檢視臉書事實查核機制:以2020美國總統初選為例 - 政大學術集成

全文

(1)International Master Program in International Studies National Chengchi University 國立政治大學國際研究英語碩士學位學程. ‧. ‧ 國. 學. An Examination on the 治 Effectiveness政 of Facebook Fact大立 checker Solutions: The 2020 U.S. Presidential Primaries sit. y. Nat. 檢視臉書事實查核機制:以 2020 n. er. io. 美國總統初選為例 al iv Ch. n U engchi. Student: Amy Yu-Lin Lin Advisor: Dr. Ming Lee. JULY 2020. DOI:10.6814/NCCU202001427.

(2) Acknowledgment I would like to express my gratitude to my thesis advisor Dr. Ming Lee, Professor of Department of Diplomacy, National Chengchi University, for his valuable guidance and advice throughout this research project. Without him, this thesis would never be accomplished. This appreciation extends to the very two professors of thesis committee, Professor Han-Pu Tung of National Taiwan University and Professor Ping-Kuei Chen of National Chengchi University, who offered kind suggestions with their academic insights.. 政治大 from parents. Their love encouraged me to pursue the master study and to go for my 立 Furthermore, like many graduate students, I am very much thankful for the support. ‧ 國. 學. dream. Finally, I would like to thank my dearest sweetheart, Nicolas, for his unconditional love, support, and encouragement to help me through all the obstacles in. ‧. writing this thesis.. n. er. io. sit. y. Nat. al. Ch. engchi. i Un. v. DOI:10.6814/NCCU202001427.

(3) Abstract The aim of this thesis is to examine the effectiveness of Facebook fact-checking program in terms of combating the spread of false news. First, this thesis would review and analyze the literature in relation to the topic of fake news in pursuits of laying out a comprehensive context with academic depths to discuss Facebook’s factchecking program and its weaknesses in later chapters. Next, it would present the general findings from the designed case study of 2020 Florida primaries and argue that Facebook’s fact-checking program exists fundamental flaws and can only serve as a contingency plan at best.. 立. 政治大. ‧. ‧ 國. 學 sit. y. Nat. n. al. er. io. Keywords: Facebook, fact-checking, fake news, false news. Ch. engchi. i Un. v. DOI:10.6814/NCCU202001427.

(4) Table of Content Chapter I: Introduction ...................................................................................................... 1 1.1. Research Motivation ............................................................................................ 1. 1.2. Research Background .......................................................................................... 2. 1.3. Propose of the Research ....................................................................................... 5. 1.3.1. Objectives of the Thesis .................................................................................... 5. 1.3.2. Research Question.......................................................................................... 11. 1.4. Research Methods .............................................................................................. 11. 1.5. Literature Review .............................................................................................. 14. 政治大 Chapter II: Understanding Fake News ............................................................................ 18 立 Scope and Limitation of the Research ............................................................... 17. 2.1. Its Origin and Drive ........................................................................................... 18. 2.2. Its Definition....................................................................................................... 20. 2.3. Its Outlets and Impacts ...................................................................................... 23. 2.4. Its Dissemination ................................................................................................ 25. ‧. ‧ 國. 學. 1.6. sit. y. Nat. io. er. Chapter III: Examination on the Effectiveness of Facebook Fact-checker Solutions: A. al. iv n C hengchi U Findings of FLDEMS ......................................................................................... 30 n. Case Study on the Florida Presidential Primaries ........................................................... 28 3.1 3.2. Findings of ROPF .............................................................................................. 33. Chapter IV: A Response to Facebook’s Fact-checker solution ........................................ 37 4.1. Analysis of Facebook’s Fact-checking Program ............................................... 37. 4.2. Suggestions for Facebook .................................................................................. 42. Chapter V: Conclusion...................................................................................................... 46 English Bibliography ......................................................................................................... 49 Chinese Bibliography ........................................................................................................ 56. i DOI:10.6814/NCCU202001427.

(5) List of Acronym CAS. Cambridge Analytica Scandal. CDC. Centers for Disease Control and Prevention. EJN. Ethical Journalism Network. FLDEMS. Florida Democratic Party. NIH. National Institutes of Health. RPOF. Republican Party of Florida. RSF. Reporters Without Border. UNESCO. United Nations Educational Scientific and Cultural Organization. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i Un. v. ii DOI:10.6814/NCCU202001427.

(6) Chapter I: Introduction 1.1. Research Motivation Inspired by the 2018 Netflix documentary, “The Great Hack,” I was drawn to. the big question of how humanity should deal with the “data right” in a digital era when “data” has become the most valuable asset on earth.1 The proliferation of data in recent decades has certainly prompted advocates of data ownership from reformers, tech tycoons, and academia (Tisne 2018). While most urge the possession of data as a fundamental Human right, some scholars, such as Tisne, argue that data ownership is a. 政治大. “flawed, counterproductive way of thinking about data” (Tisne 2018). The debate of. 立. data right, thus, carries on, and it appears to be compounded by the ubiquitous. ‧ 國. 學. information disorders on social media.. ‧. As I followed the documentary peeling off the scandal of Cambridge Analytica,. sit. y. Nat. I came to a sudden realization that the spread of fake news on social media could be. er. io. undermining the foundation of modern democracy. After all, when the intent of. al. iv n C h e n g cthehdocumentary by tainted information into doubts. Therefore, pushed me pondering i U n. information we receive is in question, it gets us putting our electoral behaviors affected. whether the current solutions delivered by social media to address the threat of fake news has performed effectively or not.. 1. “Regulating the internet giants - The world’s most valuable resource is no longer oil, but data.” The Economist, May 6, 2017.. 1 DOI:10.6814/NCCU202001427.

(7) 1.2. Research Background Human beings, social animals as we know, need interactions. With the. advancement of digital and mobile technologies, social media has obviously become a preferable platform where social needs can be fulfilled without face-to-face communication. The changes brought by social media are unprecedented; it reshaped the way people interact and communicate, and it even challenged the way how information used to be produced and consumed. A survey conducted in 2016 by the Pew Research Center noted that 6 out of 10 American adults absorb news on social. 政治大. media (Gottfried and Shearer 2016). So what would this mean when social media. 立. become the hotbeds of fake news?. ‧ 國. 學. The case of the 2016 Cambridge Analytica Scandal (CAS) would be perhaps one. ‧. of the most explicit examples illustrating concerns over the hazards brought by the dissemination of fake news on social media. According to one of the former employees. y. Nat. io. sit. of Cambridge Analytica, Brittany Kaiser, personal data was collected and studied. n. al. er. without authorization to identify “swing voters” so that personalized political. Ch. i Un. v. advertisements could be directed to them. 2 Christopher Wylie, another whistleblower,. engchi. told the press, “We exploited Facebook to harvest millions of people’s profiles. And built models to exploit what we knew about them and target their inner demons. That was the basis the entire company was built on.”3 Such revelations could be interpreted into a daunting message: Individuals are in high risks of fake news that serves benefits to certain political or ideological purposes,. 2. K. Amer and Noujaim J, “The Great Hack,” Netflix, 2019.. 3. Carole Cadwalladr, Emma Graham-Harrison, “Revealed: 50 Million Facebook Profiles Harvested for Cambridge Analytica in Major Data Breach,” The Guardian, March 17, 2018.. 2 DOI:10.6814/NCCU202001427.

