First Amendment: Speech Response 127 Harv. L. Rev. F. 351

Freedom in Systems

Free speech exists at the intersection of private and public systems of technology and organization

Response To:


Jack Balkin’s Old-School/New-School Speech Regulation1 offers a powerful analysis of how the “free” part of “free speech” is a function of the systems we inhabit as speakers and writers, listeners and readers. In this Response I explain how his masterful map of modes of new and old speech regulation expresses free speech as a function of the affordances and constraints of technical systems we inhabit, and outlines a range of strategies deployed by both public and private actors through these technical systems and the organizational systems that control them to extend and deepen speech regulation, even as these very systems offer new opportunities for free speech. That lesson is not limited to expression: Individuals are always situated in multiple overlapping systems of action that offer affordances and constraints, providing capabilities for individuals to make and pursue life plans, constraints on those capabilities, and mechanisms for others to exert power over those whose freedom we are analyzing. To consider “freedom” for beings such as we are is to analyze the anticipated lived experience at the intersections of these systems; how their interaction constrains and enables us to perceive the state of the world, conceive of possible actions, develop preferences, principles, and policies that assess the possible actions and outcomes available, and to choose and act given the system properties we inhabit. Law in general, much less constitutional law by itself, is only one of these many and diverse systems, and any conception of freedom that focuses on constitutional rights in isolation will miss its mark.

To understand freedom as a function of the systems we inhabit is, first and foremost, to take a realist view of freedom: it considers a given legal intervention in terms of its real world effects on the affordances and constraints open to individuals living in a society. New York Times Co. v. Sullivan2 was the first major legal realist First Amendment opinion and, as we celebrate its anniversary, we celebrate with it the line of cases that embraced realism as constitutional method. Following the other major realist constitutional decisions, Brown v. Board of Education3 and Shelley v. Kraemer,4 the great innovation in Sullivan was its willingness to acknowledge that an established rule of private law could become a source of constitutionally proscribed censorship. Just as a facially neutral law prescribing separate but equal facilities was, on the background of extant social practices, discriminatory in intent and practice,5 so too in Sullivan: a neutral law protecting a private interest enforced through civil litigation could be unconstitutional censorship. The critical insight was that effects in the world matter more than formal doctrinal categories, such as “state action,” and that these effects need not arise directly from the law, but can arise from its interaction with background practices in the social, economic, or technical environment. It is this core insight of Sullivan that Balkin’s paper elegantly translates to the context of speech in the networked environment of the twenty-first century.

The realism that typifies Balkin’s paper permeates this Symposium; it is a realism sensitive to the interaction among systems of action, in that it sees the effects of law always in conjunction with the effects of other systems that operate in the context as to which the legal change applies. Susan Crawford’s analysis shows how First Amendment law actually undermines, rather than protects speech, once the organizational and technical architecture of the Internet is taken into account.6 For Crawford, “free speech” in the constitutional doctrine sense threatens to prevent a regulatory response that could preserve a set of technical, organizational, and contractual practices that universally provide the actual foundations of free speech capabilities over an open, symmetrically-enabled Internet. Marvin Ammori describes how the normative and professional commitments of lawyers working at Internet companies provide a source of protection for free speech from certain classes of potential constraint.7 In his description, free speech is protected by a combination of the professional ethics system and the cultural commitments to a free and open Internet that characterize the social and educational networks of the lawyers at a core set of organizations, both translated through an organizational operational system — legal advice to shape the formal policy of certain organizations whose position in the market allows them to shape the actual capability sets necessary for free speech. What Crawford and Ammori share with Balkin is a recognition that freedom of speech in the digitally networked environment depends on the interlocking effects of multiple systems: technical, organizational, cultural, and legal. But even these legal elements are not those usually considered “constitutional.”8

What is true of free speech is true of freedom more generally. We see this most directly (with regard to privacy and dignity), in the concurring opinions in United States v. Jones,9 where the Supreme Court rejected the use of surveillance by a GPS device attached to a car without a warrant. Justice Alito’s concurring opinion, for four Justices, emphasized that before computers, “the greatest protections of privacy” were sheer difficulty and cost of maintaining long-term pervasive surveillance.10 Justice Sotomayor, in concurrence, challenged more fundamentally the collection of meta-data, which in the past was considered unproblematic, because with new data analysis techniques and ubiquitous sensors, it now “generates a precise, comprehensive record of a person’s public movements that reflects a wealth of detail about her familial, political, professional, religious and sexual associations.”11 Together, these opinions, signed by a majority of the sitting Justices, reflect an understanding of freedom not as a function purely of law, but also as a function of other extant systems: data analysis technology, the politics of budgeting, or organizational staffing and routines of police forces. Freedom cannot be understood as mere formal legal constraint on power. It is the lived reality for lives lived within the forces created by these many systems. As with the Old-School/New-School techniques Balkin analyzes, the five concurring Justices in Jones see that the human dignity and autonomy we get from privacy are a function of the intersection of multiple systems, and if constitutional law is to protect freedom, it must be designed to address the effects of the overall intersection of systems, and how they will both enable freedom for individuals and instantiate power flows against them in ways that, unchecked by law, would defeat freedom even where particular elements of the system appear on their faces to enhance freedom when considered in isolation.

