HLI102 Crop-3

\\jciprod01\productn\H\HLI\64-1\HLI102.
txt unknown Seq: 1 16-FEB-23 15:34
Volume 64, Number 1, Winter 2023
Expert Governance of Online Speech
Brenda Dvoskin*
In a world of fundamental disagreements about how social media companies should govern speech, it is
striking that nearly everyone agrees that online speech governance should be based on human rights. The
human rights project for content moderation proposes that social media platforms align their own internal
speech policies with international human rights law. It seeks, I argue, a system of expert governance: one in
which a corporate technocracy applies a set of exogenous principles imagined as objective and global.
Ultimately, this governance model shifts power to experts under the illusion of empowering the people.
To support these claims, this Article unveils the intellectual work that scholars, U.N. bodies, and the
Facebook Oversight Board are doing to portray international human rights law as an objective synthesis of
the global public interest. The Article analyzes how they have recreated several dimensions of international
law. A salient example is their new reading of the U.N. Guiding Principles on Business and Human
Rights. According to a recent interpretation, companies are expected to align their content policies with
international law. But this interpretation widely diverges from the text and the original meaning of the
instrument. The Article also examines other tools the project uses such as creating boundaries between local
facts and normative work and framing normative questions as technical challenges. Overall, the Article
provides a deep dive into the toolkit that scholars, advocates, and the Facebook Oversight Board have
developed to date to pursue a system of expert governance of online speech.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 R
I. The International Human Rights Law Project for
Online Speech Governance. . . . . . . . . . . . . . . . . . . . . . . . . 91 R
II. Justifying the IHRL Project . . . . . . . . . . . . . . . . . . . . . . . 97 R
A. The Rationales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 R
1. A New Interpretation of the U.N.
Guiding Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 R
2. Because social media companies are like states, but they
are not like states. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 R
3. Because IHRL is consented to by all states, but state
consent does not matter. . . . . . . . . . . . . . . . . . . . . . . . . . . 108 R
4. Because IHRL constrains corporate power, but its
indeterminacy is a feature . . . . . . . . . . . . . . . . . . . . . . . . 109 R
* Postdoctoral fellow, Georgetown University Law Center; Doctoral candidate, Harvard Law School;
Affiliate, Berkman Klein Center for Internet & Society. I am deeply thankful for insightful comments
from Dunstan Allison-Hope, Chinamyi Arun, Nicole Bassoff, Yochai Benkler, Elettra Bietti, Alejandro
Chehtman, Evelyn Douek, Noah Feldman, Sheila Jasanoff, Gerald Neuman, Peter Stern, Thomas Streinz,
Gali Racabi, Malcolm Rogge, and Yiran Zhang. I am also grateful to the participants at the Critical
Exploration of Human Rights Conference at the University College Dublin Center for Human Rights,
the Ideas Lunch at the Information Society Project at Yale Law School, the 2022 Global Meeting on Law
and Society, and the Science, Technology and Society Circle at the Harvard Kennedy School. Thank you
to Gabriella Papper, Deniz Aktaş, Sophia Poole, and Sara Raza for their work editing this piece. The
Article reflects developments and decisions through August 2022.
\\jciprod01\productn\H\HLI\64-1\HLI102.txt unknown Seq: 2 16-FEB-23 15:34
86 Harvard International Law Journal / Vol. 64
5. Because IHRL is a shared language and because IHRL

is like the First Amendment . . . . . . . . . . . . . . . . . . . . . . 111 R
B. Objective Justifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 R
III. Deploying the IHRL Project . . . . . . . . . . . . . . . . . . . . . . . 113 R
A. IHRL as Self-Evidently Good . . . . . . . . . . . . . . . . . . . . . . . . 113 R
B. The Relationship between Global and Regional Norms . . . . 115 R
C. Normative Indeterminacy as Technical Questions . . . . . . . . . . 119 R
D. Local Preferences as Local Facts . . . . . . . . . . . . . . . . . . . . . . . 122 R
IV. Participating in the IHRL Project . . . . . . . . . . . . . . . . . 124 R
V. The Futures of the IHRL Project . . . . . . . . . . . . . . . . . . 130 R
A. A Communal Viewpoint Developed from the Top . . . . . . . . . 130 R
B. IHRL as a Participatory Project . . . . . . . . . . . . . . . . . . . . . . 131 R
C. Disentangling Content Moderation from IHRL . . . . . . . . . . 133 R
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 R
Introduction
In April 2021, the Facebook Oversight Board (hereinafter “Board”) con-

cluded that general prohibitions on content depicting people in blackface
are incompatible with international human rights law (hereinafter
“IHRL”).1 Simultaneously, however, the Board decided that Facebook had
“met its human rights responsibilities” by adopting precisely that prohibi-
tion.2 The case originated with a video featuring two adults representing
Zwarte Piet (or “Black Pete”), the companion of Saint Nicholas in the
Dutch folklore. Zwarte Piet has been widely reported as racist because peo-
ple have traditionally portrayed it by putting on blackface.3 In August
2020, Facebook prohibited “caricatures of Black people in the form of
blackface.”4 Facebook deleted the post for violating that rule.
This decision ought to be somewhat perplexing. How did Facebook meet
its human rights responsibilities by adopting a rule incompatible with
human rights law? How did the Board explain a conclusion that, at least at
first sight, was palpably incoherent? The Board articulated various argu-
1. Case decision 2021-002-FB-UA, Oversight Bd. (Apr. 13, 2021), https://oversightboard.com/deci-

sion/FB-S6NRTDAJ/ [https://perma.cc/5PDM-VD6F] (“The Board notes international human rights
law would not allow a state to impose a general prohibition on blackface through criminal or civil
sanctions except under the conditions foreseen in ICCPR Article 20, para. 2 and Article 19, para. 3 . . .
Expression that does not reach this threshold may still raise concern in terms of tolerance, civility and
respect for others, but would not be necessary or proportionate for a state to restrict . . . .”). In all cases, I
refer to the English version of the Board’s decisions.
2. Id.
3. U.N. Comm. on the Elimination of Racial Discrimination, Concluding observations on the nineteenth to
twenty-first periodic reports of the Netherlands, ¶¶ 15-17, U.N. Doc. CERD/C/NLD/CO/19-21 (Aug. 28,
2015).
4. Guy Rosen, Community Standards Enforcement Report, August 2020, Meta (Aug. 11, 2020), https://
about.fb.com/news/2020/08/community-standards-enforcement-report-aug-2020/ [https://perma.cc/
53BN-NGFN].
2023 / Expert Governance of Online Speech 87
ments. One strand of the reasoning noted that international law protects
deeply offensive speech, suggesting that depictions of Zwarte Piet might be
protected. However, the Board determined that the international human
rights question turned on whether featuring blackface caused “objective
harm” or “subjective offense.” The Board enumerated the many experts who
had verified that content featuring blackface is discriminatory, reinforces
harmful stereotypes, can impact individuals’ self-esteem, and can contribute
to an environment of intimidation.5 Experts’ reports were considered evi-
dence of “objective harm.”6
This Article aims to explain why the Board as well as many scholars and
advocates find that IHRL is a promising framework to make decisions re-
garding the governance of online speech. It dissects the argumentative
mazes they must navigate to justify both that nonstate actors should imple-
ment international law and that the rules they propose to moderate online
content are objective applications of that framework.
Over the last decade, scholars have often compared social media compa-
nies to states due to the control they exercise over the public sphere.7 It is
therefore unsurprising that many proposals to deal with corporate power
take inspiration from state institutions.8 In that vein, what I call the IHRL
project for content moderation proposes that social media platforms align
their own internal rules to govern speech with IHRL. In other words, the
IHRL project posits that companies should adopt the international legal
framework to protect individuals’ rights to freedom of expression from state
interference in order to protect individuals from corporate power.9
Intuitively, IHRL is an attractive framework for corporate actors to gov-
ern speech on social media. It is global like social media platforms and it
offers well-respected standards to guide content moderation in these quasi-
5. Case decision 2021-002-FB-UA, supra note 1. R

6. Id. (emphasis added).
7. See generally infra Section II.A.2.
8. See Lex Gill et al., Towards Digital Constitutionalism? Mapping Attempts to Craft an Internet Bill of
Rights, 80 Int’l Commc’n Gazette 302 (2018) (discussing proposals to craft a constitution for the
internet); Kate Klonick & Thomas Kadri, Facebook v. Sullivan: Public Figures and Newsworthiness in Online
Speech, 83 S. Cal. L. Rev. 37, 38 (2019) (arguing that platforms act as legislature, executive, and judici-
ary, without any separation of powers); Noah Feldman, Facebook Supreme Court: A Governance Solution, in
Global Feedback and Input on the Facebook Oversight Board for Content Decisions Ap-
pendix, https://about.fb.com/wp-content/uploads/2019/06/oversight-board-consultation-report-appen-
dix.pdf [https://perma.cc/W37W-R5H5] (discussing the usefulness of judicial models in the context of
social media); Hannah Bloch-Wehba, Global Platform Governance: Private Power in the Shadow of the State,
72 SMU L. Rev. 27, 71-78 (2019) (drawing from administrative law to reorganize the structure of social
media companies); The Santa Clara Principles, https://santaclaraprinciples.org/ [https://perma.cc/
4MPL-STUC] (extending due process rights to users vis-à-vis platforms).
9. David Kaye (Special Rapporteur on the Promotion and Protection of the Right to Freedom of
Opinion and Expression), Report of the Special Rapporteur on the Promotion and Protection of the Right to
Freedom of Opinion and Expression, U.N. Doc. A/HCR/38/35 (Apr. 6, 2018) (hereinafter “SR Report
2018”).
public spaces.10 The IHRL project proposes to address the democratic defi-
cits of corporate online speech governance not by suggesting a more par-
ticipatory institutional structure, but by offering a set of global substantive
rules that purport to reflect the public interest, the common good, or glob-
ally shared values.11 Indeed, blurring the lines between states and corpora-
tions could lead to seeing users as political subjects with an interest in
participating in the making of policy. Instead, the IHRL project proposes
that experts implement IHRL as an already available formulation of social
interests.
The project has gained more attention and traction since David Kaye pro-
posed it in 2018 during his tenure as U.N. Special Rapporteur.12 He recom-
mended that large social media platforms adopt IHRL as their own default
rules to moderate content.13 Also in 2018, Mark Zuckerberg announced that
Facebook would create an overseeing body composed of independent experts
to review some of Facebook’s content decisions and to make recommenda-
tions on how the company could improve its content governance.14 From
2021 to the present, the Board has applied the IHRL framework to its deci-
sions. The Board has been the main executor of the IHRL project to date.15
IHRL does not assure a panacea, but it makes appealing promises. Its
main purpose is to put the public interest front and center of online speech
governance. Kaye asserts: “[i]t’s time to put individual and democratic
rights at the center of corporate content moderation.”16 The project chal-
lenges the idea that no global set of rules to regulate speech exists. In his
2018 report, Kaye stated, “[t]he founder of Facebook recently expressed his
hope for a process in which the company ‘could more accurately reflect the
values of the community in different places.’ That process, and the relevant
standards, can be found in human rights law.”17 Thus, the IHRL project
10. See Sejal Parmar, Facebook’s Oversight Board: A Meaningful Turn Toward International Human Rights
Standards?, Just Security (May 20, 2020), https://www.justsecurity.org/70234/facebooks-oversight-
board-a-meaningful-turn-towards-international-human-rights-standards/ [https://perma.cc/LTW4-
E6HS].
11. David Kaye, Speech Police: The Global Struggle to Govern the Internet 18 (2019).
12. SR Report 2018, supra note 9. See Rebecca MacKinnon, Consent of the Networked: The R
Worldwide Struggle for Internet Freedom (2012) (pioneering the importance of human rights in
the context of social media).
13. SR Report 2018, supra note 9. See Evelyn Douek, U.N. Special Rapporteur’s Latest Report on Online R
Content Regulation Calls for ‘Human Rights by Default’, Lawfare (June 6, 2018), https://
www.lawfareblog.com/un-special-rapporteurs-latest-report-online-content-regulation-calls-human-
rights-default [https://perma.cc/X3HT-5CDA].
14. Mark Zuckerberg, A Blueprint for Content Governance and Enforcement, Facebook (Nov. 15, 2018),
https://m.facebook.com/notes/mark-zuckerberg/a-blueprint-for-content-governance-and-enforcement/
10156443129621634/ [https://perma.cc/KD2H-7CWU].
15. Announcing the Oversight Board’s first case decisions, Oversight Bd. (Jan. 2021), https://over-
sightboard.com/news/165523235084273-announcing-the-oversight-board-s-first-case-decisions/ [https:/
/perma.cc/JRG9-BBZ7].
16. Kaye, supra note 11, at 18. R
17. SR Report 2018, supra note 9, ¶ 41 (internal footnotes omitted). R
claims that it can provide a basis to moderate content that already reflects
the global public interest.18
By positing that we already know what rules reflect the global public
interest, the IHRL project vests legitimacy on new “new governors”: human
rights experts.19 The Board’s members or companies’ in-house human rights
directors—those who know these universal rules—can implement these
rules on social media. The experts’ democratic credentials would proceed not
from democratic politics but from the fact that they deduce their decisions
from IHRL.
I call this project a system of expert governance. Its legitimacy comes
from assuring the public that IHRL is an objective account of the public
interest or global values, capable of producing rules and decisions that also
reflect those values. Objectivity in this context means the opposite of indi-
vidual policy preferences. It is also the opposite of politics. I do not imply
that human rights themselves are not a political commitment. The IHRL
project’s claim to objectivity stems from portraying that commitment as
universal and pre-existing to the decisionmaking moment. Thus, the system
is objective because it is supposed not to favor anyone’s specific standpoint.
As Sheila Jasanoff, a pioneer in the field of Science and Technology Stud-
ies, says, objectivity takes hard work.20 Performing as objective governors
(judges often play this role) requires construing principles as being already
available and agreed on, as well as framing decisions as reasonable deriva-
tions from them. In the case about Black Pete, the Board had to present
IHRL as the framework it was compelled to rely on and to claim that it
deduced its decision from that framework, even though the Board acknowl-
edged that the exact opposite conclusion was also possible.
The main purpose of this Article is to make that work visible. It offers an
analysis of the tools that scholars, advocates, and the Board have developed
to build that claim to objectivity. By examining how objectivity is per-
formed, this Article makes four contributions.
First, it unveils the intellectual work that has gone into presenting IHRL
as a global set of rules appropriate for governing speech online. Sometimes
the effort toward objectivity is explicit. Like in the Black Pete decision,
experts might explain the work they are doing to construe the principles
that allegedly bind them and the parameters for diverging from them. Most
often, the work is done in the dark.21 This should be unsurprising to legal
18. Id., ¶ 42.

19. See generally Kate Klonick, The New Governors: The People, Rules, and Processes Governing Online Speech,
131 Harv. L. Rev. 1598 (2018).
20. Sheila Jasanoff, The Practices of Objectivity in Regulatory Science, in Social Knowledge in the
Making 307, 308 (Charles Camic, Neil Gross & Micèle Lamont eds., 2011).
21. See McDonald v. City of Chicago, 561 U.S. 742, 805 (2010) (Scalia, J., concurring) (“Justice
Stevens abhors a system in which ‘majorities or powerful interest groups always get their way,’. . .but
replaces it with a system in which unelected and life-tenured judges always get their way. That such
usurpation is effected unabashedly. . . —with ‘the judge’s cards . . . laid on the table,’. . .—makes it even
scholars. Just as judicial courts have developed theories of interpretation that

legitimize their work,22 proponents of the IHRL project have done a great
deal of intellectual work to build the principles that ought to constrain ex-
perts’ decisionmaking process. For example, the thought leaders of the pro-
ject have argued that the United Nations Guiding Principles for Business
and Human Rights (hereinafter “UNGPs”) set the expectation that compa-
nies will align their policies with IHRL.23 This interpretation serves the
purpose of presenting this instrument as an exogenous basis for the IHRL
project. Looking at how creative this interpretation is and how widely it
diverges from previous interpretations illuminates the work done behind
principles that are portrayed as objective.24
Second, by bringing this intellectual work into the light, the Article ex-
tends long-standing critiques of courts to the Board.25 The Board presents
IHRL as a reflection of a global agreement or set of values about how to
regulate speech. That conceptualization of IHRL allows the Board to hide
away its power, that is, to perform as an enforcer of the global public inter-
est instead of as a political actor making policy. The risk, however, is that
this version of the IHRL project undermines its own purpose. While the
goal at first was to bridge the divide between public and market actors and
include the public in the governance of social media, the outcome is to
reaffirm that the participation of the public is ultimately not necessary be-
cause its interests are already embedded in IHRL.
Third, looking at how corporate actors implement international law illus-
trates the consequences of importing state institutions in the hope of ad-
dressing the problematic aspects of corporate power. When incorporated
into nonstate actors, state institutions go through radical transformation to
fit the existing structure and purposes of corporations. This often leads to
worse. In a vibrant democracy, usurpation should have to be accomplished in the dark.” (internal citations
omitted; emphasis added). I am grateful to Libby Adler for highlighting this quote.
22. See Noah Feldman, Written Statement to the Presidential Commission of the Supreme Court of the United
States Hearing on “The Contemporary Debate over Supreme Court Reform: Origins and Perspectives”, The White
House (June 30, 2021), https://www.whitehouse.gov/wp-content/uploads/2021/06/Feldman-Presiden-
tial-Commission-6-25-21.pdf [https://perma.cc/4V6A-RSXU].
23. See Evelyn Aswad, The Future of Freedom of Expression Online, 17 Duke L. & Tech. Rev. 26, 34
(2018); David Kaye (Special Rapporteur on the Promotion and Protection of the Right to Freedom of
Opinion and Expression), Mandate of the Special Rapporteur on the Promotion and Protection of the Right to
Freedom of Opinion and Expression, 1, 2 U.N. Doc. OL OTH 24/2019 (May 1, 2019); Facebook Oversight
Board: Recommendations for human rights-focused oversight, Article 19 (Mar. 27, 2019), https://
www.article19.org/resources/facebook-oversight-board-recommendations-for-human-rights-focused-over-
sight/ [https://perma.cc/399N-JUVV] (“While Facebook is not legally bound by international human
rights laws, the UN Guiding Principles on Business and Human Rights set out responsibilities that
companies like Facebook have to respect human rights. This means ensuring that their Terms of Service
and Community Standards are fully in line with human rights laws”).
24. See infra Section II.A.1.
25. See Jeremy Waldron, The Core of the Case against Judicial Review, 115 Yale L.J. 1346 (2006); Mark
Tushnet, Following the Rules Laid down: A Critique of Interpretivism and Neutral Principles, 96 Harv. L. Rev.
781 (1983).
puzzling outcomes.26 This Article examines what gets lost in translation

when corporate actors become international lawmakers.
Finally, the Article reviews many of the Board’s decisions to date. Because
the Board has become the main executor of the IHRL project, its decisions
are a useful site to see the project (or at least one version of it) at work. Thus,
the Article reviews many of the Board’s first cases, although this review is by
no means a comprehensive review of all of the Board’s work.
I proceed as follows: the first section offers a description of the IHRL
project and how it advances a system of expert governance. The second sec-
tion examines the rationales offered to justify that companies should adopt
IHRL as their corporate speech codes. These rationales show, I argue, that
the IHRL project is rooted in an ideal of objectivity and expertise. The third
section looks at the work that goes into maintaining the claim to objectivity
when adjudicating speech within the IHRL framework. It examines several
tools experts use to legitimize their own power. The fourth section digs into
the oscillation between shifting power to experts and creating new avenues
for public participation. It looks at how these two mechanisms of govern-
ance compete within the IHRL project. The last section reflects on the risks
of the IHRL project as executed to date and imagines more promising
futures.
I. The International Human Rights Law Project for Online

