Two weeks ago, Gizmodo published an article describing an anonymous former Facebook employee’s claims that Facebook news curators routinely suppressed conservative news from the social network’s “Trending Topics” module. Facebook responded; the Senate Commerce Committee sent Facebook an inquiry; Mark Zuckerberg himself invited conservatives to meet with the Facebook team. A whole lot of smart people weighed in, including the folks at Microsoft Research’s Social Media Collective (1, 2),  Harvard’s Nieman Journalism Lab, and technology scholar (and Harvard Berkman Center associate, and New York Times opinion writer) Zeynep Tufekci. And Glenn Beck just reported back from the first meeting Facebook held to discuss the issue with American conservatives.

In all the primary reporting, responses, inquiries, and commentary, there’s some common language:

Gizmodo: “…But the revelations undermine any presumption of Facebook as a neutral pipeline for news…”

Tom Stocky, Facebook Search VP: “… There are rigorous guidelines in place for the review team to ensure consistency and neutrality.”

John Thune, chairman of the Senate Commerce Committee: “…Any attempt by a neutral and inclusive social media platform to censor or manipulate political discussion is an abuse of trust and inconsistent with the values of an open Internet.”

Tarleton Gillespie, of Microsoft Research: “Algorithms are not neutral. Algorithms do not function apart from people. … Calls for more public accountability…can only proceed once we completely jettison the idea that algorithms are neutral — and replace it with a different language that can assess the work that people and systems do together.”

Nieman Lab: “Despite its claims, Facebook has never been a completely neutral platform…”

Zeynep Tufekci: “…”surfaced by an algorithm” is not a defense of neutrality, because algorithms aren’t neutral. … Without laws of nature to anchor them, algorithms used in such subjective decision making can never be truly neutral, objective or scientific.”

There are lots of excellent points here (I really recommend you take the time to read at least at those last three articles). But the language itself matters. As Ithiel de Sola Pool (of MIT Political Science fame) says in Technologies of Freedom (1983):

"The courts and regulatory agencies in the American system (or other authorities elsewhere) enter as arbiters of the conflicts among entrepreneurs, interest groups, and political organizations battling for control of the new technology. The arbiters, applying familiar analogies from the past to their lay image of the new technology, create a partly old, partly new structure of rights and obligations. The telegraph was analogized to railroads, the telephone to the telegraph, and cable television to broadcasting."

In other words, analogies matter. They form the basis for both legal and public understandings of new technologies. And there’s an implicit analogy being made here, in all this talk about neutrality–that is, to the network neutrality debate. And here, as with the network neutrality debate, neutrality doesn’t mean objectivity.

In American common parlance, the term “network neutrality” has most commonly been used in relation to debates over the ability of the Federal Communications Commission (FCC) to regulate internet service providers (alternately called broadband providers or ISPs). That debate culminated in 2015, when the FCC released the Open Internet Order and confirmed “the FCC’s legal authority to fully address threats to openness on today’s networks…including reclassification of broadband Internet access as a telecommunications service under Title II of the Communications Act” (2015). Reclassifying broadband as a telecommunications service gave the FCC legal power to enforce three primary rules, quoted below (Federal Communications Commission, 2015):

1. No Blocking: broadband providers may not block access to legal content, applications, services, or non-harmful devices.
2. No Throttling: broadband providers may not impair or degrade lawful Internet traffic on the basis of content, applications, services, or non-harmful devices.
3. No Paid Prioritization: broadband providers may not favor some lawful Internet traffic over other lawful traffic in exchange for consideration of any kind—in other words, no “fast lanes.” This rule also bans ISPs from prioritizing content and services of their affiliates.

Compare this with Mark Zuckerberg’s statement and the parallels become convincingly tight: “We have rigorous guidelines that do not permit the prioritization of one viewpoint over another or the suppression of political perspectives.”

This is, of course, an oversimplified definition of net neutrality that belies the historical and technical context from which the notions of open internet and network neutrality originate. In fact, the term “network neutrality” was coined by Tim Wu in 2003. He says the following on his website (Network Neutrality FAQ):

Network neutrality is best defined as a network design principle. The idea is that a maximally useful public information network aspires to treat all content, sites, and platforms equally…The principle suggests that information networks are often more valuable when they are less specialized – when they are a platform for multiple uses, present and future.

