Topic: Telecom, Internet & Information Policy

Misleading Project Veritas Accusations of Google “Bias” Could Prompt Bad Law

Tomorrow, the Senate’s Judiciary Committee’s Subcommittee on The Constitution will hold a hearing on Google’s alleged anti-conservative bias and “censorship.”  In a video released last month, James O’Keefe, a conservative activist, interviews an unnamed Google insider. The film, which has been widely shared by conservative outlets and cited by Sen. Ted Cruz (R-TX) and President Donald Trump, stitches a narrative of Orwellian, politically-motivated algorithmic bias out of contextless hidden camera footage, anodyne efforts to improve search results, and presumed links between unrelated products. Although the film’s claims are misleading and its findings unconvincing, they are taken seriously by lawmakers who risk using such claims to justify needless legislation and regulation. As such, they are worth engaging (the time stamps throughout this post refer to the Project Veritas video that can be viewed here).

Search algorithms use predefined processes to sift through the universe of available data to locate specific pieces of information. Simply put, they sort information in response to queries, surfacing whatever seems most relevant according to their preset rules. Algorithms that make use of artificial intelligence and machine learning draw upon past inputs to increase the accuracy of their results over time. These technologies have been adopted to improve the efficacy of search, particularly in relation to the gulf between how users are expected to input search queries, and the language they actually use to do so. They are only likely to be adopted to the extent that they improve the user’s search experience. When someone searches for something on Google, it is in the interest of both Google and the user for Google to return the most pertinent and useful results.

Board game enthusiasts, economics students, and those taking part in furious public policy debates over dinner all may have reasons to search for “Monopoly.” A company that makes it the easiest for such a diverse group of people to find what they’re looking for will enjoy increased traffic and profit than competitors. Search histories, location, trends, and additional search terns (e.g. “board game,” “antitrust”) help yield more tailored, helpful results.

Project Veritas’ film is intended to give credence to the conservative concern that culturally liberal tech firms develop their products to exclude and suppress the political right. While largely anecdotal, this concern has spurred hearings and regulatory proposals. Sen. Josh Hawley (R-MO) recently introduced legislation that would require social media companies to prove their political neutrality in order to receive immunity from liability for their users speech. Last week, President Trump hosted a social media summit featuring prominent conservative activists and conspiracy theorists who claim to have run afoul of politically biased platform rules.

The film begins by focusing on Google’s efforts to promote fairer algorithms, which are treated as attempts to introduce political bias into search results. The insider claims that while working at Google, he found “a machine learning algorithm called ML fairness, ML standing for machine learning, and fairness meaning whatever they want to define as fair.” (6:34) The implication being that Google employees actively take steps to ensure that Google search results yield anti-conservative content rather than what a neutral search algorithm would. Unfortunately, what a “neutral” algorithm would look like is not discussed.

Although we’re living in the midst of a new tech-panic, we should remember that questions about bias in machine learning and attempts to answer them are not new, nor are they merely a concern of the right. Rep. Alexandria Ocasio-Cortez (D-NY) and the International Committee of the Fourth International have expressed concerns about algorithmic bias. Adequate or correct representation is subjective, and increasingly a political subject. In 2017, the World Socialist Web Site sent a letter to Google, bemoaning the tech giant’s “anti-left bias” and claiming that “Google is “’disappearing’ the WSWS from the results of search requests.”

However, despite the breathlessness with which O’Keefe “exposes” Google’s efforts to reduce bias in its algorithms, he doesn’t bring us much new information. The documents he presents alongside contextless hidden camera clips of Google employees fail to paint a picture of fairness in machine learning run amok.

One of the key problems with O’Keefe’s video is that he creates a false dichotomy between pure, user created signals and machine learning inputs that have been curated to eliminate eventual output bias. The unnamed insider claims that attempts to rectify algorithmic bias are equivalent to vandalism: “because that source of truth (organic user input) has been vandalized, the output of the algorithm is also reflecting that vandalism” (8:14).