(8) and with the “help” of big data algorithms, it seems fake news is the newly-typed precision missile able to target any mind it wishes to bombard with. Undeniably, CAS manifested the vulnerabilities of modern democracy in the Internet era when the integrity of the information that voters access online is in question. In fact, the intent of fake news on public opinions has been confirmed to be not theoretical anymore. The Canadian research center, CitizenLab, has recently published a study on a disinformation campaign (led by pro-Iran activists) against Saudi Arabia, the U.S., and Israel, suggesting that falsehoods have been spread on social media in an. 政治大. attempt to shape individuals’ opinions.4. 立. What is happening in the digital realm is complex, and there are many different. ‧ 國. 學. actors and tactics involved to manipulate the public internet sphere, particularly around. ‧. elections. During the 2016 U.S. presidential election, there have been accusations of Russian hackers distributing sensitive information about the Democratic candidate,. y. Nat. io. sit. Hillary Clinton. 5 These Russian hackers have allegedly resulted in creating negative. n. al. er. publicity for Hilary Clinton by aggressively sharing stolen emails with contents that. Ch. i Un. v. may sabotage her reputation. What’s more, messages claiming that Hillary Clinton had. engchi. died were circulated on social media, and even in some key battlegrounds, Democrat voters were targeted by the messages indicating that the date of the election had changed (Morgan 2018).. 4. Gabriel Lim and al., “Burned After Reading, Endless Mayfly’s Ephemeral Disinformation Campaign,” The CitizenLab, May 14, 2019.. 5. Luke Hardin, “What we know about Russia's interference in the US election,” The Guardian, December 16, 2017.. 3 DOI:10.6814/NCCU202001427.

(9) Similarly, in Europe, days before the second round of the French presidential elections, the #Macronleaks, which involved the U.S. alt-right, took place.6 The emails of the candidate Emmanuel Macron were hacked, and fake news suggesting Macron had connections to offshore financial accounts were spread. These examples of “hacking, leaking and the insertion of fake information into troves of dumped documents online” demonstrate the severity of foreign interference that attempts to compromise voter’s opinions through the propagation of controversial information, and it is very likely that these cross-border manipulations are “not only in. 政治大 promote a particular world view” (Morgan, 2018). 立. state-sponsored efforts but also through motivated individuals and groups wishing to. ‧ 國. 學. Whatever causes may serve behind the behaviors to intervene in an election, the. ‧. fact that generating fake news could be a potential source of income emphasizes how immense the issue could turn out to be. For instance, Macedonian teenagers have made. y. Nat. io. sit. money by creating fake news during the 2016 U.S. presidential election campaign. n. al. er. (Allcott and Gentzkow, 2016). For every “click” on the content, it means a paycheck to. Ch. i Un. v. the fake news provider. It rounds up, creating an incentive to produce a massive amount. engchi. of fake news over social media so as to share more to earn more. Even though people might not be motivated by the willingness to influence an election or a referendum, the result of spreading fake news for economic purposes may still be the same menacing. Overall, the value of data has ushered in a new landscape where most of the power and wealth are concentrated in the few hands of internet tycoons hoarding over zillions of personal data, and their business model is paving the way for fake news providers to. 6. Akshat Rathi, “Quelle surprise! Alt-right Americans and internet bots are spreading #MacronLeaks,” QUARTZ, March 7, 2017.. 4 DOI:10.6814/NCCU202001427.

(10) make advertising revenues (Morgan 2018). Hence, there is more and more pressure on social media to develop solutions to fight against fake news as it has ruined their image and sparked concerns upon their involvement in disinformation campaigns. Facebook, in the wake of CAS, launched a fact-checking program involving non-profitable third parties aiming to mitigate the fallout of fake news.7 These third-party fact-checkers would flag posts and articles once proving the content to be false, and Facebook would then lower its visibility in the News Feed. Likewise, for those tech-giants that accommodate the world’s most personal data, Twitter, YouTube, they are also taking steps to crack down on the distribution of fake news.8. 立. Propose of the Research. 學. ‧ 國. 1.3. 政治大. 1.3.1 Objectives of the Thesis. ‧. sit. y. Nat. Since this thesis is motivated by the growing concerns regarding fake news as a. io. al. er. potential threat to democracy among policymakers and scholars, it is crucial to have a. n. closer look at the current solutions developed by social media as the validity remains unclear.. Ch. engchi. i Un. v. This thesis proposes to examine the effectiveness of Facebook fact-checker solution as Facebook is by far the largest social networking site, reaching 67% of U.S. adults, followed by YouTube with 48% of U.S. adults (Gottfried and Shearer 2016), and not to mention it was once the storm eye of fakes news scandals. Putting these facts together, the necessity to inspect Facebook’s claimed efforts in tackling fake news is self-evident.. 7. “Facebook to begin flagging fake news in response to mounting criticism,” the Guardian, December 15, 2016.. 8. Daniel Funke, “Four Major Tech Companies Take New Steps to Combat Fake News,” Poynter, July 12, 2018.. 5 DOI:10.6814/NCCU202001427.

(11) To examine the effectiveness of Facebook’s fact-checker solution, one must be well-aware of how it works, to whom it applies, and why it is designed that way. In a nutshell, the vetting mechanism that Facebook devised is a three-step program: identifying, reviewing, and action. 9 First, Facebook claims it employs a variety of signals to detect potential false information, and its third-party partners are encouraged to screen content on their own. For example, Facebook collects feedback from the user’s News Feed to see what contents have been flagged to be potentially false. Other user patterns, such as. 政治大 machine learning models are adopted to predict misinformation by accumulating the 立 comments, also serve as indicators to judge the credibility of the contents. In addition,. ‧ 國. 學. data from fact-checkers.. rate the accuracy of the contents. The rating options include:10. Nat. sit. y. ‧. Next, once spotting suspicious posts or articles, fact-checkers would review and. “False: The primary claim(s) of the content are factually inaccurate. This. io. n. al. er. rating can apply to videos, images, articles and text-only posts.”. Ch. i Un. v. “Partly False: The claim(s) of the content are a mix of accurate and inaccurate, or the primary claim is misleading or incomplete. This rating can apply to videos, images, articles and text-only posts.”. engchi. “False Headline: The primary claim(s) of the article body content are true, but the primary claim within the headline is factually inaccurate. This rating applies only to articles.” Finally, if any videos, images, articles, and text-only posts were to be rated as “False,” “Partly False,” or “False Headline,” there are five actions ready to be taken:. 9. Facebook, How Our Fact-checking Program Works, March 23, 2020. https://www.facebook.com/journalismproject/programs/third-party-fact-checking/how-it-works.. 10. Ibid.. 6 DOI:10.6814/NCCU202001427.

(12) “reduced distribution, sharing warning, sharing notification, misinformation labels, and removing incentives for repeat offenders.”11 It should be noted that in order to decrease the visibility of a debunked content, Facebook moves the content lower in News Feed, not removing it. Now that the procedure of the program has been made clear, one may assume that it applies to all individuals, entities, and organizations using Facebook. However, in Facebook’s open statement, “posts and ads from politicians are generally not subjected to fact-checking.”12 leaving public wonder why only the politicians are immune to such. 政治大. scrutiny. According to Facebook, its approach is rooted in:13. 立. ‧ 國. 學. “(…) the fundamental belief in free expression, respect for the democratic process, and the belief that, especially in mature democracies with a free press, political speech is the most scrutinized. ‧. speech there is. Just as critically, by limiting political speech we would leave people less informed about what their elected officials are saying and leave politicians less accountable for their words”.. sit. y. Nat. er. io. Putting aside whether there are causal relations between limiting political speech. al. iv n C h e n gshould at first glance, suggesting political discourse i U any false or even malicious c h tolerate n. and leaving people less informed, a noticeable logical flaw stands out in their reasoning. contents in order to safeguard democracy. What they call for “political speech” is inherently flawed because it is built upon the skeptical notions that freedom of speech allows no restrictions on expressions, along with political discourse being already automatically-monitored in democratic regimes.. 11. Facebook, “How Our Fact-checking Program Works,” March 23, 2020. https://www.facebook.com/journalismproject/programs/third-party-fact-checking/how-it-works.. 12. Facebook, “Fact-Checking on Facebook: What Publishers Should Know”. https://www.facebook.com/business/help/182222309230722. 13. Ibid.. 7 DOI:10.6814/NCCU202001427.