Speech Governance
A large consensus among companies, scholars, and advocates has emerged

around the idea that regulation coming from either states or private actors
should be based on human rights. Twitter’s guiding principles for govern-
ment regulation enunciate that the internet should be built on “the protec-
tion of human rights.”27 Facebook’s principles for online content regulation
state that “the most important elements of any system will be due regard to
each of the human rights and values at stake.”28 Companies say that human
rights should inform not only government regulation but also their own
internal rules.29 Likewise, many scholars agree that content moderation
26. See Brenda Dvoskin, Representation without Elections: Civil Society Participation as a Remedy for the
Democratic Deficits of Online Speech Governance, 67 Vill. L. Rev. 447 (2022) (exploring the distributional
consequences of corporate adoption of administrative law principles for public participation).
27. Protecting the Open Internet: Regulatory Principles for Policy Makers, Twitter, https://cdn.cms-twdig-
italassets.com/content/dam/about-twitter/en/our-priorities/open-internet.pdf [https://perma.cc/5QJW-
27QG].
28. Monika Bickert, Online Content Regulation: Charting a Way Forward, Facebook, https://about.fb.
com/wp-content/uploads/2020/02/Charting-A-Way-Forward_Online-Content-Regulation-White-Paper-
1.pdf [https://perma.cc/NHB7-9GJL].
29. See, e.g., Defending and Respecting the Rights of People Using Our Service, Twitter, https://
help.twitter.com/en/rules-and-policies/defending-and-respecting-our-users-voice [https://perma.cc/
2RYJ-T44D]; Jack Dorsey (@jack), Twitter (Aug. 10, 2018, 9:58 AM), https://twitter.com/jack/status/
1027962500438843397 [https://perma.cc/VZA3-7E43]; Miranda Sissons, Our Commitment to Human
should follow a human rights approach.30 Multiple civil society organiza-

tions focusing on online speech identify themselves as human rights organi-
zations.31 An overwhelming number of reports, proposals, and commentary
from civil society and advocates argue that it would be desirable to base
content moderation on human rights.32
The prevalence of the human rights language may only indicate a broad
consensus that content moderation must respond to the public interest. In
some cases, human rights functions as an undefined term to refer to the
Rights, Meta (Mar. 16, 2021), https://about.fb.com/news/2021/03/our-commitment-to-human-rights/

[https://perma.cc/S9AX-NNKY]; Facebook Community Standards, Meta, https://transparency.fb.com/poli-
cies/community-standards/ [https://perma.cc/A4LC-NXQS] (“These standards are based on feedback
from people and the advice of experts in fields like technology, public safety and human rights.”);
Monika Bickert, Updating the Values That Inform Our Community Standards, Meta (Sept. 12, 2019), https://
about.fb.com/news/2019/09/updating-the-values-that-inform-our-community-standards/ [https://
perma.cc/5UK6-QN38].
30. See, e.g., Nicolas Suzor, Lawless: The Secret Rules that Govern Our Digital Lives 125
(2019); Human Rights in the Age of Platforms (Rikke Frank Jørgensen ed., 2019); Dinah
Pokempner, Regulating Online Speech: Keeping Humans, and Human Rights, at the Core, in Free Speech in
the Digital Age 224 (Susan Brison & Katharine Gelber eds., 2019); Barrie Sander, Freedom of Expression
in the Age of Online Platforms: The Promise and Pitfalls of a Human Rights Based Approach to Content Modera-
tion, 43 Fordham Int’l L. J. 939, 966 (2020); Kate Jones, Online Disinformation and Political Discourse:
Applying a Human Rights Framework, Chatham House (Nov. 6, 2019), https://www.chathamhouse.org/
2019/11/online-disinformation-and-political-discourse-applying-human-rights-framework [https://
perma.cc/RRB9-XEA3]; Susan Benesch, Proposals for Improved Regulation of Harmful Online
Content, 1, 10, https://dangerousspeech.org/wp-content/uploads/2020/07/Proposals-for-Improved-
Regulation-of-Harmful-Online-Content-Formatted-v5.2.2.pdf [https://perma.cc/DP8T-5YJJ]. But see,
Brenda Dvoskin, International Human Rights Law Is Not Enough to Fix Content Moderation’s Legitimacy Crisis,
Medium (Sept. 16, 2020), https://medium.com/berkman-klein-center/international-human-rights-law-
is-not-enough-to-fix-content-moderations-legitimacy-crisis-a80e3ed9abbd [https://perma.cc/53VZ-
S8H9]; Evelyn Douek, The Limits of International Law in Content Moderation, 6 U.C. Irvine J. Int’l,
Transnat’l & Compar. L. 37 (2021); Rachel Griffin, Rethinking Rights in Social Media Governance:
Human Rights, Ideology and Inequality (Dec. 15, 2021) (unpublished manuscript) (on file with author).
31. See, e.g., Glob. Network Initiative, https://globalnetworkinitiative.org/ [https://perma.cc/
2T2Q-MGJQ]; Hum. Rts. Watch, https://www.hrw.org/ [https://perma.cc/YTR9-KKH4]; Amnesty
Int’l, https://www.amnesty.org/ [https://perma.cc/EN47-R946]; Ranking Digital Rts., https://rank-
ingdigitalrights.org/ [https://perma.cc/94G8-CRQW]; Article 19, https://www.article19.org/ [https://
perma.cc/A43R-7L39].
32. See, e.g., Corporate Speech Controls, Electronic Frontier Foundation, https://www.eff.org/is-
sues/corporate-speech-controls [https://perma.cc/DGW8-M7RM] (“we believe that they should promote
free expression and transparency and base moderation practices and policies on human rights norms.”);
Jillian C. York & Corynne McSherry, Content Moderation is Broken. Let Us Count the Ways., Electronic
Frontier Foundation (Apr. 29, 2019), https://www.eff.org/deeplinks/2019/04/content-moderation-
broken-let-us-count-ways [https://perma.cc/97VJ-ZUJE]; Elis̆ka Pı́rkov á & Javier Pallero, Twenty-Six
Recommendations on Content Governance: A Guide for Lawmakers, Regulators, and Company Policy Makers, Ac-
cess Now 1, 35 (2020), https://www.accessnow.org/cms/assets/uploads/2020/03/Recommendations-On-
Content-Governance-digital.pdf [https://perma.cc/A777-4U7T]; Parmar, supra note 10; Laura Murphy, R
Facebook’s Civil Rights Audit–Final Report, Facebook 1, 9 (July 8, 2020), https://about.fb.com/wp-con-
tent/uploads/2020/07/Civil-Rights-Audit-Final-Report.pdf [https://perma.cc/K9U4-BGK4]; Charles
Bradley & Richard Wingfield, A Rights-Respecting Model of Online Content Regulation by Platforms, Global
Partners Digital (May 2018), https://www.gp-digital.org/wp-content/uploads/2018/05/A-rights-re-
specting-model-of-online-content-regulation-by-platforms.pdf [https://perma.cc/5E8F-VFYE]; Emma
Llansó, CDT’s Comments to Facebook Oversight Board on 2021-001-FB-FBR (Case Regarding Suspension of
Trump’s Account), Ctr. for Democracy & Tech. (Feb. 11, 2021), https://cdt.org/insights/cdts-com
ments-to-facebook-oversight-board-on-2021-001-fb-fbr-case-regarding-suspension-of-trumps-account/
[https://perma.cc/W8JQ-RVWX].
interests, preferences or agreements of society as opposed to the commercial

interests of private entities.33 Other times, companies and advocates refer
explicitly to IHRL, a body of treaties and authoritative interpretations with
established—although often vague or contested—norms and conditions for
states to regulate speech.
IHRL comprises several treaties, instruments, and authoritative interpre-
tations, but Article 19 of the International Covenant on Civil and Political
Rights (hereinafter “ICCPR”) is the epicenter of the IHRL project for con-
tent moderation. This provision sets out a tripartite test to evaluate restric-
tions on freedom of expression. The test from Article 19 includes three
requirements: rules must be prescribed by law (legality), must have a legiti-
mate aim (legitimacy), and must be necessary for that aim (necessity). Suc-
cinctly, the first requirement means that restrictions on speech must be
provided by law, that is, they must be enacted within a country’s domestic
legal system.34 To meet this requirement, restrictions must also be clear and
precise enough to give appropriate notice to the public of what speech is not
allowed.35 Second, a restriction on speech must have one of the public inter-
est aims enumerated in Article 19(3): the protection of the rights or reputa-
tions of others, national security, public order, public health, or morals.36
Finally, for speech restrictions to be considered necessary, they must be the
least intrusive means to achieve their legitimate aim and they must be pro-
portionate to the interest they are designed to protect.37
The IHRL project for content moderation promises important and valua-
ble contributions.38 Scholars emphasize IHRL’s procedural mandates.39 They
argue that, in accordance with IHRL, companies have to be transparent
about how they moderate content and have to offer their users robust appeal
processes to exercise their due process rights.40 IHRL can also function as a
loose normative agreement among multiple constituencies that can serve as a
33. Rikke Frank Jørgensen, What Platforms Mean When They Talk About Human Rights, 9 Pol’y &
Internet 280 (2017) (showing how companies present themselves as public spaces). See also Amy
Kapczynski, The Right to Medicines in an Age of Neoliberalism, 10 Human. J. 79, 85 (2019) (describing the
polycentric uses of the term “human rights.”).
34. U.N. Hum. Rts. Comm., General Comment No. 34, Article 19: freedoms of opinion and expression, ¶
24, U.N. Doc. CCPR/C/GC/34 (Sept. 12, 2011) (hereinafter “General Comment 34”); David Kaye (Spe-
cial Rapporteur on the Promotion and Protection of the Right to Freedom of Opinion and Expression),
Rep. of the Special Rapporteur on the Promotion and Prot. of the Right to Freedom of Op. and Expression, ¶ 12,
U.N. Doc. A/71/373 (Sept. 6, 2016) (hereinafter “SR Report 2016”).
35. General Comment 34, supra note 34, ¶ 25. R
36. International Covenant on Civil and Political Rights, art. 19, 999 U.N.T.S. 171, 178 (entered
into force Mar. 23, 1976) (hereinafter “ICCPR”).
38. See Douek, supra note 30, at 41-50 (providing an overview of how international law can improve R
online speech governance).
39. See, e.g., Sander, supra note 30, at 988. R
40. See, e.g., Molly K. Land, Regulating Private Harms Online: Content Regulation Under Human Rights
Law, in Human Rights in the Age of Platforms 285, 288 (Rikke Frank Jørgensen ed., 2019).
starting point for designing content moderation.41 Similarly, taking inspira-

tion from an existing legal system can be a useful tool for corporations to
make decisions about online speech. For that purpose, there is no better
alternative than IHRL.42
The IHRL project is an ongoing development carried out by multiple
scholars and advocacy groups. Therefore, different versions of the project
compete.43 The main ambivalence within the project is exactly what gui-
dance companies are expected to follow.
On the one hand, companies are called to follow IHRL, which comprises
the U.N. treaties, regional treaties, international custom, and the authorita-
tive interpretations by courts, special rapporteurs, and other international
bodies. In that sense, scholars have focused their attention on translating the
substantive rules of international law to make them usable by corporate
actors.44
Within this version of the project, scholars disagree on whether the pro-
ject should focus exclusively on translating U.N. level treaties or should also
include regional human rights treaties and national interpretations of inter-
national instruments.45 David Kaye argues that regional treaties and local
interpretations of international instruments can provide important gui-
dance,46 although other authors believe it is wiser to limit the scope of the
project to U.N. treaties.47 At the same time, while in some cases the goal
has been to operationalize IHRL into granular rules that can effectively
guide speech governance, scholars have also called for multi-stakeholder con-
versations to determine what these rules should be.48
On the other hand, a variation of the IHRL project suggests that content
moderation policies should meet the requirements of legality, legitimacy,
and proportionality of Article 19(3) of the ICCPR to assess the reasonable-
ness of their rules without necessarily adhering to existing determinations of
41. Jacob Mchangama et al., A Framework of First Reference: Decoding a Human Rights Approach to Content
Moderation in the Era of “Platformization”, The Future of Free Speech (2021), https://futurefrees-
peech.com/wp-content/uploads/2021/11/Report_A-framework-of-first-reference.pdf [https://perma.cc/
BVP8-MJNQ].
42. Douek, supra note 30, at 49. R
43. See Social Media Councils: Consultation Paper, Article 19, (2019), https://www.article19.org/wp-
content/uploads/2019/06/A19-SMC-Consultation-paper-2019-v05.pdf [https://perma.cc/P6WD-28ZM]
(surveying competing proposals to align content moderation with international human rights law).
44. See, e.g., Aswad, supra note 23; Sander, supra note 30, at 969 (“Arguably the biggest challenge, R
however, resides in the translation of general human rights principles into particular rules, processes and procedures
tailored to the platform moderation context.”) (emphasis in original); id. at 971 (“[T]ranslation from the State
to the corporate context of platform moderation is likely to pose a number of challenges in practice.”)
(emphasis added).
45. See infra Section III.B.
46. David Kaye, A New Constitution for Content Moderation, Medium (June 25, 2019), https://
onezero.medium.com/a-new-constitution-for-content-moderation-6249af611bdf [https://perma.cc/2FSR-
VZ9U].
47. See, e.g., Aswad, supra note 23, at 44; infra Section III.B. (discussing the role of regional and U.N. R
instruments in the IHRL project).
48. See Aswad, supra note 23, 35. R
which governmental restrictions meet these conditions. Accordingly, the ex-

pectation is that companies should be able to justify their moderation poli-
cies as reasonable, but international law would not impose significant
restrictions on which policies are acceptable.49
This ambivalence between aligning content moderation with IHRL and
reducing IHRL to a generic proportionality test (and all possible intermedi-
ate possibilities of asking companies to adhere to some but not all interna-
tional law standards) makes the IHRL project flexible and resilient. For
example, if the objection is that IHRL is too vague to offer guidance for
content moderation, the response is that companies can look to the vast
authoritative interpretations advanced by international and local tribunals,
treaty bodies, and others. But if the objection is that those sources are inco-
herent or not appropriate for private content moderation, the response is
that companies do not need to comply with international law.
This Article traces this ambivalence and accounts for how different trends
are negotiated within the project. I pay particular attention, however, to
how the Board is carrying out the IHRL project because it is the most rele-
vant actor implementing it. Currently, the Board has 23 members with di-
verse geographical and professional backgrounds.50 Many, but not all of
these members, have strong expertise in IHRL. In response to the project’s
ambiguity, the Board’s decisions offer a concrete description of what content
moderation can look like when it is based on human rights standards.
Therefore, it could be that some of the tools, implementation challenges,
and concerns that I describe here are specific to this version of the project.
Despite the variations, the IHRL project for corporate content moderation
in all of its versions, I believe, envisions an ideal of expert and objective
governance. Performing as objective decisionmakers is a strategy commonly
pursued by judges and bodies of experts.51 The field of science and technol-
ogy studies describes various mechanisms that experts, including judges, use
to exercise power while maintaining the appearance of objectivity and demo-
cratic buy-in.52 One such mechanism is to construct exogenous and neutral
governing standards. Because these standards purport not to reflect anyone’s
49. Douek, supra note 30, at 64 (expressing concern that platforms might “clothe themselves in the R
language of IHRL and accrue legitimacy dividends merely for meeting bare minimum transparency and
justification requirements.”).
50. Meet the Board, Oversight Bd., https://www.oversightboard.com/meet-the-board/ [https://
perma.cc/EE8T-AKY6]; Evelyn Douek, What Kind of Board Have You Given Us?, Chicago L. Rev. Blog
(May 11, 2020), https://lawreviewblog.uchicago.edu/2020/05/11/fb-oversight-board-edouek/ [https://
perma.cc/Y5NF-87DS] (analyzing the structure of the Board, its founding documents, and its initial
members).
51. See generally Aziza Ahmed, Medical Evidence and Expertise in Abortion Jurisprudence, 41 Am. J.L. &
Med. 85 (2015) (exploring how medical evidence and expertise are portrayed as neutral and objective in
the context of abortion access); Michael Klarman, From Jim Crow to Civil Rights: The Supreme
Court and the Struggle for Racial Equality (2005) (analyzing how judges reconstruct the law to
reconcile their understanding of what is legally required with their moral views).
52. Sheila Jasanoff, Subjects of Reason: Good, Markets and Competing Imaginaries of Global Governance, 4
London Rev. Int’l. L. 361, 363 (2016).
standpoint, Sheila Jasanoff describes them as representing a view from no-