But there is little reason to presumptively expect organic data to generate more “truthful” or “correct” outputs than training data that has been curated in some fashion. Algorithms sort and classify data, rendering raw input useful. Part of tuning any given machine learning algorithm is providing it with training data, looking at its output, and then comparing that output to what we already know to be true.

Knight Institute v. Trump

The U.S. Court of Appeals for the Second Circuit has affirmed a district court’s ruling that Donald Trump may not block Americans from viewing or responding to his tweets. In an opinion largely mirroring that of the district court, Judge Barrington D. Parker determined that Trump’s use of his account in his capacity as President of the United States rendered it a limited public forum. As such, Trump’s decision to block the accounts of individual critics amounted to government viewpoint discrimination, rather than a private exercise of associational rights.

The case hinged upon whether or not Trump used his Twitter account in an official capacity. Initially, I expressed skepticism, writing;

The question turns upon our understanding of the purpose and character of Trump’s Twitter presence. Is @realDonaldTrump simply the private Twitter account of a man who happens to be president, or does it constitute a designated public forum, as asserted by Jaffer?

When Knight Institute v. Trump was first filed in June of 2017, @realDonaldTrump looked more like a leftover campaigning tool than an instrument of policy promulgation. Trump had inherited @POTUS, the official presidential account, from Barack Obama, and to the extent that Trump used his personal account to conduct official business, his proclamations seemed untethered from the positions officially taken by his administration.

However, over the past two years, as Judge Parker’s opinion notes in granular detail, @realDonaldTrump has been repeatedly employed to announce the policy positions, and Trump’s tweets are popularly understood to be the official positions of his administration. President Trump has used @realDonaldTrump to announce the withdrawal of U.S. troops from Syria, the selection of Dr. Christopher Waller for the Federal Reserve Board, and, just this past weekend, to reveal that the administration aims to re-include a citizenship question in the 2020 census. Parker also cites the White House’s designation of Trump’s tweets as “official statements,” and other government agencies’ affirmation of the official nature of Trump’s tweets, writing; “we note that the National Archives, the agency of government responsible for maintaining the government’s records, has concluded that the President’s tweets are official records.” Given this evidence, it would seem that @realDonaldTrump is far from a private, personal account, engaging the usual First Amendment prohibitions on content-based restrictions in public fora.

While no one in American politics utilizes Twitter quite so adroitly as Donald Trump, this decision will nevertheless bind other politicians and state officials who make use of social media, particularly those who comingle public and private functions within the same accounts. The decision cites Packingham, but only to establish that speech on social media receives no less First Amendment protection than other speech transmitted via other mediums. While the outcome of Knight Institute v. Trump prevents Donald Trump, and other public officials, from making use of Twitter’s block feature, it has no bearing on  ongoing debates about the limits of private moderation or recent attempts to categorize platform access as a “civil right”.  

One Year Later, The Harms of Europe’s Data-Privacy Law

The European Union’s General Data Protection Regulation (GDPR), which went into effect just over a year ago, has resulted in a broad array of consequences that are expensive, unintended, or both. Alec Stapp reports at Truth on the Market:

GDPR can be thought of as a privacy “bill of rights.” Many of these new rights have come with unintended consequences. If your account gets hacked, the hacker can use the right of access to get all of your data. The right to be forgotten is in conflict with the public’s right to know a bad actor’s history (and many of them are using the right to memory hole their misdeeds). The right to data portability creates another attack vector for hackers to exploit.

Meanwhile, Stapp writes, compliance costs for larger U.S.-based firms alone are headed toward an estimated $150 billion, “Microsoft had 1,600 engineers working on GDPR compliance,” and an estimated 500,000 European organizations have seen fit to register data officers, while the largest advertising intermediaries, such as Google, appear to have improved their relative competitive position compared with smaller outfits. Venture capital investment in Euro start-ups has sagged, some large firms in sectors like gaming and retailing have pulled out of the European market, and as of March more than 1,000 U.S.-based news sites were inaccessible to European readers.

