1. We See

Media provides citizens with information to make voting decisions and stimulates interest in elections. In fact, there is vast literature that explains how and to what extent voters learn from a variety of media sources including newspapers, TV, Radio, internet and so forth. In that regard, the media reinforces political interest and voting intentions.

It cannot be reasonably challenged that citizens do use and engage with the media to learn about issues and topics relevant to the political arena, such as candidates’ traits, proposals, or careers.⁵

On this matter, the evolution of technology and the development of different digital tools have gradually allowed citizens to obtain information and make up their minds before an election.

It is well known that the internet plays a significant role in people engaging with news and information around the globe. For example, in the United States, it is estimated that 37% of American adults get their news from the internet, followed by Radio (27%) and print newspapers (20%). The internet is only second to TV (57%).⁶

While TV is still a major source of information, Internet and digital platforms are thriving. In Germany, for example, it is reported that TV was the main source of news in 2013 with 82%; by 2018 it was still the major force of information, but it dropped down to 74%.⁷ Additionally, Internet has taken a paramount place as an information provider. Both social media and search engines are the relevant venues in which users consume digital news. It is important to highlight that despite the rise of social media platforms such as Twitter or Facebook, search engines remain a strong channel through which people gain access to online information.⁸ Internet users seem to trust search engines - mainly Google - almost blindly to the point that they would first question “their own ability to search properly before doubting the effectiveness of Google’s algorithm.” This applies mainly to young users, whose first experience with the internet came with Google itself.⁹

It shall not be ignored that Google, as a US-based company, is the main player in the field, with approximately 90% of the market share worldwide, leaving the remaining percentage as follows:

As shown, online search -which Google has taken over- is crucial for citizens when obtaining information regarding politics, elections, and other public issues. At the same time, search engines offer a full spectrum of digital marketing tools and techniques and so-called “ad products”, designed for political campaigns, which this paper will discuss in the next section.

The quality and content of digital information may vary across political issues and regions. This allows different kinds of media bias to take place, as well as the manipulation of public opinion.

One may observe these biases in barefaced totalitarian or undemocratic regimes. However, even in the absence of evident manipulation, market players can still exercise a tendentious influence on public opinion strengthening voters’ predispositions by “pervasive selection and filtering”.¹⁰

In this context, scholars point stress a fundamental difference when it comes to users analyzing search results¹¹:

The “visible area” is what users see immediately in the search results page, whereas all information below, which is not directly visible, is the “scrolling area.” That is, the first few results (the visible area) are the elements relevant for the user, the ones that they may engage and interact with, while the rest (the ones remaining in the “scrolling area”) would rarely be of interest to the user.

The visible area may be formed of both organic results and sponsored results. The first ones come from the search engine’s algorithm, while sponsored results are links coming from advertising, which is being paid for and appear in the results page.

However, previous studies suggest that Internet users rarely engage with sponsored results and prefer to open organic links instead. For example, one study shows that in more than 80% of the searches, participants would go first to the results identified as organic, suggesting that users tend to ignore sponsored results.¹² At the same time, usual practice points out that web users prefer to adjust their search terms instead of moving to the next results page.

Having said that, this article now goes on to explore what does the law say with regards to this issue, and what are the legal implications of the results that come out of a search query. It has been demonstrated that altering the organic results in the visible area may affect the users’ impressions, perspectives, and thoughts regarding different aspects of life, including politics and democracy. The impact of such manipulation in the light of the existing literature may be grouped up into six explanatory points.

1) As it was previously explained, higher-ranked links attract more clicks and, consequently, users tend to spend more time on websites associated with those higher-ranked search results. This is so because people trust search engines to assign higher ranks to the results best suited to their needs.¹³

On this note, Robert Epstein introduced the term Search Engine Manipulation Effect (SEME) to refer to the alteration of the ranking in the search results and its impact in elections outcomes. Epstein carried out five different experiments in two countries (three in the United States and two in India), uncovering the power SEME, and reporting the following results:

2) Changing voting preferences and shifting elections is the ultimate outcome of SEME. People’s thoughts and actions can change -according to Epstein- by simply modifying the order in which the results appear on results page. Basically, the first page of results and the order in which it is structured influence the voters’ minds.

