Fake News on the Internet: Actions and Reactions by Three Platforms. CELE Presentation to the UN Special Rapporteur on Freedom of Opinion and Expression.
Fake News on the Internet: Actions and Reactions from Three Platforms Submission to the Special Rapporteur on Freedom of Opinion and Expression, Irene Kahn, February 2021 CELE The Center for Studies on Freedom on Freedom of
Submission to the Special Rapporteur on Freedom of Opinion and Expression, Irene Kahn, February 2021
CELE
The Center for Studies on Freedom of Expression and Access to Information is an academic research center affiliated with the University of Palermo in Argentina. The Center provides technical and legal analysis on issues affecting this fundamental right, and since 2012 has been studying freedom of expression on the Internet as a specific area of research. The Center is a leading voice in the promotion and protection of freedom of expression at the national, regional, and international levels. This submission was prepared in response to the public call for input issued by the Special Rapporteur. According to the call, “The Special Rapporteur will seek to clarify how human rights law applies to disinformation, identify key issues that would benefit from further consideration by the Human Rights Council, and formulate recommendations to States and other stakeholders on the best way to tackle disinformation while protecting the right to freedom of opinion and expression.” CELE has studied and published on the issue of disinformation from various perspectives. The Center maintains a repository of laws and bills covering nine countries in Latin America that tracks, among other things, legislative efforts to address the phenomenon; and in 2020, it launched Letrachica.digital, a project that tracks changes to terms of service and community guidelines in real time. Both projects seek to understand how public and private regulations and content restrictions function and impact the exercise of freedom of expression, particularly in Latin America. This submission is based on a research project that tracked and documented the various actions taken by companies to address disinformation, especially as it relates to electoral issues. This submission addresses point 3 of the questionnaire, which reads:
What policies, procedures, or other measures have digital technology companies introduced to address the problem of disinformation?
To what extent do you find these measures to be fair, transparent, and effective in protecting human rights, particularly freedom of opinion and expression?
What procedures exist to address grievances and provide remedies for users, monitor the actions of the companies, and how effective are they? Further information, research, and analysis on this and other issues are available at www.palermo.edu/cele or by emailing cele@palermo.edu. We thank the Office of the Special Rapporteur for considering this submission. Introduction The manipulation of public opinion through lies has been a constant throughout history. However, disinformation as a modern phenomenon emerged recently in the context of electoral processes that took place amid a distorted public debate, in part due to the dissemination of false information on the Internet. Since Donald Trump’s election as President of the United States in 2016, the phenomenon has recurred in several electoral processes with varying intensity and impact. For many observers, online disinformation poses a serious threat to the future of democratic systems. Faced with this challenge, states around the world have responded in different ways. Some have sought to regulate the practice and impose on intermediary platforms—where disinformation thrives—strict obligations to moderate the content they allow. Others have sought to raise public awareness about the existence of disinformation campaigns designed to deceive. Companies, on the other hand, have sought to take action amid increasing—and often contradictory—pressures from government institutions, regulators, legislators, NGOs, and academics around the world. This is an executive summary of a report prepared by CELE that sought to track the actions taken by Google, Twitter, and Facebook regarding disinformation, covering the period from 2016 to 2020 with a special focus on Latin America. From our perspective, to understand disinformation and the actions taken by these leading intermediary companies, it is necessary to understand the historical context in which the recent phenomenon of disinformation has emerged. It is also helpful to understand how the problem and the proposed solutions to address it are part of a broader shift in public attitudes toward democracy and the Internet: while in the 1990s the Internet was viewed as a democratizing force that would topple dictatorships through the free flow of information, since the early 2000s a more pessimistic outlook has been slowly but steadily gaining traction among regulators worldwide. The Internet is no longer perceived as a democratizing force for good—it