More in Senate testimony from Pinboard founder Maciej Ceglowski via Tyler Cowen:

The plain language of the GDPR is so plainly at odds with the business model of surveillance advertising that contorting the real-time ad brokerages into something resembling compliance has required acrobatics that have left essentially everybody unhappy.

The leading ad networks in the European Union have chosen to respond to the GDPR by stitching together a sort of Frankenstein’s monster of consent, a mechanism whereby a user wishing to visit, say, a weather forecast is first prompted to agree to share data with a consortium of 119 entities, including the aptly named “A Million Ads” network. The user can scroll through this list of intermediaries one by one, or give or withhold consent en bloc, but either way she must wait a further two minutes for the consent collection process to terminate before she is allowed to find out whether or it is going to rain.

This majestically baroque consent mechanism also hinders Europeans from using the privacy preserving features built into their web browsers, or from turning off invasive tracking technologies like third-party cookies, since the mechanism depends on their being present.

For the average EU citizen, therefore, the immediate effect of the GDPR has been to add friction to their internet browsing experience along the lines of the infamous 2011 EU Privacy Directive (“EU cookie law”) that added consent dialogs to nearly every site on the internet.

On proposals to base legislation in the United States on similar ideas, see Roslyn Layton and Pranjal Drall,

Don’t Believe Too Much in Democracy, Mark Zuckerberg

Everyone dislikes the social media companies these days. Conservatives (that is, Republicans) decry alleged bias against their views. Liberals (that is, Democrats) complain about hate speech online and market domination. And that’s just the start for both sides. Of course, both sides believe “something must be done,” ranging from federal agencies controlling the platforms to breaking up the companies. Both see government as the solution for what is thought to ail social media. In his own way, Mark Zuckerberg agrees. He could not be more wrong.

Engineers and economists often focus on the tradeoffs inherent in a project. Zuckerberg has emphasized the tradeoff between the values of voice (speech) and safety (or protection from speech). At a recent appearance, he said most people do not want private companies making such fundamental tradeoffs by themselves. He continues, “I think we would be better off if we had a more robust democratic process setting the rules on how we want it to arbitrate and draw some of the tradeoffs between a lot of these values that we hold dear.”

What is a “more robust democratic process”? It could mean a process where majorities, highly organized minorities, or elected and unelected officials determine the tradeoff between speech and safety. They might do it indirectly as with Sen. Josh Hawley’s proposal to empower the Federal Trade Commission with life-or-death authority over tech companies and thereby, over their policies. All of us, including Zuckerberg, have little desire for government without constraints.

We live in a republic. The rule of the people comes with restraints on political power, not least protection for “the freedom of speech,” which places it, mostly, outside the reach of government decisions. The courts protect speech by interpreting and enforcing the First Amendment. They generally favor free speech with a few limited exceptions.

Zuckerberg wants democracy to help determine the tradeoff between speech and safety online. He may believe public involvement might accord legitimacy to online content moderation at least compared to efforts by private companies.

But what could the federal government actually do?  The courts could apply the First Amendment to the tech platforms by overturning or ignoring precedents. They would invalidate parts of Facebook’s Community Standards including rules against hate speech. If other parts of the republic – say, Congress – decided to draw a line between acceptable and unacceptable speech online, the courts would be obligated to strike down such rules. Even if the courts allow Congress to act, one side or the other of our highly polarized polity would likely see a win or a loss in that outcome leading to acrimony and distrust rather than consent and legitimacy. Government actions in the tech space cannot foster legitimacy, not now at least. 

Of course, the courts have generally found that the First Amendment does not apply to private companies like social media. If that policy status quo remains valid, we would be back to industry self-regulation. What’s a tech mogul to do?

Facebook in fact is moving forward in creating an appeals board for its content moderation. Zuckerberg notes, “We’re starting this as a project just for Facebook. But over time I could see this expanding to be something that more of the industry joins.” In other words, self-regulation just by Facebook might not work, but a private effort across companies might. Keep in mind also that we are in early days with such self-regulation; the Facebook board (or an industry-wide board) does not yet exist so it hardly can be judged a failure. Could they gain legitimacy? Perhaps. Might it fail for many reasons? Of course. But the board is an attempt to deal with immense and surprising challenges that have been evident since at least November 2016. The board deserves a chance.