In addition to the above, Epstein’s work maintains that most users show no awareness that they are viewing biased search results. Given that biased search rankings can sway the voting preferences of undecided voters without their awareness and without fair competition from opposing candidates, SEME appears to be a powerful tool for manipulating elections in this century.¹⁴ This hypothetic scenario implies that a search engine may deliberately favor one candidate, one political party or doctrine, and diminish others. While this may seem for some unfeasible or unrealistic, it is certainly possible and overall -and until now- mostly legal.

3) In the context of competition law, it must be highlighted that there are several critiques towards Google, to the extent that it may be favoring its own products, manipulating its rank to benefit itself.¹⁵ But in the context of democracy and politics, Google can technically and legally support a candidate, a campaign, or an ideology.

4) For example, if a political candidate from Tijuana, Baja California considers that Google is manipulating its search results since the first three links that appear in the screen’s visible area are news discrediting him, they would find little success in challenging that. First, because the said candidate would likely face several complications in terms of jurisdiction and international private law,¹⁶ since Google is an American company incorporated and based in California, United States. Even if it has presence worldwide, it claims to be mostly governed by American law and its courts. Secondly, even if the plaintiff manages to bring a case before a court of law, whether in Mexico, the United States or elsewhere, he would not be able to find out why the search algorithm is ranking those results in that specific order. This is because the algorithm protected under trade secret, and Google, being a private for-profit company, is under no obligation to publish or share its industrial property. Thirdly, the candidate would then realize that Google -as any other search engine- is perfectly entitled to support one political figure or another, as it enjoys freedom of speech. The American constitutional framework protects the so-called “Google Speech,” which can favor one political candidate or another. These two last elements will be discussed in more detail in the following paragraph.

5) Regarding patent protection, scholars point out that well-known search engines have always refused to fully disclose the methods and techniques by which they score and rank their search results. Naturally, Google’s search algorithms are perhaps the most famous of these secrets. Critics of search bias claim that these unknown formulas lead to a “black box effect”: users do not know neither the method through which search results are computed prior to any assigned ‘bias’ nor the adjustments search engines make purposely.¹⁷

While it is true that Google patented its first PageRank algorithm back in the late 1990’s -and as in any patent procedure, the entity seeking for patent protection needs to disclose information and after a certain period of time, that protection expires-, it is also true that, as technology has evolved, Google has made several updates and changes to the algorithm. It also makes use of other complementary algorithms to improve its search results. For instance, PageRank was the original algorithm used for Google queries, but later on the company introduced new elements to optimize their results, such as Panda, Penguin, Hummingbird and so forth. In other words, trade secret law protects all subsequent adjustments Google makes to the original algorithm.¹⁸

As Oren Bracha and Frank Pasquale point out, the proper balance between secrecy and transparency is indeed a major normative challenge in the search engine context, because on the one hand certain degree of secrecy is of legitimate interest, but at the same time -both authors stress- society has a strong interest in transparency and accountability. In various contexts people are becoming aware of the troubling aspects of a “black box society,” in which private firms are basically uncontested when locking away information despite a strong public interest in disclosure. However, the idea of a regulatory regime for Search Engines does not seem to have attracted the mainstream of legal scholars.¹⁹

6) The judicial experience has not changed this view. There are three cases relevant for this analysis: 1) Search King v. Google, 2) Langdon v. Google et al., and 3) KinderStart.com v. Google.

In Search King v. Google, the complaint was about intentional and malicious de-ranking of specific websites in Google’s search results. The court upheld Google’s argument to the extent that its PageRank system represents speech protected by the First Amendment, and “any act aimed at knowingly and intentionally modifying the ranking of websites is a legitimate expression of the freedom of speech.”²⁰

Consequently, for the Court, PageRank consists merely of “opinions on the relevance of certain websites.” As such, there is no way to prove that the ranking for a given website is false -all of it is subjective- and this is how the Court concluded that Google was entitled to “full constitutional protection.”²¹

