In a global context where political campaigning, social movements, and public discourse increasingly take place online, questions regarding the regulation of speech by social media platforms become ever more relevant. Companies like Facebook moderate content posted by users on their platforms through a mixture of automated decision making and human moderators.Footnote 1 In this content moderation process, human rights play an ambiguous role: those who struggle with marginalization may find a space for expression and empowerment, or face exacerbation of pre-existing bias.Footnote 2 Focusing on the role of human rights in Meta's content management, this essay explores how the protection of speech on social media platforms disadvantages the cultural, social, and economic rights of marginalized communities. This is not to say that speech on social media platforms is devoid of emancipatory potential, but that this potential is not uniformly or equally accessible. We see the incorporation of human rights considerations into decision-making processes as an avenue for alleviating this challenge. This approach faces obstacles from the platforms’ business models, which decenters human rights concerns, and from the limitations of liberal accounts of human rights. From within and against these constraints, human rights can be mobilized as emancipatory power in an effort to decrease marginalization.
Bias in Freedom of Speech
Norms protecting speech as a human right have developed and continue to be reiterated through practice dominated by the values and interests of those already privileged.Footnote 3 In fact, there is a general tendency to hierarchize and privilege dimensions of individual liberties, to minimize societal concerns, and to frame freedom of speech in a manner that benefits the privileged.Footnote 4 The presence of this hierarchization does not preclude the use of human rights in struggles that do not align with those interests,Footnote 5 but leads to the incorporation of bias into the way in which freedom of speech can be exercised.Footnote 6
There is a double standard for the conceptualization of what it means to exercise free speech. Those who are already privileged operate from a sphere of protection closely aligned with their practices.Footnote 7 For instance, the traditional dress of Bavarian women, as seen during Oktoberfest and involving sexualized displays of their breasts, does not interfere with their ability to engage in speech on Facebook.Footnote 8 In contrast, the Brazilian National Foundation of Indigenous Peoples had their Facebook account suspended in 2018 for a post about traditional knowledge portraying two Waimiri Atroari in their traditional dress, which leaves the nipple bareFootnote 9—the nipple being the one body part Facebook explicitly bans—in order to prevent “uncivilized” pornographic material, inconsistent with Western values.Footnote 10 Against this bias toward Western values, the Brazilian Indigenous women had to defend their dress to the extent of needing a judicial decision to have their accounts reinstated—before being able to exercise their right to speak. In short, Facebook's restrictions on nudity differ across cultures with regard to their impact on speech.
This structure has substantial marginalizing effects. In Ratna Kapur's words, “the liberal tradition from which human rights have emerged not only incorporates arguments about freedom and equal worth . . . it also incorporates civilization, cultural backwardness, racial and religious superiority.”Footnote 11 In our example, bare nipples do not accord with Facebook's standard of civilization, while expansive décolletage does. Consequently, Bavarian women are immediately free and “equal,” while Indigenous women need to take additional steps “to climb up” into that “free” space: they are pushed further to the margin.Footnote 12
Structural Bias in Social Media Content Management
Increasingly, social media platforms are not only spaces for human rights struggles to exercise free speech and document evidence of rights violations, but also spaces for the repression of rights. Popular social movements such as #BlackLivesMatterFootnote 13 demonstrate that marginalized groups can carve out a space on platforms to express their concerns. However, bias remains—often with exacerbating socioeconomic implications. For instance, Facebook uses seemingly neutral knowledge to enable gentrification: “Lookalike audience” allows one to advertise housing to those who share similar interests.Footnote 14 Thus, empowerment arising out of #BlackLivesMatter may lead to increased barriers in the housing market in regions where there is less enthusiasm about that movement.
Prejudice against marginalized groups’ content is particularly attributable to context-blindness that replicates existing structural inequalities. Content management relies on static rules that are not sufficiently attentive to meaning-defining contexts. This context-blindness regularly requires those at the margins to defend and explain how their specific context changes the meaning of their speech-act. In the Wampum belt case,Footnote 15 the Facebook Oversight Board (OB) assessed the removal of a post made by Indigenous North American artists in Canada for featuring the phrase “Kill the Indian/Save the Man” as a violation of Facebook's Hate Speech Community Standard. The post was intended to denounce the history of Native Americans having to renounce their identity in order to survive, as the applicant successfully explained to the OB, leading to readmittance of the post. The initial platform decision was blind to this specific context—Meta was unable to explain how two human reviewers failed to understand the phrase “the sole purpose is to bring awareness to this horrific story” so as to qualify the post as permitted counter-speech.Footnote 16 Despite explicit labeling of the post, the artists’ speech was misaligned with presumably “neutral” but context-insensitive platform policy, with the result that they had to carve out their space for political activity before being able to use their “freed” speech. In contrast, the display of a weapon, a statement that is arguably as supportive of violence as the word “kill,” is not automatically removed—sometimes even after the authors have been banned for illegal trade in weapons.Footnote 17
Bias in platforms’ content management tends to replicate and amplify existing vulnerabilities and inequalities with implications not only for free speech but also for socioeconomic rights Footnote 18 For instance, human rights groups criticized Facebook and Twitter for systematically silencing protests about the livelihood-disrupting conditions of Palestinians in Sheikh Jarrah.Footnote 19 Such asymmetries in take-downs have been highlighted by human rights activists more generally, to the extent that the advised strategy for investigations on human rights violations is to race against the take-down time to save the social media content off-line.Footnote 20 This adds another burden in the struggle to denounce any human rights violation.
