This is the first of a series of pieces we intend to publish on societies and elections at risk from online disinformation. Our goal is to draw the attention of the media, tech platforms and other academics to these risks and to provide a basic background that could be useful to those who wish to study the information environment in these areas.
On Saturday, January 11, 2020, Taiwanese citizens will vote for their next president. The contest is between the candidates of two parties: Tsai Ing-Wen, incumbent president and a member of the Democratic Progressive Party (DPP) and Han Kuo-Yu, the challenger representing the Nationalist Party (Kuomintang, or KMT). While the focus of Stanford Internet Observatory’s project is the upcoming Taiwanese election, we begin with a brief summary of the historical context of these political parties.
From 1945-1949, following Japan’s defeat at the end of WWII and its handover of Taiwan, the Nationalist Party (KMT) was briefly the de facto government of both China and Taiwan. However, the KMT was defeated by the Chinese Communist Party (CCP) during the Chinese Civil War, and in 1949 the KMT government fled to Taiwan. At that point, both the KMT government (in exile) and the CCP government in Beijing claimed to be the legitimate government of China, as the countries of the world split on what entity to recognize as the rightful leaders of China. In 1971, Taiwan was expelled from the United Nations, and the CCP was recognized as the ruling government of China in the UN General Assembly and the UN Security Council.
The Taiwanese government remained largely a single-party entity until 1987, when it lifted martial law and allowed competing political parties to emerge. The most significant of these was the Democratic Progressive Party (DPP), which in 2000 became the first opposition party to win the presidency. The DPP has controlled the presidency since 2016.
Taiwan's major social cleavages line up with the KMT / DPP party divide, and reflect Taiwan’s history as a territory distinct from the Chinese core provinces; it has robust ethnic diversity produced by immigration from both mainland China and the rest of Asia Pacific, and with the exception of the period from 1945-49 the island has been ruled by a distinct government since 1895. The KMT is dominated by Han Chinese who arrived with the KMT in 1949 from the mainland, are concentrated in the north. They speak Mandarin, have a history of dominating government positions, and have benefited the most from state-led economic programs and trade with China. The DPP is dominated by pre-1949 native Taiwanese who are concentrated in the southern half of the island, are ethnically Hoklo and Hakka and thus speak Hoklo and Hakka instead of Mandarin, and have historically been excluded from government and state-led economic development.
Incumbent president Tsai Ing-Wen’s administration has had fraught relations with Beijing since her election in 2016. The KMT party challenger Han Kuo-Yu, who won the July 15, 2019 primary process consisting of party member votes and multiple public polls, is the preferred candidate of the CCP in Beijing. Although he served in the legislative yuan (the Taiwanese legislature) from 1997-2002, he was thought to have left politics and was relatively unknown until his unexpected victory in the November 2018 mayoral election in Kaohsiung. This election attracted substantial attention within the region; Kaohsiung is the most important city in southern Taiwan and had been a DPP stronghold since 1998. Han Kuo-Yu’s victory was as remarkable as a far-left Democrat becoming governor of Texas or another deep-red state.
The election carries significant weight; China sees Taiwan as part of China under the One China Principle, and this election is partially a referendum on the nature of future ties with Beijing. The ongoing activities and protests in Hong Kong, which began to attract worldwide attention in June 2019, are also having an impact on the campaign, shoring up support for President Tsai Ing-wen.
Elections are a widely-recognized target for disinformation campaigns and influence operations, and the upcoming Taiwan election carries significant weight in the region; it is strategically important to Beijing. Academic observations of past elections in Taiwan have noted the presence of astroturfing and domestically-organized trolling, and cross-strait propaganda has been a long-term issue. More recently, however, investigations into the November 2018 Taiwan elections suggest the emerging presence of social media manipulation coordinated from within the PRC.
