Search Engines and Free Speech: A Historical Analysis of Editorial Analogies and the Position of Media Companies and Users in US Free Speech Discourse

Origins of the Editorial Analogy: Publishers, Platforms, and Freedom of Speech in the 1930s and 1940s

In the 1930s, the media landscape was undergoing rapid change. Consolidation and commercialization in both newspaper and radio industries was raising concerns about these industries’ ability to serve the public—and indeed, suggesting that the interests of the press and the public might be in conflict. The rise of radio as a new technology and its consolidation into a handful of national networks raised fears of the power that the new medium provided to those speakers with access (the ability to address a majority of the nation via access to one platform). Then, as now, many commentators worried about the rise of new, powerful group of cultural mediators; whereas today, these concerns are about “Silicon Valley,” in the 1930s, they were about radio executives and advertisers (McChesney 1993; Smulyan 1994; Starr 2004).

Public concern over the media in the 1930s and into the 1940s was in response to a set of political economic and sociotechnical changes. In the 1930s, local newspapers were shutting down or being bought up as part of large national chains (e.g., Hearst, Gannett, Cox, and Scripps-Howard) dubbed “absentee owners” by critics—a term that evoked the potential abuses of absentee landlords (Pickard 2015). Radio stations, which were in the 1920s primarily local media outlets, were likewise during the 1930s joining national “chains” (NBC’s Red and Blue networks, CBS, Mutual) and increasingly airing network content rather than local talent or talk. In both industries, what once had been local mouthpieces were now operated by distant corporations whose interests did not necessarily converge with those of readers or listeners (Lebovic 2016; Smulyan 1994).

Further, concern about the role of advertisers in both industries was decreasing public trust in them as sources of information and opinion. Progressive critics worried about the influence of advertisers to censor what was printed in the newspaper (see, e.g., MacLeish 1941; Seldes 1935; Rorty 1934). The influence of advertisers on the radio was even more evident (see, e.g., Durr 1944; Ernst 1926). This, coupled with Depression-era popular distrust of industry and industrialists, made it harder to reconcile the interests of media owners (increasingly, considered part of the “industrialist class”) with those of the communities they served, or the public at large. In this context, as Lebovic (2016, 55) points out:

the traditional connection between the economic and political functions of the press had become a contentious site. Whereas the capitalist basis of the press had once been the foundation for its independence and liberalism, the industrialization of the press now seemed to suggest that press economics had made it subservient to undemocratic forces.

Increasingly, it seemed to progressive critics like MacLeish (1941) and George Seldes (whose 1938 book Lords of the Press called the US press America’s House of Lords) that freedom of the press not only did not align with the interests of the public, but could be actively inimical to them.

Such critiques led to calls to regulate the press, as part of broader New Deal efforts to curb excesses of big businesses which might harm the public. From proposals that news outlets (including newspapers) should be treated more like public utilities or common carriers to limits on consolidation to the creation of government-owned news outlets to compete with commercial ones, there were a number of ideas for how the government might push the press to better serve the public (Pickard 2015; Starr 2004; McChesney 1993; Blanchard 1978). Progressives argued that these calls for regulation of the press were not a violation of freedom of speech or the press, but in fact were necessary in order to achieve it.

Proponents for regulation drew on an emergent way of defining freedom of speech and the press. They argued that freedom of the press was not simply a freedom of the press, exercised by owners of the press, but rather a freedom of the public to a diversity of ideas and to enter into robust debate. In the latter vision, freedom of expression is less centered on individual rights to speak or publish and more on the social good that expressive liberties can provide. As such, freedom of expression (freedom of speech and press) might at times require government action as well as restraint. The re-articulation of freedom of expression as a social good had its origins in a reaction to WWI-era repression of dissent to Progressive Era politics and some strains of classic liberal political theory (David 1997). But, as Sam (2016) argues, the shape and use the argument took in the 1930s was closely entwined with the structural changes in news outlets and their role within the public sphere.

The re-articulation of freedom of expression was not only about political economy, or positive v. negative liberty, but also about what free speech looked like in a structurally transformed public sphere—and, central to my argument here, who could make First Amendment claims. The rise of radio networks and the industrialization and consolidation of newspapers had changed the role of most Americans in the public sphere. No longer imagined as proto-publishers or incipient speakers, most Americans were by the 1930s positioned as members of the audience, more consumers of information and opinion than producers of them (Butsch 2007; Douglas 1987). While the idea that any citizen was potentially producer of information (or, a publisher) was substantively a fiction in the 19th century, it was at least a structural possibility for many (Lebovic 2016; see also Blanchard 1978). Changes to printing such as the growing expense and complication of printing, the consolidation of newspapers into chains and the advent of radio, which required both expensive equipment and a license to use the spectrum, meant that many fewer Americans were capable of becoming publishers, or speaking to a broad audience. ² This, and the sense that the class interests of owners and pressures of advertisers operated as a form of censorship (in particular, blocking of voices of labor, African-Americans, feminists, and those who would criticize the commercial system) seemed to foreclose the opportunity of many to speak in the mass mediated public sphere (Lebovic 2016; Starr 2004; McChesney 1993).