is often seen as a threat to democracy itself. This shift calls into question the legal standards governing the (non) liability of intermediaries, which until now have provided a solution to the problem of accountability in decentralized networks. This legal framework appears increasingly untenable. A. Background and Challenges In the beginning, as a decentralized network, the Internet was presented as the ideal space for the free flow of information and ideas of all kinds. This initial promise generated some optimism regarding its influence on the future of democracy: the Internet could serve as a democratizing tool in closed societies, since its decentralized nature would make it difficult for totalitarian governments to control their citizens. In this context, legal protections soon emerged for intermediaries that facilitated access to content produced by third parties. Section 230 of the Communications Decency Act of 1996 is the prime example of this approach. The goal of that regulation was to encourage innovation and facilitate, rather than hinder, the flow of information. Over time, however, some caution emerged. Authoritarian governments soon learned to use the Internet to increase their control over their citizens. Starting in the late 1990s, several authors warned that the Internet was heading down a troubling path that needed to be corrected. This marked the beginning of a slow but steady shift toward pessimism that would place Internet companies on the “wrong side” of the good-versus-evil divide—a trend that coincided with a pattern of ownership concentration. Various new intermediaries emerged during those years and became increasingly powerful. Google (1998), Facebook (2004), YouTube (2005), and Twitter (2006) began to concentrate a large portion of the traffic and initiated a process of re-centralization of the network. Their terms of service and community guidelines function as de facto regulations governing what information is deemed acceptable or unacceptable on each of these platforms. The power that Internet platforms actually wield became, over time, the reason why various stakeholders began to demand greater responsibility and accountability. This has pushed platforms to do much more than simply “facilitate” the free flow of information. Since the 2016 U.S. presidential election, the major platforms appear to be on the defensive. On a regular and somewhat chaotic basis, they have announced and implemented changes to their platforms aimed at combating the phenomenon of disinformation; they have testified at formal hearings before legislative bodies around the world; they have published studies and provided information as part of their efforts to offer some transparency; they have supported and led programs to strengthen quality journalism and the “verification of information.” Companies are being pushed to adjudicate what is true and what is not—a duty that used to fall on citizens in democratic societies. Suddenly, we have introduced an intermediary actor—central to the flow of information—as an arbiter of truth that cannot be influenced through democratic procedures. These changes pose new challenges: Do we maintain our commitment to broad standards of freedom of expression and reaffirm that the best response to false information is the free competition of ideas? Or should this paradigm be revised? If so, what should replace it? B. Actions and Main Findings The actions we have identified are linked to the temporal nature of the phenomenon of misinformation and to shifts in attitudes toward the Internet along the optimism–pessimism spectrum. These actions sought to address the demands of users, civil society, and governments regarding the central role that companies have assumed in the flow of information on the Internet. In a sense, these actions are a consequence of their own success. The report identified hundreds of actions: some were announced and implemented, others were only partially implemented, and finally, a significant number of actions could not be verified, at least in Latin America. We classified them into four different categories.
Awareness-Raising Initiatives. These are aimed at raising awareness about disinformation campaigns or promoting quality journalism as a presumably effective response to the phenomenon. They involve partnerships with other actors (e.g., fact-checking organizations), educational campaigns, digital and media literacy, etc. Most of these actions took place in the United States, Canada, and the United Kingdom, but they have also spread—albeit with significant limitations—to other parts of the world, including Latin America. Partnerships with fact-checkers appear to have been the preferred response by platforms to address the phenomenon of disinformation. This type of partnership does not pose significant problems from the perspective of freedom of expression. However, many challenges remain, such as limited impact on less-informed or less-educated individuals, who—in turn—are more likely to consume false information.