where.53 An alternative mechanism is to build principles that are presented
as neutral because they represent a view from everywhere.54 These principles
are imagined as the result of a participatory process to reach consensus and
agreement, so the outcome is attributed to all the participants.55 The IHRL
project relies on both kinds of objectivity. IHRL is both portrayed as self-
evidently good and as reflecting a universal consensus.56
The hypothesis of the project is that IHRL can function as an already
available account of the interests or will of global society. However, as I
hope will become apparent in the next sections, the idea of IHRL as a syn-
thesis of the global public interest is an illusion.57 While the vagueness of
IHRL shows that IHRL cannot define those interests with any precision,
attempts to construct more precise rules are equally misguided. The public’s
will is fragmented and in permanent conflict.58 Still, these efforts imagine
that what is needed is more time, thinking, and conversations to find an
intelligible, coherent, and precise enough synthesis of the public’s will. In
turn, the project assumes it is possible to hold experts accountable to the
public by asking them to follow that synthesis.
The danger of believing that it is possible to find an account of the will of
the global society is that it enables experts to assert that they are acting on
behalf of everyone (or no one). If experts were to recognize that their deci-
sions benefit only the preferences of some, it might be more immediately
apparent that it is necessary to create ample opportunities for contestation.59
Indeed, experts’ claim to objectivity rests on the closure of the public de-
bate. Therefore, when experts portray their decisions as a view from every-
where or nowhere, the need for sharing power appears as secondary.
To be sure, the IHRL project in many cases emphasizes the value of par-
ticipation. David Kaye pays special attention to broadening the opportuni-
ties for more actors to participate in the governance of online speech,
especially civil society from the Global South.60 Evelyn Aswad, a leading
scholar and member of the Board, calls for multi-stakeholder conversa-
tions.61 The Board has often recommended Facebook conduct more extensive
53. Jasanoff, supra note 20, at 313 (describing a mode of building objectivity in regulatory science R
that she describes as a view from nowhere).
54. Id. at 315.
55. Id.
56. See infra Section II.A.
57. Daniel Walters, The Administrative Agon: A Democratic Theory for a Conflictual Regulatory State, 132
Yale L.J. (forthcoming 2022) (discussing the democratic foundations of the administrative state and
arguing that what constitutes the common good is permanently contested).
58. Id.
59. See Nikolas Bowie, Comment, Antidemocracy, 135 Harv. L. Rev. 160, 160 (2021) (defining de-
mocracy as the situation in which “everyone in the community, or demos, [may] share in exercising
power, or kratos”).
60. SR Report 2018, supra note 9, at para. 54. R
61. Aswad, supra note 23, at 57. R
stakeholder engagement.62 However, as I argue in detail in Section IV, these

calls do not challenge the role of IHRL as an account of the global public
interest. Reasonable conflicts among diverse viewpoints might play a role in
refining details, but do not constitute the core of the project.
Because judicial courts often aim at performing objectivity, many of the
strategies I describe here should sound familiar to legal scholars.63 Indeed,
the Board replicates a judicial model in its quest to construct legal norms as
exogenous principles it applies in each decision. The focus of the following
sections will be on unveiling the deeply political interpretative work behind
these principles and their application. Gaining a better understanding of the
world that the project envisions, the tools it has to achieve it, and the in-
tended and unintended effects of deploying those tools may lead to a more
comprehensive evaluation of the IHRL project’s potential contributions and
risks.
Two clarifications are in order before proceeding.
First, this Article engages with how legal scholars, advocacy groups, and
the Board justify and propose to implement IHRL as a project for online
speech governance by corporate actors. The Article is, then, not a critique of
IHRL more broadly. Indeed, in some cases, a disconnect exists between in-
ternational law literature and how international law arguments are deployed
in this specific context. Here I focus only on the latter and aim at making
the disconnect explicit when relevant.64
Second, this Article is not about the intent, motivations, or state of con-
sciousness of the actors promoting what I call the IHRL project for content
moderation. When experts frame their proposals and decisions as mandated
by international law or some other exogenous source, they might be aware
that they are using legal language to cover a policy choice or might believe
that their decision is in fact mandated by an external authoritative source, or
might believe something else. Their state of mind is in no way part of this
Article’s inquiry.
II. Justifying the International Human Rights Law Project
IHRL is often invoked as the proper benchmark to evaluate content mod-

eration without explaining why that would be appropriate.65 In debates
about online speech governance, it is often assumed that human rights law is
62. See infra Section IV.B.

63. Legal realism scholars mostly accept that legal reasoning conceals policy decisions. See, e.g., Oliver
Holmes, The Path of the Law, 10 Harv. L. Rev. 457 (1897); John Dewey, Logical Method and Law, 10
Cornell L.Q. 17 (1924); Karl Llewllyn, Some Realism about Realism—Responding to Dean Pound, 44 Harv.
L. Rev. 1222 (1931).
64. See infra Section II.A.3, Section III.B.
65. See infra Section II.A.
synonymous with the public interest.66 The attractive features of the frame-
work are easy to see: its claim to universality and the high esteem in many
circles. Thus, it could be useful to guide the private governance of speech on
platforms that also intend to be global and have a great need for a legitimate
set of rules to conduct their businesses.
On a closer look, however, it is not obvious that a framework designed for
states would be appropriate for corporate actors. Accordingly, proponents of
the IHRL project hint at various justifications for their proposal as well as
modifications that they consider necessary to adjust the framework for the
corporate environment. Examining these rationales serves two purposes. On
the one hand, this section shows how the justifications most commonly of-
fered share the quest for objective and apolitical principles to function as the
foundations of speech governance. On the other hand, although these ratio-
nales are offered as objective reasons to use IHRL as the framework for con-
tent moderation, this section challenges their objectivity. It argues that
these justifications are often internally inconsistent or depart widely from
established principles of international law. In addition, I argue that the pro-
posed adjustments to IHRL are closer to the construction of a new system of
rules than to an exercise of translation.67 As a consequence, even if IHRL
were a universal framework with democratic buy-in, it is unclear that the
IHRL project for content moderation would preserve those credentials.
I explore here five common rationales for asking social media companies
to implement IHRL in their content policies: the UNGPs, a functional anal-
ogy between social media companies and states, the wide ratification of cer-
tain human rights conventions, the capacity of IHRL to rein in corporate
power, and the fact that IHRL reflects globally shared values.
A. The Rationales
1. A New Interpretation of the U.N. Guiding Principles
Perhaps the most common justification for the IHRL project is the
UNGPs. John Ruggie developed these principles during his tenure as the
U.N. Secretary-General’s Special Representative for Business and Human
Rights. The instrument provides standards to guide the conduct of states
and business enterprises. The UNGPs set the expectation that companies
66. See Douek, supra note 30, at 40 (“there has been almost no strong dissent from the proposition R
that IHRL should be adopted by companies as the basis for their rules”); Griffin, supra note 30, (describ- R
ing human rights as “the rarely-questioned moral yardstick against which all platform practices and state
regulation are measured”). Forceful opposing views to international human rights law as representative of
the public interest do exist in the broader literature outside of the specific debates about content modera-
tion. See, e.g., Samuel Moyn, The Last Utopia: Human Rights in History (2010); David Ken-
nedy, The Dark Sides of Virtue: Reassessing International Humanitarianism (2005).
67. See Sander, supra note 30, at 969, 971 (describing the IHRL project as a translation exercise). R
will respect human rights wherever they conduct their operations.68 On that
basis, scholars interpret the UNGPs as calling on social media companies to
align their policies with IHRL.69 In that vein, the Board states in every
decision that it applies international law per the UNGPs.70 But the origins
of the UNGPs show that this recent interpretation contradicts the history of
the instrument as well as its drafter’s understanding of its meaning, and it is
hard to find support for this recent interpretation in the text of the
UNGPs.71
To be sure, the UNGPs do provide guidelines that would help reduce the
harm caused by online speech. According to these principles, companies
should assess how their products impact people’s human rights in the differ-
ent markets where they operate.72 Companies should take measures to stop
any contribution to human rights violations.73 This responsibility means
that companies ought to take action regarding speech that international law
prohibits.
The UNGPs, however, do not set the expectation that companies will
align their content policies with IHRL. The UNGPs set the expectation that
companies will respect human rights, but international law does not cur-
rently recognize a human right to speak on privately owned media without
being subject to the medium’s editorial policies. Neither does it establish a
corresponding duty for media owners to host all speech that international
law protects.74 To the contrary, as Molly Land says, “users seek these plat-
forms precisely because of the choices the companies make about the infor-
68. U.N. Special Representative of the Secretary-General, Guiding Principles on Business and Human
Rights: Implementing the United Nations “Protect, Respect and Remedy” Framework, Principles 11-24, U.N.
Doc. A/HRC/17/31 (Mar. 21, 2011).
69. See, e.g., Stefania Di Stefano, The Facebook Oversight Board and the UN Guiding Principles on Business
and Human Rights: A Missed Opportunity for Alignment?, in Human Rights Responsibilities in the
Digital Age 93 (Jonathan Andrew & Frédéric Bernard eds., 2021).
70. The Facebook Oversight Board has applied human rights standards per the UNGPs in all of its
decisions. Since Facebook’s adoption of its corporate human rights, the Facebook Oversight Board has
also referenced this document as a basis for the application of international standards. See Case decision
2021-003-FB-UA, Oversight Bd., (Apr. 29, 2021), https://oversightboard.com/decision/FB-
H6OZKDS3/ [https://perma.cc/VV4K-NMRH]; see also Corporate Human Rights Policy, Meta, https://
about.fb.com/wp-content/uploads/2021/03/Facebooks-Corporate-Human-Rights-Policy.pdf [https://
perma.cc/8FBK-XFUG] (“We are committed to respecting human rights as set out in the United Na-
tions Guiding Principles on Business and Human Rights.”).
71. Coincidentally, a recent report submitted by the U.N. High Commissioner for Human Rights to
the Human Rights Council addressing the application of the UNGPs to the technology sector does not
indicate that social media should align their content moderation policies to international human rights
law. See OHCHR, The Practical Application of the Guiding Principles on Business and Human Rights to the
Activities of Technology Companies, U.N. Doc. A/HRC/50/56 (Apr. 21, 2022).
72. U.N. Special Representative of the Secretary-General, supra note 68, Principle 13. R
73. Id., Principle 11. See also Allison-Hope et al., A Human Rights-Based Approach to Content Governance,
BSR (Mar. 2, 2021), https://www.bsr.org/en/our-insights/blog-view/beyond-user-realdonaldtrump-
human-rights-based-approach-content-governance [https://perma.cc/LGC9-ELD8].
74. See Matthias C. Kettemann & Anna Sophia Tiedeke, Back up: can users sue platforms to reinstate
deleted content?, 9 Internet Pol’y Rev. 1 (2020) (showing that even in Germany, where courts have
pioneered the development of intermediaries’ duties to host certain types of content, platforms are al-
lowed to exclude lawful expressions).
mation they deliver.”75 Accordingly, because excluding lawful content is an

essential aspect of what social media companies do, IHRL does not fit well
with how platforms operate, as Emily Laidlaw points out.76
In this context, it is unclear which existing human right the companies
are called upon to respect when they are asked to regulate content in the
same terms that international law binds state regulation of expression.77 In-
deed, the scholars who conclude that the UNGPs set the expectation that
social media companies align their policies with IHRL are advancing an
interpretation of this instrument that diverges from its original meaning.
Since the 1970s, the United Nations has tried to address the negative
impact that corporations can have on human rights.78 The direct antecedent
of the UNGPs were the failed U.N. Norms on the Responsibilities of Trans-
national Corporations and Other Business Enterprises with Regard to
Human Rights 2003 (hereinafter “Draft U.N. Norms”).79 This initiative
tried to create binding standards directed at transnational corporations im-
plicated in abuses such as employing child labor, failing to provide safe
working conditions, dumping toxic wastes, and disrupting the right to bar-
gain collectively.80 The Draft U.N. Norms attributed to business enterprises
the same general obligations that states have to promote, secure the fulfill-
ment of, respect, and protect human rights under the human rights treaties
that states have ratified.81
The project was stalled following opposition from several states and the
business sector.82 The authors of the Draft U.N. Norms and the NGOs that
participated in the drafting process believed that IHRL should apply di-
rectly to business enterprises.83 The corporate sector and most states dis-
agreed.84 The U.N. Sub-Commission on the Promotion and Protection of
75. Land, supra note 40, at 292. R

76. Emily B. Laidlaw, Regulating Speech in Cyberspace: Gatekeepers, Human Rights and
Corporate Responsibility 292 (2015).
77. See SR Report 2018, supra note 9, ¶ 45. R
78. See generally Surya Deva, The UN Guiding Principles on Business and Human Rights and Its Predecessors:
Progress at a Snail’s Pace?, in The Cambridge Companion to Business & Human Rights Law 145
(Ilias Bantekas & Michael Ashley Stein eds., 2021) (mapping the efforts of the United Nations to address
the human rights duties of business enterprises).
79. Sub-Comm’n on the Promotion and Prot. of Hum. Rts., Comm’n on Hum. Rts., Norms on the
Responsibilities of Transnational Corporations and Other Business Enterprises with regard to Human Rights, U.N.
Doc E/CN.4/Sub.2/2003/12/Rev.2 (Aug. 26, 2003).
80. David Weissbrodt & Muria Kruger, Norms on the Responsibilities of Transnational Corporations and
Other Business Enterprises with Regard to Human Rights, 97 Am. J. Int’l. L. 901 (2003) (detailing the
history, purpose and content of the norms).
81. Sub-Comm’n on the Promotion and Prot. of Hum. Rts., Comm’n on Human Rights, supra note
79. R
82. See Deva, supra note 78, at 155; John G. Ruggie, Incorporating Human Rights: Lessons Learned, and R
Next Steps, in Business and Human Rights: From Principles to Practice 64, 65 (Dorotheé Bau-
mann-Pauly & Justine Nolan eds., 2016).
83. Weissbrodt & Kruger, supra note 80, at 906. R
84. John G. Ruggie, Just Business: Multinational Corporations and Human Rights 47–55
(2013).
Human Rights approved the norms in August 2003, but the Human Rights
Council (then Commission) refused to endorse them.85
In that context, the Human Rights Council created John Ruggie’s man-
date in 2005. The original mandate of the Special Representative of the
Secretary-General for Business and Human Rights was to “identify and clar-
ify standards of corporate responsibility and accountability” for businesses
concerning human rights and to elaborate the role of states in effectively
regulating corporations.86 Ruggie “saw no reason to replicate the debate”
that had bogged down the U.N. Draft Norms.87 Thus, he introduced some
modifications to the previous framework to gain support from states and
corporations.88
In Ruggie’s vision, the U.N. Draft Norms “had serious foundational
flaws, such as intermingling state and corporate obligations while providing
no boundaries for the latter.”89 Accordingly, in the UNGPs, he distin-
guished states’ obligations from corporations’ responsibilities. Unlike the
U.N. Draft Norms, which would have imposed legal obligations on corpora-
tions, the UNGPs set expectations that enterprises should meet voluntarily.
However, the distinction was not merely that states were bound by interna-
tional law and businesses should follow it voluntarily. Instead, Ruggie intro-
duced a substantive distinction between the content of states’ duties and
businesses’ responsibilities.90
According to Ruggie’s framework, companies should look at international
treaties not as a source of rules that apply to them but as an enumeration of
recognized rights.91 Thanks to this differentiation, even states that had not
ratified core U.N. human rights treaties endorsed the UNGPs. Notably,
China and the United States ratified them even though China has not rati-
fied the International Covenant on Civil and Political Rights, and the
United States has not ratified the International Covenant on Economic, So-
cial and Cultural Rights.92
Corporations’ main responsibility is to “avoid infringing on the human
rights of others and should address adverse human rights impacts with
which they are involved.”93 Some emblematic cases help to clarify the com-
85. Comm’n on Hum. Rts., Report on the Sixtieth Session, ¶41, U.N. Doc. E/CN.4/2004/L.73/Rev.1
(2004).
86. Office of the High Comm’r for Hum. Rts., Secretary-General to appoint a special representative on the
issue of human rights and transnational corporations and other business enterprises, U.N. Doc E/CN.4/RES/2005/
69 (Apr. 20, 2005).
87. Ruggie, supra note 82, at 65. R
88. Id.
89. John Gerard Ruggie, The Social Construction of the UN Guiding Principles on Business and Human
Rights, in Research Handbook on Human Rights and Business 63, 71 (Surya Deva & David
Birchall eds., 2020).
90. Ruggie, supra note 84, at 81-127. R
panies’ responsibilities. By 1990, Nike had outsourced its entire production.