The structural positioning of most citizens as “listeners,” or consumers of information, then, transformed the average citizen’s relation to freedom of speech. If freedom of speech and the press only applied to acts of public oration and publication, and most Americans were positioned as listeners rather than speakers—their role in the public sphere seemed increasingly outside the scope of the First Amendment. ³ The First Amendment would, as contemporaneous leftist commentators and justices pointed out, leave many Americans behind, favoring primarily media owners (Barbas 2011; Seldes 1935; Rorty 1934).

The progressives’ re-articulation of freedom of speech and press as a social good responded to this shift. In this reframing, freedom of speech did not only or primarily protect the rights of individual media owners but rather protected the ability of members of the public to receive a broad array of information and points of view (David 1997; Graber 1991). ⁴ As Morris Ernst, a leftist legal advocate and early leader of the ACLU, argued “Freedom of speech—at least in the realm of radio—is in the ear of the listener” (Ernst 1926). It was a freedom to not to have one’s voice heard, but rather a freedom to hear diverse points of view from labor, feminists, African-American and immigrant speakers. And, in order to realize this freedom, the interests of media owners might need to be downplayed or curtailed in order to protect the interests of the public as audience. In this way, the idea of freedom of speech as a social good provided a way of interpreting the First Amendment that would apply not only to the expression of owners of media outlets, but also to the interests of the audience (Lebovic 2016; Pickard 2015; Benjamin 2001).

In these debates, then, the meaning of freedom of speech was discussed and revised within the contexts of not only social and political change but also within the context of sociotechnical and political economic change in the infrastructure of public communication. It was a re-interpretation of freedom of the press and speech that shifted the locus of participation from the act of speaking to acts of listening and reading, or more broadly, to attention and engagement. It was, in other words, a re-thinking of what freedom of speech meant in a public sphere in which the relationships of speakers and audiences were transformed by new technologies and industry organization (consolidation into chains). ⁵

Crucially, this re-articulation acknowledged the social and technical bifurcation of speakers and listeners associated with the industrialization of the press and the rise of radio as well as the conflicts of interest between speakers and listeners highlighted by this bifurcation. In earlier discussions of freedom of speech, the interests of the speaker and audience, publisher and reader could be assumed to coincide. Within the political and technological context of the 1930s and 1940s, this no longer held (Lebovic 2016). The role of advertising, the dearth of competition, and the sheer size of media outlets cast doubt on the workings of the market. Further, the alignment of the press with industry—and a powerful elite of “industrialists” who were, in the context of the Great Depression and New Deal progressivism, understood to operate at the expense of average citizens—discursively highlighted an antagonistic relationship between the press and the public. How to resolve this tension—via the market or via regulation—was a subject of debate (McChesney 1993).

Progressives used the argument that freedom of speech meant the public had a right to diverse information to argue for curbs on the ability of media owners to use their platforms to sway public opinion, and to argue against monopolistic practices (e.g., refusing competing papers’ entry to the Associated Press [AP], the expansion of large radio networks) and to push back when newspapers fired journalists for unionizing. In response, the newspaper industry turned to the First Amendment. The American Newspaper Publishers Association (ANPA) worked hard to align freedom of the press with the freedoms (economic and political) of owners (Benjamin 2001; Lebovic 2016). These rights, the ANPA argued, included not only decisions about what ideas or information to print, but also business decisions, such as who to hire and fire. In other words, freedom of the press meant freedom from economic regulation—at the time, mainly from the New Deal labor laws enabling unionization and from anti-trust laws. By the late 1930s, “the language of press freedom was monopolized by publishers who used it to protect their own autonomy” (2016, 84)—an autonomy that was as much economic as political. ⁶

These are the roots of the editorial analogy. It was an argument that aligned freedom of the press with the freedom of publishers (more than editors) and was used to argue for both ideological independence and freedom from structural, economic regulation. Newspaper owners had limited success employing this tactic against New Deal labor regulation and anti-trust law: in AP v. NLRB (1937), the Court narrowly decided that freedom of the press did not include freedom from labor regulation and in AP v. US (1945), that freedom of the press did not mean freedom to engage in anticompetitive practices. More recently, in Miami v. Tornillo (1974) the Court ruled that any content regulation, even a right of reply that would increase the amount of information or diversity of ideas in circulation, is a violation of freedom of the press. It was an understanding of the First Amendment in which operators of a press are the only ones that may claim speech rights in print communication (Stein 2007; Horwitz 1991). In this way, current interpretations of the law subsume discussions of the social good to the freedom of editors, or at best treat editorial freedoms as proxies for the expressive interest and rights of the public. For many tech companies today, this is the core and the lure of the editorial analogy.