Changes to Code. These actions modify the platforms’ code, altering recommendation mechanisms and content visibility. Platforms are increasingly introducing algorithm-based timelines that—supposedly—seek to provide users with content that is more “relevant” to them, determined by consumption patterns that the platforms themselves track and exploit. Within this category, we found actions aimed at detecting false information, either with the help of users (e.g., by expanding reporting mechanisms) and/or through artificial intelligence. We also found actions aimed at providing more context for information, especially on issues where disinformation was detected and deemed a problem, such as elections and the COVID-19 crisis that began in late 2019. These efforts have faced significant challenges due to their heavy reliance on context and issues related to scalability and replicability. Additionally, there are disparities in the information available to counter misinformation or conflicting information across different jurisdictions. This, in turn, creates new challenges regarding predictability and uniformity in the application of social media rules and self-regulation standards. Companies have also sought to give a more powerful voice to “professional” journalists, media outlets, and fact-checkers by prioritizing their content and visibility. This approach, which relies on the platforms’ own free speech rights, is based on a problematic underlying assumption, however: that traditional media are trustworthy and serve as tools to combat disinformation rather than amplifiers of such a phenomenon. As multiple recent studies have shown, this is not necessarily true. Finally, companies have also sought to address the problem indirectly by restricting “behaviors” (rather than content) that disinformation agents use to advance their campaigns, such as the use of bots and the massive distribution of content on encrypted platforms. This provides scalability and opportunities for automation that other measures do not.
Changes to policy and moderation measures. This category encompasses actions aimed at changing the rules that define what speech is acceptable on each platform. These are actions that may impact their business model, as they often involve self-imposed limitations without necessarily changing the platforms’ code. Until now, these actions were driven by companies’ refusal to control the information their users share through their services—a stance rooted in the optimistic paradigm of the 1990s. However, increasing pressure on companies to exercise their moderation powers more decisively has led them in that direction, especially regarding electoral issues and—following the COVID-19 crisis—matters of public health. These changes are significant and demonstrate that companies are flexible, with the ability to adapt their policies to the growing pressures to play a more relevant role in controlling the information that flows through their services. Still, this adaptability and susceptibility to pressure also reveal the arbitrariness and instability of the rules, the lack of transparency and predictability in their application, and the increasing susceptibility of self-regulation to various social and political pressures, to the detriment of freedom of expression.
Transparency and public relations. These are actions aimed at establishing the companies’ position both internally and in the face of external pressures: on the one hand, from political sectors that can regulate them through legislative changes; on the other, from social sectors that exert pressure on them. Advertisers who view certain practices with concern are another group that pays close attention to how platform policies evolve. These actions generally demonstrate that companies are willing to address external concerns and accommodate often contradictory demands. They also show companies taking a stance regarding their role in the free flow of information—a role they have often, until now, been reluctant to assume as “arbiters of truth.” Furthermore, under pressure, companies have often provided information at the request of authorities, such as, for example, regarding the influence of Russian intelligence operations on their platforms. Finally, within this category, we should note that the creation of procedures and mechanisms to develop adequate implementation policies and criteria that satisfy both users and regulators is one of the main developments in recent years. In April 2018, Facebook published its internal moderation criteria, expanded its internal appeals process for its decisions, and promised responses within 24 hours. By then, Twitter had already launched its Trust and Safety Council and recently announced an expansion of its powers in line with the body announced by Facebook. In 2018, Facebook announced the creation of the Oversight Board, a council of prominent figures tasked with helping the company make sound moderation decisions consistent with the principles of freedom of expression; the board issued its first decisions in January 2021. C. A Changing World The pessimistic shift regarding the Internet and the re-centralization of the network call into question the legal solution reached so far to address content produced by third parties. While the non-liability of intermediaries was taken for granted until a few years ago, that no longer seems to be the case. The increasing pressure on major platforms to exercise their moderation powers more vigorously has led them down the path of public forums—they no longer seem to enjoy an absolute prerogative to decide which speech is allowed on their services and which is forbidden. External criteria, including international human rights standards, seem to be playing an increasingly significant role. This is happening, however, in a context in which regulatory action by state actors appears only as a threat. Aside from the laws adopted in Germany and France, no country in the West has addressed the challenge of disinformation through legislative action. Our research suggests that companies act under pressure but within the exercise of their own self-regulatory powers. In that sense, our report suggests that the pattern of findings documented in CELE’s 2017 report is becoming more pronounced.
