487
Chapter 2
INFORMATION OPERATIONS AGAINST
HONG KONG PROTESTERS IN 2019
Hong Kong was retroceded to China in 1997 after an agreement between the United
Kingdom and the PRC that preserved the existing political system in Hong Kong for 50
more years (in conformity with the Basic Law, Hong Kong’s “constitution”). This arrange-
ment is commonly known as the “One Country, Two Systems,” and the island became a
“Special Administrative Region” (特别行政区–SAR) with its own judicial and political
system. The PRC’s authority is represented by the Liaison Office of the Central People’s
Government in the Hong Kong SAR (中央人民政府驻香港特别行政区联络办公), aka
the Liaison Office.
Throughout the winter of 2019 a bill on extraditions to China angered the population,
which began to protest against it. The movement intensified as the police repressed them.
The June 16 protests gathered a record two million people according to the organizers, but
confrontations with the police became more violent. And tensions culminated during
the summer and fall of 2019. In this context, Beijing conducted a number of influ-
ence and disinformation operations against protesters using three narrative lines: sup-
port for the police and the Hong Kong government, attacks against the reputation of
protesters, and accusations of foreign interference.1-2
I. Beijing’s difficulties in imposing its narratives on the
protests
On the ground, the Chinese regime proved unable to anticipate the size of an opposi-
tion or to understand the movement’s evolution and the meteoric decline of executive head
Carrie Lam’s standing in the public opinion. The failure of the Liaison Office’s representa-
tive was also obvious. In this regard, images of protesters surrounding the Liaison Office
and trampling the PRC’s flag were seen as a clear affront to the CCP. Surprised by the
scale of the protests and the inability of Lam’s government to restore order, the CCP was
forced to react quickly. The Party’s margin for maneuver was slim however since Hong
Kong remains largely autonomous and a direct coercive action would only have come at a
high political price, detrimental to the CCP’s public image – and completely demonetizing
1. This chapter focuses on the events of 2019. For a general analysis of United Front activities in Hong Kong, see
Sonny Shiu-Hing Lo, Steven Chung-Fun Hung, and Jeff Hai-Chi Loo, eds., China’s New United Front Work in Hong Kong:
Penetrative Politics and Its Implications (London: Palgrave Macmillan, 2019).
2. Iain Robertson, Targeting the Anti-Extradition Bill Movement: China’s Hong Kong Messaging Proliferates on Social Media,
Atlantic Council’s Digital Forensic Research Lab (Dec. 2020), 10.
488
the rhetoric of the “One country, Two systems” in the eyes of the people of Hong Kong
but also in Taiwan.
Faced with this humiliation that rapidly tested the Chinese regime’s capacity to restore
its authority – in China strictly speaking as much as in Hong Kong or within its population
– Beijing’s response turned to the informational sphere.
Beijing’s counter-attack faced three challenges. Its discourse had to address three tar-
get audiences simultaneously: (1) mainland public opinion, which was the easiest to
channel, as information is tightly controlled behind “the Great Chinese Firewall” with
the Propaganda Department setting the tone and Chinese media falling behind; (2) Hong
Kong’s public opinion, and more generally the Chinese diaspora, especially in North
America and Europe; and (3) the non-Chinese-speaking international public opinion.
The challenge for China was to retain a minimum of its capital for sympathy: the
protests indeed attracted the attention of foreign media to the situation in Hong Kong and
naturally to the management of the crisis by the Party. However, the Chinese government
was facing a complex international environment, under growing criticisms and hostility with
the “trade war” it faces with the U.S. administration, the backlash against its colonization
policy in Xinjiang and its increasingly brazen constraints on public liberties for instance.
Beijing’s real margin for maneuver was slimmer than appeared at first glance.
