Perspective Map
Platform Moderation and Free Expression: What Each Position Is Protecting
In January 2021, four things happened in quick succession that revealed a philosophical fault line running beneath every argument about speech and moderation. Twitter permanently suspended Donald Trump's account following the Capitol attack, citing incitement to violence. Within days, critics in Africa and South Asia pointed out that BJP politicians in India had for years posted content calling Muslims "termites" and "viruses" without consequence, and that Facebook's hate speech systems routinely flagged Arabic-language political protest while leaving Hindi anti-Muslim rhetoric untouched. Meanwhile, civil liberties organizations that had spent decades defending the speech of Nazis and Klan members were split internally: staff of color drafted a memo arguing that the organization's absolutist stance on content moderation was, in effect, a choice about whose dignity mattered. And in Ethiopia, the government suspended internet access as armed conflict intensified in Tigray — a state power restricting speech that the free expression argument was never designed to address.
The four camps that emerged from those weeks were not arguing about the same thing. The disagreement was not a debate about platform governance design — about whether Section 230 should be reformed or whether the EU's DSA framework is better than US self-regulation. It was a disagreement about first principles: about what "free expression" actually protects, about what harm speech can do and to whom, about whether the liberal tradition's tools for thinking about speech are adequate to a world the tradition was never built to address. Those four camps have names, and each is protecting something real.
What free expression absolutists are protecting
The strongest version of the free expression absolutist position draws on John Stuart Mill's argument in On Liberty (1859) that the only legitimate basis for restricting speech is demonstrable harm to others — not offense, not discomfort, not the social costs of tolerating a position someone finds wrong. Mill's argument rested on an epistemological observation: because no one can be certain they possess the truth, suppressing an opinion always risks suppressing something true. Even a false opinion that is never answered makes the truth "a dead dogma rather than a living truth" — it prevents us from knowing why we believe what we believe. The absolutist case for protecting expression, on this view, is not that speech never causes harm. It is that the machinery required to suppress speech always causes worse harm: a government empowered to define what is harmful is a government empowered to suppress whatever threatens it.
They are protecting the epistemic argument that no institution can be trusted to define harm accurately. FIRE (Foundation for Individual Rights and Expression), in its defense of campus speech codes and platform bans alike, argues that the category of "harmful speech" is inherently elastic — it expands to cover whatever the institution finds politically inconvenient. The historical record of this elasticity is not subtle: sedition laws used to suppress labor organizing, obscenity standards used to suppress LGBTQ identity, "public order" provisions used to imprison civil rights activists. The contemporary expansion of "harmful speech" to include "disinformation," "hate speech," and "extremism" simply adds new categories to a lineage of expansive definitions that consistently served established power more than marginalized dissent. Absolutists argue that the right response to bad speech is more speech — counter-speech, context, education — not the concentration of definitional authority in institutions with every incentive to protect themselves.
They are protecting the chilling effect argument. The risk of content moderation is not primarily that it will remove speech that should remain — that can be corrected — but that it will prevent speech from being produced at all. A speaker who knows that their words might be flagged, demonetized, or removed will self-censor into a narrower range of expression than they would otherwise choose. This chilling effect falls hardest on the most vulnerable speakers: journalists reporting on state violence, human rights advocates documenting atrocities, satirists targeting powerful institutions. The overfiring problem in content moderation — where legitimate speech is removed at scale because automated systems cannot distinguish it from the harmful speech it resembles — is, on this view, not a technical problem to be solved by better ML classifiers. It is the predictable consequence of building suppression systems at all.
What harm-based governance advocates are protecting
In 2012, the philosopher Jeremy Waldron published The Harm in Hate Speech (Harvard University Press), which challenged the premise that speech is harmless unless it directly incites violence. Waldron's argument was about dignity: not the subjective feeling of being offended, but the objective social standing — the basic assurance of inclusion — that members of targeted groups carry into every public interaction. When a Black person drives past a billboard displaying a white supremacist slogan, they have not been "offended." Their standing as an equal member of society — their assurance that the community recognizes them as someone who belongs here — has been publicly attacked. This is a real harm, Waldron argued, and societies have both the right and the obligation to protect against it.