(13) These perceptions Facebook possesses are not only rooted in pure ideals lack of evidence but also contradicted. Counter-arguments that “the effectiveness of the press as a check on power was significantly compromised” (Allcott and Gentzkow 2016) are not novel in the context of modern democracy, and in effect, a wake-up call for Facebook. If a well-functioned democracy is running through inspection on political speech via free press as Facebook believes, shouldn’t have it already shed some light to Facebook that the freedom of speech is bestowed with boundaries, because if freedom of speech is literally as free as it is, why would democracies bother to censor political discourse in the first place? The more one tried to get to the bottom of the. 政治大. reasoning; the more illogical and tangled one would find the strings of thoughts.. 立. ‧ 國. 學. Furthermore, one may argue that the freedom of speech should be more than tolerating controversial discourse. After all, true freedom does not come along without. ‧. limitations. If freedom were to be practiced without restrictions, chaos would likely be. sit. y. Nat. expected. And if Facebook believes by protecting the expressions of politicians is the. n. al. er. io. best way to keep people updated with the necessary information and in the mean time. v. to put politicians accountable, then why only protecting political speech? Is Facebook. Ch. engchi. i Un. implying other sources of speech are not as important to be taken into consideration to know about “their elected officials,” and that they do not need to be accountable for what they publish? Their argument can hardly make sense, and such narratives generate suspicions to their intentions of establishing the fact-checking program.. 8 DOI:10.6814/NCCU202001427.

(14) Nevertheless, Facebook includes politicians at every level, meaning “candidates running for office, current officeholders - and, by extension, many of their cabinet appointees - along with political parties and their leaders.”14 It further points out:15 “When a politician shares a specific piece of content - i.e., a link to an article, video or photo created by someone else that has been previously debunked on Facebook - we will demote that content, display a warning and reject its inclusion in ads. This is different from a politician’s own claim or statement. If a claim is made directly by a politician on their Page, in an ad or on their website, it is considered direct speech and ineligible for our third party fact checking program — even if the substance of that claim has been debunked elsewhere.. 政治大. Ads about social issues, elections or politics are held to a higher transparency standard on Facebook.”. 立. ‧ 國. 學. In other words, if the politicians were the creators of fake news, Facebook would spare them with mercy for the sake of free speech, but if they were “the sharers” of any. ‧. false content, Facebook would step in and take action. Obscured by the seemingly. Nat. sit. y. ridiculous logic, some observers started to doubt Facebook’s resolution to make. n. al. er. io. amendments and argue that if Facebook truly wishes to avert fake news and hate speech. i Un. v. from spreading, it should adjust its business model, which relies heavily on micro-. Ch. engchi. targeting for advertising profits.16 Not surprisingly, some journalists working as factcheckers for Facebook have terminated their partnership with it for they have lost trust in Facebook, pointing out Facebook “has repeatedly refused to release meaningful data. 14. Facebook, “Fact-Checking on Facebook: What Publishers Should Know”. https://www.facebook.com/business/help/182222309230722. 15. Ibid.. 16. Tenz Shih (施典志), Fangxin,Facebook bu hui weile duikang jia xinwen er zijue cai lu/Frederic Filloux (Rest assured, Facebook will not kill itself to fight against fake news / Frederic Filloux), Tuna Business Review, August 30, 2018.. 9 DOI:10.6814/NCCU202001427.

(15) about the impacts of their work” and basically used them as “crisis PR.” 17 Brooke Binkowski, a former managing editor of Snopes, a fact-checking site that has partnered with Facebook for two years, stated that Facebook is not “taking anything seriously … They clearly do not care.”18 In conclusion, if Facebook attempts to counter fake news through the implementation of a fact-checking program which already incurred skepticism and criticism without adjusting its business models, we have every reason to testify the very approach and to strive for its betterment and improvement, for it might not have enough. 政治大 aims to verify the validity of Facebook’s fact-checking program in terms of stopping 立. time for the humanity to address the fake news crisis before it escalates. Thus, this thesis. serve as a contingency plan at best.. The objectives of this thesis to be achieved are as follows:. Nat. sit. y. ‧. ‧ 國. 學. fake news from dissemination and argues that it exists fundamental flaws and can only. -. To provide possible policy suggestions based on the findings of the study.. al. iv n C h e social More importantly, to pressure to act accordingly based on the hi U n g cmedia n. -. er. To contribute to the literature about the evaluation of fact-checker solutions.. io. -. findings.. 17. Same Levin, “'They do not care': Facebook factchecking in disarray as journalists push to cut ties,” the Guardian, December 13, 2018.. 18. Ibid.. 10 DOI:10.6814/NCCU202001427.

(16) 1.3.2 Research Question To reach the above objectives, this thesis would have to address the following questions: Can Facebook fact-checking program work effectively in terms of preventing false news from spreading?. 1.4. Research Methods As this thesis is expected to evaluate the effectiveness of Facebook’s fact-checking. program, I propose to conduct a case study on the 2020 presidential elections in the U.S. 政治大. to observe whether Facebook has avoided fake news distribution successfully. Since. 立. four years ago Facebook was the center of the political vortex stirred with fake news. ‧ 國. 學. scandals, it only stresses the significance to focus the study on the context of U.S. ‧. elections, especially when the upcoming presidential elections is a unique occasion to see how fake news might spread (or not) again during political campaigns.. sit. y. Nat. n. al. er. io. The case study would lockdown to Florida, given the limited resource, research. v. labor, and time constraints. Knowing it is unlikely to cover all the activities from each. Ch. engchi. i Un. state, Florida was picked for the subject of the study because it is a “swing state.”19 As mentioned earlier in the case of CAS, swing states are the primary targets of fake news and disinformation campaigns during the previous elections. By assuming fake news would proliferate again, it is only logical to study one of the crucial swing states. The study on Florida would be carried out during the selected key event attached to the elections. The selected key event would be the primaries, in which presidential. 19. Tom Muse, “Which Swing States Are in Play in the 2016 Election?,” ThoughtCo Aug. 21, 2019.. 11 DOI:10.6814/NCCU202001427.

(17) candidates are elected by the parties. 20 Through primaries, states elect their delegates to the national party convention so the party could nominate their final candidate. Therefore, one can reasonably predict that fake news might materialize to influence the results of the primaries. Also, one important reason for choosing Florida is that its presidential primaries (of both Democrats and the Republicans) would fall on March 17, 2020, a date after Super Tuesday (March 3, 2020) has occurred.21 It is speculated that substantial changes or major tactics in political campaigns would take place after Super Tuesday, for the. 政治大 delegation) would be determined. It is highly probable fake news pinpointing certain 立. results of the primaries in 16 states (which is up to around 40 percent of the total. ‧ 國. 學. figures would be more conspicuous and be transmitted wilder.. ‧. How will the case study be executed now that the event and subject have been settled? This thesis would look into two Florida Facebook pages to detect any possible. y. Nat. io. sit. fake news dissemination activities in a given time: two weeks before and after the. n. al. er. primaries. The time designed in the study is mainly arbitrary with assumptions that 1). Ch. i Un. v. fake news would go spiral when the election days are near, and 2) prominent candidates. engchi. for both parties would come out after Florida primaries and fake news targets would be more precise and distinct.. 20. United States Government, “Presidential Primaries and Caucuses”. https://www.usa.gov/election#item-37162.. 21. Sarah Almukhtar, Jonathan Martin and Matt Stevens, “2020 Presidential Primary Election Calendar,” The New York Times, Updated June 12, 2020.. 12 DOI:10.6814/NCCU202001427.