Then, in Langdon v. Google et al., the situation was slightly different. In this case, the plaintiff Christopher Langdon was running two websites, one exposing the alleged fraud perpetrated by North Carolina Officials, and another highlighting atrocities executed by the Chinese Government. He sought online advertisements for his websites in different search engines, such as Google and Yahoo. They all refused to run them. Langdon argued that such refusal translated into a violation of his First Amendment right. The Court ruled in favor of Google and the other search engines since they were entitled to editorial discretion in “deciding whether to publish, withdraw, postpone, or alter content” as they see fit. Even more, the Court clarified the legal nature of Google and other search engines: a private, for-profit company that “uses the internet as a medium to conduct business.”²²

Finally, in KinderStart.com v. Google, the ranking for KinderStart dropped to zero. It brought similar claims as the ones previously discussed, which failed to prove Google’s responsibility. KinderStart claimed that Google’s search engine was a public forum because everyone online could access Google’s website (or any number of thousands of other websites having a “Google Search Box” as provided by Google) and everybody use its engine without any payment or charge. According to KinderStart, Google had willfully and continuously dedicated the engine for public use and public benefit. As previously mentioned, the Court did not uphold this claim, basing its reasoning on the lack of precedents to support such contentions. Even more, the question of when a search engine may eventually become a public forum has remained unsolved.²³

As can be seen, there is an increasing need to regulate this complex phenomenon. But acknowledging the need of rules and principles is not the end of the story. Where should that regulation come from? From the market, statutory law, or government regulatory authorities? The idea of market discipline sounds attractive for scholars supporting the laissez faire, laissez passer economic model. It can be summarized as follows:

Fortunately, market forces limit the scope of search engine bias. Searchers have high expectations for search engines: they expect search engines to read their minds and infer their intent based solely on a small number of search keywords. Search engines that disappoint (either by failing to deliver relevant results, or by burying relevant results under too many unhelpful results) are held accountable by fickle searchers. There are multiple search engines available to searchers, and few barriers to switching between them. As a result, searchers will shop around if they do not get the results they want, and this competitive pressure constrains search engine bias. If a search engine bias degrades the relevancy of search results, searchers will explore alternatives even if searchers do not realize that the results are biased.²⁴

This idea would be undisputed if the market had diverse competitors providing search engines. Unfortunately -as it was shown in the first part of this article- the data proved that one specific search engine holds the vast majority of the market share. As Bracha & Pasquale highlight:

The market discipline argument is based on two key premises: robust competition in the search market and users’ responsiveness to abuse. Unfortunately, both premises are highly problematic […] It is unclear whether search engines fall under the strict definition of a natural monopoly, but they exhibit very similar characteristics. [Additionally] In many, if not most cases, consumers lack both the incentive and even the ability to detect such manipulation or determine its reasons. Given the lack of transparency of the search algorithms, search consumers simply cannot reverse engineer the hundreds of factors that go into a ranking, and they have little incentive to compare dozens of search results to assess the relative efficacy of different search engines.²⁵

This leads us to the concluding remark that market forces alone cannot deal with the risk of search engine manipulation and prevent major harms to democracy. Nevertheless, this issue cannot be addressed alone but rather as part of a wider context in which digital platforms can affect our thinking, our political views and civic engagement. The next two aspects are even more complex and require a deep analysis and reflection.

2. We Think

Digital platforms know what users do, and who they are. They also manipulate their consumer habits: what to buy, where to go, what to watch, and so forth. Lilian Edwards states that in the last two decades we have seen a major change in online marketing, since it moved from broadcasted ads, where millions of people would see the same content delivered by email spam or banner ads on websites, to the more complex targeted ads based on users’ behavior (Online Behavioral Advertising, or OBA).²⁶ As Edwards summarizes: the theory goes that since these ads are “tailored to individual desires,” recipients are more likely to read them, to click through to actual websites and engage with that content.²⁷

For some it may not be completely clear how the “targeting” and “tailoring” take place. Where do Big Tech companies gather information about users, to then analyze it and then capture their attention? In this regard, it is appropriate to bring to this discussion Shoshanna Zuboff’s explanation on the matter:

Nothing is too trivial or ephemeral for this harvesting: Facebook ‘likes,’ Google searches, emails, texts, photos, songs, and videos, location, communication patterns, networks, purchases, movements, every click, misspelled word, page view, and more. Such data are acquired, datafied, abstracted, aggregated, analyzed, packaged, sold, further analyzed, and sold again. These data flows have been labeled by technologists as “data exhaust.” Presumably, once the data are redefined as waste material, their extraction and eventual monetization are less likely to be contested.

