Earlier this week, President of the United States Donald Trump took to Twitter to rant and rave against Twitter, which fact‐checked a claim the president made about mail‐in ballots. Today, he will issue an executive order targeting social media websites. Last week, The Wall Street Journal reported that President Trump is considering establishing a White House commission to investigate anti‐conservative bias within the best‐known Internet companies such as Facebook, Twitter, and Google. These news stories aren’t a surprise. President Trump and many of his allies have been critical of Silicon Valley for years, alleging that social media companies are engaged in a campaign aimed at silencing conservative viewpoints. This claim is weak, with little empirical support, and conservative responses have revealed an unfortunate enthusiasm for government regulation of speech and a consistent misunderstanding of the law at the heart of online content moderation debates.
The Claim: Silicon Valley is Engaged in Anti‐Conservative Censorship
Allegations that prominent social media companies have been attempting to stifle and/or remove conservative content are varied and target different companies. Summarizing these arguments is difficult. Some allegations take aim at social media, with Facebook, Twitter, and YouTube being among the most popular targets. Others are more focused on Google’s search engine. The claims vary from allegations of “shadow banning,” tweaked algorithms, “de‐ranking,” to outright deletion of accounts. For the purposes of this article I will assemble all of these allegations under a broad claim:
The most famous U.S.-based Internet companies are engaged in a campaign focussed on limiting access to American conservative‐oriented content.
Lack of Evidence for The Claim
I have addressed allegations of anti‐conservative bias before. Readers interested in why I’m not convinced by James O’Keefe’s claims about Google’s bias or Dennis Prager’s complaints about YouTube’s treatment of PragerU can read my analysis here and here. In a recent working paper for The C. Boyden Gray Center for the Study of the Administrative State at George Mason University’s Antonin Scalia Law School I discuss other claims of bias, including those associated with Google’s search engine made by Robert Epstein of the American Institute for Behavioral Research and Technology (AIBRT). All of the claims of Silicon Valley anti‐conservative bias suffer from one or both of the following two major issues:
- The plural of anecdote is not data: Those paying attention to the ongoing content moderation debate already know that it is common for the conservative activists to highlight a collection of content moderation decisions and to present them as if they constitute evidence of anti‐conservative bias. In a social media environment as large as the one Internet users can access it would be strange if we didn’t occasionally observe false positive and false negatives that could collectively be presented as evidence of bias. Twitter users post hundreds of millions of tweets each day. In one hour, YouTube users upload more than 400 hours of content. Internet users across the world conduct tens of thousands of Google searches per second. Those looking for persecution can find it if they are looking. In fact, although these days most discussions about politically‐motivated content moderation focus on conservatives we shouldn’t forget that the political left has its own bias claims. For example, in 2017 the Chairperson of the International Editorial Board of the World Socialist Web Site claimed in a letter to Google and Alphabet executives that “Google is manipulating its Internet searches to restrict public awareness of and access to socialist, anti‐war and left‐wing websites.” Twitter’s purported “shadowbanning” of conservatives in July of 2018, really a problem with the site’s search feature, also affected the hosts of the left‐wing podcast Chapo Trap House.
- The studies suffer from either bad methodology or no methodology at all: One problem Silicon Valley critics have is that the firms they would like to research are understandably wary of sharing content moderation data or details of their search algorithms. The burden of proof is on the party making the positive claim and the lack of data to analyze makes definitively making the argument that Silicon Valley companies engage in anti‐conservative censorship difficult. Nonetheless, some have tried. For example, AIBRT’s Robert Epstein, who Sen. Ted Cruz (R-TX) invited to testify before the United States Senate Judiciary Subcommittee on the Constitution, wrote a paper arguing that Google’s search results in the months leading up to the 2016 presidential election favored Hillary Clinton over Donald Trump. As I explain in my C. Boyden Gray Center paper, Epstein’s study examined search results from fewer than one hundred people in twenty four states and did not control for the fact that Google search results can change depending on the device used and the location of the person making the query. Epstein also rejected data from users with Gmail addresses and used crowd‐sourcing to determine bias. Such a paper should not be used by activists as evidence of anti‐conservative bias. The methodology yields a study that cannot provide useful results. In Epstein’s defense, while his methodology is flawed it is at least a methodology. Most claims of anti‐conservative bias are based on nothing more than the anecdotes I mentioned above.
I am sure that many readers are not convinced by 1) or 2) and continue to maintain that left‐wing bias in Silicon Valley has created a social media environment hostile to conservative views. To those readers I only ask that you consider what an “unbiased” social media landscape would look like, and if they have a theory about what an “unbiased” Google search result or an “unbiased” Twitter would look like to let me know (my email: email@example.com). The absence of a conservative outline of what an “unbiased” social media platform would look like has been conspicuous by its absence in the ongoing content moderation debate.