Platforms have embraced a more robust moderation role, a development that the COVID-19 crisis appears to have encouraged. Over the past two years, companies have moved forward with measures such as information localization, government media tagging, blocking of foreign media advertisements, suspension of political advertising, and the labeling of political content, among others. These actions take place against a backdrop of uncertainty. There is still much we do not know about disinformation campaigns: how they operate, who is behind them, and what their effects are. Platforms’ actions are carried out within that context. The push toward the automatic detection of disinformation is bold precisely because it is a difficult issue to identify, one that is fraught with deep disagreement regarding what it is and what the appropriate response of a democratic society should be. The actions that provide more “context” operate within the traditional paradigm of freedom of expression and, for now, seem to be more prominent than direct acts of censorship, which appear to be measures of last resort—such as, for example, the deplatforming of Donald Trump in January 2021. Viewed as a whole, the actions analyzed in our report seem to follow one another somewhat chaotically in an effort to respond to growing pressures and unease.
The development of corporate bodies that are more or less “independent” from the companies themselves—such as Twitter’s Trust and Safety Council or Facebook’s Oversight Board—appear to be promising governance innovations insofar as they are capable of achieving some degree of legitimacy in the eyes of the external actors who exert pressure on companies. It is, however, too early to tell whether their efforts will be successful.
Our analysis shows that several measures announced in the United States or in Europe have not yet been implemented in Latin America. There is an “implementation gap” that is problematic in and of itself.
Transparency regarding moderation and other actions is still lacking. Transparency reports are difficult to analyze and present information in formats that are hard to read or overly aggregated. As a general rule, we are unaware of specific cases of moderation except for those covered by the mass media. To address this gap, platforms should be open to studies conducted by independent actors and academics. At the same time, moderation rules should be clearer, and their application should be consistent. This scenario poses profound dilemmas regarding the role of intermediaries in the flow of information on the Internet. The processes of concentration and re-centralization that characterized the pessimistic shift in attitudes toward the Internet cast doubt on whether the principle of intermediary liability exemption is sufficient to address multiple current problems, ranging from disinformation to the algorithmic generation of new content. Regulatory innovation will occur, but we are less certain about the form it will take. From our perspective, the current trend toward self-regulation is not likely to last long, but traditional, nation-state regulation will not be the answer either. With initiatives such as the Global Network Initiative in place, we see a scenario of co-regulation—such as the one recently described by Marsden, Meyer, and Brown—as a much more likely turn of events in the not-too-distant future. The trend toward “re-centralization” may also come under increased scrutiny if current antitrust investigations, both in the United States and in Europe, move forward. In a less-concentrated Internet and a more decentralized network—more similar to the original model—moderation would be less effective. Information would circulate in a somewhat more chaotic manner, and there would be no simple ways to exercise the control that these central actors are required to exert today. Although this is possible, it is also unlikely. Incentives appear to be aligned toward maintaining and protecting the central role that the actors analyzed here have achieved. For governments, it is easier to control the flow of information when there are central actors with the capacity to exercise control than when such actors are absent, when there are too many of them, or when they do not handle a significant portion of the traffic. The alignment of these interests with the private interests of some of the most powerful corporations in the world suggests that the current regulatory approach is designed to treat these platforms (and this level of concentration) as a given phenomenon. To the extent that this characteristic of the Internet persists, and to the extent that diverse actors continue to demand concrete actions from the platforms against threats to democracy perceived as serious, greater transparency regarding the criteria used for moderation seems a reasonable demand. In this context, platforms assume a role that is increasingly similar to that of “public forums.” This conclusion, we argue, is perhaps not only avoidable but also undesirable. By: Ramiro Álvarez Ugarte @ramiroau