Subjectivities are converted into objects that repurpose the subjective for commodification […] Populations are the sources from which data extraction proceeds and the ultimate targets of the utilities such data produce.²⁸

This new economic logic, that Zuboff has dubbed as “Surveillance Capitalism”, brings with it a new tool for political advertising: micro-targeting, which usually involves, as explained above, monitoring people’s online behavior, and using the collected data, to catch the user’s attention and display political advertisements tailored to their views.

Additionally, Jacob Silverman sheds light on the power of digital platforms and their political effects. He claims that social media increases the voting turnout. While increasing political participation seems plausible, it may also trigger serious concerns on how this knowledge might be repurposed if, for instance, Facebook/Meta would encourage people in some districts to vote while “saying nothing to others.”²⁹ This one company possesses vast power to sort the information people see and is capable to nudge them towards certain behavior. As Silverman enquires “could it influence the fate of elections, not to mention specific policies?” or even worse “would we ever know if it did?”³⁰

One piece of literature is of particular interest for this analysis. In 2015, one year before the US Presidential Election that would trigger the infamous Cambridge Analytica scandal, Zeynep Tufekci recounted the following:

In 2010, a massive experiment (performed without being noticed by any of the sixty-one million subjects, none of whom were asked for permission), Facebook demonstrated that it could alter the U.S. electoral turnout by hundreds of thousands of votes, merely by nudging people to vote through slightly different, experimentally manipulated, get-out-the-vote messages. In this experiment, some messages geared toward Facebook users appeared stand-alone while other, more potent ones, were socially embedded, showing a “your friend voted” extra nudge.

[…]

Facebook has stated explicitly that they had tried to keep their 2010 experiment from skewing the election. However, had Facebook not published the results, and had they intended to shape the electorate to favor one candidate over another, the algorithmic gatekeeping enabled through computational agency would have been virtually unnoticeable, since such algorithmic manipulation is neither public, nor visible, nor easily discernible.³¹

Again, this problem reiterates the power that digital platforms hold and how they can affect not only what users see, but also trigger their behavior and therefore our collective political decisions.

There are some legal instruments across specific jurisdictions, particularly in Europe, to fight this possible abuse, such as the General European Data Protection Regulation (GDPR).³² It should be noted that the Mexican legal system has certainly similarities with the GDPR with one notorious difference: the legitimate interest exception. That is, the use of personal data of an individual without their express consent. The Mexican legislation on data protection requires express consent from the subject to process that data.

However, this is not precisely promising, since the ways in which “consent” is expressed in the digital age are evidently vague, abusive, and almost unnoticed. This is characterized by “pre-formulated declarations of consent,” or “clickwrap” contracts, that hide extensive privacy policies, well-known for “taking a disproportionate amount of time to go through and require reading comprehension abilities at university level,”³³ and this translates into millions of users clicking “I accept” or “Yes” on declarations of consent that they do not actually understand or make sense of, and thus feeding the economic logic of data extraction, commodification and personalized advertisements and publications that restrain their way of thinking.

3. We Say

Freedom of speech has changed with the advent of Internet, but more particularly with social media platforms, such as Facebook or Twitter. This has even led to the Supreme Court of Justice of the United States to say that “the vast democratic forums of the Internet in general, and social media in particular” are “the most important places […] for the exchange of views,” but as some scholars have pointed out, the digital age has also imposed users of social media an unprecedented regime of private censorship.³⁴

But how did we get here? It is necessary to bring Lilian Edwards’ explanation to understand this issue:

Content often carries with it legal liability, which may be civil or criminal […] how far should online intermediaries be responsible for this content, or contrarily, how far should responsibility stay with the original content author or provider?