Responses to this alleged bias reveal a misunderstanding of law and have resulted in policy proposals that would increase government power over speech and (ironically) result in less conservative online content.
Congress passed Section 230 of the Communications Decency Act in 1996. Written by Reps. Chris Cox (R-CA) and Ron Wyden (D-OR), the law aimed to solve the “Moderator’s Dilemma” that had emerged in the early 1990s thanks to a few court cases dealing with alleged defamatory content posted by third party users of online services.
In Cubby v. CompuServe (1991), a federal judge found that CompuServe — an online service provider — could not be held liable for alleged defamatory content posted by a third party newsletter. The judge found that CompServe, which did not moderate the newsletter’s automatically uploaded content, was the distributor of the third party content not unlike a news vendor or public library rather than a publisher.
A few years later, a New York Supreme Court judge considered a similar case, Stratton Oakmont v. Prodigy Services (1995), and came to a different holding. In that case, the judge ruled that Prodigy Services, which did engage in content moderation, could be considered the publisher of third party content.
These two cases gave rise to the “moderator’s dilemma”: either engage in content moderation and be considered the publisher of third party content or take a hands‐off approach to third party content and be treated like a distributor. Both options are awful for anyone interested in building a website where users can upload content. A hands‐off approach might save an Internet service provider from legal liability, but it also means users may upload pornography, photos of murders, and images of animals being tortured to death. To anyone trying to make a family‐friendly service, such a move has obvious downsides. Taking steps to remove such content may be desirable, but not if it means that your service is considered the publisher of third party content. Such an approach would require review every third party piece of content before it appeared on the service. Needless to say, this effectively prohibits an Internet with useful third party upload functions and services.
Section 230 solves the moderator’s dilemma by providing interactive computer services with a sword and a shield. The shield, Section 230(c)(1), explicitly rejects the Prodigy holding, stating that “[n]o provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.” This protects websites big and small from being considered the publisher of the vast majority of content users upload. The sword, Section 230(c)(2), states that interactive computer services cannot be held civilly liable for “any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected.”
In sum, interactive computer services are not publishers of the vast majority of third party content (see Section 230 (e) for exceptions) and they cannot be held liable for removing third party content, even if that content is protected by the First Amendment.
Section 230 is at the heart of the present policy debates surrounding allegations of online anti‐conservative bias. If you have been paying attention to these debates you may have heard a few false claims about the law that I’d like to address before turning to serious issues about proposed Section 230 changes.
Section 230 is Not a Subsidy
You may have seen some commentators portraying Section 230 as a subsidy. The argument usually goes something like this: Absent Section 230 Silicon Valley companies would have to spend millions (if not billions) of dollars on lawsuits, so we should consider Section 230 to be a multi‐million (or multi‐billion) dollar subsidy. This view is flawed for a few reasons.
The most obvious issue with this view is that it distorts what a subsidy is. Subsidies are usually found in bills, where handouts to particular industries or companies are outlined. Section 230 does no such thing. There are laws on the books that protect individuals and companies from lawsuits, but they’re not described as subsidies. See for example anti‐SLAAP legislation.
Leaving aside the semantic issue, this view is particularly misguided when used by conservatives concerned about market incumbents such as Google and Facebook. Section 230 protects those interested in competing against Google and Facebook from having to raise money for potential lawsuits, which will only increase in number as their competing service grows in popularity. For instance, it’s unlikely the video‐sharing site BitChute, which portrays itself as a YouTube competitor, or Gab, a social media site for the far‐right, would ever have got off the ground if their founders had to prepare for a deluge of lawsuits. Full30, a firearms friendly alternative to YouTube, is more vulnerable to suit than YouTube. If conservatives are worried about platform bias, they should recognize that changes to Section 230 will fall hardest on competitors. It would be more accurate (though still flawed) to describe Section 230 as a subsidy for social competitors rather than a subsidy for market incumbents.
Portraying Section 230 as a “handout” or “special privilege” for “Big Tech” is also misleading, as there is no size requirement for Section 230 protections. Section 230 protects the comments section on a retiree’s baking blog as much as it protects Facebook.
Section 230, Platforms, and Publishers
Before turning to misunderstandings of free speech writ large I want to briefly address one other Section 230 myth: the “platform” v. “publisher” distinction.
It is not uncommon to see someone claim that when a social media company engages in content moderation it doesn’t deserve Section 230 liability protection because it is “acting like a publisher” or not behaving like a public forum.
Anyone who reads Section 230 will see that Section 230 does not state that the act of moderating content makes an interactive computer service a publisher. Section 230 merely states that an interactive computer service is not the publisher of most third party content and is free to moderate content.
