistically, I was specifically interested in the use of metaphor from the Bible, as this was a key part of the disagreement, and my analysis traced the use of different metaphors over time to see how their uses changed, depending on the context.

For this project, I included three kinds of language data: my transcripts of the YouTube videos, text produced by the users posting videos in description boxes of videos and video tags, and comments posted by users. In the case of the description box videos, the question of copyright was relatively straightforward. The video page owner produced the text and could be assumed to be solely responsible for its content. The comments, however, were less straightforward in terms of ownership. Although the commenter produced the content and was able to remove the content, the comment exists on the video page of another user’s video, who controls and curates the content, with the ability to remove comments. Finally, I produced the transcripts of the videos from watching the video, following the transcription conventions I had chosen. Given that I produced the transcripts, I argued that I retained the copyright on them, as they were materially different from the videos.

An additional difficulty emerged in this research: the volatility of the videos I was observing. Early on, it became clear that users would post content to the site, which they might remove themselves or which might be removed by YouTube for a violation of the YouTube Terms of Service. The main video that led to the drama I analysed (Pihlaja 2014) was removed within several days of being posted. It was never clear if the user had removed the video himself or if it had been removed by YouTube. Problematically, the video had then been uploaded again by other users, sometimes in a ‘remixed’ fashion, but also sometimes as a ‘mirror’ of the original, with no editing and clearly violating copyright law. Ultimately, most videos (with the exception of two or three) were removed as users quit YouTube, closed accounts, and/or YouTube removed them. The removed videos produced a challenge both in terms of copyright and ethical standards in reproducing content from them (cf. Spilioti forthcoming). I argued that I held the copyright of the transcripts. However, the description box and comment content, which had been publicly available at the time I gathered the videos and comments, but may eventually have been removed either during the analysis stage or after I had published articles, was more problematic as the content was no longer published and publicly available. I argued, therefore, that my fair use of them related to the time that I accessed and copied them, following fair use and dealing rules. Like a copy of a book that might eventually be destroyed, reproducing this content was still, in my view, appropriate under the license I had originally made them.

More problematic, however, was the ethical consideration of reproducing something that a user may have explicitly attempted to remove, despite my own transcripts. If a user had said something and then publicly retracted what they had said, I reported this clearly in my own writing and cited the date I accessed the video. In my research of publicly available interaction among moderately popular YouTube users (videos in the drama had several thousand views and hundreds of comments, large numbers for 2009), the users present themselves as public figures. They refer to their large number of fans and subscribers and explicitly work to publicize their channel by encouraging others to subscribe to and share their content. Their personalities and positions are important to the process of understanding and describing their actions. Any attempt to anonymize the videos would also have been unsuccessful, as the videos were easily searchable once they were described. Anonymizing them was impossible and did not respect their own positioning towards the online content, which they were clearly aware was being viewed by large numbers of users, including those making public commentaries about their opinions. Indeed, that was the point of making the videos and engaging others: to have their positions discussed, as we will see in the Facebook example below when the video creator repeatedly asks users to share and comment on the video. While the users may not have expected academics to analyse their writing, the same could be said of much copyrighted material in books and newspapers. The creator might not specifically foresee them being analysed in an academic context, but their use in academic analysis is not unreasonable or unethical.

Changes to YouTube, and particularly the integration of Google’s social network site Google+ with the YouTube commenting function, mean that YouTube comments are now linked to profiles beyond YouTube channels. Although this initially meant users’ real names were being associated with comments, this ‘real-name’ policy was eventually abandoned (Ruesch and Märker 2012) leaving a mix of screennames and clear pseudonyms (like ‘Barack Obama’ or ‘Peter Pan’) and ‘real names’ on the site comments. Whether to anonymize comments may then become a question of individual interactions. If, for example, a user’s persona of ‘Barack Obama’ affects the subsequent discourse, then anonymizing the usernames at the point of data collection could have a negative effect on the subsequent analysis. It was also not clear that anonymizing comments, in the case of YouTube comments, would succeed in avoiding causing harm to users, especially if the comments could be found through a search anyway. In my own research, I chose to keep the user names, be they users’ real names, pseudonyms or screennames, while analysing the data. When choosing which comments to publish and how to cite them, I favoured examples where users had a sustained online presence and their argument was essential in order to represent and illustrate the key arguments in the analysis. In doing so, I respected available content as intellectual property, but also avoided problematic cases such as one-off comments from users with their real names.