Wu’s definition points not at regulatory policy but at the philosophies shared by the original framers of the internet, closely related to the “end-to-end principle” in network design (Saltzer et al., 1981), David Isenberg’s “stupid network” (1998), and Jonathan Zittrain’s account of the “generative internet” (2006). It has nothing to do with objectivity and a lot to do with cyberlibertarian ideologies of openness and free access.

The FCC’s Open Internet Order speaks explicitly about broadband internet providers, and Wu’s definition uses “public information network” and “platform” synonymously with the internet. But we live in an age of many platforms, not just the monolithic “internet.” Facebook is a social network platform, as is Twitter or Snapchat. Arguably, email lists or newspapers are platforms for networks as well, insofar as they are a means through which people connect with each other and form communities. So if neutrality doesn’t mean objectivity, the questions become: in what cases does it make sense to assess these platforms with some idea of neutrality? And how can leverage notions of net neutrality to develop a generalizable framework for assessing the neutrality of any other networking platform?

Without debating the validity of the “open internet” philosophies in which traditional definitions of net neutrality are rooted, I want to extend the principle of net neutrality so that it might apply not just to the internet, but more generally to social networks, platforms, and infrastructures both online and offline. To do so, I propose an alternate definition of neutrality, framed neither as a matter of regulatory policy nor as a matter of network design, but instead as a matter of infrastructural power. Under this alternate definition, I establish a framework for assessing the neutrality of other network platforms like Facebook or email, in the hopes of providing a tool which might inform broader questions of citizenship, civic duty, user rights, and governance in networked communities.

From Platforms to Infrastructure

I have chosen to leverage the notion of “infrastructure” in describing networking platforms because infrastructure is associated with an idea of basic necessity, of existing technology that users take for granted and build on top of. Not all networks are infrastructure; for instance, the comments section of a particular YouTube channel, or a forum for aggregating and discussing Star Trek slash fiction, might appropriately be termed platforms on which users connect with each other and form communities. Nevertheless, neither the comments section nor the Star Trek forum are usually viewed as broadly necessary for daily life, and therefore would not be called “infrastructure.” Networking technologies often shift from platforms to infrastructure. Consider an excerpt from Wu’s testimony to the House Committee on the Judiciary Telecom and Antitrust Task Force (2006):

[Network neutrality] is an issue that affects people directly. Once upon a time the internet was a kind of toy, used by hobbyists, scientists, and geeks. But today it’s something different: it has become part of America’s basic infrastructure…It’s an infrastructure that people and firms depend on for everyday activities, whether planning weddings, managing investments, or running a small business.

An implication in Wu’s argument is that the more we depend on a platform (that is, the more infrastructural a platform is), the more it matters to negotiate and preserve the neutrality of that platform. When the internet was “a kind of toy,” neutrality didn’t matter too much; but now that the internet is “an infrastructure that people and firms depend on,” discussions of neutrality become more and more pertinent.

What does it mean for infrastructure to be neutral? We have seen that internet-centered definitions of network neutrality circumscribe the notion that the internet should be agnostic to content and users; recall Wu’s statement that “a maximally useful public information network aspires to treat all content, sites, and platforms equally.” In this sense, network neutrality is a matter of ensuring basic democratic equality (or at least of “equality before the law”). But, to speak technodeterministically, it is impossible to engineer totally neutral infrastructure, and often engineers explicitly make nonneutral infrastructural decisions. Langdon Winner argues this point in “Do Artifacts Have Politics”, illustrating his argument with a discussion of New York highway overpasses (1980):

Robert Moses…had these overpasses built to specifications that would discourage the presence of buses on his parkways…Automobile-owning whites of “upper” and “comfortable middle” classes, as he called them, would be free to use the parkways for recreation and commuting. Poor people and blacks, who normally used public transit, were kept off the roads because the twelve-foot tall buses could not get through the overpasses. One consequence was to limit access of racial minorities and low-income groups to Jones Beach, Moses’s widely acclaimed public park.

Equality (and therefore neutrality) is necessarily conditioned on the technological affordances of the infrastructure, whether they were implemented by intention or by necessity. Highway infrastructure treats everyone equally, as long as that “everyone” only includes people using vehicles that have the technical ability to drive on highways. That condition implies equality for a certain class of automobiles, and discrimination against other vehicles like bicycles, airplanes, or large buses, as Winner describes.