(18) Following this train of thought, the Facebook pages of the Republican Party of Florida (RPOF) 22 and the Florida Democratic Party (FLDEMS) 23 are chosen for the study because they each represent the two major U.S. political parties (Republicans and Democrats). To date (until January 2020), RPOF is followed by 210,907 people while FLDEMS, 234,140. Duly noted that Facebook does not censor posts from politicians, and neither political groups, false contents shared by these Facebook pages, however, are still technically subject to the program. Once spotting any debunked false content shared by. 政治大 the evidence to argue that the program does not work effectively because it did not do 立 these Facebook pages that still remains unlabeled with a warning or intact, I would have. ‧ 國. 學. what it promised to do.. ‧. Finally, if necessary, I would continue to study the visibility; namely the number of time of a post being “liked” and “shared,” for “Like” and “Share” are displayed on. y. Nat. io. sit. users’ NewsFeed, which contributes to a significant part of the propagation of fake news.. n. al. er. It is thus essential to include the effect of such a parameter in order to examine the. Ch. i Un. v. effectiveness of the fact-checker solution as a whole. To sum up, the study would. engchi. scrutinize the contents posted by the selected pages from March 3 to April 1.. 22. See Republican Party of Florida’s Facebook Page https://www.facebook.com/FloridaGOP/. 23. See Florida Democratic Party’s Facebook Page https://www.facebook.com/FlaDems/. 13 DOI:10.6814/NCCU202001427.

(19) 1.5. Literature Review Over the past years, scholars have been focusing on the study of fake news due to. the impact it has brought. Research on fake news can be organized into different categories based on their goals and approaches. The first category of fake new studies consists of the research aiming to define fake news. Indeed, fake news are not something new or unprecedented. Even before the apparition of social media, traditional news media already showcased unverified or misleading information (Shu and al. 2017). However, with the rise of entertainment. 政治大 some scholars consider satirical imitations of real news, such as TV shows and satirical 立 media and social media, fake news has become more diverse and complex. For instance,. ‧ 國. 學. newspapers, as a form of fake news as they distort the reality and can alter people’s judgment (Balmas 2012, Berkowitz and Schwartz 2016). Other scholars fixate on the. ‧. notion of “fake”. For instance, Wardle contends that “fake” can fall into seven different. sit. y. Nat. categories: satire or parody (attend to fool the audience but not to cause harm), false. n. al. er. io. connection (i.e., when the headlines do not support the content), misleading content. v. (aiming to frame an issue or individual), false context (the information is genuine but. Ch. engchi. i Un. not the context), imposter content (the information is genuine but has been impersonated), manipulated content, and fabricated content (the content is entirely fake and aims to harm) (Wardle 2017). The second category comprises the studies focusing on the influence of fake news on citizens’ knowledge, opinion, and political trust. These studies reveal that fake news disturbs people’s perceptions of the political system and can generate a feeling of distrust among citizens toward the political elites (Balmas 2012, Brewer and al. 2013). Through her analysis of the 2006 Israeli elections campaign, Meital Balmas demonstrated that voters who were exposed to fake news tended to distrust candidates 14 DOI:10.6814/NCCU202001427.

(20) and the political system (Balmas 2012). Her study also points out that this trend increases when viewers are more accustomed to seeing fake news than real news (Balmas 2012). In their analysis of news media coverage of a satirical show discussing American politics, Brewer and al. denoted that the way traditional media address fake news also impacts people’s opinion (Brewer and al. 2013). As their study points out exposure to fake news (even via traditional media) generates political distrust and misconceptions of decision-making processes (Brewer and al. 2013). Other researchers study fake news’ impact among the different age groups. For example, based on her study of 61 American teenagers, Regina Marchi pointed out that the main reasons. 政治大. behind fake news’ success among the younger generation is their preference for fake. 立. news’ subjectivity and mistrust toward mainstream news media’s objectivity (Marchi. ‧. ‧ 國. 學. 2012).. The third category encompasses the mechanisms behind the creation and. sit. y. Nat. propagation of fake news. For example, Shu and al. pointed out that not only new. n. al. er. io. technologies have lowered the cost to produce fake news, they also levitated the spread. v. (Shu and al. 2017). The scholars further stress the importance of social bots, which are. Ch. engchi. i Un. social media accounts controlled by a computer (Shu and al. 2017). According to a study of the University of Oxford, during the 2016 US elections, the social media Twitter was flooded by 19 million bots publishing content in support of either candidate. 24 Still there are other research trying to explain how fake news spread through a new network of alternative sources of information called the “Fifth Estate,” which is formed by blogs, social media pages, satirical shows, and other alternative sources of information (Dutton 2009, Berkowitz and Schwartz 2016). Finally, some. 24. Oxford Internet Institute, “Resource for Understanding Political Bots,” The Computational Propaganda Project, November 18, 2016.. 15 DOI:10.6814/NCCU202001427.

(21) studies point out that fake news spread through partisan media as they use them to push the popularity of certain issues such as the economy, international relations, or religion (Vargo and al. 2017). Besides, according to Vargo and al., during the 2016 US election, partisan media relied on fakes news to attack the other candidate (Vargo and al. 2017). Finally, since fake news has become a major concern, many researchers have been putting effort to define the best approaches to identify and fight them. Some scholars recommend a linguistic approach to develop automatic fake news detection (Wang 2017). This idea derives from the use of “clickbait,” a common characteristics of fake. 政治大 Regarding the way of preventing the propagation of fake news, some propose to combat 立. news to attract the attention of audience by sensational headlines (Chen, and al. 2015).. ‧ 國. 學. them by reducing their economic attractiveness (Bakir and McStay 2017). For example, convincing major advertising associations to stop supporting websites that display fake. ‧. content might be a solution to turn down the economic incentive behind the production. sit. y. Nat. of fake news (Bakir and McStay 2017). Among the studies in pursuit of preventing the. n. al. er. io. propagation of fake news on social media, one of the possible solutions proposed is to. v. identify and block the accounts that share fake news and try to persuade other users of. Ch. engchi. i Un. their veracity (Shu and al. 2018). Another solution would be to maximize the spread of real news on social media to mitigate users’ exposure to fake content (Shu and al. 2018). Still others suggest monitor social bots as a way to combat fake content at an early stage (Shao and al. 2017). Since a few social bots suffice to amplify the spreading of fake news, taking these accounts down could drastically reduce their propagation (Shao and al. 2017).. 16 DOI:10.6814/NCCU202001427.

(22) 1.6. Scope and Limitation of the Research The scope of the research would focus on the discussion regarding Facebook’s. capacity to resolve fake news. Sources, incidents, studies, and data relative to the debate would be considered as essential to attain the objectives of the research. Since the research is set out to examine the effectiveness of Facebook factchecking program during the election periods, it is not dedicated to analyzing the impact of fake news on individual’s behavioral change, nor does it aim to explain the nexus between fake news and public opinions (as most of the literature has addressed the. 政治大. questions already). Instead, it touches upon the topic of whether Facebook can fight. 立. fake news with its fact-checking program.. ‧ 國. 學. Furthermore, what is important to mention here is that despite this thesis seeks to. ‧. argue that Facebook’s fact-checker solution is a contingency plan at best and should. sit. y. Nat. yield to meaningful efforts via the structural changes in its business models, the findings. al. er. io. do not apply to all the other fact-checking programs adopted by different actors. Any. v. n. implications made from the thesis should be limited to social media only.. Ch. engchi. i Un. Last but not least, due to the time and resource constraints, the study simply goes through the activities on two Facebook pages during a specific event and is limited to reach for further observation in the following major events closer to the final elections held in November. As a result, it is highly recommended and encouraged for later academic fellow members to conduct broader and more in-depth research to perfect the flaws and weaknesses that this paper has shown.. 17 DOI:10.6814/NCCU202001427.