In the case of dealing with problematic content, researchers must act in a manner that is in line with the ethical frameworks of their institution, any available learned association guidelines, and their own ethical stances related to what is personally acceptable to them. For researchers uncomfortable with the potential issues involved in reproducing copyrighted content, it is advisable simply to avoid reproducing them. In many cases, researchers can avoid exact quotes or direct references to users, although applied linguists may feel that they need to be able to analyse and thus publish attested examples of language use. For researchers, the principles of reporting honestly and accurately, as well as avoiding subjective judgements, are important to keep in mind when reporting on the interaction of users. Although unable to completely remove the risk of harming users by reproducing content in a fresh context of academic research, researchers attempting to ‘do no harm’ can mitigate concerns by acting and writing in a way that does not target users for ridicule or critical comment. As with any project, researchers must always reflect on why they are carrying out their work and for whom, and whether the potential good outweighs or justifies the potential harm to themselves or to participants.

3.2 Public Facebook content and real names

While the first case study dealt with YouTube pages which are public-facing and intended to engage diverse, general audiences, the second case study deals with Facebook, which has a spectrum of privacy settings, allowing content to be viewed by only a small number of friends, or anyone with an internet connection, as well as many settings in between. The comments on Facebook videos present more significant challenges in terms of making decisions about reproducing content and deciding what exactly should be treated as ‘copyrighted material’. Facebook historically has been more oriented in terms of its design towards interaction among ‘friends’ whom a user accepts and allows to view content on their ‘wall’. How a researcher deals with content on the site must then take into account the expectations of users and the value of presenting content as belonging to individual users often using their real first and last names.

In this second case study, I looked at inter-religious dialogue among different users on Facebook and YouTube: I used the public Facebook videos and comments from a popular YouTube Evangelist, Joshua Feuerstein. Feuerstein came to prominence on Facebook and YouTube in 2014 when his video ‘Dear Mr. atheist’ was responded to by the popular YouTube user theamazingatheist. Feuerstein’s videos are short, often about inflammatory subjects in the American socio-political environment, such as gun control, abortion, and feminism, and they are meant to be shared, with Feuerstein encouraging users to always ‘Share if you care’ at the end of his videos. Feuerstein uses Facebook almost exclusively as in 2015 his platform to publicise his videos, with nearly two million likes on his page. In addition to inflammatory content, Feuerstein also posts videos including ‘uplifting’ content, aimed at Christian users to encourage people who are facing difficulty in their lives. The comments sections on his page are then a mix of arguments about controversial subjects, users’ tagging of friends, thanks to Feuerstein, and support for one another.

For this project, I was interested in how users positioned themselves in relationship with one another. I used corpus-led discourse analysis to identify broad themes on each channel and then looked at how narratives about the world and others were used in order to take certain positions. I looked at the way in which Feuerstein uses Facebook in comparison to the work of other ‘evangelical’ users from atheist and Muslim perspectives. While I am interested in the development of arguments on his page, Feuerstein’s page is not often oriented towards debating particular issues. Instead, users often simply offer simple responses like ‘Amen!’ to Feuerstein’s videos or use the tagging function to add the name of an additional user who has not chosen to comment on the video, but whose attention is being drawn to the video by another user. In this case, although users might not necessarily have the assumption that their comments are private, they are oriented towards the specific community around Feuerstein. At the same time, the practice of tagging others suggests that care must be taken not to involve users who might unwillingly be appearing on the page, simply because their privacy settings allow them to be tagged. Although some comments may also fall into the category of users attempting to publicise their own page and position on Facebook, it is difficult to provide evidence that this is the case from what has been posted on the site.

While it seemed clear that fair use allowed me to copy the content for my own research, whether or not to cite the users when I chose to represent their comments in my writing was more complicated. Initially, following the practice I developed with YouTube comments, I felt compelled to cite users by their names and with their content, assuming that since they have posted on a public domain, they are aware of the consequences of their posting. However, the value of this approach seemed limited, particularly in this project where I was less interested in mapping the development of discourse over time on the page, but rather interested in the overall content of the comments. Knowing, for example, the name of the user commenting ‘God bless you Bro. You are doing a great work for the Lord.’ was less important to me as a researcher than the content of the comment. I chose instead, where I quoted comments in my own writing as examples, to anonymize the comments, following the suggestion of Wilkson and Thelwall (2011).