In fact, there are times when a publisher can enjoy Section 230 protection. For example, The Wall Street Journal is a newspaper owned by a corporation. If The Wall Street Journal publishes an oped that defames someone the victim can sue the author of the oped and The Wall Street Journal. As well as publishing a dead tree newspaper, The Wall Street Journal also runs a website, which includes a comments section. This comments section is an “interactive computer service” covered by Section 230. If someone posts defamatory content in the comments section the victim can sue the user who posted the comment, but not The Wall Street Journal.
It is important to also note that Section 230 does not protect an interactive computer service from its own content. Remember, Section 230(c)(1) states (italics mine): “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.” Facebook might not be liable for the vast majority of content Facebook users post, but Facebook could be held liable for information posted by Facebook employees on Facebook’s newsroom page.
Similarly, Twitter is liable for its fact‐checking of President Trump’s tweet. Such content does not enjoy Section 230 protection. Nonetheless, many Section 230 critics, including Sen. Josh Hawley (R-MO) took advantage of the news of the day to use Twitter’s fact‐check as an opportunity.
Silicon Valley as the Public Forum
Conservative activists unhappy with Silicon Valley’s content moderation have sometimes turned to First Amendment law to argue that Facebook, Youtube, Twitter, etc. should be considered a “public forum.” This approach is confused and reveals a fundamental misunderstanding of how free speech works.
Some Section 230 critics, including Sen. Cruz (R-TX), have argued that the basis of Section 230 is that interactive computer services would be a “neutral public forum.” I am unsure of the genesis of this neutrality myth, but one candidate for a point of origin is the text of Section 230. It is true that Section 230(a)(3) reads: “The Internet and other interactive computer services offer a forum for a true diversity of political discourse, unique opportunities for cultural development, and myriad avenues for intellectual activity.”
This is true today, as it was in 1996, and does not impose any obligation on interactive computer services to implement a politically neutral content moderation policy.
More recently, Sen. Marco Rubio (R-FL) took to Twitter to falsely claim that “The law still protects social media companies like @Twitter because they are considered forums not publishers.” This is not true. Reps. Cox and Wyden wrote Section 230 in part to encourage content moderation, to allow Internet companies to remove content they considered incompatible with their business model.
Perhaps the most notable proponent of the view that social media constitutes a public forum is Dennis Prager, who runs the conservative education website Prager University. Prager University — commonly referred to a PragerU — has its own website as well as a YouTube channel with almost 2.5 million subscribers. The channel is best‐known for its short animated videos, which aim to educate viewers on issues of interest to conservatives. PragerU sued Google, which owns YouTube, alleging that YouTube is a state actor because it performs a public function.
PragerU’s complaint stems from the fact that YouTube placed many of its videos in Restricted Mode, which prevents users who have opted into the setting from viewing such videos. Although some activists may like to point out that this is evidence of anti‐conservative bias, the fact is — as the trade association NetChoice pointed out — a far higher percentage of videos on liberal channels such as The Young Turks have their videos in Restricted Mode.
A Ninth Circuit panel of three judges, two of whom were nominated by Republican presidents, made short work of PragerU’s argument:
[…]PragerU argues that YouTube is a state actor because it performs a public function. It is true that a private entity may be deemed a state actor when it conducts a public function, but the relevant function “must be both traditionally and exclusively governmental.” Lee v. Katz, 276 F.3d 550, 555 (9th Cir. 2002). This test is difficult to meet. It is “not enough” that the relevant function is something that a government has “exercised … in the past, or still does” or “that the function serves the public good or the public interest in some way.”
Halleck, 139 S.Ct. at 1928–29. Rather, the relevant function must have been “traditionally the exclusive prerogative of the [s]tate.” Rendell‐Baker v. Kohn, 457 U.S. 830, 842 (1982) (internal quotation marks omitted). Indeed, “[w]hile many functions have been traditionally performed by governments,” Flagg Bros., Inc. v. Brooks, 436 U.S. 149, 158 (1978), the lean list of the “very few” recognized public functions includes “running elections,” “operating a company town,” and not much else, Halleck, 139 S.Ct. at 1929 (internal quotation marks omitted); see, e.g., Terry v. Adams, 345 U.S. 461, 468–70 (1953) (elections); Marsh v. Alabama, 326 U.S. 501, 505–09 (1946) (company town). The relevant function performed by YouTube — hosting speech on a private platform — is hardly “an activity that only governmental entities have traditionally performed.”