Its overseas sourcing factories employed over 24,000 workers in Asia.94
Workers claimed problematic practices such as not being allowed to leave
the premises except on Sundays and, even then, needing authorization from
management.95 Life magazine published a photograph of a twelve-year-old
boy stitching soccer balls.96 Nike’s initial defense was that they did not own
the factories.97 In response, the UNGPs set the expectation that companies
should conduct human rights due diligence to learn how their operations
may contribute to or be involved in human rights abuses, and address this.
For example, in this case, Nike would have been expected to recognize that
child labor and other abusive practices infringe on human rights that inter-
national treaties identify and to avoid engaging in those abuses.98
Facebook’s involvement in the genocide in Myanmar also illustrates cor-
porate failure to meet this expectation. The report of the independent inter-
national fact-finding mission on Myanmar found that “Facebook has been a
useful instrument for those seeking to spread hate, in a context, where, for
most users, Facebook is the Internet.”99 The report stressed that Facebook’s
inability to provide country-specific data about the spread of hate speech on
its platforms made it difficult to assess the adequacy of Facebook’s response
to the situation.100 This type of risk assessment is the kind of due diligence
that, under the UNGPs, companies are expected to undertake in order to
evaluate and prevent their participation in human rights violations.
The IHRL project for content moderation interprets the UNGPs in a new
fashion. Under its original meaning, social media companies are expected
not to infringe on individuals’ rights to freedom of expression (and any other
rights) to the extent that international law already recognizes those rights.
The IHRL project proposes that companies ensure a new right to freedom of
expression on privately-owned social media companies.
The project collapses the substance of states’ duties and businesses’ re-
sponsibilities. Even though they propose adaptations,101 IHRL advocates call
on companies to look at the text of international treaties and the interpreta-
tions that the U.N. Human Rights Committee and U.N. Special Rap-
porteurs have made of them as a set of rules that companies should apply on
94. Jennifer Burns & Debora Spar, Hitting the Wall: Nike and International Labor Practices, Case 9-700-
047, Harv. Bus. School Case Collection (2002).
95. Id.
96. Sydney Schanberg, Six Cents an Hour, Life (Mar. 28, 1996).
97. Id.
98. See Ruggie, supra note 84, at 3-6, 17, 69. R
99. Human Rights Council, Report of the independent international fact-finding mission on Myanmar, ¶ 74,
U.N. Doc. A/HRC/39/64 (Sept. 12, 2018).
100. Id.
101. See, e.g., Benesch, infra note 127. R
their platforms.102 This is the approach adopted by the U.N. Draft Norms,
which the Human Rights Council rejected and from which the UNGPs
explicitly diverged.103
At the same time, this new interpretation of the UNGPs alters the inter-
play between states’ duties and corporate responsibilities. Arguing that
every content moderation decision implies an IHRL issue means that states
have a corresponding duty to regulate each instance of content modera-
tion.104 The IHRL project’s interpretation of the UNGPs creates a human
right to freedom of expression on social media platforms vis-à-vis the owners
of such platforms.105 If individuals have that right, states have a correspond-
ing duty to protect it.106 However, there is little support, if any, for the idea
that states have a duty to protect the exercise of all IHRL-protected expres-
sion on private platforms.107 Indeed, no state requires companies to host all
speech that international law protects.108 The IHRL project for content
moderation asks from corporations what international law does not require
states to protect.
Another innovation refers to the different levels of responsibility between
large and small firms. According to the UNGPs, all firms are expected to
respect all internationally recognized rights.109 Does that mean that all com-
panies, including the New York Times or the Mandarin Oriental Hotel
Group,110 need to follow Article 19 of the ICCPR when deciding what
speech they will host? This interpretation of the UNGPs would be absurd
and hard to reconcile with the independence of the media and the protection
of their editorial freedom.111 In response, IHRL advocates sometimes rely on
Principle 14 of the UNGPs to distinguish among the responsibilities of
prominent social media companies and other business enterprises.
Principle 14 of the UNGPs establishes that the corporate responsibility to
respect human rights applies to all enterprises. However, the means through
102. See, e.g., Aswad, supra note 23; Parmar, supra note 10; Side-stepping rights: Regulating speech by R
contract, Article 19 (2018), https://www.article19.org/wp-content/uploads/2018/06/Regulating-speech-
by-contract-WEB.pdf [https://perma.cc/ZB6R-LCLB] (hereinafter (“Side-stepping rights”).
103. See supra n. 57-68 and accompanying text. R
104. See Land, supra note 40, at 292 (making a similar point stating that “it cannot be the case that R
every content moderation decision made by every digital platform should be subject to human rights
scrutiny.”).
105. See supra Section II.A.1.
107. See, e.g., Balkin, infra note 169, at 2025 (explaining why the best alternative to the current R
autocracy is not imposing the duty to carry all lawful speech).
108. See Daphne Keller, Who Do You Sue? State and Platform Hybrid Power Over Online Speech, 1902
Hoover Inst. Aegis Paper Series, 12 (2019) (surveying laws that impose the legal duty on platforms to
host speech, none of which includes the duty to host all lawful speech).
110. The Mandarin Oriental group is one of 35 hotel companies and hundreds of other companies
that have adopted a human rights policy and committed to respect human rights. See Companies, Bus. &
Hum. Rts. Res. Ctr., https://www.business-humanrights.org/en/companies/ [https://perma.cc/X68U-
9SWS] (tracking companies’ human rights policies).
111. See General Comment 34, supra note 34, ¶ 16. R
which enterprises meet that responsibility may vary according to their size,
sector, operational context, ownership, and structure.112 Some authors who
support the IHRL project indicate that this principle gives companies flexi-
bility in how they apply human rights norms to their operations.113 Accord-
ingly, Principle 14 could justify that smaller platforms with specific
purposes, more limited resources to moderate content, or specific targeted
audiences may depart from international guidance to regulate speech.
This distinction in the UNGPs was originally intended to recognize that
some firms have more extensive resources to assert influence. Opportunities
to exert more influence come with greater responsibility for protecting
human rights. Going back to the Nike example, a much smaller business
that outsourced part of its production may not have the resources to audit in
depth the whole supply chain it takes part in, or may not have the necessary
influence to bring about change. In the words of David Weissbrodt and
Muria Kruger, “[t]his nuanced approach does not lower the standards for
any business; it simply ensures that those with greater power and influence
will also have greater responsibilities.”114
Companies may devote different amounts of resources to their human
rights due diligence, but the content of the rights that companies must
respect is always the same.115 According to the interpretation advanced by
the IHRL project, however, large and small companies have widely different
substantive responsibilities. Large companies should adopt international law
as their own internal rules. In contrast, small companies may adopt rules
that directly contradict international standards for speech regulation.
Finally, a note on the project’s ambivalence between asking companies to
implement IHRL and calling on companies only to take inspiration from
IHRL.116 It could be that the UNGPs require that companies will imple-
ment international law, but not as if they were a state.117 Perhaps the
UNGPs do not set the expectation that companies follow IHRL, but do set
the expectation that companies ground their policies in human rights or use
human rights as “an overall framework for decision-making and ac-
tion[.]”118 For example, perhaps companies need to balance the right to free-
dom of expression and the right to equality, but the rule they reach as a
result of that balancing exercise might contradict international law as in the
Black Pete case. As I discuss in detail in the next subsection, it is unclear
why the new rule would be called IHRL. The value of labeling these new
standards as IHRL might reside in preserving the legitimacy credentials of
113. See, e.g., Benesch, infra note 127, at 95; Aswad, supra note 23, at 39. R
114. Weissbrodt & Kruger, supra note 80, at 911. See also Ruggie, supra note 84, at 101, 114. R
115. Id.
116. See supra Section I.
117. See, e.g., Land, supra note 40, at 305-306 (offering a thoughtful description of how private com- R
panies might follow Article 19 of the ICCPR in order to meet the expectations set by the UNGPs).
118. Allison-Hope et al., supra note 73. R
IHRL, even if the connection between the new rule and IHRL is thin at
best.
Ruggie’s interpretation of the UNGPs is not necessarily the best one. It
may be that IHRL advocates have created a new way of reading the princi-
ples that advances worthy policy objectives like expanding users’ rights to
freedom of expression vis-à-vis private actors. My aim is to show both the
urge to frame policy positions as mandated by legal documents as well as the
intensive intellectual work behind the purportedly objective principles
under which experts govern online speech.
2. Because social media companies are like states, but they are not like states
Scholars raise the concern that companies that control giant social media
platforms are displacing governments as the main speech regulators. In the
words of Richard Ashby Wilson and Molly Land:
Governments are no longer the primary regulators of speech.
Their regulatory capacity has been far outstripped by some of the
largest companies in the world . . ., which together regulate the
speech of 3.7 billion active social media users. . . . In a reversal of
the historic roles, private corporations have even become the de
facto regulators of government speech[.]119
Scholars conceptualize what giant platforms do as state functions. Nadine
Strossen states, “the Platforms wield censorial power of a magnitude that in
the past only governments have exercised.”120 Julie Cohen says, “[d]ominant
platforms’ role in the international legal order increasingly resembles that of
sovereign states.”121 Daphne Keller agrees, stating that, “platforms can take
on and replace traditional state functions, operating the modern equivalent
of the public square or the post office, without assuming state
responsibilities.”122
In most aspects, platforms are nothing like states. Platforms do not col-
lect taxes, manage prisons, hold elections, etc.123 The role of technology
companies is hardly unprecedented. Other highly concentrated media indus-
tries have controlled the public sphere in the past.124 However, the analogy
can be a useful rhetorical device to highlight the immense power platforms
wield over public discourse and collective affairs.
119. Richard Ashby Wilson & Molly Land, Hate Speech on Social Media: Content Moderation in Context,
52 Conn. L. Rev. 1, 5 (2021).
120. Nadine Strossen, United Nations Free Speech Standards as the Global Benchmark for Online Platforms’
Hate Speech Policies, 29 Mich. State Int’l L. Rev. 307, 324-5 (2021).
121. Julie Cohen, Law for the Platform Economy, 51 U.C. Davis L. Rev. 133, 199 (2017).
122. Keller, supra note 108, at 2-3. R
123. See Benesch, infra note 127 (highlighting the obvious but fundamental fact that “Facebook is not R
a country.”).
124. Paul Starr, The Creation of the Media: Political Origins of Modern Communica-
tions (2004).
Describing platforms as states can have profound normative implications.

For example, one could imagine users as citizens or political subjects with an
interest in participating in policy making. The IHRL project emphasizes a
different implication of this analogy: companies should meet the require-
ments that states must meet to restrict expression under human rights trea-
ties. This emphasis on adhering to existing legal frameworks for states (as
opposed to other possible derivations from the analogy between platforms
and states) is also more compatible with the functioning of the Board. The
Board operates under a judicial model. That background helps explain the
gravitational force attracting the Board to a system of legal norms.
Because platforms are different from states, scholars propose adaptations
to the international law framework. Proponents of the IHRL project ac-
knowledge that the first two requirements set out by Article 19 of the
ICCPR—legality and legitimacy—need to be “translated.”125
Regarding the legitimacy prong, supporters of the IHRL project struggle
to define what aims content moderation could legitimately pursue. Article
19 of the ICCPR enumerates legitimate aims for governmental restrictions
on speech: respect for the rights or reputations of others, the protection of
national security, and public order, public health, or morals.126 However,
these aims are both over-inclusive and under-inclusive for platforms. On the
one hand, Susan Benesch argues that firms are not well positioned to deter-
mine, for example, national security goals.127 On the other hand, Evelyn
Aswad asks if it would be legitimate for companies to pursue the aim of
maximizing profit.128 Aswad believes IHRL does not provide a conclusive
answer here. In her version of the IHRL project, multi-stakeholder conversa-
tions should define the legitimate aims for corporate IHRL. In practice
though, this specific discussion about commercial aims may not be that rele-
vant. Companies will usually find it easy to develop a public-interest ratio-
nale to restrict speech without invoking business reasons.129
Suppose the IHRL project concedes that the aims defined by Article 19
are ill-suited for private speech regulation and accepts, instead, any public-
interest purpose. In that case, the IHRL project would probably approve any
aim that companies enunciate as legitimate. In fact, in all decisions to date,
the Board has found that Facebook’s community standards have a legitimate
aim.
125. See Sander, supra note 44. R

126. ICCPR, supra note 36, art. 19(3). R
127. Susan Benesch, But Facebook’s Not a Country: How to Interpret Human Rights Law for Social Media
Companies, 38 Yale J. on Reg. Bull. 86, 106 (2020).
128. Aswad, supra note 23. R
129. See, e.g., Facebook Community Standards, supra note 29 (including a policy rationale for each com- R
munity standard).
The requirement of legality has also been partially eroded. This prong
demands that restrictions to freedom of expression are “provided by law.”130
That is, there must be a legal basis for restrictions. In other words, only
bodies authorized to make law may impose such limits.131 The legality
prong has also been interpreted to require that restrictions be subject to
public comment and that independent judicial officials oversee their
implementation.132
The Human Rights Committee has argued that the requirement refers
not only to the body authorizing the restriction but also to its clarity and
precision. A law must be formulated with sufficient accuracy so that indi-
viduals can regulate their conduct accordingly.133 This aspect of the require-
ment has been very generative for the IHRL project. In the context of
content moderation, the provision is understood to mean that moderation
practices must be clear and transparent.134
The first part of the legality requirement, which refers to the entity that
may restrict individuals’ freedom of expression, is impossible for companies,
as they exist today, to meet. Again, seeing companies as states could be a
first step towards imagining participatory institutions. Instead, the legality
part of the test has been narrowed down to the transparency requirement.
Both transformations risk leaving too little of IHRL standing. What is
left is a test that says that restrictions on speech should be necessary to
achieve a legitimate goal. Without any guidance about which aims are per-
missible or who should issue the restrictions, what is left is a common-sense
idea that rules should reasonably serve a reasonable purpose.
Overall, IHRL is thought of as a constraint on state power that can also
constrain corporate power. On closer examination, IHRL requires the exis-
tence of state institutions to function effectively. Corporations do not have
internal systems authorized to pass laws in the terms of Article 19 of the
ICCPR, nor do they have institutions capable of making decisions about
national security or the adequate balance between human rights. IHRL can-
not make up for the lack of these structures; it can only demand that when
these institutional actors are available, they must regulate speech in accor-
dance to Article 19 of the ICCPR and any other relevant provisions. The
130. ICCPR, supra note 36, art. 19(3). See also Gilad Abiri & Sebastián Guidi, From a Network to a R
Dilemma: The Legitimacy of Social Media, Stan. Tech. L. Rev. (forthcoming 2023) (manuscript at 32-3)
(available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4230635) (making a similar point).
131. General Comment 34, supra note 34, ¶ 24; SR Report 2016, supra note 34, ¶ 12. R
132. SR Report 2016, supra note 34, ¶¶ 12, 13; SR Report 2018, supra note 9, ¶ 7; David Kaye R
(Special Rapporteur on the Promotion and Protection of the Right to Freedom of Opinion and Expres-
sion), Promotion and Prot. of the Right to Freedom of Op. and Expression, ¶ 6(a), U.N. Doc. A/74/486 (Oct.
2019) (hereinafter “SR Report 2019”).
134. See Sander, supra note 30, at 971; Benesch, supra note 127, at 103; Aswad, supra note 23, at 46. R
However, calls for transparency in content moderation are sound, long-standing, and widely accepted. It
is unclear whether the IHRL project has added any force or meaning to these calls. See, e.g., The Santa
Clara Principles, supra note 8. R
tension then lies in the fact that even if companies do what states ought to
do, corporations do not have the institutions that are critical pieces to oper-
ate the IHRL machinery.
3. Because IHRL is consented to by all states, but state consent does not
matter
Nearly everyone seems to be well aware that no universal, or even local,
agreement on how to regulate speech exists. Even international law scholars
deeply question the universality of international law.135 However, when jus-
tifying the IHRL project, advocates remind us that IHRL is the only glob-
ally-adopted framework. As Evelyn Aswad says, “[c]ompanies need not
recreate the wheel in developing speech norms that have worldwide legiti-
macy if they base their content moderation policies on international human
rights standards.”136
In some accounts of the IHRL project, its claim to universality comes
from state consent, particularly the vast adoption of U.N. treaties.137 As
former ACLU president Nadine Strossen puts it, “almost every single coun-
try in the world is a party to the ICCPR and the ICERD.”138 Proponents of
the IHRL project also invoke the wide acceptance of the UNGPs as one of
the main reasons to acknowledge their legitimacy. Strossen and Aswad em-
phasize that the Human Rights Council endorsed the UNGPs
unanimously.139
From these statements, one could assume that states would be a good
source for identifying a community’s values or agreements. However, IHRL
advocates regard states with deep suspicion. When Aswad analyzes a propo-
sal to renegotiate and clarify some international treaties, she rejects that
possibility (for understandable reasons): “an international negotiation to
regulate speech on platforms, including content moderation, is undesirable
because it would no doubt be dominated by powerful countries with weak
records on freedom of expression that would seek to roll back international
speech protections.”140 Instead of opening up avenues for states to deliber-
ate, Aswad proposes that we rely on experts’ interpretations that emphasize
the interpretations of U.N. treaties that broaden protections for freedom of
expression.141
135. See Anthea Roberts, Is International Law International? (2017).