We know the other path forward. The concrete alternatives are the courts enforcing the First Amendment in private forums or Congress, state officials, and the president drawing lines between permitted and suppressed speech, lines that the courts then invalidate (we hope). Neither option seems likely or desirable. The leaders of social media may regret the political and social dangers and complexities of content moderation. The tradeoff among values is hard, no doubt. But their task cannot be avoided. No one else, least of all government, can or should do the job.

For more on the limits of government regarding online speech, see my recent policy analysis.

Hawley Contra Reagan

You don’t need a weatherman to know which way the wind blows…

-Bob Dylan, a long time ago

The winds of illiberalism are blowing strongly here and abroad and in both major political parties. What had seemed to be gusts and squalls signifying little prior to 2015 now threatens a major storm. Sen. Josh Hawley’s Ending Support for Internet Censorship Act, introduced yesterday, seems to be the latest potential disaster of this hurricane season.

Sen. Hawley won election to the Senate decrying bias against conservatives by online platforms. He now proposes to act to “correct” this presumed bias. Forty years ago, his job would have been easier. The federal government claimed ownership of the airwaves and the power to license their use by broadcasters. The Federal Communications Commission (FCC) awarded licenses to broadcasters that promoted the public interest not least in the “fairness” of their political coverage. Government officials used the leverage created by licensing to harass and censor political speech.

The federal government does not own the internet, so a new FCC seems unlikely. In fact, Congress has favored the private over the government in regulating the internet. In 1996, Congress decided to immunize internet companies against the actions of their users and to empower them to remove content for various reasons, including it being “objectionable.” The platforms could remove such material without liability “whether or not it was constitutionally protected.” Congress, not the courts, decided the First Amendment did not apply to private forums for speech.

Sen. Hawley and his allies argue Section 230 subsidizes social media. They also say that the companies originally agreed to moderate their platforms in a politically neutral way in exchange for such subsides. In fact, Congress decided social media, unlike publishers, could not control ex ante the speech of their users. Holding platforms to a publishing standard could encourage too much moderation of speech to avoid liability, thereby restricting the growth of this new industry. Congress also empowered the companies to remove speech to fight obscenity on the new medium. Section 230 has nothing in common with Congress deciding to subsidize growing (or not growing) various crops.

However flawed, the subsidy argument does get Sen. Hawley to a licensing regime for social media. Subsidies should be granted in exchange for something, and why not the original “political neutrality” Sen. Hawley claims the companies promised and failed to deliver? The license this time is called an “immunity certification” granted by the Federal Trade Commission for two years to large internet content providers. This certification establishes that an internet company “does not moderate information provided by other information content providers in a manner that is biased against a political party, political candidate, or political viewpoint.”

The meaning of those terms may matter less than the process established by the bill to certify this “political neutrality.” The companies must convince four out of five members of the FTC that their content moderation is not politically biased. That fourth vote will decide whether companies operating social media platforms are granted Section 230 immunity.

Absent that certification, a social media company could continue in business. It might avoid the liability related to suppressing speech by letting everything on the platform. But the companies don’t do that now, primarily for business reasons: many users would be repelled by a platform permitting extremism. Another response to losing “immunity certification” would be moderating too much; content moderators might suppress any speech that could lead to liability and without protections, this suppression itself would lead to more lawsuits. The result would be at best a boring service hostile to speech and again, unattractive to users. In other words, losing certification might well do lasting damage to the economic value of a platform. To avoid such losses, a company would have to do whatever is necessary for to gain that fourth vote (and certification) at the FTC. We should not give that much power over business and speech to a federal agency.