Internet law scholarship has long been sensitive to the kinds of integrated systems effects that I outline here as central to the definition of practical freedom. Among Internet law scholars, Lawrence Lessig’s synthesis of the law and social norms literature12 and the values in design literature13 has made it commonplace to think of “regulation” (traditionally thought of in legal scholarship as a matter of law) as being the function of the constraints that law, technology, markets, and social norms place on our behavior.14

The next generation of scholarship on the integrated effects of multiple systems needs to expand along three dimensions. First, systems — the regular interactions of objects (including people) and processes that provide a degree of predictability to connect perceived states of the world, behaviors, and outcomes — need to be understood at lower levels of abstraction.15 If we are trying to understand the actual operative elements of systems that constrain or afford action, we need to describe systems at the level of abstraction that best reflects how they in fact create affordances and constraints given the diversity of actual, instantiated systems in the world. An open farmers’ market is a very different system than the market in large-scale communications infrastructure. To say “the market” regulates is to abstract from the enormous difference between these different kinds of markets. If the first generation scholarly project was to expose the fact that law was not the sole regulatory system one needed to consider in designing for a human interaction, the second generation needs to operate at the level of specificity that best describes the dynamics that make a system effective as a determinant of behavior. We begin to see this approach in the papers in this Symposium. The market for telecommunications home delivery is the functionally-operative “market” in Crawford’s paper; the market in cloud computing or handhelds, and its interaction with markets in general purpose decentralized personal computers, is the relevant market for some of Balkin’s concerns; the social practices of a particular generation of lawyers, and their intersection with their peer age group among Free and Open Source Software (FOSS) developers and young engineers in Silicon Valley startups is the relevant level for Ammori, given the roles of these actors in the particular market for speech-enabling Internet capabilities. Knowing the particulars of the relevant market or social dynamic offers purchase on the practical effects of any given legal or other change on values we care about in a way that more abstract analysis does not.

Second, affordances and constraints are inevitable characteristics of systems. Framing systems interaction as “regulation” focuses too much on the constraints side, and may risk missing the paired effect Balkin emphasizes here — that the very systems that enable new forms of speech also enable new forms of surveillance and censorship, and vice versa. Thus, although the shift from publishers to platform providers permits government to focus on a smaller set of actors through which to conduct surveillance and censorship, it also, perhaps first and foremost, enabled a tremendous decentralization of speech capabilities among users of these platforms.

Third, the model of power is too one-dimensional. The emphasis on regulation focuses a unidirectional flow of power: from the state, or the engineer designing a system, to the users. But the concern for freedom for individuals, situated as we are in multiple overlapping systems, is the flow of power and the relative ways in which different configurations allow different actors to impose their will upon each other and dodge efforts to exert such power. If Facebook prohibits hate speech but excludes humor, including humor “that many people may find to be in bad taste,” from the definition,16 it is deploying technical and organizational power, alongside the social dynamic that has made Facebook an unavoidable utility for millions of people, to shape the expressive uses of millions of people in a central part of their social and expressive lives. This may be a good idea or a bad idea, depending on one’s normative position as to the relative values of hate speech and humor, but it is undoubtedly power. And though it lacks “state action” and will not be recognized as a First Amendment violation, it likely imposes stricter constraints than a state, at least under the American First Amendment, ever could. Moreover, power need not be in the hands of organizations or centralized platforms at all, whether state or private. Where users try to silence content of which they disapprove by using platform affordances that let them flag content as inappropriate or violating copyright, they are certainly deploying a combination of technical affordances, organizational practices of the platform provider, and law (the terms of use) to impose censorship; their power is diffuse and largely avoids rule of law constraints, but it is power nonetheless. As we consider the implications of various system configurations for freedom as lived experience, we must account for a more diverse range of sources of power than simply the state.

To conclude, perhaps the most important contribution that Internet and society scholarship offers constitutional scholarship is that the tremendous technological, organizational, economic, and cultural perturbation that the Internet has wrought has led those of us studying the Internet to focus on the deep interconnection between all these systems, and the extent to which these interconnected effects, more than any distinct set of legal rights, shapes the very fabric of freedom as lived experience. As we celebrate a half century since Sullivan, we should take its basic realist insight to heart. Only if we develop an understanding of freedom as lived experience, in the actual systems we inhabit, will we be able to diagnose the real constraints and affordances we need to be free individuals, capable of developing a life plan, alone and in connection with others.

* Jack N. and Lillian R. Berkman Professor of Entrepreneurial Legal Studies, Harvard Law School; Co-Director, Berkman Center for Internet and Society, Harvard University.

1. 127 Harv. L. Rev. 2296 (2014).

2. 376 U.S. 254 (1964).

3. 347 U.S. 483 (1954).

4. 334 U.S. 1 (1948).

5. See Brown, 347 U.S. at 494 & n.11.

6. See generally Susan Crawford, First Amendment Common Sense, 127 Harv. L. Rev. 2343 (2014).

7. See generally Marvin Ammori, The “New” New York Times: Free Speech Lawyering in the Age of Google and Twitter, 127 Harv. L. Rev. 2259 (2014).

8. An earlier, more technology-focused conception of this understanding is in Eben Moglen, Liberation by Software, Guardian (Feb. 24, 2011),, archived at

9. 132 S. Ct. 945 (2012).

10. Id. at 963 (Alito, J., concurring in the judgment).

11. Id. at 955 (Sotomayor, J., concurring).

12. See generally Richard McAdams, The Origin, Development, and Regulation of Norms, 96 Mich. L. Rev. 338 (1997) (reviewing the law and social norms literature).

13. See generally Langdon Winner, The Whale and the Reactor 19–39 (Do Artifacts Have Politics?); Batya Friedman & Hellen Nissenbaum, Bias in Computer Systems, 14 ACM Transactions On Information Systems 330–47 (1996), archived at

14. See generally Lawrence Lessig, Code (1999).

15. See Yochai Benkler, Networks of Power, Degrees of Freedom, 5 Int’l J. of Comms. 721, 724–38 (2011).

16. What Does Facebook Consider to Be Hate Speech?, Facebook, (last visited June 4, 2014), archived at