137. See, e.g., id. at 35.
138. Strossen, supra note 120, at 329. See also Benesch, supra note 127, at 89 (“No source of rules for R
speech regulation is as widely known or formally adopted as international human-rights law. That law’s
most relevant instrument, the International Covenant on Civil and Political Rights (ICCPR), has been
ratified by nearly ninety percent of the countries in the world.”).
139. Strossen, supra note 120, at 355-56; Aswad, supra note 23, at 38. R
141. Evelyn Aswad, To Protect Freedom of Expression, Why Not Steal Victory from the Jaws of Defeat?, 77
Wash. & Lee L. Rev. 609, 648-649 (2020).
When IHRL advocates discuss the divergences between U.N. and re-
gional treaties, they disregard state consent as an essential factor in deter-
mining which rules should govern speech. Interestingly, they acknowledge
that all regional systems diverge to various extents from U.N. standards.142
If state consent were the source of IHRL’s universality, these conflicts of
norms should be read as an opportunity to determine which commitment
more accurately reflects the state’s intent, values, or norms. However, pro-
moters of the project often assert that companies should follow only U.N.
guidance. For example, Aswad states that “regional human rights instru-
ments (and monitoring bodies) are not international human rights instru-
ments (and monitoring bodies).”143 Global expectations, she explains, are
reflected in international instruments, not regional ones.144
If it is the wide ratification of international treaties that makes these
norms a reflection of global values, it is unclear why the ratification of con-
tradictory treaties would be dismissed as irrelevant. A way out of this ten-
sion is to see human rights’ universality as prior to state consent and not
dependent on it. The wide ratification of human rights would only confirm
their universal status, rather than constitute it. In that case, IHRL may be
universal because it reflects global ethical principles that align with the
public interest. The next two subsections discuss this alternative claim to
universality.
4. Because IHRL constrains corporate power, but its indeterminacy is a

feature
The IHRL project’s central promise is that it can align corporate govern-
ance with the public interest. However, IHRL is also justified as an appro-
priate framework because it does not mandate specific policies.
IHRL could put the public interest at the center of speech regulation if it
reflected clear ways of delimiting the scope of conflicting rights that are
either globally shared or ethically correct. In response to concerns that IHRL
is too vague or internally contradictory,145 some scholars argue that IHRL
provides precise enough standards on many valuable points. Evelyn Aswad
and organizations such as Article 19 have unpacked what IHRL demands
142. See, e.g., Evelyn Aswad & David Kaye, Convergence & Conflict: Reflections on Global and Regional
Human Rights Standards on Hate Speech, 20, Nw. J. Hum. Rts., 165 (2022).
144. Id. See infra Section III.B. (addressing how IHRL advocates try to iron out this contradiction
between state consent as the source of the legitimacy of IHRL and IHRL’s indifference toward state
consent).
145. See generally Amal Clooney & Philippa Webb, The Right to Insult in International Law, 48 Colum.
Hum. Rts. L. Rev. 1 (2017); Douek, supra note 30, at 37; Brenda Dvoskin, Why International Human R
Rights Law Cannot Replace Content Moderation, Medium (Oct. 8, 2019), https://medium.com/berkman-
klein-center/why-international-human-rights-law-cannot-replace-content-moderation-d3fc8dd4344c
[https://perma.cc/S7AE-N3AT].
from corporations.146 David Kaye’s report on hate speech as Special Rap-

porteur offered clear principles that companies should follow.147
Despite these efforts to unpack IHRL into detailed rules, IHRL has areas
of indeterminacy even in its most granular versions, and its most well-estab-
lished principles admit exceptions. Hate speech bans provide a good exam-
ple. Kaye explains that bans on statements that deny well-established
historical atrocities such as the Holocaust or the Armenian genocide are in-
compatible with IHRL. According to the Special Rapporteur’s 2018 report:
“offensive interpretation of a religious tenet or historical event . . . is not to
be silenced under Article 20 (or any other provision of human rights
law).”148 The same report states that “[l]aws that penalize the expression of
opinions about historical facts are incompatible with Article 19 of the Cove-
nant, calling into question laws that criminalize the denial of the Holocaust
and other atrocities and similar laws, which are often justified through refer-
ences to hate speech.”149 Kaye immediately adds that “the application of any
such restriction under IHRL should involve the evaluation of the six factors
noted in the Rabat Plan of Action.”150 Therefore, even prohibitions that
seem clear enough to guide companies on highly controversial questions ad-
mit exceptions under vague circumstances.
The technical singularities of social media increase the indeterminacy of
IHRL. The organization Article 19 explains that most corporate policies on
hate speech are too broad to be compatible with IHRL because they ban
expressions that do not incite violence or illegal action.151 However, Kaye
clarifies that “[a]cross a range of ills that may have a more pronounced im-
pact in digital space than they might offline . . . human rights law would
not deprive companies of tools.”152 Rather, “it would offer a globally recog-
nized framework for designing those tools and a common vocabulary for
explaining their nature, purpose, and application to users and States.”153
While some see IHRL as limiting bans on offensive content that does not
incite illegal action, Kaye tells us that these bans could be acceptable in the
online context.
146. See Side-stepping rights, supra note 102. R

147. SR Report 2018, supra note 9; SR Report 2019, supra note 132. See also Michael O’Flaherty, R
International Covenant on Civil and Political Rights: interpreting freedom of expression and information standards
for the present and the future, in The United Nations and Freedom of Expression and Information
55, 82 (Tarlach McGonagle & Yvonne Donders eds., 2015).
148. SR Report 2018 supra note 9, ¶ 10. R
149. Id., ¶ 22. See also Aswad & Kaye, supra note 142. R
150. SR Report 2018 supra note 9, ¶ 22; Human Rights Council, Ann. Rep. of the U.N. High Comm’r R
for Hum. Rts. Addendum, U.N. Doc. A/HRC/22/17/Add.4 (Jan. 11, 2013) (outlining a six-part threshold
test taking into account (1) the social and political context, (2) status of the speaker, (3) intent to incite
the audience against a target group, (4) content and form of the speech, (5) extent of its dissemination
and (6) likelihood of harm, including imminence).
151. Side-stepping rights, supra note 102, at 19. R
152. SR Report 2018, supra note 9, ¶ 43. R
153. Id.
At the same time, IHRL’s indeterminacy is praised as a positive feature.

Because IHRL is, like any legal system, indeterminate, its adoption does not
impose answers. Instead, it offers a margin of maneuver for companies to
choose their house rules154 and a process for actors to engage in conversation
and to reach those answers.155 However, it is unclear how IHRL would facil-
itate conversations between different actors. If actors are not included in the
discussion, it is more likely due to power and resource differentials rather
than the need for a common language to understand each other.156
The two sides of the debate emphasize potentially positive aspects of
IHRL: it allows a diversity of regulatory approaches and it provides some
parameters for acceptable rules. However, a tension exists between arguing
that IHRL offers global standards that companies can voluntarily adhere to
and arguing that IHRL’s main advantage is that it can create the necessary
conditions for a multiplicity of actors to decide what those standards should
be.
5. Because IHRL is a shared language and because IHRL is like the First
Amendment
IHRL is portrayed as “a shared language” or least a common denomina-
tor.157 It is thought to provide a baseline for (unidentified) actors to be in
conversation and reach more granular agreements with each other.
At the same time, U.S. scholars argue that calling on companies to adopt
U.N. standards is the most promising avenue to bring content moderation
closer to First Amendment doctrine.158 They believe this strategy is auspi-
cious because U.N. sources would be more readily accepted globally than
the First Amendment doctrine, even though the two share many vital
elements.159
Similarly, as previously discussed, when U.N. authoritative interpreta-
tions conflict with decisions originating from European bodies, proponents
of the IHRL project assert that U.N. solutions should prevail in the context
of content moderation.160 It is problematic to claim that a legal framework
is open-ended, and either reflects or invites the search of collective answers,
while arguing that U.N. solutions should always preempt regional or local
preferences.
154. SR Report 2019, supra note 132, ¶¶ 43, 48; Article 19, supra note 43, at 13. R
155. Sander, supra note 30, at 967-68. R
156. Blayne Haggart & Clara Iglesias Keller, Democratic legitimacy in global platform governance, 45
Telecomm. Pol’y 1, 11-12 (2021); David Kennedy, The International Human Rights Movement: Part of the
Problem?, 15 Harv. Hum. Rts. J. 101, 109 (2002).
157. Michael Ignatieff, Human Rights as Idolatry, in Tanner Lectures on Human Values 349
(2000) (articulating the value of human rights as a “shared vocabulary” in a broader context).
158. Strossen, supra note 120, at 333 (“Notwithstanding widespread assumptions about the excep- R
tionally speech-protective nature of U.S. free speech law, careful comparison of the U.N. approach to that
of the U.S. demonstrates that the two share more key elements than has generally been recognized.”).
159. Id.
160. See, e.g., Aswad, supra note 23, at 57. R
B. Objective Justifications
These tensions, contradictions, and innovative interpretations show that

the invoked reasons for corporations to apply international law are not objec-
tive. However, by portraying the reasons as such, it is possible to justify why
the global society should perceive the governance of online speech by corpo-
rate actors and bodies of experts as advancing the public interest. Human
rights experts’ democratic buy-in allegedly comes from the framework that
constrains them and that they implement. The decision to use this specific
framework, in turn, must itself be a reflection of the common good or some
general consensus. Each of the justifications examined above reflects this
kind of discourse: experts shall apply IHRL because an exogenous instru-
ment like the UNGPs mandates this application or because IHRL has been
collectively agreed on as the appropriate framework to govern speech on
social media.
The UNGPs, through hard work, have been reconstructed as an exoge-
nous instrument, already endorsed unanimously by the Human Rights
Council, that recommends that corporations align content policies with
IHRL. In all the other justifications, the premise is that the “global society”
already knows what rules should regulate speech.
In that sense, in the second justification (platforms are like states and we
already know how states ought to restrict speech), the debate shifts from
how society wants companies to moderate content to how the standards we
supposedly already recognize as appropriate to govern speech can be opera-
tionalized. That is, this rationale is premised on the view that a broad agree-
ment exists about how speech should be governed. Accordingly, what is
needed is an adjustment to a new context that experts can carry out.
State consent is another expression of the ideal of objectivity: interna-
tional law expresses what everyone—or every state—has agreed on. In this
case, objectivity draws not on the rightfulness of the principles but on every
state’s participation. When IHRL advocates reference states’ consent, they
imagine that the participatory procedure has already occurred.161 What is
left are standards for speech regulation that can be implemented in different
contexts. In the last two rationales, the ideal of objectivity can also be traced
to an imagined consensus among global constituencies. IHRL is described as
a shared language among societies globally.
Not all the reasons supporting the IHRL project, however, are premised
on this ideal. When advocates say that IHRL is a good framework because it
resembles the core principles of First Amendment jurisprudence,162 these
advocates champion their normative preferences. Even in this case, advocates
161. Aswad, supra note 141 (arguing that this participatory procedure should not be reopened to R
negotiate more granular rules applicable to social media platforms because the outcome could be less
protective of freedom of expression than the current interpretations made by U.N. agencies).
162. See supra Section I.D.
believe that their normative preferences will be more convincing if framed as

neutral principles. When Nadine Strossen compares the First Amendment
to U.N. treaties, she emphasizes that the two “share more key elements than
has generally been recognized.”163 However, Strossen is right to believe that
it is strategic to frame advocacy efforts in terms of U.N. rules because they
have a claim to universality that the First Amendment lacks.
Overall, to the extent that the IHRL project is concerned with making
online speech governance more aligned with the public interest, the project
imagines a governance model where a group of people implement a set of
principles that reflect the interests of the global society. This section ex-
amined how advocates justify or imagine the availability of that set of prin-
ciples. The next section analyses how experts and especially the Board
perform as objective executors of those principles.
III. Deploying the IHRL Project
This section explores how decisionmakers, and especially the Board, navi-
gate conflicts of norms and areas of indeterminacy within international law
while preserving the project’s claim to objectivity. When IHRL does not
yield a specific outcome or experts are faced with multiple possible solu-
tions, experts invoke interpretative theories, technical facts, or other strate-
gies to conceal their policy choices. Decisions are justified as a logical
deduction from higher principles or as the necessary consequence of objec-
tive facts. This section looks at four tools that the IHRL project has so far
developed for expert governance, while paying close attention to the Board’s
first decisions.
A. IHRL as Self-Evidently Good

How is it that, despite so much disagreement about how platforms should
moderate content, so many people believe that their views about online
speech are based on human rights?164 It may be that because this concept is
so ill-defined, it can justify many policy proposals and decisions. All sides of
any debate about restrictions on speech are defending a human right. In-
deed, most interests usually invoked around speech adjudication can find an
anchor in a human right: freedom of expression, safety, dignity, non-dis-
crimination, equality.
Because of that open texture, “human rights” can be used as a synonym
for the common good. A policy report by the organization Article 19 offers a
clear example. This report describes the steps social media companies should
take to comply with international freedom of expression standards.165 When
163. Strossen, supra note 120, at 333. R

164. See supra notes 21-26 and accompanying text. R
165. Side-stepping rights, supra note 102. R
analyzing platforms’ internal hate speech rules, the report finds that plat-
forms’ rules usually restrict speech that IHRL protects.166 According to the
report, companies do not offer robust speech protections for commercial rea-
sons. The authors hypothesize that these rules enable platforms to grow their
user base and accommodate advertisers’ interests.167 Overall, the report as-
sumes that there are two options: rules driven by business interests and rules
that align with IHRL. At no point does the report question the idea that
IHRL and the public interest are equivalent.
The comparison between conversations about platforms adopting First
Amendment doctrines and IHRL as a default illuminates how IHRL is dis-
cussed as undoubtedly good. Most scholars who support First Amendment
doctrine as a generally appropriate normative framework to govern speech
are resistant to asking companies to apply it as a default rule for content
moderation.168 Scholars acknowledge that if platforms hosted all First
Amendment–protected speech, platforms would be worse for almost all
users.169 However, similar concerns have not been raised by IHRL support-
ers, even though if social media were to allow all the speech that IHRL
protects, users and companies would also have to tolerate all types of unde-
sirable speech.170 IHRL has both a non-contestable character and a higher
level of indeterminacy that seems to incentivize people to accept it as a
framework and work out exceptions rather than rejecting the whole frame-
work because it does not perfectly track their normative preferences.
Because of its simultaneously undefined and uncontestable character,
human rights language can be invoked to justify a decision without many
further explanations. If someone proposes that platforms should adopt a cer-
tain policy because it aligns with the First Amendment, it is very likely that
other advocates will raise serious objections to that justification. If someone,
however, argues in favor of a rule because it follows human rights law, they
are unlikely to be challenged on the basis that their framework is inade-
quate, even if they still have to persuade others about the merits of the
specific rule.
166. Id. at 16.

167. Id.
168. See Dvoskin, supra note 26. R
169. See, e.g., Jack Balkin, Free Speech is a Triangle, 118 Colum. L. Rev. 2011, 2025 (2018) (“[T]he
best alternative to this autocracy is not the imposition of First Amendment doctrines by analogy to the
public forum or the company town.”); Keller, supra note 108, at 13; Jonathan Peters, The Sovereigns of R
Cyberspace and State Action: The First Amendment’s Application-Or Lack Thereof-To Third-Party Platforms, 32
Berkeley Tech. L.J. 989 (2018); Rebecca Tushnet, Power Without Responsibility: Intermediaries and the
First Amendment, 76 Geo. Wash. L. Rev. 986, 988 (2008); Christopher Yoo, Free Speech and the Myth of
the Internet as an Unintermediated Experience, 78 Geo. Wash. L. Rev. 697, 700 (2010).
170. See Keller, supra note 108. R
B. The Relationship between Global and Regional Norms

A recurrent problem that IHRL supporters face is the tension between the
U.N. system and regional systems of human rights. Discussing the feasibil-
ity of the IHRL project, scholars Susan Benesch and Barrie Sander emphasize
that human rights law can be confusing and inconsistent.171 In particular,
Evelyn Douek points out that regional treaties sometimes contradict each
other or U.N.-level instruments.172
IHRL advocates acknowledge these conflicts. David Kaye mentions that
the European Court of Human Rights has adopted a deferential approach
toward states that ban blasphemy or criminalize genocide denial “in contrast
to trends observed at the global level.”173 Aswad adds that most regional
systems have points of tension with U.N. treaties.174 The Cairo Declaration
on Human Rights in Islam states that everyone shall have a right to express
their opinions freely “in such manner as would not be contrary to the princi-
ples of the Shari’ah.”175 The Association of Southeast Asian Nations’ human
rights declaration limits freedom of expression in ways that are inconsistent
with the Universal Declaration of Human Rights.176 Recently, Aswad and
Kaye have analyzed in depth the convergences and divergences of regional
human rights systems and U.N. norms regarding hate speech standards and
have concluded that regional norms often diverge from U.N. standards.177
Human rights tribunals often navigate these conflicts through jurisdic-
tional rules. For example, even though international tribunals may be in
dialogue with each other, the Human Rights Committee is tasked with in-
terpreting the ICCPR while the Inter-American Court first looks at the
American Convention on Human Rights.178 This is also true for other kinds
of tribunals: World Trade Organization tribunals normally give priority to
WTO agreements, while human rights norms take precedence before human
rights courts.179 Similarly, national authorities might give more importance
171. Benesch, supra note 127, at 91; Sander, supra note 30, at 977. R
172. Douek, supra note 30. See also Dvoskin, supra note 145. R
173. SR Report 2019, supra note 132, ¶ 26 (referring to the U.N. level). R
174. Aswad, supra note 136, at 634, n.94. R
175. Org. of Islamic Cooperation, The Cairo Declaration on Human Rights in Islam, Annex to Res. No.
49/19-P, art. 22 (Aug. 5, 1990).
176. Victoria Nuland, Press Statement on the ASEAN Declaration on Human Rights, U.S. State Depart-
ment (Nov. 20, 2012), https://2009-2017.state.gov/r/pa/prs/ps/2012/11/200915.htm [https://perma.cc/
ZC2W-WNTC] (“While part of the ASEAN Declaration adopted November 18 tracks the [Universal
Declaration of Human Rights], we are deeply concerned that many of the ASEAN Declaration’s princi-
ples and articles could weaken and erode universal human rights and fundamental freedoms as contained
in the UDHR.”).
177. Aswad & Kaye, supra note 142. R
178. Antoine Buyse, Tacit Citing: The Scarcity of Judicial Dialogue between the Global and the Regional
Human Rights Mechanisms in Freedom of Expression Cases, in The United Nations and Freedom of
Expression and Information: Critical Perspectives 443 (Tarlach McGonagle & Yvonne Donders
eds., 2015) (describing these dynamics).
179. Int’l Law Comm’n, Fragmentation of International Law: Difficulties Arising from the Diversification
and Expansion of International Law, ¶¶ 165-71, U.N. Doc. A/CN.4/L.682 (Apr. 13, 2006).
to some international norms than others because their domestic constitution