The regime’s objective was to spread counter-narratives that could be heard and
accepted by all three audiences. The channels and languages used to diffuse these nar-
ratives were numerous. On the one hand, in the Chinese-speaking world, Beijing essentially
used sealed-off ecosystems like WeChat, Weibo and Douyin (the continental version of
TikTok), in which only content that conform to the mainland’s laws and rules can circulate
(→ p. 196). These platforms are widely used by overseas Chinese and constitute an efficient
way to influence the diasporas. On the other hand, to shape international public opinion
Beijing mostly used U.S. platforms such as Facebook, Twitter, YouTube, banned in China,
and the Chinese TikTok. Social media occupied an important role in the Chinese commu-
nication strategy regarding Hong Kong but traditional media were not neglected either.
II. The Chinese media’s narrative
The Chinese government’s spokespeople often complain about the foreign media cover-
age of the PRC. This coverage is often considered to be too negative and too ignorant of
the successes of the Communist Party to focus on “smear” campaigns instead. This tactic
or posture as a victim, which consists in criticizing the coverage of China in western
media, is the main – and most commonly used – method to influence a public opinion. The
target audience is just as much that of China as that of the diaspora and sometimes even
certain fringes of the international public opinion (the far left and the far right in European
liberal democracies for example).
On the continent the Party can rely on its well-oiled propaganda apparatus and
its network of censors to steer the public’s attitude to the protests. David Bandurski from
the China Media Project analyzed the methods and narratives used to achieve this goal on the
continent.3 Along with big official daily newspapers, he noted the importance of numerous
WeChat accounts massively posting “low quality” articles or junk propaganda. These articles
3. David Bandurski, “Hong Kong through China’s Distorted Lens,” ChinaMediaProject (24 Jul. 2019).
489
serve to inflame public opinions through carefully selected and powerfully divisive content
(desacralized flag, imputing acts of violence to protesters or even “proof ” of foreign inter-
ference– most notably from the CIA). This content, largely spread on the web, can be lik-
ened to political “spam” aiming to drown a public opinion in manipulated content.
Examples of articles posted around October 8, 2020 on WeChat after an Internet
search for “the bandits of Hong Kong” (香港暴徒).4
In Hong Kong, however, the Chinese government could not directly censor or influ-
ence the protesters’ media narrative. The Liaison Office did have relays in public opin-
ion inasmuch as it had acquired publishing houses or formed close links with a number
of daily newspapers and tabloids beforehand,5 but the CCP could not control flows
of information on social media as it did on mainland platforms. Besides, Beijing
announced its decisions, at least officially, to let the Hong Kong government handle the
protests. All the declarations from the Liaison Office, or from the general secretary of the
CCP Xi Jinping, displayed the “CCP’s confidence in the Hong Kong government.” The
Party-State’s actions kept a low profile and strived to unite the “loyalist” forces with a
dual strategy: forcing the hand of Hong Kong’s institutions (firms like Cathay Pacific,
the Mass Transit Railway or the Ministry of Justice for instance) but also playing the long-
term game of letting the opposition movement wither away, collapse or be discredited
because of the protesters’ violence and radicalization.
The channels used to get this type of messages through varied. Aside from the declara-
tions of official representatives and organized protests against certain TV channels (illus-
trated by the 41-page letter sent in late August 2019 to prominent international newspapers
and press agencies),6 the task of amplifying the message was mostly assigned to offi-
cial Chinese media.7 For instance, a Google search for “Hong Kong media coverage
biased” returned several videos posted mainly in August 2019 by CGTN, CCTV or even
China News.
4. This is only a sample among hundreds of proposed search results. In his article, David Bandurski and his team
have listed more than 200 published in July. See “暴徒以镭射笔照射军营 _驻港部队首次 ‘警告’” (“Rioters Shine
Laser Pens Toward the barracks, Troops in Hong-Kong Give a First Warning”) (10 Jul. 2019).
5. “中聯辦掌控聯合出版集團擁三大書局兼壟斷發行 議員指涉違《基本法” (“The Liaison Bureau Takes
Control of a Publishing House and Owns Three Big Libraries, this is a Monopoly Problem. The Members of Legco
Point Out a Violation of Basic Law”) (9 Apr. 2015).