The Infrastructural Reach of the Modern State

Our brief detour has established that net neutrality is most pertinent when platforms are not only networked, but also infrastructural. We have also seen that neutrality is rooted in basic notions of equality, and that infrastructure often (and necessarily) limits equality to a class of users which satisfies certain conditions. That equality on a network is necessarily limited motivates a generalizable definition of network neutrality premised not on objectivity or equal treatment, but instead on infrastructural power. Infrastructural power is a term coined by sociologist Michael Mann in his analysis of state power (1980):

[The state] stores and can recall immediately a massive amount of information about all of us; it can enforce its will within the day almost anywhere in its domains; its influence on the overall economy is enormous…The state penetrates everyday life more than did any historical state. Its infrastructural power has increased enormously. If there were a Red Queen, we should all quail at her words – from Alaska to Florida, from the Shetlands to Cornwall, there is no hiding place from the infrastructural reach of the modern state.

Mann’s description of the state is prescient of networked platforms like Facebook, which indeed “stores…a massive amount of information about all of us” and “penetrates everyday life more than did any historical state.” To apply this notion of infrastructural power specifically to networked platforms, I build on Mann’s definition to specify that an actor has infrastructural power over a network to the extent that they can (1) modify, (2) augment, or (3) control access to that network. Under this framework, assessments of network neutrality are no longer deliberations of the (necessarily limited) technical equality of the network; instead, network neutrality is a measure of the distribution of infrastructural power. We can call a network “more neutral” if infrastructural power is distributed among its users, and “less neutral” if infrastructural power is more centralized in one or few entities.

Case Studies in Internet History

The usefulness and simplicity of this framework becomes clear when it is applied to pivotal events in the development and preservation of network neutrality. We can begin by revisiting the FCC’s Open Internet Order, which equated neutrality with three rules: no blocking, no throttling, and no paid prioritization. Under our new definition, the Open Internet Order is a regulatory limit on the infrastructural power of broadband providers. The “no blocking” rule prohibits broadband providers from controlling access to the internet. The “no throttling” and “no paid prioritization” rules limit broadband providers’ ability to modify and augment the internet, prohibiting them from implementing “fast lanes” to differentiate preferred users from general users. Therefore, the Open Internet Order preserves network neutrality in that it assures that broadband providers do not hold a monopoly on infrastructural power. Instead, it gives the FCC power to regulate broadband providers, therefore distributing infrastructural power between broadband providers and the FCC.

Going farther back in time, we can consider the landmark 1968 Carterfone ruling (13 F.C.C.2d 420), where the FCC ruled that Carterfone radio devices (and any other device) could be connected directly to the AT&T telephone network as long as they did not cause harm to the system. In other words, the FCC guaranteed third-party users the infrastructural power to augment the AT&T network with Carterfone devices, and later on with other third-party devices such as the answering machine, the fax machine, and the dial-up modem.

Generalizing to Other Networks

It’s a simple matter to apply this framework to other networking systems. Social network corporations like Facebook and Twitter wield great infrastructural power over their platforms, in that Facebook and Twitter engineers hold unilateral power to modify, augment, or limit access to their platform. This infrastructural power is sometimes wielded in highly visible ways. For instance, Facebook engineers famously tweaked Facebook’s algorithms to display more positive content to some users and more negative content to other users in the “emotional contagion” study (Kramer et al. 2014), exercising their ability to modify the Facebook platform without user consent. More recently, beginning in mid-2015, Twitter suspended 125,000 accounts “for threatening or promoting terrorist acts, primarily related to ISIS” (Twitter, 2016), exercising their ability to control access to their network (the matter of whether Twitter should actually have the judicial ability to ceremonially execute accounts that they deem ISIS-related is a separate and more complicated discussion of internet governance).

By contrast, infrastructural power over email (which is a federated system) is distributed among a number of email clients (Gmail, Microsoft Outlook, Mozilla Thunderbird) and a broad community of users. Strictly speaking, nothing prevents a user from accessing the email network as long as they have the technical ability to do so; notably, journalist Chris Kirk engineered his own personal email client and documented the process in a Slate article. Similarly, if we say that the English language is a networking platform (insofar as it provides an infrastructure for communication), very little prevents someone from speaking (and therefore accessing the “networking platform” created by the English language), as long as they are able to speak English. English is also highly modifiable and augmentable; new language is invented daily, and no centralized organization or actor has the ability to formally regulate new words and meanings. So the English language is neutral where Orwellian Newspeak is not; infrastructural power over the former is diffuse, while infrastructural power over the latter is centralized in an authoritarian state (not to say that there aren’t still more and less powerful actors over English–people who write dictionaries and grammar stylebooks might have more infrastructural power than the average citizen. But even if a language authority tells you to not split infinitives or to liberally use the Oxford comma, nothing forces you to comply in the various spaces, speeches and platforms of daily discourse.)