While the audience-centric interpretation of free speech had less success when applied to newspapers, it was famously influential in broadcast regulation. Arguments that the interests of listeners needed to be taken into account in applications of the First Amendment shaped regulation—and the way judges and justices understood the relation of this regulation to the First Amendment in the 1940s (Pickard 2015). While the right of the public to receive information was not officially recognized until 1969, in Red Lion v. FCC, the question of just whose rights were protected by the First Amendment was central to debates about radio regulation in the 1940s. These debates included including in the pivotal 1943 case, NBC v. US, which upheld the ability of the Federal Communications Commission (FCC) to refuse to renew radio licenses based on programing (NBC v. US 1943). The conflicts of interest among listeners, local station operators, and broadcast networks—and the question of whose expressive interests were primary under the First Amendment—were integral to the arguments in the case (see Appellant’s Brief, NBC v. U S 1943). These were in fact the terms on which the case was initially decided in district court (for the southern district of New York) by famed liberal judge Learned Hand. Hand decided for the FCC arguing that freedom of speech belonged first, to the listening public, second, to the local stations, and only third and last to the networks ( NBC v. US 1942 , 946). The Supreme Court offered a more technical rationale, including spectrum scarcity, for upholding the FCC’s regulatory discretion, dodging questions of whose interests should prevail. Much of the radio regulation in the following decade (e.g., the “Mayflower doctrine” attempt to limit the ability of radio owners to use their stations or networks as a platform to promote their own views) can be characterized as attempts to balance the power of radio speakers and the public as listeners, whose interests were understood to be in conflict (Pickard 2015).

This history highlights several features of the debates around the application of the First Amendment to radio and newspapers in the 1930s and 1940s as useful comparison points for today. First, conflicts over whose interests should be served in legal interpretations of the First Amendment were clearly delineated. The framing and articulation of the debate were organized around antagonisms between media owners’ (and among local and national owners) versus media audiences’ interests, and discussion of which interests should be given preference. Second, the debates around the application of the First Amendment to radio in the 1930s and 1940s were about how to interpret the law and precedent in light of political economic and sociotechnical changes that restructured the possibilities of public communication. Advocates and jurists came up with new arguments about freedom of speech, including the recognition of activities like listening as part of this freedom.

The Editorial Analogy Today: Monopolizing the First Amendment

In many ways, it is the outcome of these debates that Google and other search engines wish to avoid today. This is clear in the white paper Google commissioned on the First Amendment status of search engines, which comes close to calling out the Fairness Doctrine (and broadcast regulation) by name, arguing against proposals to regulate companies like Google in the name of “what some people see as ‘fairness.’” (Volokh and Falk 2012, 3). The FCC policies known as the Fairness Doctrine required broadcasters to provide balanced coverage of matters of public importance. Google and others have gone to great lengths to argue that any such rules do not apply to them, often by arguing that the company is more like a newspaper than a broadcast or cable operator. They are, in fact, following in the footsteps of the newspaper industry in the 1930s and 1940s. ⁷ To claim to be like a newspaper in this case is to claim that the search engine is the only legally relevant speaker in a web search.

Volokh and Falk based a large portion of their argument on analogizing Google and its employees’ design of search algorithms to the judgments of editors about what stories to place above the fold (or travel guides’ decisions about which events and business to feature). The question of whether and how—or perhaps which types of—algorithmic systems involve judgments akin to those involved in editorial work is a compelling one. Yet, too exclusive a focus on the nature of algorithms does not highlight the way in which the editorial analogy is deployed opportunistically, as a tactic to evade regulation. Comparisons of algorithms to editorial decisions may be made in good faith, but they need not be so in order to be effective.