6. “Hong Kong Government Supporters’ Protest Targets ‘Biased’ Journalists at Public Broadcaster RTHK’s
headquarters,” South China Morning Post (24 Aug. 2019).
7. “China Sends 41-page Letter to Tell Foreign Media to Stop Biased Reporting,” Mothership, (23 Aug. 2019).
490
Google search September 26, 2019.8
Another example of this type of content designed to spread Beijing’s narrative was ana-
lyzed by David Bandurski.9 On August 1, 2019, the PLA broadcast a video from a training
of its Hong Kong garrison on Chinese public television. This video, which showed PLA
soldiers training to disperse protesters by force, was widely taken up by the “pro-Beijing”
opinion as a guarantee that the “motherland” would not abandon the “loyalists,” but also by
the protesters themselves to illustrate the threat China represented. The video was shared
by several official accounts (PLA, China Daily) in English and Chinese, totaling tens of
thousands of views. It is still available today. In fact, the threat of an armed intervention
was used recurrently as part of a Chinese strategy of gradually imposing more pressure
on protesters, as testified by the announcements made on August 12,1018,11and 28.12
The state media’s strategy seemingly favored shocking content that could catch the
attention, spark a debate and generate views. For example, a TV clip compared Hong
Kong protesters with jihadist terrorists for the way they presumably instrumentalized
children (see the screenshot below).13 Posted on Facebook in mid-September, this video
was later broadcast by the main Chinese media channels (Global Times, China Daily, etc.).
8. The first contradictory search results arrived in sixth position but the majority are on the second page. The search
results for “Hong Kong Protest” were more varied and included little or no content emanating from state television.
Moreover, YouTube signaled channels funded by a state with a banner as in the examples quoted here.
9. David Bandurski, “War Games in Hong Kong,” China Media Project (1 Aug. 2019).
10. “Chinese Armed Police Truck Convoy Rolls into Shenzhen as Hong Kong Enters Another Week of Protests,”
South China Morning Post (12 Aug. 2019).
11. “Chinese Armed Police Stage Another Riot Drill Across Hong Kong Border as Protests Enter 11th Week,”
South China Morning Post (18 Aug. 2019).
12. Austin Ramzy, “Chinese Military Sends New Troops into Hong Kong,” The New York Times (28 Aug. 2019).
13. See https://www.facebook.com/watch/?v=2495151990809474 (last accessed 08/10/2019).
491
III. The Example of Guancha
The official social media accounts of certain Chinese media outlets played a central role
in spreading anti-protest narratives and similar iconographies.14 “RealGuancha” in partic-
ular played a prominent role. Guancha (观察) is a popular Chinese news website in the
style of Jimian (界面), Sina (新浪) or Global Times ((环球时报). If Global Times was already
well-known on Western social media (particularly Twitter) for its nationalist remarks and
clear-cut stances, the Guancha platform had not been noticed outside of China at that
point. But its Twitter account, created in March 2019, turned out to be very active and
virulent in its coverage of the protests.
The Twitter account @realGuancha mostly shared content in Chinese but not only.
Posts from October 7 (see above) showed the destructions of shops and boutiques belong-
ing to mainland franchises and brands. These undeniable acts of violence were depicted
as a generalized pillaging and even as “pogroms” (targeting mainland Chinese resi-
dents). On the screenshot on the left, the sentence “[if] you commit a crime, is an apology
enough?” was a reference to the initiatives of moderate protesters who wanted to apologize
for the violence.
Hence, the @realGuancha account published content intended to divide, like the con-
tent circulated on WeChat or Weibo. Even if it remained a small account (more than 32,000
14. See https://twitter.com/realGuancha.
492
subscribers on March 21, 2021), its growth during this period was significant: it went from
2,000 subscribers on October 6, 2019 to 3,543 on the 19th – a 77% increase in two weeks.
Extremely active, it posted no fewer than 750 messages during the same period (from
1,800 tweets on October 6 to 2,550 on the 19th) for an average of 57.7 messages a day.