They are protecting the empirical reality that speech causes injury — not metaphorically but in measurable, documented ways. Mari Matsuda and Charles Lawrence, in their foundational critical race theory contributions to Words That Wound (Westview Press, 1993), drew on the documented psychological and social harms of racist speech: elevated cortisol responses, reduced cognitive performance in academic settings, hypervigilance, PTSD-like symptoms in communities subjected to sustained harassment campaigns. These are not abstract harms. They affect the capacity of targeted individuals to participate in the very democratic discourse that Mill's marketplace of ideas requires. An account of free expression that ignores this asymmetry — that treats the speaker's liberty as the only value at stake — systematically produces environments in which some people are freer than others to speak.
They are protecting the distinction between individual offense and collective dignity. Harm-based governance advocates do not argue that speech should be restricted whenever someone is offended. Waldron is explicit: offense — the subjective experience of discomfort — is not a sufficient ground for restriction. The target is speech that degrades the basic standing of a group, that functions as "group libel" in the way that individual defamation functions: by attaching false, damaging claims to an identity that the target cannot shed. You can choose to leave a bad neighborhood; you cannot choose to leave your race. The harm-based position is not paternalism about feelings. It is a substantive claim about what democratic equality requires: that people be able to participate in public life without their right to be there being publicly contested.
What communitarian speech norm theorists are protecting
Stanley Fish's There's No Such Thing as Free Speech, and It's a Good Thing, Too (Oxford University Press, 1994) made an argument that sounds paradoxical and turns out to be clarifying: all speech is already regulated by community norms, always and inescapably. The claim of the free expression absolutist — that there exists a neutral position from which speech can be evaluated without reference to contested values — is false. The decision to protect a particular kind of speech over another kind is itself a substantive value judgment, not a procedural one. When American courts protect Nazi marches on the grounds that content-based restrictions are impermissible, they are not applying a neutral principle. They are applying a principle drawn from a specific community's political tradition, one that weights speaker autonomy above listener dignity and procedural neutrality above substantive equality.
They are protecting the recognition that the liberal free speech framework does not describe a neutral architecture — it describes one community's substantive norms, dressed as universal rules. The asymmetry Fish identifies is not merely philosophical. In practice, the "neutral" framework consistently produces outcomes that track power: the speech that receives the most protection (political advocacy, religious expression, academic inquiry) is the speech most valued by dominant groups; the speech most vulnerable to restriction (street protests that block traffic, "threatening" language in labor organizing, political speech coded as uncivil by dominant standards) is disproportionately the speech of marginalized communities. Communitarian theorists argue that this pattern is not a bug or an inconsistency. It is what community-norm governance looks like when one community's norms are declared universal and others are declared particular — or absent.
They are protecting the possibility of community-governed speech that is transparent about its values rather than hiding them behind procedural claims. The alternative to the liberal framework is not censorship — it is governance that names its values explicitly. A platform that says "we amplify speech that strengthens democratic deliberation and restrict speech that degrades the standing of historically targeted groups" is governing speech by community norms, as all speech governance does. What makes it different from the liberal framework is not that it is more value-laden — it is that it is honest about being value-laden. Communitarian theorists argue this honesty is preferable to a false neutrality that smuggles in substantive value judgments while denying doing so.
What anti-colonial critics are protecting
In 2022, scholars publishing in International Political Sociology introduced the concept of the "settler coloniality of free speech" — the argument that the dominant free speech framework is not merely culturally contingent but historically colonial: that it was developed during the era of Western imperial expansion, was applied selectively to exclude colonized peoples from its protections, and now circulates globally through US-headquartered digital platforms as though it were a universal standard, carrying its original exclusions with it.
They are protecting the historical record of what the free speech tradition actually protected and what it left unprotected. Nanjala Nyabola's Digital Democracy, Analogue Politics (Zed Books, 2018) documents how platforms calibrated for US political norms systematically fail communities in Africa and the Global South: content moderation policies trained on English-language toxic content miss toxicity in Swahili, Amharic, and Tagalog; platforms that protect political speech against government removal in the US comply with government removal requests in authoritarian states because the business logic requires it; the global content governance that flows from Silicon Valley reflects the political culture of one country in one moment. The free speech framework that platforms claim to defend was built for a particular polity and operates on the assumption that the threat to speech is governments and the protection is private companies — an assumption that inverts the political reality for much of the world, where platforms cooperate with authoritarian surveillance and governments are not the primary threat to citizen speech but the primary tool of its suppression via platform compliance.