There are two prongs of influence operations that may present a threat to the integrity of the Taiwan 2020 election. The first is media manipulation, primarily via narrative laundering and propaganda campaigns that facilitate the placement of both strategic persuasion and disinformation narratives into local press. China has extensive experience in the dissemination of propaganda, and the world’s most extensive domestic propaganda apparatus, covering every conceivable form of media. While it has had propaganda capabilities since the Maoist era, the current approach began in 1980 with the creation of the Central Committee Outward Propaganda Small Group (中共中央对外宣传小组), which began in 1991 the State Council Information Office / International Propaganda Office (中共中央对外宣传办公室 / 国务院新闻办公室). An August 2018 report by France’s Institute for Strategic Research (IRSEM, Ministry for the Armed Forces), provided a survey of the CCP’s capabilities and goals:
Effort on the ideological front has two objectives: first, to shape the internal political space and maintain the Party’s legitimacy (through censorship and disinformation); second, to influence international opinion and wage the “information war” in favor of Chinese interests. (p. 58)
Today, China controls more than 3,000 public television channels in the world, over 150 pay TV channels, around 2,500 radio stations, about 2,000 newspapers and 10,000 magazines and more than three million internet sites. (p. 59)
Most recently, Chinese media has been amassing substantial English-language audiences on Facebook and Twitter to convey its point of view on world affairs to English-speaking peoples. This activity includes paid promotion, such as the recent purchasing of promoted Tweets to spread its point of view on the Hong Kong conflict to the world. This is despite the fact that both platforms are not accessible from within China without the use of circumvention technologies.
The second type of interference is covert social influence via the use of bots and fake persona accounts, which can be used to disseminate propaganda in a seemingly peer-to-peer capacity, to amplify memes, articles, or videos, or to create the appearance of grassroots consensus among a community. China has a well-documented covert social influence capability; its inward-facing digital commenter brigade, known as the “50 cent army”, was launched in 2004. The effort consists of hundreds of thousands—some estimates reach as high as 2 million—of conscripted posters who comment on local Chinese social media and news articles to bolster the CCP and its leaders and policies, or to simply distract real participants in the conversation from controversial topics. On the surface, these comments appear to be the creation of ordinary people. While attribution is not always conclusive, there have been debates about the extent to which similar tactics are now being deployed in more outwardly-focused arenas, including potentially suspicious Reddit activity on posts related to China, a disinformation campaign culminating in the suicide of a diplomat in September 2018, and persona accounts (“sockpuppets”) promoting state-sponsored narratives about Hong Kong protestors on Twitter and Facebook.
Most directly, there is early evidence suggesting the use of Facebook persona accounts in Taiwanese elections—specifically in the November 2018 mayoral race in Kaohsiung. On the surface, Han Kuo-yu’s upset victory initially appeared to be a well-run digital campaign by a charismatic populist candidate who was able to generate significant digital engagement and dominate the conversation on social media platforms. However, recent investigations claim to have detected the presence of a social media manipulation campaign that may have played a role in his rise, the source of which journalists assessed as a “seemingly professional cybergroup from China”.
During the campaign, a Facebook group named “Han Kuo-yu Fans For Victory! Holding up a Blue Sky!” grew to be a convening place for Han’s supporters, amassing 61,000 members. Writing in Foreign Policy, journalist Paul Huang described it as a place for fans to promote “talking points, memes, and very often fake news attacking Han’s opponent Chen, the DPP government, and anyone who said a bad word about Han.” Mr. Han acknowledged the efforts of the “netizens” supporting him, saying, “I don’t know you, but I thank you”. Shortly after the election, it emerged that three of the original creators and administrators of the Han Facebook Page had LinkedIn profiles suggesting that they were employees of Tencent; however, these accounts had several characteristics that suggested they were fake (duplicated profile pictures, no contacts, language patterns more indicative of non-Taiwanese residents).
Throughout the 2018 midterm election campaign, President Tsai repeatedly attempted to alert the Taiwanese public to the presence of interference: “There are those people who mistakenly think that if you simply shout falsehoods loudly, they’ll become real,” she wrote on Facebook. Her post describes an “Olympics” to disrupt Taiwan’s elections, with a two-pronged strategy: first, spreading fabrications and fake news stories; second, flooding the social media ecosystem and targeting anyone who “spoke up for Taiwan” with smears and attacks.
It is possible that online boosterism for Han in 2020 will be comprised of a complicated mix of legitimate domestic support, influence from the PRC and the actions of private actors. Such a mix would pose challenges for both Taiwanese authorities and US tech platforms looking to distinguish between “authentic” and “inauthentic” activity.
The media environment in Taiwan is robust; there are radio channels and newspapers representing most viewpoints, as well as a variety of cable television options. 80% of Taiwanese residents watch cable TV.
Although the media in Taiwan is considered to be one of the freest in Asia, there are popular papers that serve as conduits for mainland influence. In 2008, the paper with the 4th largest circulation, the China Times, was sold to a billionaire businessman who considers unification an inevitability. It came to be seen as sympathetic to the CCP; a recent Financial Times report asserts that the Taiwan Affairs Office (TAO) of the CCP is in daily contact with the editors of the China TImes, and may be exercising editorial control. In addition, several popular online news sites write in a tabloid style, resulting in the potential for factual inaccuracies or fake news stories. The Taiwanese media ecosystem is largely unprofitable, and doesn’t have resources to devote to high-quality investigative journalism.