incorporates some human rights treaties and not others, for example.180
Overall, different authorities give priority to the norms that are, for various
reasons, more relevant to their decisionmaking process.
This strategy is not available to corporate actors because they are not part
of any treaty-based system. It is therefore unclear what rules, if any, should
take priority in guiding their internal speech governance. The Board has so
far applied U.N. treaties and regularly references the opinions and reports
issued by U.N. bodies.181 When Facebook and Twitter have referenced both
U.N. and regional treaties, Aswad has called on them to be cognizant of the
tensions between those instruments and to prioritize the protection that the
U.N. affords to freedom of expression.182
IHRL advocates have looked for principles within international law to
justify the preeminence of U.N. norms. International law, however, offers
little help, as it contains only a few and not robust principles to solve con-
flicts of norms.183 A report of the International Law Commission (hereinafter
“ILC”), finalized by Martti Koskenniemi, addressed the fragmentation of
international law.184 The report’s starting point is that “[w]hereas domestic
law is organized in a strictly hierarchical way, with the constitution regulat-
ing the operation of the system at the highest level, there is no such formal
constitution in international law and, consequently, no general order of pre-
cedence between international legal rules.”185
The report by the ILC identifies legal techniques capable of resolving
normative conflicts by putting rules in relationship with each other.186 A
few things are somewhat clear, such as the fact that the United Nations
Charter takes priority over other treaties.187 But the relationship between
global and regional human rights treaties is not one of them. The report
itself warns that looking for formal unity in a pluralistic and complex global
society is “pointless.”188 Instead, fragmentation and normative conflicts re-
flect the differing preferences and projects of actors in a heterogeneous
world. Identifying plausible ways to deal with these conflicts, however, may
be valuable and even necessary for tribunals adjudicating conflicts.
180. Joana Harrington, The Democratic Challenge of Incorporation: International Human Rights Treaties and
National Constitutions, 38 Victoria U. Wellington L. Rev. 217 (2007).
181. See, e.g., Case decision 2022-003-IG-UA, Oversight Bd. (June 13, 2022), https://over-
sightboard.com/decision/IG-2PJ00L4T/ [https://perma.cc/CZD5-VMA4] (citing Communication 488/
1992, Toonen v. Australia, Human Rights Committee, 1992; Resolution 32/2, Human Rights Council,
2016; UN Special Rapporteur on freedom of opinion and expression, reports: A/HRC/38/35 (2018) and
A/74/486 (2019); UN High Commissioner for Human Rights, report: A/HRC/19/41 (2011)).
183. See Int’l Law Comm’n, supra note 179, ¶ 26. R
184. Id.
185. Id., ¶ 324.
186. Int’l Law Comm’n, supra note 179, ¶ 410. R
187. U.N. Charter art. 103.
Most relevant to the problem of regional and universal treaties is the tool
of lex specialis.189 This technique prioritizes the most specific rule when two
legal provisions are applicable and no clear hierarchical relationship exists
between them. The closer connection to the particular context may more
adequately reflect the interests and consent of the parties involved.190 How-
ever, this technique did not make it into the Vienna Convention on the Law
of Treaties from 1960, which identifies other methods for treaty interpreta-
tion.191 Some authors dismiss this technique as irrelevant or impractical.192
In any case, using lex specialis as a plausible tool to put global and regional
rules in relation to each other is incompatible with the general rule that
IHRL advocates propose (i.e., that U.N. norms always take priority).
Because international law has not traditionally dictated that U.N. norms
should be preferred to regional norms, proponents of the IHRL project have
focused on alternative interpretations of the relationship between human
rights treaties. For example, Aswad’s reply to Douek’s objection about in-
consistencies between regional and global treaties is that “[s]uch a concern
inappropriately conflates IHRL with separate bodies of law embodied in re-
gional human rights instruments.”193 Excluding regional human rights trea-
ties from international law is highly uncommon. Describing a norm as
“international” refers not to its global application, but to the source of legal
authority upon which the norm exists.194 That is, regional human rights
treaties are part of international law “because they have been recognized as
an international legal obligation through established international legal
process.”195
IHRL advocates also argue that “[r]egional human rights norms cannot,
in any event, be invoked to justify departure from international human
rights protections.”196 But this says nothing about the hierarchy of interna-
tional norms or which rules companies should use to govern speech. This
only says that a U.N. body would consider a state to be in violation of its
international duties even if its conduct is permissible under another interna-
tional norm. A regional tribunal could reach a similar conclusion if a state
189. Id., ¶¶ 103-07.

190. Id., ¶ 206; Hugo Grotious, De Jure belli ac pacis. Libri Tres, in The Classics of International
Law 428 (James Brown Scott ed., 1925) (“Among agreements which are equal . . . that should be given
preference which is most specific and approaches most nearly to the subject in hand, for special provisions
are ordinarily more effective than those that are general.”).
191. Vienna Convention on the Law of Treaties (May 23, 1969), arts. 31-33.
192. See, e.g., Daniel P. O’Connell, International Law 253 (2nd ed. 1970).
194. Douglas Lee Donoho, International Human Rights Law 9 (2017).
195. Id.
196. SR Report 2019, supra note 132, ¶ 26. But see Kaye, supra note 46 (arguing that companies R
should draw guidance from the European Court of Human Rights, the Inter-American Court for Human
Rights, the emerging jurisprudence of regional and sub-regional courts in Africa, national courts in
democratic societies, treaty bodies that monitor compliance with their norms, and the work of U.N. and
regional human rights mechanisms).
adopted a rule that is compatible with a U.N. norm but incompatible with
its duties under a regional treaty.
It is common for IHRL advocates to argue that the U.N. system acts as a
floor or minimum of rights that companies must respect. Companies should
only apply regional norms when these norms expand rights. Along those
lines, Aswad argues that inconsistencies between international and regional
mechanisms do not render the U.N. human rights regime incoherent: “[i]t
simply means that the U.N. system provides more protections for speech
than the regional systems.”197 Discussing the idea of referencing regional
human rights treaties in the decisions of the Board, Co-Chair Catalina
Botero stated that the Board could apply regional treaties in the future if the
treaties expanded the level of protection of freedom of expression.198
Usually, the expansion of one right bears costs on other rights. For exam-
ple, in regulating prior restraint on speech, the American Convention on
Human Rights is more protective of speech than any other system.199 But
this protection of speech comes at the expense of less protection for other
rights such as privacy or safety.200 One can only see the Inter-American Con-
vention as expanding rights if one assumes that the right that ought to be
expanded is freedom of expression and not others.
In the context of the Board, this preference to expand freedom of expres-
sion over other rights has normative support. The Board’s Charter refers only
to “human rights norms that protect freedom of expression.”201 As a general
rule for how companies should adopt international human rights, however, a
normative basis is lacking. The principle is articulated in neutral terms
(“apply the rule that offers higher protection”) but it hides a clear normative
orientation (“always choose the most free-speech protective rule”).
Finally, IHRL advocates may find a basis for the preeminence of U.N.
treaties in the UNGPs. The UNGPs only refer to the International Bill of
Rights, which consists of the Universal Declaration of Human Rights, the
ICCPR, and the International Covenant on Economic, Social, and Cultural
Rights.202 Therefore, Aswad argues that companies ought to look exclusively
at global treaties. For example, she considers that Twitter’s statement about

198. Online Event: The Decisions of Facebook’s Oversight Board – Implications for the Global South, particu-
larly in Latin America, The Dialogue (May 18, 2021), min. 55:00, https://www.thedialogue.org/events/
online-event-the-decisions-of-facebooks-oversight-board-implications-for-the-global-south-particularly-
in-latin-america/ [https://perma.cc/92BB-QKSK].
199. American Convention on Human Rights, art. 13, Nov. 22, 1969, O.A.S.T.S. No. 36, 1144
U.N.T.S. 123 (“[t]he exercise of the right [to freedom of expression] shall not be subject to prior censor-
ship but shall be subject to subsequent imposition of liability . . . [with the exception that] public
entertainments may be subject by law to prior censorship for the sole purpose of regulating access to
them for the moral protection of childhood and adolescence.”).
200. Dvoskin, supra note 145. R
201. Oversight Board Charter, art. 2(2), Oversight Bd., https://oversightboard.com/governance/
[https://perma.cc/43ZW-2CMU].
looking at U.S. law and the European Convention on Human Rights “de-
parts from the UNGPs, which provide that companies should seek to align
their operations with IHRL rather than domestic laws (like the U.S. Bill of
Rights) or regional law (such as the European Human Rights
Convention).”203
As argued above, this is a new interpretation of the UNGPs.204 John Rug-
gie, the author of the UNGPs, believed companies should look at the Inter-
national Bill of Rights and the ILO’s Declaration on Fundamental Principles
and Rights at Work as an authoritative list of recognized rights.205 How-
ever, this did not mean that companies should follow the law stemming
from U.N. treaties and their authoritative interpretation. In addition, dis-
cussing the relationship between global and regional norms, Ruggie con-
cludes that “no homogenous hierarchical meta-system is realistically
available within the international legal order to resolve the problem of in-
compatible provisions among different bodies of law.”206
Looking at the efforts to build a coherent and uniform system of interna-
tional rights makes the work toward objectivity visible. A clear relationship
between global and regional norms creates the illusion that there is a system
of norms readily available to constrain corporate power. However, asking
companies to behave as if a coherent universal system of human rights ex-
isted gives corporations unprecedented powers. Corporations are put in
charge of deciding which rights should be prioritized when international
norms conflict, or how conflicts of norms between the European system of
human rights (or any other regional system) and the U.N. should be
adjudicated.
C. Normative Indeterminacy as Technical Questions

Scholars agree that IHRL has areas of significant indeterminacy.207 Some
scholars argue that technological innovation exacerbates this indeterminacy:
the governance of speech on social media platforms poses novel questions
that international law has not had the chance to address yet.208 In addition,
some believe that the speed and reach of the distribution of online expres-
sion make social media content exponentially more harmful.209 Therefore,
social media may require rules that give more weight to the values of dig-
nity, safety, or equity. The question for the IHRL project is then, given the
technical specificities of social media, should companies interpret IHRL dif-
203. Aswad, supra note 23, at 44; see also Aswad & Kaye, supra note 142, at 56. R
206. Id. at 65.
207. See, e.g., Danielle Citron, Extremist Speech, Compelled Conformity, and Censorship Creep, 93 Notre
Dame L. Rev. 1035, 1063 (2018); Douek, supra note 13, at 66. R
208. See, e.g., Douek, supra note 30, at 56, 72. R
209. Danielle Citron, Cyber Civil Rights, 89 B.U.L. Rev. 61, 63 (2010).
ferently than states and international bodies do when they restrict speech
offline?
The Board has started to deal with this question. In some cases, it allowed
Facebook to adopt rules that, in its own view, directly contradict what
IHRL prescribes. Even in these cases, the Board justified these rules as an
appropriate implementation of human rights in part because online speech is
different from offline speech. A series of decisions evaluated Facebook’s ban
on a list of racial slurs210 and Case Decision 2021-002-FB-UA reviewed
Facebook’s prohibition on content depicting blackface.211
The Board had to justify why, in its view, a state would violate interna-
tional law if the state issued a ban on blackface or on racial slurs, but
Facebook was meeting its human rights responsibilities when issuing those
same bans. In the case about blackface, the Board hinted at four rationales:
U.N. experts and other authorities determined that blackface creates objec-
tive harm;212 Facebook has a human right responsibility to promote equal-
ity;213 it is hard to evaluate the intent of the speaker on social media;214 hate
speech, even when it does not have an intent to discriminate or to incite
violence, can create a discriminatory, harassing, and degrading
environment.215
Consider the last two rationales. One could challenge them on factual
terms. It might well be that racial slurs and people in blackface create a
discriminatory environment regardless of whether the expression is online or
off, and the intent of the speaker may always be hard to judge. It is also
unclear that these types of expressions when they take place at one’s univer-
sity, in the workplace, or other close social environment are less harmful
than when they occur on social media. But let us accept these distinctions
between online and offline speech. Perhaps digital racial slurs and people in
blackface create an even worse discriminatory environment because of their
reach and quantity. What should we make of that difference?
The Board gave three different answers. First, the general ban on black-
face that Facebook adopted is incompatible with international law.216 In the
Board’s words, “international human rights law would not allow a state to
impose a general prohibition on blackface through criminal or civil sanc-
tions, except under the conditions foreseen in ICCPR Article 20, para. 2 and
Article 19, para. 3.” In the Board’s view, the post discussed in this case
210. Case decision 2021-011-FB-UA, Oversight Bd. (Sept. 28, 2021), https://oversightboard.com/
decision/FB-TYE2766G/ [https://perma.cc/UE9F-FHP7]; Case decision 2020-003-FB-UA, Oversight
Bd. (Jan. 28, 2021), https://oversightboard.com/decision/FB-QBJDASCV/ [https://perma.cc/DC5H-
STWU].
211. See supra Introduction (discussing this decision).
212. Case decision 2021-002-FB-UA, supra note 1, at 16. R
213. Id.
214. Id.
215. Id.
216. Id. at 14.
“[w]ould fall within this category of protection from state restriction.” Sec-
ond, in this new medium, speech produces different consequences, and new
answers are possible.217 Third, the prohibition is compatible with interna-
tional law. The Board concluded that “Facebook followed international gui-
dance and met its human rights responsibilities in this case.”218
Notice the tension between asserting that states would breach interna-
tional law if they adopted this rule (that is, IHRL does prescribe what rules
are acceptable) and that the new medium creates a situation in which new
normative answers are possible (that is, IHRL does not prescribe what rules
are acceptable).
The Board had two options to avoid that tension. First, given the open
texture of IHRL, the Board could have argued that IHRL allows prohibi-
tions on blackface, at least in the online context.219 However, that would
have meant committing to the view that states can also issue these regula-
tions. What seems to animate the Board’s reasoning is the belief that states
and companies should actually govern speech differently. If that is the case,
the Board also had a second option. It could have stated that IHRL was not
the appropriate framework in this case because states and corporations are
different. However, that would have meant losing the legitimizing force of
IHRL.
Ultimately, the Board adopted a third choice: states and corporations
ought to regulate speech differently and both options are compatible with
IHRL. In order to preserve the claim to objectivity, the Board framed this
third option as determined by technical facts.
The strategy here was to transform an unanswered question into one that
can be answered with technical facts. The Board’s move was the following:
international law does not allow general bans on content depicting black-
face, but online speech is more harmful and harder (perhaps impossible) to
adjudicate on a case-by-case basis. The Board’s conclusion could have been
that IHRL provides no guidance to decide the case. Instead, the Board de-
cided that these technical facts make Facebook’s rule proportional and com-
patible with IHRL. Through these steps, the Board framed an open
normative question as a question about how to translate international law
principles in a new technical context, as if the technical context could deter-
mine the answer to the question. As a result, the Board did not appear to be
creating new human rights standards, but rather appeared to be an objective
translator.
Labeling new norms designed for platform governance as IHRL is a clear
expression of the ideal of expert governance. Even if these new norms oppose
IHRL norms, preserving the human rights label helps to maintain IHRL’s
217. Id. at 16 (assessing the cumulative effects of speech on social media and the challenges of enforce-
ment at scale).
218. Id. at 14.
219. Thanks to Evelyn Douek for raising this point in a personal conversation.
claims to legitimacy, universality, and democratic buy-in. This import of

credentials from some initial principles (such as IHRL) to new normative
outcomes is the essence of how expert governance works.
D. Local Preferences as Local Facts

Differences across geographical contexts bring into question the univer-
sality of human rights.220 This tension between the universal character of
human rights and local divergences is an old dispute in the field.221 Scholars
sometimes see the variations in normative solutions as a positive develop-
ment.222 Jack Goldsmith and Tim Wu value the different attitudes toward
proper speech regulation as a reflection of differences in culture and taste.223
More generally, the ILC finds that regional law better reflects the interests
of the affected constituencies.224
The challenge for the IHRL project is how to account for the contradic-
tions across cultures and societies while maintaining the claim that IHRL is
shared across communities. A tool used to overcome this challenge is to
reconstruct the variations across societies as problems of fair implementa-
tion. Indeed, the IHRL project often reduces the divergences between nor-
mative preferences to differences in the local facts. In turn, this move shifts
the problem from the unavailability of universal normative solutions to a
problem of lack of sufficient knowledge about the situation on the ground to
ensure the correct application of global rules.225 In order to address the im-
plementation problem, the IHRL project emphasizes the importance of that
local expertise. Local experts can make local contexts explainable and visible
so that the global executor of rules can govern fairly.
This technique is not exclusive to the IHRL project. The broader project
of creating a set of global rules often tries to incorporate different local pref-
erences through expertise. Trusted partner programs are a prominent mecha-
nism to incorporate local expertise in global content moderation.226 Local
NGOs assist in monitoring social media and applying companies’ commu-
220. Siva Vaidhyanathan, Antisocial Media: How Facebook Disconnects Us and Under-
mines Democracy 27 (2018).
221. Kennedy, supra note 156. R
222. See, e.g., Dina Shelton, The Promise of Regional Human Rights Systems, in The Future of Interna-
tional Human Rights 351, 356 (Burns H. Weston & Stephen P. Marks eds., 1999); Malcom Evans,
The Future(s) of Regional Courts on Human Rights, in Realizing Utopia 261 (Antonio Cassese ed., 2012).
223. Jack Goldsmith & Tim Wu, Who Controls the Internet? Illusions of a Borderless
World 150 (2006).
225. See Stephen Humphreys, Theatre of the Rule of Law: Transnational Legal Inter-
vention in Theory and Practice (2010) (exploring how rule of law promotion programs reframe
political preferences as questions of expertise).
226. Naomi Appelman & Paddy Leerssen, On “Trusted” Flaggers, Yale-Wikimedia Initiative on
Intermediaries & Information (July 12, 2022), https://law.yale.edu/sites/default/files/area/center/isp/
documents/trustedflaggers_ispessayseries_jul2022.pdf [https://perma.cc/NWP4-UXE8]; Brenda Dvos-
kin, Social Media Platforms and Civil Society in Latin America: A View from the Nonprofit Organizations
(Centro de Estudios en Tecnologı́a y Sociedad, Working Paper No. 20, 2020).
nity guidelines. These programs provide local civil society organizations a