Nor is YouTube anything like the company town at issue in the Supreme Court case Marsh v. Alabama (1946). In that case, the Court ruled that the First Amendment applies in towns that are owned by a company. As the Ninth Circuit opinion correctly states:
In Marsh, the Court held that a private entity operating a company town is a state actor and must abide by the First Amendment. […] But in Lloyd Corp. and Hudgens, the Court unequivocally confined Marsh’s holding to the unique and rare context of “company town[s]” and other situations where the private actor “perform[s] the full spectrum of municipal powers.” Lloyd Corp., 407 U.S. at 569; see also Hudgens, 424 U.S. at 518–20. YouTube does not fit the bill. Unlike the company town in Marsh, YouTube merely operates a platform for user‐generated video content; it does not “perform all the necessary municipal functions,”
Analogizing YouTube to Chickasaw, Alabama (the company town at issue in Marsh) is a conceptual mistake. There are many places for Internet users to find video content. YouTube is hardly the only website offering video‐sharing services. To analogize social media writ large to Chickasaw is to make an even bigger conceptual mistake. Unlike Chichasaw, the Internet’s social media environment isn’t run or regulated by one company. The social media environment is international and includes large and small companies. There is a difference between popular firms and monopolies. Those who do not like Twitter can use Gab, and those who are not fans of Google have alternatives.
Complaining about being ejected from the most popular party in town doesn’t entitle you to demand that cops show up to force the host to let you in. Hosts of less popular parties are free to open their doors to you.
Don’t Forget the First Amendment
Fortunately the First Amendment to the U.S. Constitution is a formidable barrier to President Trump and his planned commission and executive order doing anything of policy relevance, as Mike Masnick correctly noted at TechDirt:
the government simply cannot force platforms to moderate in a certain way. That would violate the 1st Amendment. So even if a panel is formed, it couldn’t actually do anything to change things, beyond just being an annoying pest.
The freedom of speech protects those who host speech as much as it does those who wish to speak. At its core, the freedom of speech relies on the freedom of association. No country in the world is better than the United States when it comes to protecting the freedom of speech. Unfortunately, many conservatives seem intent on undermining the United States’ enviable position as the world’s foremost defender of free speech.
Some conservatives claim to be supporters of free speech and the First Amendment all while seeking to undermine both and ignoring the fact that Section 230 is better than the First Amendment. As Santa Clara University School of Law Professor Eric Goldman has noted, Section 230 provides defendants with valuable procedural benefits and litigants with more predictive certainty over potential suits than they do not have under the First Amendment.
Section 230 Reform Will Result in More Government and Less Speech
Some conservative activists have convinced themselves that the largest social media companies are engaged in a campaign of anti‐conservative discrimination. There is no compelling evidence of such a campaign, but even if there were it would not justify amending Section 230 or establishing a White House commission.
Proposals to rein in Big Tech vary. Sen. Hawley would like to make Section 230 protections for large interactive computer services contingent on politically neutral content moderation, as determined by the Federal Trade Commission (FTC). Other proposals, such as Rep. Gosar (R-AZ)’s “Stop the Censorship Act” aims to limit content moderation to “unlawful content,” leaving social media sites free to those who wish to engage in trolling or to spread of racist and pornographic content. Rep. Gosar’s bill is at least informative insofar as it reveals that many Section 230 critics fundamentally misunderstand the law. Whoever drafted Rep. Gosar’s bill overlooked the fact (as Prof. Eric Goldman explained) that courts routinely find that Section 230’s shield, not its sword, preempts suits over content removal.
These kinds of proposals would result in more government regulation of speech. A scheme such as Sen. Hawley’s that relies on four out of five FTC commissioners deciding which large social media sites deserve Section 230 protection is one we should expect to be abused for partisan aims. In addition, if implemented the anti “Big Tech” bill would result in unintended casualties. Sen. Hawley’s bill would only affect interactive computer services with 30 million U.S. monthly average users, 300 million global monthly average users, or global revenue of at least $500 million a year. Services that have nothing to do with social media but nonetheless rely on Section 230 would be affected. Did Sen. Hawley’s staff not consider that Dropbox, Etsy, and Amazon would also be dragged into this mess?
If Internet companies were faced with a regulatory environment in which content moderation is reserved to legal content they are back in the “moderator’s dilemma” I outlined above. The result would either be less speech across the board — harming conservatives and everyone else with views to share — or an Internet that increasingly looks more like PornHub and 8chan.
As conservatives continue to argue for infringements on free speech be sure to keep an eye out for the common myths I’ve outlined. Social media companies are not modern public fora. Section 230 is not contingent on politically neutral content moderation, is not a subsidy or special privilege, nor does it make a distinction between publishers and platforms.
It is bizarre to consider that “conservative” proposals these days include commissions, executive orders, and bills that would grow the power of an alphabet soup agency, allow for more spread of pornography, and infringe on free speech. But that’s where we are. Welcome to modern American conservatism.