The EC Electronic Commerce Directive 2000 (or E-Commerce Directive or ECD) alongside the Digital Millennium Copyright Act (DMCA) in the USA effectively established the ideas of limited liability and “notice and take down” (NTD) as the template for intermediary responsibility, a solution which had remarkable reach for over a decade and remains the pattern of many Organization for Economic Co-operation and Development (OECD) laws […]

Underneath it all, perhaps, lies the feeling, among both users and traditional state governments, that the giant “GAFAM” platforms (Google Amazon Facebook Apple Microsoft) now exert power greater than any prior private companies and possibly than some elected governments […]

Since 2000, automated content curation has become steadily more sophisticated and prevalent, especially in relation to copyright “take downs”, and with the rise of machine learning (ML), automated blocking has begun to look more feasible, even in heavily contextual and cultural areas, such as “fake news” and indecency, albeit with more successful application to images than text […] the public has arguably begun to comprehend that algorithms used by platforms to distribute, moderate and filter content can and do incorporate value-judgements and inherent bias, and can be used to apparently modify public emotions and opinions. The dawning horror at this covert manipulation of everything from buying choices to democratic decision making may have finally killed off once and for all the notion of platforms as innocent intermediaries.³⁵

In that regard, scholars Frederik Stjernfelt and Anne Mette Lauritzen provide an interesting perspective on this issue. They argue that today’s freedom of speech is under pressure. This is because Big Tech companies have amassed political power and they can lay down the rules of public conversation and access to information, leaving private individuals and even governments with little or no rights of defense against their decisions”.³⁶

In 2018, a group of human rights organizations and academics launched a set of core principles to limit and frame the increasing powers of Tech giants on moderation of user-generated content: the ^{Santa Clara Principles}.³⁷

Later that year, one of these Tech giants, Facebook/Meta, in a desperate attempt to cope with the increasing demands for greater accountability and transparency to users about how it controls the flow of speech in the web, decided to launch an external independent appellate body: The Oversight Board.

The idea behind the board is to give users a way to challenge content removal decisions on Facebook or Instagram. More specifically, “if you have already requested that Facebook or Instagram reviews one of its content decisions and you disagree with the final decision, you can appeal to the board”.³⁸

Nevertheless, there are some issues that trigger concern about the Oversight Board’s functioning: I) independence, and II) self-regulation. As to the first, the Oversight Board is run by a Facebook/Meta-funded trust that exercises administrative powers, such as “the ability to enter into Board member contracts and service agreements, remove and appoint members and staff issue payment and compensation, and […] research expenses”³⁹, therefore, Members of the Board -the ones who deliberate which content remains online- are factually being paid by Zuckerberg’s company. While some may argue that funds come from Facebook/Meta but it is the Board, a legally and technically independent body, the one who manages the payroll, the truth is that there is also a lack of intellectual independence, since the only method of external input for selecting Board members “is a public portal through which members of the public, Facebook, and Board members can submit candidate recommendations,”⁴⁰ leaving no room for external candidates proposed directly by human rights organizations, activists, or academics. In relation to the second point, the Board is Facebook/Meta’s own creation to self-control its own decisions, that is, it is trying to solve by itself the extensive critiques regarding abuse on content moderation practices and criteria, establishing its own sort of judicial branch. In a certain way, Facebook/Meta is just transferring the power of decision from its corporate executives to the Members of the Board, all within the atmosphere of its own corporate governance.⁴¹

In a very interesting critique, ^{Aldo Iannotti della Valle} explains that the Oversight Board arose as a para-jurisdictional body in the context of the “erosion” characterized by the State -as a political entity- losing jurisdictional power over the cyberspace, allowing these Big Tech companies to evolve from self-regulation to self-jurisdiction schemes.⁴²

Regardless of the launch of Oversight Board, the truth is that Big Tech corporations exercise power that affect the way we express ourselves in the digital age and it is still unclear how to build an appropriate framework that allows these companies to function but at the same time limits their abuse of power.

This resonates with Giovanni De Gregorio’s extensive research on Digital Constitutionalism, since he states that online platforms, while establishing the standard of free speech and shaping democratic culture on a global scale, disclose little information about their practices and procedures on content moderation, making it all “opaque or lawless.” While there are other interesting proposals -whose analysis exceeds the research limits of this article-, such as the Manila principles or the Internet Governance Forum Dynamic Coalition on Platform Responsibility, users still must deal with “discretionary and voluntary mechanisms,” since there is no binding force on online platforms.⁴³