privileged channel to report content that companies should take down or
appeal decisions when content has been eliminated incorrectly.227 The direc-
tor of a Latin American organization says in an interview that Facebook
often consults them about the meaning of a slang phrase or the details of an
ongoing protest, but the opportunities to challenge the rules of the com-
pany, which they see as over-censorial, are practically nonexistent.228
The Board is particularly emphatic about the importance of local con-
texts. When deciding cases, it often cites expert reports that explain what
rules the situation on the ground requires.229 The interest in understanding
the local situation adequately is a step worth celebrating.230 However, like
trusted partner programs, the Board is more invested in learning local facts
rather than recognizing the variations in preferences about how to balance
free expression and other values.
The Board usually looks at local specificities under the proportionality
analysis of Article 19 of the ICCPR. This proportionality test is an interest-
ing site to locate different possible balances between rights (in the case
about Black Pete, between the right to freedom of expression and the right
to non-discrimination). Indeed, whether a speech restriction is proportional
to achieve a specific goal is fundamentally a matter of normative taste. A
proportionality analysis demands answering questions such as how many
false positives are tolerable to prevent certain harm.231 For example, a ban on
nudity can pursue the goal of preventing all cases of nonconsensual distribu-
tion of intimate images. The ban will also affect content that is uploaded
consensually. A more nuanced rule could affect less content but would be
less effective in countering all instances of nonconsensual uploads since er-
rors will be unavoidable. Whether the most expansive ban is proportional to
the goal will depend on how a society values the protection of nudity online
or how essential a society considers the avoidance of nonconsensual distribu-
tion to be.
227. See, e.g., EFHR welcomed into Trusted Partner Channel of Facebook, European Foundation of
Human Rights (Jan. 29, 2018), https://en.efhr.eu/2018/01/29/efhr-welcomed-trusted-partner-channel-
facebook/ [https://perma.cc/8M2Y-A4AU].
228. Dvoskin, supra note 226, at 8-9. R
229. See, e.g., Case decision 2021-010-FB-UA, Oversight Bd. (Sept. 27, 2021), https://over-
sightboard.com/decision/FB-E5M6QZGA/ [https://perma.cc/GCE6-75XA]; Case decision 2020-006-FB-
FBR, Oversight Bd. (Jan. 28, 2021), https://oversightboard.com/decision/FB-XWJQBU9A/ [https://
perma.cc/4K9L-K2GA].
230. Chinmayi Arun, Rebalancing Regulation of Speech: Hyper-Local Content on Global Web-Based Plat-
forms, Medium (Mar. 28, 2018), https://medium.com/berkman-klein-center/rebalancing-regulation-of-
speech-hyper-local-content-on-global-web-based-platforms-1-386d65d86e32 [https://perma.cc/XUD4-
3K9W] (stressing the importance of local contexts for platform governance).
231. See generally Evelyn Douek, Governing Online Speech: From “Posts-As-Trumps” to Proportionality and
Probability, 121 Colum. L. Rev. 759 (2021); New York Times Co. v. Sullivan, 376 U.S. 254 (1964)
(arguing, although not in these terms, that actual malice was the appropriate standard of review because
it avoids false positives, even if that means allowing some false negatives).
However, this is not the kind of local context that the Board considers
necessary to consider. For example, in Case Decision 2020-006-FB-FBR, the
Board analyzed Facebook’s removal of false information about COVID-19.232
The balance between free expression and public health in the context of the
COVID pandemic is a good example of where people and communities hold
different views. Platforms, accused of contributing to the spread of misinfor-
mation, have aggressively targeted public health misinformation at least
partially in response to the preferences of some segments of the public.233
In the post at issue, a user had stated (incorrectly) that a cure for COVID-
19 was available, that the drug was harmless, and complained that the
French authorities were not making the drug available in France.234
Facebook explained to the Board that, following the opinion of consulted
experts, it decided to take down all content stating that a cure for COVID-
19 is available because other users may believe it and may disregard precau-
tionary health guidance or may self-medicate as a consequence.235 The Board
disagreed that this was the right balance between freedom of expression and
public health. The user was questioning a governmental policy and calling
for a change.236 The Board reasoned that the protection of this kind of politi-
cal discourse was fundamental.237 In addition, the Board was not certain that
the post could contribute to imminent harm.238
In this case, the Board highlighted that the drugs referenced in the post
were not available in France without a prescription.239 It added that, “the
alleged cure has not been approved by the French authorities and thus it is
unclear why those reading the post would be inclined to disregard health
precautions for a cure they cannot access.”240 As in other decisions, the
Board was very interested in looking at the offline context of a post, but this
context was limited to facts that experts can explain and the Board can con-
sider when making its own normative choice.
IV. Participating in the IHRL Project
At times, the IHRL project calls for a conversation to develop new

rules.241 As discussed above, one strong justification for the IHRL project is
232. Case decision 2020-006-FB-FBR, supra note 229. R

233. See, e.g., Guy Rosen, An Update on Our Work to Keep People Informed and Limit Misinformation About
COVID-19, Meta (Apr. 16, 2020), https://about.fb.com/news/2020/04/covid-19-misinfo-update/
[https://perma.cc/42GN-6N83]; COVID-19 Medical Misinformation Policy, Google (May 20, 2020),
https://support.google.com/youtube/answer/9891785 [https://perma.cc/MTZ7-ATWJ].
234. Case decision 2020-006-FB-FBR, supra note 229, at 3-4. R
235. Id. at 8.
236. Id.
237. Id. at 10.
238. Id. at 8.
239. Id.
240. Id. at 8-9.
241. Douek, supra note 30, at 47. R
its alleged potential to facilitate the conversation among multiple stakehold-

ers.242 Evelyn Aswad calls for multi-stakeholder initiatives to refine some
aspects of the IHRL framework.243 David Kaye insists on the importance of
including more actors in the conversation.244
These proposals for multi-stakeholder conversations or consultation
mechanisms envision a project of objective governance different from the
pursuit of exogenous standards discussed so far. These initiatives aim at
building what Sheila Jasaoff calls a view from everywhere.245 In this version,
legitimacy depends on all views having adequate representation. Many au-
thors theorize that legitimacy stems from procedures that ensure that all
positions are heard. Ronald Dworkin argued that the legitimacy of laws
depends on everyone’s opportunity to manifest their opposition to them.246
Owen Fiss contended that the purpose of the right to freedom of expression
is genuine collective self-determination.247 Therefore, free speech regulations
ought to ensure that all viewpoints are heard.248 In Jürgen Habermas’s work,
rules of participation and public reasoning create an ideal speech situation
that makes the outcome of a deliberation worthy of respect by all parties,
regardless of the substantive content of the result.249
Unlike the claims that IHRL is universal because it is the outcome of a
participatory process that took place in the past (e.g., “states have ratified
human rights treaties” or “IHRL is the outcome of a global consensus”),
this version of the IHRL project calls for a future participatory process to
draft new rules. It is unclear in what institutional setting that conversation
may occur or how IHRL may facilitate it. On the one hand, the IHRL pro-
ject may contribute substantively to these dialogues. It demands trans-
parency and public justifications for corporate practices.250 Transparency
requirements would enable more informed conversations. On the other
hand, the calls for transparency are strong enough not to gain much from
the added support from the IHRL project. IHRL advocates also argue that it
would provide a “common conceptual language” to be in conversation.251
The test set out by Article 19 of the ICCPR could serve as a reasoning
process to guide the conversation. However, the test itself, stripped from

244. SR Report 2018, supra note 9. R
245. Jasanoff, supra note 53, at 315. R
246. Ronald Dworkin, Foreword, in Extreme Speech and Democracy (Ivan Hare & James Wein-
stein, eds. 2009).
247. Owen Fiss, Free Speech and Social Structure, 71 Iowa L. Rev. 1405, 1411 (1986).
248. Id. at 1421 (“To assess the validity of the state intervention the reviewing court must ask,
directly and unequivocally, whether the intervention in fact enriches rather than impoverishes public
debate.”).
249. Jürgen Habermas, Moral Consciousness and Communicative Action 43 (Christian Len-
hardt & Shierry Weber Nicholsen trans., 1990) (1983).
250. Land, supra note 40, at 288; Douek, supra note 30, at 48. R
251. Sander, supra note 30, at 967. R
other requirements, is so thin that it is unclear how it would play a role in

channeling debates.
IHRL may enable a wider conversation about content governance because
it requires that corporations create new participatory mechanisms. IHRL
advocates derive this conclusion from their new interpretation of the
UNGPs, in particular the principles that refer to human rights due
diligence.
Principle 17 of the UNGPs prescribes that “[i]n order to identify, pre-
vent, mitigate and account for how they address their adverse human rights
impacts, business enterprises should carry out human rights due dili-
gence.”252 Human rights due diligence should “[i]nvolve meaningful con-
sultation with potentially affected groups and other relevant stakeholders, as
appropriate to the size of the business enterprise and the nature and context
of the operation.”253
The Board interprets these principles to mean that Facebook is expected
to consult with civil society and other stakeholders when developing new
content moderation rules or practices.254 Initiatives to achieve legitimacy
through these procedures exist everywhere. All major companies now have
different forms of engaging external stakeholders in their internal rule-mak-
ing process.255
The Board has shown an interest in mechanisms for civil society organiza-
tions to participate. The Board provides opportunities for the public to sub-
mit comments in all cases in a similar fashion to amici curiae.256 In some
instances, it has established rudimentary forms of dialogue with civil society
organizations, directly citing reports or addressing concerns expressed by
253. Id.
254. See, e.g., Case decision 2021-011-FB-UA, supra note 210, at 10. R
255. See, e.g., Vanessa Pappas, Introducing TikTok Content Advisory Council, TikTok (Mar. 18, 2020),
https://newsroom.tiktok.com/en-us/introducing-the-tiktok-content-advisory-council [https://perma.cc/
DL34-C3M7]; Arjun Narayan Bettadopur, Introducing the TikTok Asia Pacific Safety Advisory Council,
TikTok (Sept. 22, 2020), https://newsroom.tiktok.com/en-sg/tiktok-apac-safety-advisory-council [https:/
/perma.cc/5DMU-YGNM]; Julie de Bailliencourt, Meet TikTok’s European Safety Advisory Council, TikTok
(Mar. 1, 2021), https://newsroom.tiktok.com/en-gb/tiktok-european-safety-advisory-council [https://
perma.cc/83VD-NNWU]; Introducing the Twitch Safety Advisory Council, Twitch (May 14, 2020), https://
blog.twitch.tv/en/2020/05/14/introducing-the-twitch-safety-advisory-council/ [https://perma.cc/YS3G-
YCAK]; Nick Pickles, Strengthening our Trust and Safety Council, Twitter (Dec. 13, 2019), https://
blog.twitter.com/en_us/topics/company/2019/strengthening-our-trust-and-safety-council.html [https://
perma.cc/ZB4M-T759]; Patricia Carter, Announcing the Twitter Trust & Safety Council, Twitter (Feb. 9,
2016), https://blog.twitter.com/en_us/a/2016/announcing-the-twitter-trust-safety-council.html [https://
perma.cc/UW3E-WSSS]; Stakeholder Engagement, Meta, https://www.facebook.com/communitys-
tandards/stakeholder_engagement [https://perma.cc/KL63-KP64].
256. See, e.g., Case decision 2020-006-FB-FBR, supra note 229, at 15 (acknowledging that the board’s R
recommendations to Facebook drew on public comments the board received).
civil society.257 The Board has valued Facebook’s efforts to engage civil
society.258
As discussed above, the UNGPs refer to companies’ responsibility to re-
spect human rights that international law recognizes.259 Accordingly, when
the UNGPs discuss human rights due diligence, they envision that compa-
nies will make an effort to audit the impact of their operations on human
rights.260 For example, under this framework, Facebook is expected to un-
derstand how its platforms are used in different countries. If it is the case, as
it has been in Myanmar, that the platform is a key tool in organizing crimes
against humanity, Facebook should be aware of this and take active steps to
prevent or remedy its involvement in human rights abuses.261
The Board’s interpretation goes well beyond that responsibility. When
Facebook engages civil society around the world to gather feedback on how
to amend its community guidelines, it proposes different potential rules and
hears advantages or concerns about each one of them.262 The kind of stake-
holder engagement that companies carry out and that the Board promotes
does not have the sole or even central purpose of identifying potential
human rights infringements. Rather, companies engage civil society to in-
corporate their preferences among multiple human-rights-respecting policy
options.263 The assumption is that all discussed alternatives respect human
rights.
Even if divergent from the text of the UNGPs, the Board’s interpretation
can be very useful. The main stated goal of the IHRL project is to align
private speech regulation with the interests of the people. This reading of
the human rights due diligence responsibility may strengthen companies’
consultation with the public during the policy development process. Along
those lines, in Case Decision 2020-006-FB-FBR concerning a post contain-
ing inaccurate information about COVID-19 treatments, the Board recom-
257. See, e.g., Case decision 2021-011-FB-UA, supra note 210, at 11 (citing a report elaborated by the R
NGO Media Monitoring Africa).
258. See, e.g., Case decision 2021-002-FB-UA, supra note 1, at 11. R
260. See Ruggie, supra note 84, at 99. R
261. See id. at 100 (describing the responsibilities of multinational corporations operating in countries
where they are at risk of becoming complicit in egregious human rights abuses).
262. See generally Matthias Kettemann & Wolfgang Schulz, Setting Rules for 2.7 Billion. A (First) Look
into Facebook’s Norm-Making System: Results of a Pilot Study (Hans-Bredow-Institut Working Paper, 2020)
(describing the steps in Facebook’s stakeholder engagement process).
263. See, e.g., Standing Against Hate, Meta (Mar. 27, 2019), https://about.fb.com/news/2019/03/
standing-against-hate/ [https://perma.cc/BRV3-D9A7]; Monika Bickert, Enforcing Against Manipulated
Media, Meta (Jan. 6, 2020), https://about.fb.com/news/2020/01/enforcing-against-manipulated-media/
[https://perma.cc/C2YX-PB7U]; Vanessa Pappas, Combating Misinformation and Election Interference on
TikTok, TikTok (Aug. 5, 2020), https://newsroom.tiktok.com/en-us/combating-misinformation-and-
election-interference-on-tiktok [ https://perma.cc/6XJH-VDER]; Twitter Safety, Updating Our Rules
Against Hateful Conduct, Twitter (July 9, 2019), https://blog.twitter.com/en_us/topics/company/2019/
hatefulconductupdate.html [https://perma.cc/PE5X-WD4L]; Vijaya Gadde & Del Harvey, Creating New
Policies Together, Twitter (Sept. 25, 2018), https://blog.twitter.com/official/en_us/topics/company/2018/
Creating-new-policies-together.html [https://perma.cc/7EAB-DX6B].
mended that “Facebook should conduct a human rights impact assessment

with relevant stakeholders as part of its process of rule modification” per
Principles 18 and 19 of the UNGPs.264 In Case Decision 2021-006-IG-UA
concerning a post discussing the solitary confinement of Abdullah Öcalan,
the Board recommended that Facebook “ensure meaningful stakeholder en-
gagement” to review its policy on dangerous individuals and organizations,
including through a public call for input.265 Notice that in this case, the
Board did not refer to any specific IHRL obligation. Perhaps the Board can
make this recommendation in a similarly convincing fashion regardless of
whether the Board refers to the UNGPs.
In the future, the Board’s interest in stakeholder engagement may diverge
further from the UNGPs’ meaning and evolve into more precise standards
for stakeholder consultation. For example, in Case Decision 2021-011-FB-
UA, regarding a racial slur in South Africa, the Board appreciated
Facebook’s consultation with external stakeholders to draft an exception to
its hate speech policy for insults when used self-referentially. The Board also
pointed out that the external stakeholders represented diverse geographical
regions.266
In some cases, the Board hinted timidly that consultation with civil soci-
ety organizations could even lead the Board to accept rules that diverge from
interpretations made by the Human Rights Committee or U.N. Special
Rapporteurs. In Case Decision 2020-007-FB-FBR, regarding an alleged
veiled threat, the minority was willing to defer to Facebook’s decision be-
cause Facebook had consulted with regional and linguistic experts and had
worked with a local partner to identify and adjudicate the content. In Case
Decision 2021-002-FB-UA, regarding Facebook’s ban on content depicting
blackface, the Board took into account that the rule was the outcome of a
process that “involved extensive research and engagement with more than
60 stakeholders, including experts in a variety of fields, civil society groups,
and groups affected by discrimination and harmful stereotypes.”267 The ma-
jority considered this stakeholder consultation to be in line with “interna-
tional standards for on-going human rights due diligence” and cited
Principles 17(c) and 18(b) of the UNGPs.268
Despite the Board’s interest in stakeholder engagement initiatives, the
view from nowhere has, so far, prevailed in the Board’s deployment of the
IHRL project. In the case regarding Facebook’s general prohibition on con-
tent depicting people in blackface, the Board appreciated the process of en-
gaging stakeholders.269 However, when justifying its decision to uphold the
264. Case decision 2020-006-FB-FBR, supra note 229, at 13. R

265. Case decision 2021-006-IG-UA, Oversight Bd. (July 8, 2021), https://oversightboard.com/deci-
sion/IG-I9DP23IB/ [https://perma.cc/J2SD-6GZP].
268. Id.
269. Id. at 11.
ban, the Board did not reference this process. Instead, the Board decided
that the divergence was justified because many experts had found that
images depicting blackface are discriminatory and harmful. Among other
reasons discussed above, the Board relied on the fact that “[n]umerous
human rights mechanisms have found the portrayal of Zwarte Piet to be a
harmful stereotype.”270 But not any harmful stereotype would satisfy the
Board. The Board cited many reports and concluded that these expert find-
ings provided “sufficient evidence of objective harm to individual’s rights to
distinguish this rule from one that seeks to insulate people from subjective
offense.”271 This passage defines a boundary between objective knowledge
and subjective feelings. Here, objectivity comes not from the stakeholder
engagement process but from the human-rights experts’ findings.
Another example in the same direction comes from Case Decision 2021-
011-FB-UA concerning the use of a racial slur in South Africa.272 The Board
analyzed Facebook’s decision to delete a post that used a racial slur in the
context of discussing wealth and racial dynamics in South Africa. The Board
had already decided other cases dealing with bans of specific terms, always
finding that even though prohibiting the use of particular words would
breach IHRL if adopted by a state, Facebook’s ban was compatible with
IHRL.273
This case offers an additional glimpse of how the Board divides tasks
among U.N. experts and civil society organizations. On the one hand, the
Board stated that banning racial slurs is incompatible with IHRL. Still,
Facebook can adopt that ban because the “[U.N.] Special Rapporteur indi-
cates that entities engaged in content moderation like Facebook can regulate
such speech.”274 On the other hand, the Board stated that Facebook should
consult affected groups and human rights experts, as it did in this case, to
review its policies and update the list of banned slurs.275 The consultation in
this case contributed to Facebook’s understanding of the meaning of the
term in the context where it is used.276 In this sense, engagement with local
civil society is used mainly as an implementation mechanism to ensure the
correct understanding of racial slurs, while the rule-making process is kept
in the hands of a small group of experts.
270. Id. at 14.