Taiwanese citizens are quite active online. Taiwan has extremely high social network penetration; as of January 2019, multiple social media marketing surveys indicate that 95% of residents access the internet daily, and 89% of residents visit social networking sites. Facebook and YouTube are easily accessible within Taiwan, and are the two most popular social platforms among residents. The Japanese mobile messaging app Line ranks 3rd. Line’s offering is similar to Chinese powerhouse WeChat; it offers a digital wallet, food delivery, taxi-hailing services, and other convenience tools in addition to social communication.
Malign narratives can be deployed over a long time horizon, often with the goal of shifting societal positions; this can take the form of manufacturing consensus around a topic, or exacerbating societal divides and eroding trust. The below narratives are far from the only ones we will investigate, but they are topics of debate within Taiwan that may be leveraged by inauthentic actors:
Ties to China (KMT supports stronger ties, particularly economic)
The protests in Hong Kong
LGBT rights and gay marriage
The US-China trade war
Economic growth, especially in the south
Chinese interference, which is likely to be positioned as a DPP excuse for poor electoral performance (this has become an ongoing concern, as President Tsai Ing-Wen has raised it on numerous occasions)
The other type of narrative manipulation is conducted in rapid response to a discrete event or provocation. We will also be observing emerging hashtags and anomalous distribution of content in response to breaking news or situations with the potential to galvanize audiences. One example of disinformation spread in response to a contentious moment occurred during the November 2018 campaigns, when a DPP candidate was falsely accused of wearing an earpiece to receive answers during the debates; this led to extensive chatter about the accusation on social apps.
The 2020 Taiwan election is extremely important to the future of democracy in Taiwan; the Hong Kong protests have increased local concerns about future relations with Beijing.
Activities in 2018 suggest that there is Chinese interest in leveraging disinformation capabilities to interfere in the democratic process of Taiwan’s presidential election. Both the incumbent president and Taiwan’s National Security Bureau—the country’s primary intelligence agency— have already issued warnings about Chinese ‘50 cent army’ influence operations and information warfare, pointing to several YouTube channels by name and attributing them to Beijing.
While the impact of propaganda and social influence operations remains difficult to quantify, the November 2018 election did see substantial losses for DPP and President Tsai Ing-wen.
Newly-attributed influence operations targeting Hong Kong reinforce the concern that Beijing is willing to deploy its influence capabilities to spread disinformation about regional political conflicts.
SIO intends to monitor the Taiwan 2020 presidential election throughout the remainder of the campaign.
Many popular applications in the Taiwanese social media ecosystem are opaque to external monitoring and little academic work has been done to create alternative techniques appropriate for these platforms.
None of the major social media ad networks (Google, Facebook and Twitter) have launched their political ad transparency projects in Taiwan. There is no mechanism to browse, study or analyze political advertisements on these popular platforms targeted at the Taiwanese population.
Estimating the impact of malign narratives is complicated by the fact that Taiwanese residents are accustomed to propaganda from Beijing, and in fact expect it. Prior efforts to discredit the DPP and promote the KMT have previously had a limited impact, if not backfired, since 1996.
Computational Propaganda in China: An Alternative Model of a Widespread Practice, Robert Gorwa, Oxford Internet Institute
Computational Propaganda in Taiwan: Where Digital Democracy Meets Automated Autocracy, Nicholas J. Monaco, Jigsaw/Oxford Internet Institute
Chinese Cyber-Operatives Boosted Taiwan’s Insurgent Candidate, Paul Huang, Foreign Policy
Taiwan election: KMT’s Han accuses Tsai of spying, KG Chan, Asia Times
Decoding China’s 280-Character Web of Disinformation, James Palmer, Foreign Policy
King, Gary, Pan, Jennifer, and Margaret Roberts. 2013. “How Censorship in China Allows Government Criticism but Silences Collective Expression.” American Political Science Review 107(2): 1-18
King, Gary, Pan, Jennifer, and Margaret Roberts. 2017. “How the Chinese Government Fabricates Social Media Posts for Strategic Distraction, not Engaged Argument.” American Political Science Review Volume 111, Issue 3, pp. 484-501
China using fake news to divide Taiwan, Chien Li-chung, Chung Li-hua and Jonathan Chin, Taipei Times
Made-in-China fake news overwhelms Taiwan, I-fan Lin, Global Voices