They are protecting the recognition that Mill's epistemic argument was never applied symmetrically. Mill himself worked for the East India Company and argued — in texts that co-existed with On Liberty — that "despotism is a legitimate mode of government in dealing with barbarians." The philosophical tradition that generated the free expression argument also generated justifications for colonial suppression of the speech, culture, and self-governance of colonized peoples. Anti-colonial critics do not argue that the free speech tradition is worthless. They argue that a tradition with this genealogy cannot be applied globally as though it has no genealogy — cannot be used to structure speech governance for two billion users across dozens of political and cultural contexts as though it were universal law rather than one culture's political inheritance. The question is not whether free expression matters. It is whether the specific institutional form through which it is currently defended — First Amendment absolutism as default global speech governance — can be made adequate to the political realities it was never designed to address.
They are protecting the claim that what counts as "harmful speech" is itself produced by unequal power — and that frameworks that ignore this will systematically protect dominant speech while suppressing marginalized speech. Joy Buolamwini's work at the MIT Media Lab on algorithmic bias, and Safiya Umoja Noble's Algorithms of Oppression (NYU Press, 2018), document how the systems platforms use to identify harmful content — trained on data produced by communities that are not representative of their global user base — consistently misclassify Black, Indigenous, and Global South speech as violating community standards while leaving equivalent speech by dominant groups untouched. This is not a technical error to be corrected by diversity in training data. It reflects a deeper problem: harm is a contested concept, and the institutions that define it for global platforms are not globally representative.
Where the real disagreement lives
The four positions are not arguing about the same question. Identifying this is the beginning of genuine disagreement rather than the end of it.
The epistemology question. The absolutist argument and the harm-based argument are partly a dispute about facts: whether speech causes the kinds of direct, measurable harm that justify restriction. Waldron and Matsuda say yes; Mill's tradition says no, or not enough. But they are also arguing about what counts as evidence: the absolutist discounts psychological harm and collective dignity as non-harms or second-order harms that should yield to the primary value of speaker autonomy; the harm-based advocate holds that this discount is itself a substantive value judgment — choosing to weight one kind of harm over another. Neither is applying a neutral framework to disputed evidence. Both are making choices about which harms matter most.
The neutrality question. The deepest split is between those who believe the liberal free speech tradition can be applied universally (absolutists and, in a different register, harm-based advocates who work within liberal theory) and those who believe the tradition is itself particular (communitarian and anti-colonial critics). This is not a debate that can be resolved by appealing to the liberal tradition — because the dispute is about whether that tradition is the appropriate frame. Resolving it requires work that is explicitly cross-cultural and historically honest about the conditions under which current frameworks developed.
The power asymmetry question. Communitarian and anti-colonial critics share the insight that liberal free speech doctrine produces outcomes that track power — that the "neutral" framework is not neutral in its distribution of protection. Absolutists respond that the alternative — empowering institutions to make substantive distinctions about what speech is good and what is harmful — is even more likely to produce outcomes that track power, because the institutions given that authority will be the ones with the most power. Both observations are empirically plausible. The disagreement is about which risk is larger: the risk of harmful speech flowing freely, or the risk of suppression-machinery being captured by those with incentives to suppress dissent.
The scale question. None of these frameworks was developed for the conditions of global digital platforms. Mill wrote for a community of literate English subjects engaging in face-to-face and print-mediated discourse. CRT developed within US legal academia addressing US constitutional doctrine. The anti-colonial critique was developed before the current configuration of US-headquartered platforms governing global speech. The question of whether any of these frameworks is adequate to governing speech at the scale of two billion users across hundreds of linguistic and political communities simultaneously is, genuinely, open.
See also
- Who belongs here? — the framing essay for the membership and dignity conflict inside this map: the speech-values dispute is never only about abstract liberty, but about whether targeted people can enter shared digital space with their standing intact, whose vulnerability counts as politically real, and what kind of community a platform is implicitly choosing to build.