271. Id. at 16 (emphasis added).
273. Case decision 2020-003-FB-UA, supra note 210; Case decision 2020-007-FB-FBR, Oversight Bd. R
(Feb. 12, 2021), https://oversightboard.com/decision/FB-R9K87402/ [https://perma.cc/9VT9-XDYP].
275. Id.
276. Id.
V. The Futures of the IHRL Project
Overall, what is problematic, if anything, with expert governance? This

section asks if the IHRL project might indeed help us bridge the democratic
deficits that David Kaye identifies. Because the IHRL project is still incipi-
ent, the answer might depend on how the IHRL project evolves. Therefore,
this section answers that question by imagining what I believe are the most
auspicious futures for the IHRL project. First, it might be that human
rights experts persuade many people that their decisions are good. Second, it
might be that the participatory aspects of the project open up space for more
actors and flatten the power relationships among governance actors. Finally,
I try to imagine what disentangling content moderation from IHRL could
look like. I suggest that the execution of the IHRL project has so far under-
mined its own stated goals, and that this disentanglement might offer a
more promising path.
A. A Communal Viewpoint Developed from the Top

IHRL skeptics, myself included, have raised the concern that the main
effect of the IHRL project may be to legitimize corporate decisions without
imposing severe constraints on their power.277 From the perspective of the
Board or the advocates of the IHRL project, having human rights experts
with perceived legitimacy to govern speech might be a good result. If using
the IHRL framework brings about legitimacy either to the Board or to com-
panies framing their policies in human rights language, this outcome could
be considered a victory.
Even from the perspective of the public at large, it is not obvious why
this outcome of increased legitimacy would be problematic. If the public
perceives experts as legitimate governors, it might be because the public
finds the experts’ actions and decisions convincing. Invoking IHRL will
likely help the experts make their decisions more compelling. It is possible
that, over time, experts persuade a large number of people that the solutions
that they offer are reasonable. In that case, the Board’s legitimacy might
come from the public’s approval and agreement with the Board’s decisions.
In the long term, it could be that the Board’s decisions reflect a communal
viewpoint developed from the top.278
Some of the risks of making IHRL the relevant and legitimate language
of online speech governance are analogous to the risks that the human rights
movement faces more generally. If IHRL is perceived as the yardstick for
online speech governance, there might be less intellectual energy to concep-
277. See, e.g., Douek, supra note 30, at 63; Dvoskin, supra note 145. R
278. See Sheila Jasanoff, Designs in Nature: Science and Democracy in Europe and the
United States 152-55 (2005) (explaining and illustrating the idea of a communal viewpoint developed
from the top).
tualize other normative possibilities.279 Relatedly, Barrie Sander highlights

that the inevitable risk of the IHRL project is that it may legitimize minor
improvements “at the expense of undertaking more structural” changes.280
Looking at transparency and due process through the lens of human rights
carries the risk of focusing too strongly on formal procedures while ignoring
the power differential among users and civil society to make use of these
mechanisms.281
More important is IHRL’s ambition to reflect global preferences. Ulti-
mately, the faith or lack thereof in the IHRL project might reside in
whether one believes that speech regulation should aim to reflect a broad
global consensus—a communal viewpoint—or that speech regulation is
mainly a permanent site of disagreement. Daniel Walters, reflecting on the
relationship between administrative agencies and democracy, criticizes agen-
cies for “trying to convince participants that the agency’s proposal is good
for everyone.”282 Instead, he argues, agencies ought to “forthrightly ac-
knowledge that the proposal may not be good for everyone.”283 Likewise,
performing as executors of some objective notion of the will of the global
society denies the distributive effects of experts’ decisions and the fact that
some viewpoints will be necessarily disadvantaged. In turn, that claim to
universality conceals the need for stronger mechanisms to share power
among diverse constituencies.
B. IHRL as a Participatory Project

The promise of a broader conversation with the power to shape content
moderation facilitated by IHRL has yet to be realized. I am skeptical that a
shared language, common framework, or proportionality test can meaning-
fully impact the power distribution among the actors that can or do partici-
pate in online speech governance. However, existing institutions may
leverage the IHRL project to create or strengthen institutional initiatives to
increase the number of opportunities for participation. Additionally, it
could be that engaging in IHRL discourse helps these institutions build
their credibility, which they might in turn use to pursue projects unrelated
to IHRL.
The principal executor of the IHRL project has been the Board. So far, it
has found experts’ reports, especially from U.N. bodies, to be the most
promising source to build its own legitimacy. As the Board matures, it
279. See Kennedy, supra note 156, at 108; Samuel Moyn, Humane: How the United States R
Abandoned Peace and Reinvented War (2021) (arguing that international humanitarian law di-
verts from pursuing more radical projects).
281. See Ganesh Sitaraman, The Puzzling Absence of Economic Power in Constitutional Theory, 101 Cor-
nell L. Rev. 1445, 1500 (2016); Nico Krisch, The Pluralism of Global Administrative Law, 17 Eur. J.
Int’l. L. 247, 276 (2007).
282. Walters, supra note 57, at 56. R
283. Id. (emphasis in the original).
might be willing to lend some power to stakeholder engagement initiatives.

The Board could show more deference to Facebook’s rules when these rules
are the product of consultations with civil society. The Board should develop
standards for such consultations in terms of how open they need to be,
whose participation should be ensured, how transparent the process needs to
be, and how the company needs to consider the received input.284 The Board
could ask for a report of the disagreements identified in the stakeholder
engagement process, and which stakeholders were favored by the outcome of
the policy-drafting process. The Board may ground these standards in its
innovative interpretation of the UNGPs on human rights due diligence dis-
cussed above285 or use the legitimacy that human rights discourse brings to
undertake these participatory projects.
The Inter-American Commission on Human Rights has recently launched
a multi-stakeholder dialogue on making content moderation policies “com-
patible” with IHRL standards.286 This initiative contemplates multiple op-
portunities to seek input from experts, the general public, and other
regional and international endeavors. It could be a significant effort to set
authoritative standards for content moderation developed outside of compa-
nies, borrowing legitimacy from participatory procedures and using the
IHRL framework to strengthen the authority and persuasiveness of the sub-
stantive outcomes.
This broadening in institutional opportunities is a positive development.
On the one hand, the Commission’s initiative will not necessarily address
the background conditions that explain why civil society across geographical
areas and ideological viewpoints has different capacities, resources, and lev-
erage to influence online speech governance.287 As a result, it might create
more formal participatory mechanisms but it might also be insufficient to
ensure that those mechanisms are used equitably.288 On the other hand, the
initiative can turn up the volume of Latin American voices, which have so
far been neglected in global debates.
Ultimately, these participatory initiatives will have to be evaluated based
on their own merits, not on how closely they follow IHRL guidance. Their
main weakness is that their processes are designed to find convergence
around online speech norms. The risk is confining participation within nar-
row bounds as a ritualized phase of policy development or a one-time
dialogue.
284. See Dvoskin, supra note 226. R

285. See supra Section IV.B.
286. Americas Dialogue on Freedom of Expression Online, Americas Dialogue, https://
www.americasdialogue.org/en/americas-dialogue/ [https://perma.cc/F68V-E88R].
287. See Dvoskin, supra note 226, at 10. R
288. See K. Sabeel Rahman, Policymaking as Power-Building, 27 S. Cal. Interdisc. L.J. 315, 353
(2018) (arguing that participatory mechanisms need actors willing and able to make full use of them).
C. Disentangling Content Moderation from IHRL
The IHRL project provides a framework for experts to present their deci-
sions as objective to the public and, in turn, gain the public’s acceptance.
This is a tempting offer. But resisting the comfort that legal language pro-
vides and coming to terms with one’s power can lead to decisions that are
more transparent, less likely to be framed as universal answers, and easier to
contest.
If newly empowered experts are too concerned with proving their objec-
tivity, they may miss a valuable opportunity for plural and experimental
governance. Indeed, the most interesting recommendations the Board has
made to date were not grounded in IHRL.289 Moving away from the IHRL
project may prompt experts to provide better justifications for their deci-
sions, involve more actors, and make space for more meaningful inclusion of
normative disagreements. Imagining how the Board could have reached its
decision in the case about Black Pete illustrates these points.
First, IHRL is highly praised for offering a framework to justify decisions
to the public.290 Likewise, one of the core purposes of the Board is to deliver
transparent and well-justified decisions.291 But in the Black Pete case, the
Board’s adoption of the IHRL project actually harmed the coherence and
clarity of its decision.
Throughout the Article, I have discussed how the Board dealt with tech-
nical facts,292 U.N. experts,293 and stakeholder engagement in the Black
Pete case.294 An additional point to consider is how the Board assigns corpo-
rate actors a fundamental role in the protection of human rights that, in the
Board’s view, is a role barred to states. In the Black Pete case, the Board
concluded that “international human rights law would not allow a state to
impose a general prohibition on blackface”295 and at the same time asserted
that “Facebook followed international guidance and met its human rights
responsibilities in this case.”296
289. See, e.g., Case decision 2021-009-FB-UA, Oversight Bd. (Sept. 14, 2021), https://
www.oversightboard.com/decision/FB-P93JPX02/ [https://perma.cc/NR3T-WM65] (recommending
Facebook “. . . conduct a thorough examination to determine whether Facebook’s content moderation in
Arabic and Hebrew, including its use of automation, have been applied without bias.”); Case decision
2020-004-IG-UA, Oversight Bd. (June 28, 2021), https://oversightboard.com/decision/IG-7THR3SI1/
[https://perma.cc/PNY8-B7HX] (recommending Facebook “[i]mplement an internal audit procedure to
continuously analyze a statistically representative sample of automated content removal decisions to re-
verse and learn from enforcement mistakes.”); Case decision 2021-006-IG-UA, supra note 265 (recom- R
mending Facebook open a public call for inputs on how to update its Dangerous Individuals and
Organizations policy).
291. Feldman, supra note 8, at 102. R
292. See supra Section III.C.
293. See supra Section IV.
294. Id.
296. Id.
The reasons why the Board thought that this was an appropriate rule for
Facebook to adopt were, again, that blackface produces objective harm, that
the real intent of the speaker on social media is hard to determine, that this
kind of speech can create a discriminatory, harassing, and degrading envi-
ronment, and that Facebook should promote equality.297 It is hard to under-
stand why Facebook would be expected to prevent the harms caused by this
kind of content under its human rights responsibility but a state would not
be allowed to take analogous steps under the same legal framework.
All of the circumstances that justified Facebook’s decision also apply to
state regulation. Instances of discriminatory speech, even if not intended to
cause harm, are pervasive in many contexts and create discriminatory envi-
ronments.298 If experts’ reports find that people in blackface create “objec-
tive” harm, then the harm is obviously independent of whether Facebook or
governments are the ones to address it. States have a greater responsibility
than Facebook to promote equality. Ultimately, the need to justify a rule
that diverges from IHRL as an adequate application of that framework
makes the decision more confusing, not more transparent.
Second, being transparent about what is driving the Board’s decisions
would enable more effective participation by the public. In the comments
received by the Board in this case, Professor Sejal Parmar from the Univer-
sity of Sheffield was the only commenter who provided an analysis of the
decision under IHRL, and particularly Articles 19 and 20 of the ICCPR.299
Her comment cites some of the same reports the Board referenced, although
she highlights that the reports do not call for a total ban on portraits of
Black Pete. Ultimately, the Board decided to diverge from IHRL. In this
case, telling the public that IHRL will determine the decision does not en-
hance the public conversation, as the project promises, but rather under-
mines it. Indeed, it becomes less clear to the public what reasons the Board
will consider to be weighty when making a decision.
The Board instead could have explained why Facebook’s rule aligned with
the Board’s normative approach. For example, the most important factor
driving the decision could have been the support that the prohibition had
among consulted stakeholders. If the Board acknowledges that this factor is
driving its decision, then the Board can openly set requirements that a con-
sultation process needs to meet in order to persuade the Board to give defer-
ence to the consultation’s findings.
Third, if the Board is successful in promoting IHRL as the correct moral
yardstick for content moderation, other companies might believe they need
297. Id.
298. See Chester Pierce, Psychiatric Problems of the Black Minority, in American Handbook of Psy-
chiatry 512 (Silvano Arieti ed., 1974) (coining the term “microaggressions” and studying their impact
on mental health).
299. Public Comment Appendix for Case decision 2021-002-FB-UA, Oversight Bd. (Apr. 13, 2021),
https://www.oversightboard.com/sr/decision/2021/002/public-comments [https://perma.cc/3GB5-
BD37].
to converge with the Board in their decisions. When the Board says that
Facebook is meeting its human rights responsibilities by adopting a general
ban on blackface, the Board makes it harder for other companies to adopt a
different rule. More fundamentally, the Board’s proposal for how to assign
roles in online speech governance greatly increases the power of corporate
actors (including the Board’s own power). In the Board’s interpretation, a
great objective harm exists to social equality that only companies (and the
Board) should remedy. Framing its decisions as the product of its members’
views would help the Board avoid this kind of general and universalizing
stance of what online speech governance should look like.
Finally, abandoning the ideal of objectivity would help the Board be more
explicit about its political commitments. As Walters says about administra-
tive agencies, there is a lot to gain from the Board admitting that its deci-
sions reflect some people’s preferences, not everyone’s.300 Instead of putting
an end to the debate, making normative preferences explicit calls for in-
stances for those with different preferences to contest the decision in the
future. In expert governance, the quest for objectivity is an obstacle for
meaningful participation. Ultimately, putting politics front and center may
more effectively facilitate the public conversation, transparency, and public
justifications that the IHRL project envisions.
Conclusion
The IHRL project for content moderation promises to rein in corporate

power on behalf of the public interest. It rests on the idea that IHRL can
function as an objective synthesis of the global public interest. The main
goal of this Article has been to examine the tools that experts have devel-
oped to build that claim to objectivity.
300. Walters, supra note 57, at 56. R


HLI102 Crop-3

Uploaded by

Copyright:

Available Formats

HLI102 Crop-3

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HLI102 Crop-3

Uploaded by

Copyright:

Available Formats

\\jciprod01\productn\H\HLI\64-1\HLI102.

txt unknown Seq: 1 16-FEB-23 15:34

Volume 64, Number 1, Winter 2023

Expert Governance of Online Speech

86 Harvard International Law Journal / Vol. 64

5. Because IHRL is a shared language and because IHRL

In April 2021, the Facebook Oversight Board (hereinafter “Board”) con-

1. Case decision 2021-002-FB-UA, Oversight Bd. (Apr. 13, 2021), https://oversightboard.com/deci-

2023 / Expert Governance of Online Speech 87

5. Case decision 2021-002-FB-UA, supra note 1. R

88 Harvard International Law Journal / Vol. 64

2023 / Expert Governance of Online Speech 89

18. Id., ¶ 42.

90 Harvard International Law Journal / Vol. 64

scholars. Just as judicial courts have developed theories of interpretation that

2023 / Expert Governance of Online Speech 91

puzzling outcomes.26 This Article examines what gets lost in translation

I. The International Human Rights Law Project for Online

A large consensus among companies, scholars, and advocates has emerged

92 Harvard International Law Journal / Vol. 64

should follow a human rights approach.30 Multiple civil society organiza-

Rights, Meta (Mar. 16, 2021), https://about.fb.com/news/2021/03/our-commitment-to-human-rights/

2023 / Expert Governance of Online Speech 93

interests, preferences or agreements of society as opposed to the commercial

94 Harvard International Law Journal / Vol. 64

starting point for designing content moderation.41 Similarly, taking inspira-

2023 / Expert Governance of Online Speech 95

which governmental restrictions meet these conditions. Accordingly, the ex-

96 Harvard International Law Journal / Vol. 64

standpoint, Sheila Jasanoff describes them as representing a view from no-

2023 / Expert Governance of Online Speech 97

stakeholder engagement.62 However, as I argue in detail in Section IV, these

II. Justifying the International Human Rights Law Project

IHRL is often invoked as the proper benchmark to evaluate content mod-

62. See infra Section IV.B.

98 Harvard International Law Journal / Vol. 64

1. A New Interpretation of the U.N. Guiding Principles

2023 / Expert Governance of Online Speech 99

100 Harvard International Law Journal / Vol. 64

mation they deliver.”75 Accordingly, because excluding lawful content is an

75. Land, supra note 40, at 292. R

2023 / Expert Governance of Online Speech 101

102 Harvard International Law Journal / Vol. 64

panies’ responsibilities. By 1990, Nike had outsourced its entire production.

2023 / Expert Governance of Online Speech 103

104 Harvard International Law Journal / Vol. 64

2023 / Expert Governance of Online Speech 105

106 Harvard International Law Journal / Vol. 64

Describing platforms as states can have profound normative implications.

125. See Sander, supra note 44. R

2023 / Expert Governance of Online Speech 107

108 Harvard International Law Journal / Vol. 64

135. See Anthea Roberts, Is International Law International? (2017).

2023 / Expert Governance of Online Speech 109

4. Because IHRL constrains corporate power, but its indeterminacy is a

110 Harvard International Law Journal / Vol. 64

from corporations.146 David Kaye’s report on hate speech as Special Rap-

146. See Side-stepping rights, supra note 102. R