(23) Chapter II: Understanding Fake News This section reviews and analyzes the literature regarding fake news in the hope of laying out a comprehensive context with academic depths to discuss Facebook’s factchecking program, its weaknesses, and latent concerns in this and later chapters. 2.1. Its Origin and Drive The origin of fake news articles during the 2016 elections could be traced back to. over 100 pro-Trump websites where most of them were run by teenagers in the small. 政治大 Veles drew in handsome advertising revenue (which used to go to traditional media) 立. town of Veles, Macedonia. 1 Motivated by monetary incentives, those teenagers in. ‧ 國. 學. when their news articles going viral on social media and prompted users to click to the original sites (Allcott and Gentzkow, 2017). One of the young founders behind these. ‧. websites, Boris (an alias used by the founder), earned nearly $16,000 from his two. y. sit. n. al. er. io. is $371.2. Nat. pro-Trump websites within four months while the average monthly salary in Macedonia. i Un. v. Examples of youngsters supplying fake news could also be found in. Ch. engchi. Endingthefed.com, a site run by a 24-year-old Romanian man that was responsible for four of the ten most popular fake news stories on Facebook.3 However, the young man claimed that he started the site to help Donald Trump’s campaign, attributing his deeds to be mostly ideologically-driven (Allcott and Gentzkow, 2017).. 1. Samanth Subramanian, “Inside the Marcedonian Fake News Complex,” WIRED, February 15, 2017.. 2. Ibid.. 3. Tess Townsend, “The Bizarre Truth Behind the Biggest Pro-Trump Facebook Hoaxes,” INC.COM, November 21, 2016.. 18 DOI:10.6814/NCCU202001427.

(24) The above examples illustrate the factors pushing for generating fake news could be either financial or ideological, or even complicated by both if one does not rule out the possibility. Knowing what motivates the actors to distribute fake news on social media is crucial. Without identifying the root cause, we won’t be able to come up with adequate solutions. For Facebook, audaciously as it promised to stop the sprawling fake news, how could it manage to accomplish such goal while ignoring the financial and ideological drives and failing to submit a political speech to the fact-checking program? Didn’t. 政治大 there was a massive portion of politically-affiliated fake news contents produced to 立 these fake news websites mushrooming during the elections already speak the fact that. ‧ 國. 學. make money and/or to interfere with the elections? It bottled my mind when Facebook decided not to scrutinize political speech in their fact-checking program because. ‧. political discourse could be virtually the most significant source of fake news.. y. Nat. io. sit. Suppose politicians take advantages of these fake news and post the exact same or. n. al. er. similar false contents on Facebook to serve his or her benefits. The next possible. Ch. i Un. v. scenario would be the post dodges the censorship of the program because it is executed. engchi. under the banner of “speech right,” which is disproportionally endowed to politicians exclusively. Facebook needs to give us a concrete explanation of why political speech is bestowed with immunity when there are proofs of fake political news generated by ideological and financial purposes, and there are chances that they might be exploited by politicians to win them an upper hand. In conclusion, failing to offer a convincing explanation, the program trusted politicians with an honor system that transcended them to a superior rank requiring no inspection. And if the program does work, it inevitably creates new problems alongside. 19 DOI:10.6814/NCCU202001427.

(25) 2.2. Its Definition What severe the hardship to tackle fake news is that it often finds itself overlapping. with other information chaos, such as misinformation and disinformation. It will be troubling if we do not steer clear of the ambiguity. Defined by David M. J. Lazer et al, fake news is “fabricated information that mimics news media content in form but not in organizational process or intent” (Lazer et al. 2018). Which is to say, fake news may appear in accordance with our daily perceptions of news, but its intention may not mean to inform truth or educate the public.. 政治大. Similarly, according to the Ethical Journalism Network (EJN), fake news is. 立. “information deliberately fabricated and published with the intention to deceive and. ‧ 國. 學. mislead others into believing falsehoods or doubting verifiable facts.”4 There is no. ‧. denying when it comes to fake news, one can easily stumble into more than one narrative describing the concept of fake news.. sit. y. Nat. er. io. To disengage the already misused term from the definition void, we can look upon. al. v. n. the UNESCO’s publication “Journalism, fake news & disinformation: A handbook for. i n C U h einspiration journalism education and training” for ngchi. “It can be helpful, however, to propose that misinformation is information that is false, but the person who is disseminating it believes that it is true. Disinformation is information that is false, and the person who is disseminating it knows it is false. It is a deliberate, intentional lie, and points to people being actively disinformed by malicious actors. A third category could be termed mal-information; information, that is based on reality, but used to inflict harm on a person, organisation or country.”. 4. “Fake News: Disinformation, Misinformation and Mal-Information,” Ethical Journalism Network.. 20 DOI:10.6814/NCCU202001427.

(26) Brushing through the rhetoric and meanings of each term, I noticed it could be concluded that there are two main variables employed to distinguish the disparities and that they could be further developed into a finer classification. The two variables, after summarized, would be “the authenticity of the story” and “the intent of the disseminator.” To shore up the classification, I transformed the above statement into a 2x2 square chart (see Figure 1) and assigned the terms to the corresponding blanks. As a result, the chart shows “mal-information” to be a true story but with deliberate intent, “disinformation” to be a false story and with deliberate intent, “ideal. 政治大 “misinformation” to be a false story but without deliberate intent. 立. news/information” to be a true story and without deliberate intent and finally,. ‧ 國. 學. Figure 1: Identification of “Mal-Information, Information, Disinformation, and. ‧. io. n. al. False story. Mal-information. Ideal news/ Information. y. Without deliberate intent. sit. Nat True story. With deliberate intent. er. Misinformation”.. i n CDisinformation U hengchi. v. Misinformation. Resonating both David M. J. Lazer et al. and EJN’s defining elements of fake news, the “intent of the publisher” and “fabricated information” (Lazer et al. 2018). Figure 1 suggests that the term, “fake news,” in both of their narratives falls into the same category where disinformation stands. Here, by concluding fake news to have matched the same notion as disinformation, I do not attempt to urge everyone to recognize fake news as disinformation only; rather, I wish to point out it is because fake news has been 21 DOI:10.6814/NCCU202001427.

(27) used as an umbrella term in many different situations covering multiple meanings, it needs to be addressed precisely. Take Facebook for example, if it desires to combat fake news, shouldn’t it pinpoint the idea of fake news among all the forms of information disorders so it can prioritize its efforts to correct the online chaotic state step by step? Fortunately, also having recognized that “the term ‘fake news’ has emerged as a catch-all phrase to refer to everything … The overuse and misuse of the term ‘fake news’ can be problematic because, without common definitions, we cannot understand or fully address these. 政治大 articles that purport to be factual, but which contain intentional misstatements of fact 立. issues,”5 Facebook publicly adopted the term “false news” to identify its target: “news. ‧ 國. 學. with the intention to arouse passions, attract viewership, or deceive.” 6 Referring to Figure 1, “false news,” being factual but containing intentional misstatements,. ‧. corresponds to the idea of mal-information.. y. Nat. io. sit. Now, the fact that false news shares the same concept with mal-information despite. n. al. er. different wordings gave rise to two interesting yet important questions. First, compared. Ch. i Un. v. with the other two narratives of major scholars and renowned NGO, Facebook has. engchi. implied to put their interests in fighting mal-information, rather than disinformation. The question, then, is why Facebook would target mal-information as its chief concern? And second, did Facebook’s fact-checking program perform effectively in terms of fighting mal-information?. 5. Jen Weedon, William Nuland, and Alex Stamos, “Information operations and Facebook,” Facebook April 27, 2017.. 6. Ibid.. 22 DOI:10.6814/NCCU202001427.

(28) Even though the aim of Facebook is clear thanks to its coined term definition of “false news,” it does not explain to us why they adopted false news and defined it in a fashion different from the general narratives of fake news. Could it be false news (possibly perceived mal-information as described in UNESCO) a major source of damage on their platform after attentive calculation and observation, an easy target to get over with, or even an innocent description that did not harbor too many thoughts from Facebook? Whichever it is, if it really wishes to address the havoc wreaked by fake news, endeavors to make every decision serving for a purpose are only necessary.. 立. Its Outlets and Impacts. 學. ‧ 國. 2.3. 政治大. By conventional wisdom, fake news would mostly manifest itself in online social. ‧. platforms, and it often raises concerns over democracy in political contexts because. sit. y. Nat. individuals may fall prey to malicious manipulations. Interestingly, based on recent. n. al. er. io. studies, social media being the major outlet for fake news requires to be discussed.. i Un. v. One recent evidence pointed out that 62 percent of the U.S adults absorb news on. Ch. engchi. social media (Gottfried and Shearer, 2016), while the findings in a 2017 research suggest “social media was an important but not dominant source of election news, with 14 percent of Americans calling social media their most important source” (Allcott and Gentzkow 2017). As mentioned previously, CAS underlined the vulnerability of the U.S. elections, and by inference, democratic elections in general. Based on studies following the 2016 elections, “the most popular fake news stories were more widely shared on Facebook than the most popular mainstream news stories,” and that “the most discussed fake news 23 DOI:10.6814/NCCU202001427.