As with YouTube’s integration of Google+, it is also important to consider not only the state of the technology at a certain time when making decisions about expectations of privacy, but also how the technology has changed over time. This is particularly important in the case of Facebook, which has changed significantly over the years, from being an invite-only, closed community, to encouraging users to have a publicly facing profile. Facebook algorithms push content and comments to other friends in your network, often without the creator’s awareness that the site has done this. It is fair to assume that some figures like Josh Feuerstein are using Facebook to spread their messages and are likely not confused about the public nature of their names. As discussed above, Feuerstein regularly implores users to spread his videos to have the widest reach possible, but users who might not have this expectation should be treated with deference. Simply because a user has posted on a Facebook page does not mean they intend or expect for their comment to be reproduced. Researchers must exhibit judgement in these cases, erring on the side of caution in what they choose to publish.

3.3 Comments on porn videos

While the second case study dealt with comments on public Facebook pages (which may or may not be intended for a broad audience), the third case study deals with comments on adult videos. These represent an immediate, physical response to online content, an argument I have made in a corpus-led discourse analysis project, focusing on comments made on adult videos on the popular porn-hosting website, Pornhub (Pihlaja 2016). Given that Pornhub models its interface and site language around YouTube, this study’s aim was to compare and contrast YouTube video comments with those on Pornhub, looking specifically at frequently occurring words in the corpus of Pornhub video comments and attempting to discover if and when impoliteness arose.

The project analysed 22,562 comments taken from the 100 most-viewed videos on Pornhub. Whereas the projects I have thus far discussed were focused on the content of the discourse activity and how it affected communities of users, this research had a decidedly different aim. The article looked to discover what, if any, differences can be identified between the two sites, particularly around antagonistic and impolite interaction, and offer suggestions about why differences may be present, based on language use and interaction. I used corpus linguistics to identify frequently used words and the analysis focused on seeing the extent to which antagonistic interaction was present. I was less interested in tracing specific moments of interaction between users, but rather in offering a snapshot of the corpus of comments as a whole, attempting to describe what people do when they comment on porn videos. Although the project did touch on moments of interaction among users, it did not involve close discourse analysis of interactions among users over an extended period. Instead, I was largely interested in the extent to which negative language was absent in interaction.

In some ways, the comments on Pornhub are more similar to those on YouTube than Facebook, with usernames and avatars rarely, if ever, featuring full names of users or images of their faces. Instead, Pornhub represents a site of fantasy, where unlike Facebook, user profiles are not clearly connected to an offline identity. The profiles that users make for the ‘community’ section of the site did not, in the research I did, appear to generate interaction on the site between users. In contrast to YouTube, users did not appear to be interacting with one another in any way, beyond telling others the name of a particular performer. Indeed, the findings of the study showed that although YouTube was a site of intense interaction among users around a variety of issues, Pornhub was decidedly not. Instead, the comments section provided users with the affordance to respond to the videos with short comments, often praising the performers or expressing pleasure. There was little, if any, interaction among users in the comments section of the videos. Moreover, the screennames that users chose were often oriented towards the sexual nature of the site (and could have potentially provided a further site of analysis), showing that users were decidedly oriented towards fantasy on the site.

Considering again the aims of the study and the content of the comments, like those on Facebook, the comments I chose to reproduce were exemplary of larger trends. For example, one comment I included, ‘Wonder what her name is…hmm’ was exemplar of a larger trend of users asking the names of performers. The fact that the comment was made by a particular user on a particular video was less important to the analysis of the trend. Moreover, I chose to publish quotes from comments that also represented examples of users employing violent language, while avoiding a complete reproduction of the comment. The intention behind publishing the comment in my own writing and the analysis it represented was not focused on a particular user at a particular time, but on larger trends in the dataset. All examples were indicative of larger trends, and I did not look specifically at individual users. Although usernames were published publicly with the comments, I chose to anonymize the users in the article when I referred to specific comments.

While the comments were anonymized, the text is still searchable, as they are posted on public pages. Indeed, although I have argued for anonymizing the comments on public Facebook pages and the Pornhub videos, in both cases a reader interested in the context of a particular comment in a research article or book could trace the comment back to a particular user, if the comment remains online. The researcher must do their best to balance these different concerns, taking into account all the complex issues related to the reproduction of content.

Having considered three different contexts, I now conclude with a discussion of the different data in relation to one another and the important guiding principles that must be followed in considering reproduction of creative content.