So in assessing the neutrality of any networking technology, we now have two guiding questions:

1. to what degree is this platform necessary, and therefore infrastructural?
2. who are the actors holding infrastructural power over this platform?

I have refrained from any discussion of whether, and in what cases, it is actually appropriate to decentralize infrastructural power in networked platforms and therefore increase neutrality. Mann’s paper on state power emphasizes a distinction between “infrastructural power” and “despotic power”; the two are related but separable, so a state with high infrastructural power does not necessarily mean it is despotic (just consider American democracy). Moreover, despotism is not necessarily always undesirable: Tim Wu says in his Net Neutrality FAQ that “On a private network [such as a cable TV network], discrimination…gives the network its utility. By definition it is closed to outsiders, and that’s what makes it useful.” The traditional newspaper is highly despotic over its content–its editors and writers select what gets published and what doesn’t–but that’s not a problem because we don’t expect newsrooms to be editorially “neutral.”

Which brings us back to Facebook Trending Topics.

Zuckerberg, in stating his intent to meet with “leading conservatives” and continue investigating the Trending Topics team, indicates Facebook’s continued pursuit of neutrality (or at least the appearance of neutrality). The Commerce Committee is complicit in this pursuit. Of the three main players–Gizmodo, Facebook, and the Senate Commerce Committee–both Facebook and the Commerce Committee behave as if neutrality is something desirable and achievable. Only Gizmodo suggests the possibility that true neutrality may not be attainable at all, arguing instead that we need to start thinking of Facebook “like a traditional newsroom, reflecting the biases of its workers and the institutional imperatives of the corporation.” There are two metaphors in play: Facebook as neutral network, and Facebook as editorial newsroom. The one which wins out (if one wins out at all) will have profound implications for regulatory policy concerning social networks. If as a public, we decide that Facebook is and should be a neutral network, that conclusion invites federal oversight and regulation under common carriage laws. If instead we agree that Facebook is more like a newsroom, Facebook becomes an independent member of the fourth estate with all the protections of free speech, and all the biases of editorialism.

The danger is that these meanings haven’t crystallized yet. Facebook still insists that it is content-neutral, and is trying to convince the Commerce Committee that the process of choosing Trending Topics is indeed unbiased. But Gizmodo’s report perhaps flags a shift in public understanding that Facebook (and all our other social media) isn’t and can never completely be the neutral, nonpartisan, open forum that it purports to be. And this ambiguity–this lack of resolution in what Facebook fundamentally is (and what it isn’t)–has implications not only for Trending Topics but also for Facebook as a whole, and every other social network, when it comes to how we use, regulate, and publicly understand them. There’s also the issue of transparency–why did it take a whistleblower for Facebook to disclose that humans are a part of the algorithm at all? The opacity of social networks’ internal workings has always been problematic, and it’s been flagged with regards to content moderation in ways that apply here as well. And I intentionally say “humans are a part of the algorithm” to speak to the point that Silicon Valley has an endemic labor problem regarding the mistreatment and invisibility of so-called “clickworkers”–it’s discussed in the content moderation article I linked above, as well as by Gizmodo in a separate article about how poorly the Trending Topics news curators were treated, as well as here, and here.

To conclude with some words from Tarleton Gillespie:

“The myth of the social network as a neutral space is crumbling, but it’s still very powerful. For Facebook to finally say, ‘Yes, we construct social life online. We construct public discourse’ — that would be so important, but for them, dangerous.”

---

This article was adapted from two essays that I wrote for CMS.701, Current Debates in Media, as part of my course requirements for the Comparative Media Studies undergraduate major. When I tell people I’m a CMS major, most of them are confused by what exactly “comparative media studies” is; I think this article is representative of the kinds of discussions that happen in a typical CMS classroom. It’s an excellent department and if you’re an incoming student who’s made it this far down, I highly encourage you to take at least one CMS class while you’re at MIT.

My thoughts here are by no means complete; if you have responses, critiques, or things to say, I’d love to hear them! Comment below or hit me on Twitter.