As I suggested in discussing the origins of the editorial analogy, it works to position the media operator as the only legitimate claimant of speech rights. ⁸ So, the use of the editorial analogy positions Google as the only relevant speaker in search for legal purposes. Yet, the question of whose expressive interests are most central to the normative ends or precedent of the First Amendment is not clear cut (see, e.g., Bracha and Pasquale 2008; Grimmelmann 2007). Not only scholarship but also the actions and rhetoric of Google itself testify to this murkiness. Google is far from the only speaker involved in web searches, as the company itself has made clear in its branding (which emphasizes user speech in its emphasis of relevance) and in its own efforts to evade liability for search results, for example in anti-Semitic and racist search results (Noble 2018; Vaidhyanathan 2011; see also Bilić 2016). When faced with the less savory results of search, Google has claimed that it is a mere conduit for the speech of others; search results are the speech of the website creators (and offensive autocomplete suggestions in search queries are the expression of search users). On the other hand, when a competitor alleges that Google is cooking its rankings in order to quash competition, Google claims that search results represent the company’s opinion. It argues that the algorithmic selection and ranking of results is equivalent to an editor deciding what stories to place in a newspaper, and what priority to give them in terms of placement (see Volokh and Falk 2012).

In response to the comparisons of algorithmic outcomes to editorial judgments, many critics have argued newspapers are a bad analogy and that other media, like cable, are a more fitting comparison. Search engines, like cable operators, make editorial decisions about what channels to carry and thus have speech rights, but there are some limits; they can be required to carry local broadcast stations (Bracha and Pasquale 2008). Still others suggest that search engines might also be like supermarkets or law schools hosting military recruiters (Whitney 2018). Too often, these analogies are made in terms of surface or technical resemblance divorced from social context and concerns—for example, whether or not search engines, like cable operators, have the technical power to silence some speakers or whether search engine results are neutral or objective. ⁹ Such a focus on indicators and technical features can substitute for substantive ethical and political discussion of the matter being assessed (Star 1999). In comparisons like those listed above, it is easy to forget that cable operators are required to carry local broadcasts not primarily due to the technical capacity of cable wires or operators, but in an effort to balance the speech rights of cable operators and audiences (Horwitz 1991).

The contemporary deployment of the editorial analogy as if it were a purely technical question (of whether algorithms are neutral, of what type of “channel” search algorithms are), then, occludes many political and social stakes of the debate. This framing is only possible if we forget the ways in which the analogy, and free speech law more broadly, has developed in reaction and relation to specific social arrangements and relations (often given shape by specific communication technologies). The editorial analogy is not a simple extension of the regulatory regime applied to print, but rather was a way of settling a set of conflicts about how to apply an existing law in a context where many of the background assumptions about communication embedded in it no longer applied. The analogy was the outcome of a political debate, in which there were winners and losers. And its logic and legitimacy rested on a presumption that the social role of editors (defined in part by professional norms, and the potential damage to one’s reputation for diverging from those norms and market forces) tended to align their judgments with democratic processes and the public interest. The editorial analogy attempts to superimpose this set of presumptions—and the regulatory space defined by the winners (editors and publishers) in these debates—on a different set of social and political, not to mention technological, concerns today.

Reframing debates about search engine speech – and cases like those mentioned at the beginning of this article (Search King v. Google and Zhang v. Baidu)—less in terms of resemblance to an older channel (i.e., cable, broadcast, or pint) and more in terms of the different speakers and audiences and their various roles and interests brings a different set of issues to the fore. In the Search King case mentioned in the introduction involving allegations of anti-competitive practices, the court determined that search results were subjective because they were “produced by an algorithm unique to Google,” and thus a form of opinion—and thus that Google was a speaker protected by the First Amendment (Search King v. Google 2003, 11). If search results were opinions, then Google’s manipulation of these results could not be considered interference with a competitor’s business. (This was a prime case of First Amendment opportunism, bringing in free speech claims to gain an easier victory and a broader protection from liability than would an argument about business practices.)

In Zhang v. Baidu, the case in which pro-democracy activists alleged Chinese search engine Baidu violated their civil rights, the court found that the search engine Baidu could remove pro-democracy results from its US searches—and indeed that the ability of the company to decide to suppress or highlight political content fell within the “principle on which our political system and cultural life rest” (Zhang v. Baidu 2014, 441). The individuals who made the case against Baidu were alleging violations of civil liberties (an argument that was a legal long shot); what I want to highlight about the decision is the way that the court was able to simultaneously ground its decision in freedom of speech as a core freedom and simply ignore the claims of the users, or citizens interested in either promoting or finding information. What might have been a complex case in which the court parsed the claims of the different speakers and listeners and considered the public interest was presented as a simple case of editorial liberty, and hung on whether Baidu was a speaker. The search engine’s suppression of pro-democracy results was proof that the company made “editorial judgments” and thus its actions were fully protected by the First Amendment, beyond the claims of the plaintiffs (Zhang v. Baidu 2014, 437–438). (The rationale in the case is even more striking and paradoxical if one views Baidu as acting not only as a private company but also as an extension of the Chinese government in its suppression of pro-democracy results; Lakier 2018). The decision in Baidu dramatically illustrates one way that speech rights have been distributed in disputes about search engine regulation. In Baidu, as well as Search King, the decisions and freedoms of companies take center stage. The use of this regulatory framework advances the economic interests of search engine companies as political rights. It also hinges the public interest in a diversity of quality information and ideas on the business models and branding efforts of a few large companies.