(29) stories tended to favor Donald Trump over Hillary Clinton” (Silverman, 2016). It is not that hard then to understand why there was a bulk of implications shouting, “Donald Trump would not have been elected president were it not for the influence of fake news” once connecting all the dots. “Of the known false news stories that appeared in the three months before the election, those favoring Trump were shared a total of 30 million times on Facebook, while those favoring Clinton were shared 8 million times.” 7 Having also implied a potential correlation between the victory of Trump and the prevalence of pro-Trump. 政治大 stories that favor their preferred candidate, especially if they have ideologically 立 fake news, the research further points out that “people are much more likely to believe. ‧ 國. 學. segregated social media networks.” That is, if one were to receive information from a biased social media network, one would be more likely to accept information that is. ‧. inherently compatible with his knowledge or beliefs.. y. Nat. io. sit. This is saying that a piece of fake news via different social media platforms may. n. al. er. incur different kinds and degrees of perceptions that would wind up impacting the day-. Ch. i Un. v. to-day democracy. The importance of social media, like Facebook itself, to take heed. engchi. of being politically neutral, is then heightened. If Facebook could not remain unbiased with political issues, the integrity of the fact-checking program would conceivably be doubted in return. By and large, fake news could be everywhere and does not limit to social media. News media, websites, and even giveaway pamphlets are all possible outlets sending. 7. Ibid.. 24 DOI:10.6814/NCCU202001427.

(30) fake news. As for interpreting the impacts of fake news, there is no denying one should be very cautious. It is essential to look into the “medium-to-long–run” effects on political behaviors as they are “essentially nonexistent in the literature” (Lazer and al. 2018). When studying the consequences, one should always be cautious because encountering or sharing a piece of fake news is not the same thing as reading or being affected by it. Despite some have contended that political campaigns and advertisements, let alone fake news, may have limited effects on citizen’s judgments (Kalla and. 政治大. Broockman 2018), the latent risks of implicit endorsement by sharing fake news over. 立. social media should never be neglected.. ‧ 國. 學. With a click away, social media networks enable users to share links with any. ‧. counterpart located in whichever corner of the world, transport numerous shuttles of. n. al. er. io. sit. y. Nat. traveling information, and are, as a result, the chief contributors to the risks.. 2.4. Its Dissemination. Ch. engchi. i Un. v. As the internet levels the playing ground for new competitors by lowering the cost of entry, the traditional business models of news media that sustain heavily on print and broadcast are suffering a major setback. Technologies that enable information to be delivered more rapidly and on a larger scale approve individuals to practice the same business. Eventually, it is not shocking to see unverified contents traveling at a fleeting pace throughout online platforms.. 25 DOI:10.6814/NCCU202001427.

(31) In fact, an interesting yet worrying study points out that fake news “diffused significantly farther, faster, deeper, and more broadly than the truth in all categories of information” (Vosoughi, Roy, and Aral 2018). Also, it shows while “robots accelerated the spread of true and false news at the same rate,” it is the humans, not robots, are more likely to spread fake news (Vosoughi, Roy, and Aral 2018). The latent risks provoked by robots, however, could not be underestimated. Driven by the logic of the business model, which attempts to draw attention in the exchange of astronomic advertising revenue (Harari, 2018), fake news is spread through. 政治大 analyzing personal data, it allows data tycoons to identify an individual’s preferences, 立 the help of big data algorithms to pinpoint its target audience. By collecting and. ‧ 國. 學. so they could further similar direct content to their users. For those exposed to identical viewpoints, they are more likely to be like-minded to form “echo chambers” or “filter. ‧. bubbles,” where they are often isolated from contrary perspectives (Allcott and. io. sit. y. Nat. Gentzkow, 2017) and give rise to polarized camps and online hatred.. n. al. er. In this sense, for those tech giants that accumulate most personal data on earth,. Ch. i Un. v. Facebook, again, for example, they are creating what Harari called “information. engchi. dictatorship” that must be addressed before another CAS takes place (Harari 2018). Hence, the discussion upon the ownership of data is pressing: Should it belong to individuals, governments, corporates, the human collectives, or should we look upon the property of ownership in a brand-new way? Described as the attention economy—the resale of human attention—by Tim Wu from Columbia Law School, the business model that rallies eye balls and sells them to advertisers has raised profound questions about the news people access: Have attentions come before contents already (Morgan, 2018)? To hit more views, to trend more 26 DOI:10.6814/NCCU202001427.

(32) contents, the use of automated accounts or bots has brought about the amplification of fake news distribution. Activities by fake accounts to increase the likelihood of something trending on social media has squeezed out the chance of legitimate news being found by internet users (Morgan, 2018). The attention-based business model, as a result, is contended by observers as the inherent DNA flaw which destines the bleak future of the fact-checking program.8 It makes one further wonder if it is why Facebook resorts to fact-checking, for it does not concern to make adjustments of business models at all.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. 8. Ch. engchi. i Un. v. Tenz Shih (施典志), Fangxin,Facebook bu hui weile duikang jia xinwen er zijue cai lu/Frederic Filloux (Rest assured, Facebook will not kill itself to fight against fake news / Frederic Filloux), Tuna Business Review, August.. 27 DOI:10.6814/NCCU202001427.

(33) Chapter III: Examination on the Effectiveness of Facebook Factchecker Solutions: A Case Study on the Florida Presidential Primaries Although this thesis would not see through the end (the final election day on November 2), the journey of the 2020 presidential elections has nonetheless reached halfway by the time this thesis hit the results of its findings. Ever since the horn of elections aired with the sound of the Iowa caucuses on February 3, the war between the two biggest parties have not yet been blessed with smooth and peaceful events—several re-schedules of the primaries were arranged due to the coronavirus outbreak. The riots. 政治大. caused by the death of George Floyd1 further blurred and complexed the outlook of the. 立. elections.2. ‧ 國. 學. There is no way to tell how this round of political chessboard would develop. One. ‧. thing for sure though, by mid-April, Joe Biden (Democrat) and Donald Trump (Republican) were regarded as the finalists for the presidential run. The road leading to. y. Nat. io. sit. the determination of party nominee for Democrats was not at all flat and clear in. n. al. er. contrast to GOP, which underwent fewer competitions in party nomination as it is the. Ch. i Un. v. ruling party, and the incumbent is prone to succeed a second term. Prominent candidates. engchi. such as Michael Bloomberg, and Elizabeth Warren, dropped out of the run after Super Tuesday (actually it was Thursday) on March 3, and Bernie Sanders stopped his campaigns later in April, leaving Joe Biden the presumptive candidate against Donald Trump. 3. 1. Kashish Parpiani, “George Floyd protests and the 2020 US presidential election,” Observer Research Foundation, June 4, 2020.. 2. Thomas B. Edsall, “The George Floyd Election,” The News York Times, June 3, 2020.. 3. Wikipedia contributors, "2020 Democratic Party presidential primaries," Wikipedia, The Free Encyclopedia.. 28 DOI:10.6814/NCCU202001427.