Sen. Hawley’s bill seeks to undermine an older American conservatism. Not long ago, the Reagan legacy spoke powerfully to conservatives. Of course, the Reagan administration ended the Fairness Doctrine and in general, supported markets free of government regulation. Some now believe that times have changed and the ideals of 1980 no longer should guide conservative thinking. If that was “then,” Sen. Hawley is definitely “now.” But if conservatives are turning their backs on timeless ideals like free markets and free speech, what could conservatism mean beyond political necessity and the rage of transient majorities?

One final point. Sen. Hawley’s bill might lead to a good outcome: in invalidating the law, the Supreme Court might recognize clearly that social media curation deserves strong First Amendment protections. On the other hand, we also might end up with a Red Lion decision for the internet, that is, a decision supporting state oversight of speech and business on social media. Far better for Congress to take up its constitutional duties with this bill.

Other free speech advocates reject this licensing of social media for many good reasons (see Eric Goldman’s collection of links). This debate is just beginning. We can thank Sen. Hawley for making its stakes abundantly clear.

The Fairness Doctrine Was Terrible for Broadcasting and It Would Be Terrible for the Internet

Skepticism of big tech companies is surging on both sides of the political spectrum, from Democratic Senator Elizabeth Warren calling for breaking up Amazon to Republican Senator Josh Hawley advocating rules that would prohibit online viewpoint discrimination. This wave of techno-progressivism finds its latest expression in Slate journalist April Glaser’s article, “Bring Back the Golden Age of Broadcast Regulation.”

Glaser argues that the problems of internet discourse—eg hate speech, haphazard content moderation, and conspiracy peddling—are so trenchant that government intervention is warranted. She calls for applying the rules that once governed mid-twentieth century radio and television broadcasting to the internet, the most important of which was the mandate that broadcasting be done in the “public interest, convenience, and necessity” as laid out in the 1934 Communications Act. Inspired by that mandate, reform-minded progressives at the Federal Communications Commission (FCC) enacted the Fairness Doctrine in 1949, which required broadcasters to provide multiple points of view when discussing political disagreements.

Glaser’s proposal is light on details about how exactly broadcast rules would be adapted for the internet, but it is heavy on assurances that any such regulations would be “light-touch.” Those who worry that inviting the feds to just “do something” could lead to violations of free speech need not be concerned. As Glaser argues, “For decades, radio and television followed regulations—hardly heavy-handed ones—meant to ensure they served the information needs of their audiences and did not actively harm political discourse.”

That would be lovely, if it were true, but not a single part of that statement is correct. The belief that government regulation of internet content providers will be effective and minimally-invasive is rooted in a poor understanding of the history of broadcast regulation. That history actually suggests the opposite, that these regulations will be ineffective, highly-intrusive, and will create significant unintended consequences.

Why Section 230 Is Unstable

Everyone interested in social media should read Jeff Kosseff’s The Twenty-Six Words that Created the Internet. It provides an excellent history of Section 230, the legal foundation of social media. That might sound boring, but Kosseff makes it work by mixing stories and analysis without vitiating either. I agree with Kosseff that, problems notwithstanding, the benefits of Section 230 have outweighed its costs.

Given that, I don’t look forward to future editions of his book since they may document the “fall” of Section 230. In other words, the law is likely to be amended to limit the protections offered internet platforms. We have already seen changes meant to combat sex trafficking. Section 230’s most serious persistent vulnerability, however, comes from a mismatch between its statutory language and the larger world the law inhabits.

Our political world is divided between (let’s call them) Progressives and Conservatives. Progressives see society as a struggle between designated oppressor and oppressed groups. No one – not government, not the tech companies – should be neutral between these groups; everyone should favor the oppressed. Since the speech of oppressors is causally essential to the harms experienced by the oppressed, government and the platforms should suppress that speech to help the oppressed. Remember, no one is neutral in this endless struggle.

As it happens, given the extent of Progressive cultural authority, the language of Section 230 favors the Progressive cause. Here’s an edited version of Section 230(c)(2) that clarifies this point:

No provider or user of an interactive computer service shall be held liable on account of any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be…objectionable, whether or not such material is constitutionally protected…[emphasis added]

So if content moderators think speech offending the oppressed is “objectionable,” they can banish it from their service.