The point I want to make here is not that Google or Baidu should be obligated to run or feature content from every website creator, nor that search engines should attempt to be neutral conduits. There are good economic and normative reasons for search engines to exclude some sites, and to make judgments about what constitutes quality results. ¹⁰ Rather, I want to point out that the use of the editorial analogy in debates about the First Amendment and search engines (and other technologies) covers over the varied expressive interests involved, recognizing only the rights of owners. Perhaps more importantly, the analogy is ill equipped to address the bigger questions of how search engines and other examples of algorithmic mediation reshuffle the forms of communication and social relationships that constitute the public sphere.

How to address these different interests and forms of communication in a way that meaningfully protects, or even cultivates, public discourse is not an easy task, and doing so well may take some years. One initial step is to refocus our conversation on the social contexts and social actors involved in search (and other algorithmically mediated communication). Doing so brings to the fore arguments about power, agency, and the type of social actors and relations that ought to be protected under the First Amendment (see, e.g., Baker 2007; Bracha 2014; Grimmelmann 2007). For example, Baidu pitted the interests of users in accessing a variety of political information against the editorial decisions of search engines. Cyber harassment cases pit the interests of some members of the public against individual content creators and, at times, the preferences of search engines (Citron 2015). And privacy protection proposals—as in the European Union’s right to be forgotten—pit the privacy interests of members of the public (and potential users) against the interests of other users in accessing information as well as against the search engines’ interests in providing information and avoiding the complication of processing take-down requests. Like many other examples of social interaction online, these are not simple problems with straightforward solutions. Declaring search engines akin to editors evades complexity in the hope that the companies will either act as trustees of the public interest, diagnosing what informational policies best serve the public interest and enacting them, or that the companies’ interests will align with those of the users or public through the alchemy of market mechanisms.

More fundamentally, the current framing does not address the way that the very mechanics and social dynamics of public communication are, once again, undergoing rapid change. The regulatory debates in the 1930s were not only about formal politics, or the abstract legal-political debate about positive v. negative liberties, or even about narrow economic interests. They were also and centrally about how to understand—and adjust existing conceptions of—freedom of speech in the face of technological, economic, and experiential changes to the dominant means of communication. That is, scholars and advocates seeking to understand or shape the contemporary and future First Amendment regimes for emerging media should take lessons from the way that social and technological change shaped regulatory debates in the last century. Commentators and legal practitioners in the 1930s worked to re-articulate our understandings of freedom of speech from one tuned to a 19th century socio-technical arrangement, in which the primary “speakers” were publishers and editors of small presses and in which, in theory, anyone could become a pamphleteer or publisher. In its place, they found ways of articulating freedom of speech not only around speakers (or proto-speakers), but also around audiences (non-dialogic listeners) and a different set of activities (listening, or receiving information). Media scholars and advocates might take inspiration from these debates to ask questions about how the meanings and mechanisms of public discourse are changing, and what the most pressing barriers to freedom of speech are today. The answers are unlikely to be the same ones as those developed in the 1930s; many of the problems are different—today, opportunities to speak or to access information abound but the logics that mediate the amplification of our speech or filter the information we receive are opaque and seemingly beyond our control (Ananny 2016; Gillespie 2014).

What history teaches in this case is to attend to the way that new technology may reshape practices of public discourse and the activities that constitute participation in the public sphere—or the very nature of public speech. This reshaping calls for different applications or understandings of freedom of speech. Like radio in the 1930s, search engines and similar algorithmic outputs (which produce statements based on the design decisions and values embedded in algorithms, the input of users, and the words or content produced by the online crowd) are novel forms of communication that disrupt existing practices and norms of public communication. In the case of search engines, these algorithms mediate our access to information and the public sphere in ways are not yet addressed in policy, regulatory or legal contexts. In the 1930s, advocates and jurists crafted news ways of thinking and making law around freedom of speech to fit a new media and social environment that changed the nature of public discourse and the structural role of the public in that discourse. The technological and social changes under way today call for a similar flexibility and engagement with the meaning of freedom of speech.