(34) Not surprisingly, the campaign activities and posts on social media mirrored the political turnout. In total, for the studied period, 37 posts have been published by RPOF and 77 by FLDEMS. It can be argued that the number of posts from both pages came in distinct figures (almost twice the different) because by then the primary target for GOP to campaign against remained vague (Should it be Biden or Sanders?) — it was just not yet to pull out the big gun. So there were fewer posts from RPOF than FLDEMS for democrats have more obvious prey to hunt. However, the discussion on the campaign strategies and tactics is not the focal point here. I won’t go further study the factors behind the number gap despite a simple possible shortage of staffers in RPOF could be the reason at play.. 立. 政治大. ‧ 國. 學. What I did found interesting to mention outside the focal point is that although there is no specific figure targeted by ROPF in the studies period, after Sanders dropped. ‧. out of the run and endorsed Biden on April 8 (followed by Obama on April 14) 4, posts. io. sit. y. Nat. pertaining questionable contents aiming for Biden mushroomed.. n. al. er. In short, the thesis found out possible mal-information along with misinformation,. Ch. i Un. v. and disinformation from both of the political campaigns. For those questionable false. engchi. news shared by the pages (some have been debunked by fact-checkers), they remain intact. What’s worse, no single warning label was detected with these contents. This proves that Facebook fact-checking program could not deter the spread of false news because it couldn’t even manage to do what it promised to do.. 4. Ibid.. 29 DOI:10.6814/NCCU202001427.

(35) 3.1. Findings of FLDEMS During the period, most of the controversial contents in need of fact-checking. revolve the incompetence of Florida governor, Ron DeSantis and Trump administration in response to the coronavirus and their attempts to eliminate Affordable Care Act (ACA), taking up almost 30% of the total posts (23 out of 77). The contents of those posts, despite not entirely false, fell short of comprehensive perspectives and were obviously manipulated in a certain fashion to fit Democrats’ benefits. For example, on March 6, FLEDEMS posted a link of an article from its own. 政治大. website (floridadems.org) called “Ahead of Latest Trump Visit, Florida Democrats. 立. the following potentially-misleading message:. 學. ‧ 國. Launch New Campaign Slamming Trump for Failed Health Care Policies” that contains. ‧. “A Donald Trump victory means an end to the Affordable Care Act, an end to pre-existing condition protections and billions cut from Medicare. Democrats have a record of expanding affordable healthcare and, unlike the President, we have plans for every single. er. io. al. sit. y. Nat. American to access quality care.”5. n. iv n 6 C In fact, according to Factcheck.org, h e n git chashbeen i Uverified that a) it is true the. president’s proposed budgets have included funding cuts to the Centers for Disease. Control and Prevention (CDC)— but Congress has not enacted those cuts. b) The president did propose to reduce the CDC budget but not completely to defund it. c) Even though Trump did propose to cut funding of National Institutes of Health (NIH) funding, but lawmakers instead have enacted increases.. 5. See the whole text on flordiadems.org: https://www.floridadems.org/news/ahead-of-latest-trump-visit-floridademocrats-launch-new-campaign-slamming-trump-for-failed-health-care-policies.. 6. Lori Robertson, Jessica McDonald and Robert Farley, “Democrats’ Misleading Coronavirus Claims,” FactCheck.org, March 3, 2020.. 30 DOI:10.6814/NCCU202001427.

(36) Therefore, jumping into the conclusion that “a Donald Trump victory means an end to the Affordable Care Act” without offering the facts that Congress enacted increase but any cuts could be considered a piece of mal-information. To one’s astonishment, the same content was posted again later on March 10, only this time with a new link of an article (from the same website) that basically only went under few editorial modifications. Considering how Facebook defined false news, this mal-information certainly constituted the elements of Facebook’s asserted false news. So technically speaking, weren’t these posts the very targets that needed to be addressed?. 政治大 The post on March 7, however, has reached around thirteen thousand shares and 立. ‧ 國. 學. the post on March 10, around 1.2 thousand shares, while both of them remain unlabeled with any warning. It is tricky because one can contend that these posts are from the. ‧. political group, and by rules, they are not subject to the fact-checking program. Still,. sit. y. Nat. technically these twisted messages were shown in the shared links of articles, so they. n. al. er. io. should be examined by fact-checking. The fact that these posts are still intact and enjoy. v. undisturbed visibility in the News Feed suggests Facebook failed their commitment in. Ch. engchi. implementing the fact-checking program.. i Un. Despite there was no obvious disinformation shown in the studied period, the questionable posts were mainly mal-information. Most of them appeared to mislead the public with partial facts, as shown above. Here’s one but not least example of a shared article published on March 11 that dodged censorship: “Health care... This president promise you would be able to keep your care. It will be better and cheaper, he said. And he broke that promise...This president campaigned, and he said ‘I will not cut Medicare, I will not cut Social Security.’ … He proposes to cut Medicare. He proposes to cut Social Security. He proposes to cut 31 DOI:10.6814/NCCU202001427.

(37) Medicaid further. That is a broken promise... The trail of broken promises just runs so far and so deep, and that is why we need to move forward.” Again, the Democrats are free to criticize the proposals, but without pointing out the result that Congress, in fact, enacted an increase in the budget instead of decrease, it is unfair. It can be seen as a false accusation or attack even by saying, “that is a broken promise.” This kind of wrong message has been delivered repetitively throughout the studied period, and I would stop naming the rest of them due to the redundancy. Besides the over-proliferated content pertaining to how Trump or DeSantis poor. 政治大. management over the crisis, there is another interesting phenomenon shown on March. 立. 31. A series of pictures7 stating “In 2020 Hispanic women only earned 54% of the. ‧ 國. 學. wages that men earned for equal work,” “In 2020 American Indian women only earned. ‧. 57% of the wages that men earned for equal work,” “In 2020 black women only earned 62% of the wages that men earned for equal work,” and “In 2020 women still earned. y. Nat. er. io. in the comment sections.. sit. 80% of the wages that men earned for equal work,” created a wave of public opinions. al. n. iv n C h e nofgthe Users were demanding the sources i U shown in the pictures. This c hfigures. demonstrates even public mass could sense a possible source of misinformation, but sadly these pictures somehow sustain their visibility without issues. Again, it inevitably cast doubts on Facebook fact-checking program that what exactly can it do with false news.. 7. See pictures at: https://www.facebook.com/FlaDems/photos/a.288549429974/10151223935994975/?type=3&theater.. 32 DOI:10.6814/NCCU202001427.

(38) 3.2. Findings of ROPF Although ROPF published fewer posts, it was not short of posts with controversial. information. Before the study began, it already showcased fake news at first glance. For example, RPOF shared a picture (Picture 1) that claims “during Obama’s last year in office there were, missiles launches (to North Korea) every 24 days”. This allegation was false and debunked by Check Your Fact, Facebook fact-checker partner.8 What’s more, the post still finds its way to be shared one year later. Picture 1 – Facebook’s post from the Republican Party of Florida comparing Presidents. 政治大. Barack Obama and Donald J. Trump, published on February 28, 2019.. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i Un. v. Since FLDEMS was mostly shooting arrows at President Trump and Governor DeSantis’s inadequate response to the coronavirus, most of the posts on ROPF were on 8. Emily Larsen, “Fact Check: Did North Korea Test Missiles Every 24 Days Under Obama?,” Check Your Fact March 8, 2019.. 33 DOI:10.6814/NCCU202001427.