The incorporation of human rights considerations into a platform's architecture is a possible avenue for alleviating the potential biases of decision making in social media.Footnote 21 However, there are significant obstacles for such emancipatory use, as will be detailed below.
Human Rights Are Not the Main Priority
Most social media platforms’ policies and rules, particularly that of Facebook and its OB, explicitly refer to the human rights responsibilities of platforms in speech regulation. Nevertheless, the application of human rights standards remains secondary to Meta's articulated values. Those values focus—far from proclamations of liberty and equality—first and foremost on “giving voice,” not on ensuring equal and free speech.Footnote 22 This is well aligned with the corporation's business model: “more speech” means more profit.Footnote 23 In this setting, the protection of (other) human rights is regularly secondary to the corporation's priorities.Footnote 24 Assessing the Myanmar violence in 2017, Amnesty International pointed to Facebook's business model—geared toward leaving content online, even if it fuels hatred—as a major reason for the belated and inadequate content review that allowed for a vast range of human rights violations.Footnote 25 Leaked internal documents demonstrate that Facebook knew that a subsequent change in its algorithm in 2018 increased divisive and violent speech leading to human rights violations.Footnote 26 So, instead of reassessing its structure in order to reflect human rights, Facebook exacerbated what had caused human rights violations before, knowing, as said files testify as well, that content-moderation was unable to keep hate speech in check.Footnote 27
In this regard, the OB's role in highlighting ambiguities in Meta's rules and policies in consideration of human rights provisions could be viewed in a positive light,Footnote 28 but its lack of diversity, limited impact, and financial dependence on Meta limit this optimism.Footnote 29 In fact, that the OB reviewed 176 recommendations out of over 2.5 million cases qualifies not even as a drop in the ocean.Footnote 30 In sum, Facebook's structure is less attuned to a fairly balanced marketplace of ideas than to a marketplace for revenue creation through advertisement.Footnote 31
Limitations in Human Rights
To the extent that human rights considerations are implicated in the framework for decisions on free speech in social media, the process has to contend with the bias implicit in human rights. That is, there is a tendency to hierarchize and privilege dimensions of individual liberties with minimal consideration to societal dimensions that cannot always be addressed through protections for individual liberty.Footnote 32 In fact, the focus on speech exacerbates the tendency to overemphasize individual liberty, to the detriment of considerations about social, economic, and cultural rights. In the example of the Indigenous women in Brazil, Facebook's community guidelines do not provide for a group right to culture, but require that the post is taken down—that freedom of speech is reclaimed and reassessed in light of the interest to talk about cultural rights to dress in traditional clothing.Footnote 33
Furthermore, most biases on platforms are manifestations of existing socioeconomic hierarchies and marginalization that cannot be addressed through the individualistic lens of human rights.Footnote 34 The marginalization of Indigenous communities, amplification of disinformation against minorities in Myanmar, and silencing of Palestinians are all rooted in pre-existing inequalities. The root of such marginalization is more profound than an incident of individual rights violation,Footnote 35 and the resultant harm, such as perceived and reported discrimination, is a symptom of a deeper structural problem such as racism and other forms of prejudice.Footnote 36
It is, thus, crucial to pay attention to potential obstacles such as limitations within the human rights system itself and platforms’ business models, which decenter human rights concerns. With those priorities in mind, the effort to incorporate human rights considerations into platforms’ system of speech regulation may become a positive step toward countering bias.
Human Rights to Counter Bias
Structural bias in human rights and content management need not be mutually reinforcing. In fact, the emancipatory potential of human rights can play a part in resistance against prejudice in and beyond platforms’ content management. There is a conceivable space for human rights work and struggle on social media platforms. The transnational and accessible character of social media networks is enabling global cooperation around the exposure of human rights violations by counter-power movements.Footnote 37 For instance, the movements during the Arab Spring relied on communication via Facebook.Footnote 38 While this appraisal has been celebrated as a Facebook revolution, movements less aligned with Western values received less favorable treatment. In other words, as long as the social movement promotes what the West deems “universal,” social media platforms provide a welcoming space to promote their cause.Footnote 39
For those adversely affected, in order to overcome the limits of human rights in addressing structural bias on platforms, a focus on human rights’ emancipatory power is promising. Human rights claim-making can push beyond the structural constraints of rights and can organize and unify people in their struggle against bias.Footnote 40 Returning to our example of the Brazilian Indigenous women, their cultural rights provided them with a tool to combat exclusion. For Facebook, a next step toward equality and inclusion would mean, on the one hand, to push these cultural rights further into the structure of Facebook's content management so that take-down protocols are more careful in the assessment of cultural rights, and on the other hand, to expand and protect the digital space in which Indigenous communities can discuss, promote, and transform their cultural rights.Footnote 41
Conclusion
Free speech is a central object of concern both in a human rights framework and in the regulation of content on social media platforms. It can become a site of contention when hegemonic values encounter positions and inputs from those at the margins challenging said values. Certain kinds of speech are more robustly protected than others, thereby marginalizing those whose practices are at odds with dominant speech acts. In other words, the way in which freedom of speech is protected, both as a human right and in social media, is biased. While there are emancipatory, unifying elements in a human rights framework that carry the potential for using it as a framework to better regulate social media, without careful consideration there is a risk of merely reinforcing and amplifying bias when selectively protecting free speech. Most importantly, typical human rights formulations are unable to capture structural bias, and therefore tend to neglect or even exacerbate such problems. From within and against those constraints, these platforms’ potential as a space for the struggles of human rights movements—mobilizing the emancipatory power of human rights—may provide a path to counter bias, and toward positive transformations of rights and social media platforms for marginalized peoples.
Open access