- Who gets to decide? — the framing essay for the authority dispute this map opens onto: once platforms function as de facto public squares, the speech-values conflict becomes inseparable from the question of who has legitimate power to set and enforce the rules for shared digital life.
- Platform Accountability and Content Moderation: What Each Position Is Protecting — the companion to this map: where this map examines the values collision about what speech should be governed, that map examines the institutional design question of who should govern it and how; the four positions in each map are distinct — you can be a harm-based governance advocate and prefer democratic accountability, or platform self-governance, as your institutional mechanism; separating the values question from the institutional question is necessary to see what each debate is actually about.
- Free Speech on Campus: What Each Side Is Protecting — a concentrated version of this same values collision in an institutional setting with more defined stakes; campus debates reveal the communitarian and harm-based tensions more sharply than platform debates because the affected community is specific, the power relationships are local, and the institution's mission introduces a third value (intellectual inquiry) that neither speech absolutism nor harm-based restriction captures cleanly.
- Social Media and Democracy: What Each Position Is Protecting — the upstream question about what platforms do to democratic discourse before any moderation decision is made; the platform speech values debate is partly a response to empirical claims made in this debate, particularly the claim that algorithmic amplification of emotionally activating content shapes democratic epistemology in ways that no speech governance framework yet adequately addresses.
- Reparations: What Each Position Is Protecting — a structurally similar debate where the liberal procedural frame (equal treatment, no group rights, present-tense evaluation) collides with the historical accountability frame (group-based injustice requires group-based remedy, history shapes present conditions); the anti-colonial critique of free speech and the reparations debate share a root: the question of whether liberal frameworks developed during and after slavery and colonialism carry those conditions invisibly forward.
- Surveillance Capitalism: What Each Position Is Protecting — the economic model that makes all four positions harder: if platforms are monetizing behavioral data generated by engagement with content, then the free expression framework, the harm-based framework, and the communitarian framework are all being applied to a speech environment that is also a surveillance environment, where "expression" is also "data production" and the interests at stake include not just speech values but privacy, economic autonomy, and the integrity of attention.
Further reading
- John Stuart Mill, On Liberty (1859) — the foundational text of the free expression absolutist tradition: Mill's argument that the only legitimate basis for restricting speech is direct harm to others, and his epistemological case that suppressing ideas always risks suppressing truth, remain the clearest statement of the position; chapter 2, on liberty of thought and discussion, is the specific source of the "marketplace of ideas" framework that still structures most First Amendment jurisprudence; reading it alongside the literature on Mill's work for the East India Company — and his arguments about "despotism" as appropriate governance for colonized peoples — makes visible the tradition's genealogy that the anti-colonial critique targets.
- Jeremy Waldron, The Harm in Hate Speech (Harvard University Press, 2012) — the most rigorous philosophical challenge to the absolutist position from within the liberal tradition: Waldron argues that hate speech degrades the basic social standing — the "assurance of inclusion" — that targeted group members carry into every public interaction, and that this is a real harm to a public good, not subjective offense; his distinction between dignity harm and offense is essential for understanding what the harm-based position actually claims and what it does not; his engagement with Ronald Dworkin and C. Edwin Baker — both of whom mounted serious liberal defenses of absolute speech protection — makes this one of the most rigorous internal debates in liberal philosophy of speech.
- Mari J. Matsuda, Charles R. Lawrence III, Richard Delgado, and Kimberlé Williams Crenshaw, Words That Wound: Critical Race Theory, Assaultive Speech, and the First Amendment (Westview Press, 1993) — the foundational CRT text on hate speech regulation: Matsuda's argument that racist speech should be legally actionable when it targets a historically oppressed group with a message of racial inferiority, Lawrence's argument that the First Amendment itself requires attention to the Fourteenth Amendment's equal protection guarantee, and Crenshaw's intersectional analysis of how speech regulation interacts with race and gender are the necessary background for the harm-based position in platform debates; the book's critique of First Amendment orthodoxy as itself a form of racial ideology is the direct precursor of the communitarian and anti-colonial positions.