(39) the contrary about the feats of these very two “heroes.” Contents celebrating the accomplishments over the disease control and successful push for better health care, and criticizing Democrats over their general impediment amounts to roughly 32% of the total posts (13 out of 37). Among these posts, although they tend to be self-boasting or hostile, they embodied in a feature with a mixture of facts and exaggerations to butter up Republican’s or Trump’s image. Strictly speaking, they hardly constitute a piece of disinformation or mal-information. However, some of their photos could be widely considered offensive, especially. 政治大 became the presumptive candidate, malicious photos with questionable contents 立. after Bernie Sanders dropped out of the run, and former vice-President Joe Biden. ‧ 國. 學. mushroomed in quantity. For instance, on May 19 (see Picture 2) and May 20 (see Picture 3), there were photos related to Biden.9. ‧. The allegation surfaced when a woman named Tara Reade, accused Biden of. y. Nat. io. sit. sexually assaulting her when she was a staffer on a Senate team in 1993, and prompted. n. al. er. reports that he made women uncomfortable. 10 Such accusation raised questions for. Ch. i Un. v. Democratic voters who previously said women’s complaints should be listened to and. engchi. treated as credible.11 Following the debate over what it means to “believe women,” the rallying cry of the #MeToo movement, Biden has encountered escalated sexual assault allegations against him. Still, the claim remains unverified, and it was over the line for ROPF to produce pictures with asserted blame on Biden before the whole incident even comes to a clear. 9. See Repulican Party of Florida Facebook’s page: https://www.facebook.com/FloridaGOP/photos/.. 10. “Transcript: Into Joe Biden and the Women's Vote,” ABC News, May 23, 2020.. 11. Holly Honderich, “Joe Biden assault claim: What does 'believe women' mean now?,” BBC News, May 12, 2020. 34 DOI:10.6814/NCCU202001427.

(40) end. What could ROPF campaign instead was to urge Biden to come clean, not throwing a vicious attack on the internet platform. The other photo (which has been debunked as false) was attacking at Biden’s claims on travel bans. As shown below, it suggests Biden opposed Trump’s travel restrictions on China. The picture said, “Biden called travel ban xenophobic,” and “Biden is good for China.” But given the findings from Factcheck.Org, Biden “was not even talking about China”—he was talking about the travel restrictions imposed on Europe.12 So the campaign ad attacks Biden on his opposition to the China restrictions is highly disputed.. 立. 政治大. Even though the above examples were not extracted during the studied period,. ‧ 國. 學. there is a point that must be made, and therefore it deserves to go a little extra. The fact. ‧. that once Democrats gave birth to their presumptive candidates, the fire shot at this figure multiplied, and it also came with a wave of disinformation. Unfortunately,. y. Nat. io. sit. Facebook already claimed that campaign ads are “held by higher transparency” so are. n. al. er. not subject to the program but given there were myriads of debunked false news. Ch. i Un. v. contents in the ads, for what concrete reasons that political ads are not subject to the. engchi. program? It is only reasonable to mention then that the aftermath of the exclusion of politicians in the fact-checking program and putting faith in politicians’ news ethics and morality was just about to begin.. 12. Rem Rieder, “Trump Campaign’s Misleading Ad Attacking Biden on China,” FactCheck.org, April 10, 2020.. 35 DOI:10.6814/NCCU202001427.

(41) Picture 2 – Facebook’s post from the Republican Party of Florida caricaturing Democrat’s candidate Joe Biden. Posted on May 19, 2020. 立. 政治大. ‧ 國. 學. Picture 3 – Facebook’s post from the Republican Party of Florida caricaturing. ‧. Democrat’s candidate Joe Biden. Posted on May 20, 2020. n. er. io. sit. y. Nat. al. Ch. engchi. i Un. v. 36 DOI:10.6814/NCCU202001427.

(42) Chapter IV: A Response to Facebook’s Fact-checker solution This chapter as a whole answers the research questions and encourages Facebook to act on certain moves to help themselves. While the first section is dedicated to an analysis of the program, the second one proposes several suggestions to improve the current situation regarding the dissemination of fake news on the social media.. 4.1. Analysis of Facebook’s Fact-checking Program. 政治大 that the fact-checking program proposed by Facebook exists fundamental flaws which 立. Based on the discussions and findings in the previous chapters, the thesis argues. ‧ 國. 學. are deeply rooted in the business model they sustain to amass a substantial amount of advertising revenue. Thus the program could only serve as a contingency plan at best.. ‧. The fundamental flaws in the program consist of two main dimensions: design and. y. Nat. io. sit. implementation. First, there are at least three pending questions in regard of the program. n. al. er. design that requires Facebook to make a concrete stand: 1) The mystery of false news. Ch. i Un. v. definition, 2) The puzzle of the exclusion of the political speech, 3) The question of. engchi. whether there is any untold connection between the above two items. It seems very suspicious to exclude politicians from the program and, at the same time, appoint false news to be their addressing target. Why? As mentioned previously, the logic upholding Facebook’s mentality is self-contradicted. On one hand, they heralded the merry of speech right, but on the other hand, sublimed political speech and called it “the most scrutinized speech there is.” So with this poorly-connected logic, they justified their exemption of political discourse from the program. It is as if they tried to tell people they did not scrutinize posts from politicians because they cherish 37 DOI:10.6814/NCCU202001427.

(43) freedom of speech, and other people (whom I do not know) in the democracies would do it for them automatically anyway. If the above speculation was what they are truly after, what they did with the definition of false news could make one sense then, but just not a decent one. Per their own definition, false news could be known as facts with misleading intentions, which by UNESCO standards, mal-information. By claiming their bold commitment to deter the spread of false news did leave an image that they are trying to take on the herculean tasks because it is not easy to do fact-check and clarify one’s intention at the same time.. 政治大 intend to carry such burdens in the first place. 立. But to peel off the layers of their words one by one, it is quite obvious that they did not. ‧ 國. 學. The rules they set for themselves in this game have freed them from the troubles. ‧. to scrutinize the major source of false news—posts of politicians. According to the findings in this thesis, there was a great deal of mal-information (or in Facebook’s. y. Nat. io. sit. words, false news), appearing in the selected Facebook pages. If Facebook did intend. n. al. er. to fight the spread of false news, why would it miss the supposed hotbeds of false news. Ch. and protect them with the shields of speech right?. engchi. i Un. v. These conflicts, of course, have shattered the trusts of their fact-checker partners and even generated antipathy toward Facebook. The old-time fact-checker partners criticized that fact-checking program is merely a PR function to save Facebook’s reputation after being harshly criticized its role in CAS. Taking their business model into accounts, observers believe Mark Zuckerberg won’t take up serious measures against fake news because it would mean cuts to their financial funnels. They argue Facebook needs to adjust the business model in order to combat the spread of fake news.. 38 DOI:10.6814/NCCU202001427.

(44) In addition, the flaws in their program design underlie great concerns as well. For one thing, to detect and lower the visibility of false news aren’t easy works because it involves a certain degree of subjective judgment to tell which contents contain fragments of facts with misleading intents and which not. A case in point would be a Politico article1 titled, “Trump rallies his base to treat coronavirus as a hoax,” which was later debunked as a false statement by Check Your Fact, a third-party fact-checking partner of Facebook.2 Check Your Fact explained that Trump was referring to “the alleged ‘politicizing’ of the coronavirus by Democrats” as their new hoax, not coronavirus itself. 3 While some contended what Trump said was extremely ambiguous. 政治大. and that debunking the content of Politico was merely too arbitrary, it was no doubts. 立. that Facebook’s fact-checking is a powerful tool, which makes it an arbiter of truth.. ‧ 國. 學. Another similar yet contentious example4 could also be seen in a recent online. ‧. post from President Trump, saying “when the looting begins, the shooting begins,”. sit. y. Nat. which lots of people translated into “a call for violence in nationwide protests over the. n. al. er. io. death of George Floyd.” Some Facebook employees believe the company should take. v. action on the controversial post, the CEO, Mark Zuckerberg, however, reacted. Ch. engchi. i Un. otherwise. It is shocking to witness Facebook not to take up any measure while Twitter had put a warning label and flagged it as violent. The non-action from top management has caused outrage from Facebook employees, resulting in at least two employees resigned.. 1. Nancy Cook and Matthew Choi, “’Trump rallies his base to treat coronavirus as a ‘hoax’,” Politico, February 28, 2020.. 2. Adi Robertson, “Facebook fact-checking is becoming a political cudgel,” the VERGE, March 3, 2020.. 3. Ibid.. 4. Elizabeth Dwoskin, “Zuckerberg defends decisions on Trump as Facebook employee unrest grows,” The Washington Post. June 2, 2020. 39 DOI:10.6814/NCCU202001427.