Following years of accusations of negligence or at least wilful blindness to radical right actors' use of their sites for recruitment and the dissemination of hate, Facebook/Meta and several other social media platforms have advertised new policies for better moderating content. With almost 3 billion current users, the act of regulating Facebook may be likened to a Herculean, perhaps even Sisyphean, task. As New York University's Stern Center for Business and Human Rights observed, each day, Artificial Intelligence (AI) flags upwards of 3 million pieces of unique content for potential community standards violations. With a prior white paper admitting that, conservatively, these detection mechanisms making an error in at least 10 per cent of cases, it stands to reason that hundreds of thousands of pieces of violating content remain on the site unchallenged. Meanwhile, crack downs on content have been flatly decried by the extreme right, validating accusations of the existence of a undemocratic elite class that works to stifle free speech. More to the point here, though, is the reality that these facts and figures leave out another critical area for those attempting to map hate on Facebook and to evaluate moderation successes: content that is successfully constructed to avoid AI detection.
Recently, when trolling public Facebook groups that teeter on the edge of extremism and community standard violations, variations on a meme appeared again and again. The consistent element of the meme is as follows: a banner at the top of the image that reads 'In order to stay out of Facebook jail I will be posting memes in code,' with a large photograph of US President Joe Biden beneath it. Underneath the picture of the president are the words 'Is A...' with two stock images below them. In essence, the meme is one designed to compare President Biden to various insulting items. For instance, one shows stock photo of a Summer's Eve douche and then a photograph of a paper bag. In another, there is a picture of a bike pedal and then pictures of nail files, a riff on the accusation of pedophilia.
Admittedly, the memes are extremely rudimentary. They amount to mere acts of name-calling--much like the kind of the found on the school yard--with no serious argument to them, no attempts to convince a sceptical reader of the merits of any political claim. Equally basic is the design--the images of President Biden are often the same image of him delivering a speech, while the other features photographs are low-quality, hyper-pixelated stock images. Basic as they may appear, they have drawn quite an audience. One such meme comparing him to a douchebag on one Facebook group alone garnered more than 550 comments (all either likes, loves, or laughter), 299 shares, and 28 comments. Meanwhile, one on another Facebook group likening him to a pedophile earned almost 470 positive responses, 500 shares, and 50 comments.
Although the sizeable audience for these memes in itself supports claims that memes have some enduring influence on social media, it is also important to discuss the implicit facets of these memes, namely, 'Facebook jail' and how to avoid it. Aside from just insulting the president, these posts have the more subtle outcome of reinforcing a hostile mentality between Facebook user and Facebook (and, by extension, with those who run Facebook). Clearly, the phrase 'in order to stay out of Facebook jail I will be posting memes in code' builds on preconceived notions of what Facebook moderates and casts the company's community standards in a negative light. It also evidences the tactic of using image codes by fringe communities seeking to stay on more mainstream platforms despite the risk of detection.
However, on closer inspection of the memes and of Facebook's community standards, it appears that the content these memes claim would result in a hiatus in Facebook jail may not traverse that threshold. It neither threatens violence nor criticises President Biden on the basis of his race, gender identity, sexual orientation, class, ethnicity, ability status, or religion. The memes do not feature false news or peddle in conspiracy theories. Instead, they are merely pictorial depictions of political mudslinging. While likening the sitting president to a 'cock-sucker' (as one other meme did) hardly advances political discourse or promotes civil cross-party conversation, Facebook's policies are not aimed to enforce that level of 'civility.'
This being the case--and one relatively easy for the posters of these memes to discover (not least given that they purport to know what would see them thrown in 'Facebook jail')--the question becomes a matter of why go through the effort of making these memes and staging the message in this way. To some extent, it may be attributed to the desire to speak to their perceived audience, which is often considered youth, who have a perceived propensity towards political memes, and those who may feel disaffected by the state without feeling up-in-arms about government policies. Further, I believe that these memes are clear attempts of the more critical, longer-term plan to sow distrust between Facebook users and moderators, which is itself a cypher for 'the people' against 'the elites.' If the posts are uploaded and then eventually taken down, it reinforces the user's point. If they are not, readers are made to feel as though their community had outsmarted the free-speech-hating moderators of Facebook.
The anecdote that a picture is worth one thousand words may slightly over-exaggerate the power of any one of these given memes (not least ones that feature 3 photographs). That said, they also cannot be ignored, no matter how pixelated the stock-photos. The anti-Biden meme discussed here represents one of many attempts at amusing, coded, pictorial propaganda by those on the fringes of Facebook. The site is now faced with the question of how to moderate this kind of content if at all, how to prevent moderation policies from being re-characterised in a negative light so as to breed and feed conspiracy theories (in this instances mostly populist and anti-Semitic theories), and how to train AI to more successfully intervene even as the proverbial mice they attempt to catch grow, learn, and react constantly.