- Stanley Fish, There's No Such Thing as Free Speech, and It's a Good Thing, Too (Oxford University Press, 1994) — the sharpest statement of the communitarian critique of liberal neutrality: Fish argues that all speech is already governed by community norms, that the claim of neutral proceduralism is a form of ideological mystification, and that the question is not whether speech will be regulated but whose norms will do the regulating; his argument that free speech absolutism is not a neutral position but a substantive political choice with real winners and losers remains provocative and underengaged in platform governance debates; the book functions best as a diagnostic tool for identifying what the absolutist position is actually doing rather than a governance program.
- Nanjala Nyabola, Digital Democracy, Analogue Politics: How the Internet Era is Transforming Politics in Kenya (Zed Books, 2018) — the most grounded account of what platform speech governance looks like from outside the Western liberal context: Nyabola documents how Kenyan activists, journalists, and ordinary users navigate digital platforms calibrated for US political culture, showing how speech governance decisions made in California ripple through communities with entirely different political histories, power structures, and relationships between speech, safety, and authority; her analysis refuses the naive optimism of "tech for democracy" and the equally naive pessimism of "tech as imperialism" to produce something more demanding: an account of the specific ways that governance frameworks embed the political culture of their origin and the work required to address that without discarding the real gains that digital access provides.
- Darcy Leigh, "The Settler Coloniality of Free Speech", International Political Sociology 16, no. 3 (2022) — the most direct academic treatment of the colonial genealogy of current free speech frameworks: the article traces how liberal free speech doctrine developed alongside colonial governance, was applied selectively to exclude colonized peoples, and now circulates globally through US-headquartered platforms as a universal standard; the "settler coloniality" framing draws on Aníbal Quijano's concept of coloniality — the persistence of colonial power structures after formal decolonization — to argue that the current free speech architecture perpetuates colonial-era exclusions in new institutional forms; essential for understanding the theoretical basis of the anti-colonial position.
- Safiya Umoja Noble, Algorithms of Oppression: How Search Engines Reinforce Racism (NYU Press, 2018) — the most-cited empirical account of how automated content systems encode and amplify existing social biases: Noble's analysis of how Google's search results systematically produced racist, sexist, and dehumanizing representations when queried about Black women extends to content moderation systems — the same training data problem, the same feedback loops that reward engagement over accuracy, the same structural position of Black, Indigenous, and Global South communities as subjects of automated systems rather than participants in their design; her framework for understanding algorithmic bias as the product of social context rather than technical error is necessary for the anti-colonial critique of platform speech governance.
- Evelyn Douek, "The Free Speech Blind Spot: Foreign Election Interference on Social Media", in Defending Democracies: Combating Foreign Election Interference in a Digital Age (Oxford University Press, 2021) — a practical account of how the absolutist position, when translated into platform design, produces systematic failures that none of its defenders intended: Douek shows how the First Amendment's default protection of speech interacts with the scale of platform governance to produce outcomes that are neither free nor well-governed; her argument that platform speech governance requires a new conceptual framework — neither First Amendment absolutism nor state speech regulation — is the most useful practical bridge between the philosophical positions this map describes and the institutional design questions addressed in the companion platform accountability map.
- Kate Klonick, "The New Governors: The People, Rules, and Processes Governing Online Speech", Harvard Law Review 131, no. 6 (2018) — the sociological account of how platform trust and safety teams actually make speech governance decisions: Klonick's interviews reveal that the people who govern speech at scale are not applying any coherent philosophical framework — neither Mill's harm principle nor Waldron's dignity standard nor Fish's communitarian norms — but navigating competing institutional pressures, legal constraints, and operational realities; her account of how internal policy develops across edge cases is essential context for evaluating all four theoretical positions, none of which was designed for the operational realities of governing speech at billions of posts per day.
- Joy Buolamwini and Timnit Gebru, "Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification", Proceedings of Machine Learning Research 81 (2018) — the empirical foundation for the argument that automated content systems systematically fail non-white, non-Western users: Buolamwini and Gebru demonstrate that commercial AI systems trained on non-representative data consistently misclassify darker-skinned faces at far higher error rates than lighter-skinned ones; extended to content moderation, this means that harm detection systems — trained on data from communities that overrepresent English-speaking Western users — will systematically overflag legitimate speech from underrepresented communities and underflag harmful speech from dominant ones; the paper operationalizes the anti-colonial critique at the level of algorithmic design rather than philosophical genealogy.