According to the Accountanalysis application, a peak of activity was registered on October
1, 2019, during the national holiday of the PRC, with 633 messages.
Guancha was present not only on Twitter: it was just as active on YouTube, with at
least a hundred videos uploaded between October 3 and 6 on its YouTube channel.15 All
these videos showed images of violent protesters and highlighted their acts of violence.
Quality and length varied but they all seemed to originate from different sources, having
been filmed by Hong Kong locals and mainland Chinese citizens in Hong Kong. Very few
of these videos seemed to have been the object of a journalistic approach, or even edited,
and on the whole they resembled a compilation of amateur videos rather than a careful
montage.
These posts were frequently posted, almost akin to a saturation attack, and it led peo-
ple to believe that the objective was not to inform. Guancha aimed to flood platforms
with a counter-discourse favorable to the “loyalists.”
Jimmy Sham’s “self-mutilation”
On October 16, the activist Jimmy Sham, who had organized numerous peaceful protests that
rallied several hundred thousand Hongkongers, was targeted and beaten by a group of men
armed with hammers and iron bars. On October 19, Guancha posted a video entitled “Jimmy
Sham’s self-mutilation” (岑子杰的苦肉计 – “kurouji” which means “to wound oneself to
gain the enemy’s trust”).16 The video shows the blood-stained scene of the attack with Jimmy
Sham lying on the ground. Then Sham was shown at the hospital the day after the attack, smil-
ing. The innuendo was clear: if he had been as gravely wounded as he claimed how could he
only have had a little plaster on his forehead and be seen smiling, as if he had never felt better?
15. Guancha News 观察者网, https://www.youtube.com/channel/UCJncdiH3BQUBgCroBmhsUhQ/video
16. https://www.youtube.com/watch?v=b1kd88RgYMk.
493
The case of Jimmy Sham’s “self-mutilation” is only one example among dozens of vid-
eos uploaded on October 19 at an industrial pace – by midday ten videos had been
posted in two hours. They covered news related to the movement while, of course, offer-
ing an “alternative” point of view. For example, on “the list of the cockroaches’ crimes”
(曱甴罪行錄CASE 01),17 we found videos showing the “truth” behind the case of a young
15-year-old girl whose lifeless body was found in the port. These videos insinuated that it had
just been a suicide and that other explanations were mere rumors to defame and humiliate
the police.18 Despite these suspicious messages, Guancha was not one of the 210 accounts
suspended by YouTube.19 According to YouTube, the accounts affiliated to Guancha
used VPNs to post videos on its platform. Even if the firm announced that this discov-
ery “was consistent with recent observations and actions related to China announced by
Facebook and Twitter,” it was not able to establish a clear link between these activities
and the PRC, unlike for the two other platforms.
Screenshot of a video from the “cockroach crimes” series (曱甴罪行錄CASE 01). The video only garnered
70 views and 8 “likes” with no comments, which is about average for videos like this one.
IV. The use of social networks with an international audience
A. TikTok
As we explained in part three (→ p. 200), TikTok is censored and instrumental-
ized by Chinese authorities. For instance, The Washington Post observed that a search for
“#hongkong” on the application gave no result tied to the protests.20 And “violence in
Hong Kong” (香港暴力) mainly returned videos from official media which unanimously
condemned the protesters’ “terrorism.”
17. “曱甴罪行錄CASE 01” (“Cockroach crime CASE 01”) (19 Oct. 2019), https://www.youtube.com/
watch?v=YY_6vnGgjsY.
18. “视香港15岁少女陈彦霖母亲：女儿是自杀，望停止造谣” (“See the Mother of Chen Yanlin, the
Young 15-Year-Old Girl from Hong Kong: She Committed Suicide, Hoped of Stopping the Rumors, I Hope that
the Disinformation Will Cease”) (Oct. 2019),; https://www.youtube.com/watch?v=IxXE5u-WIn4; “暴徒炒作
「少女自殺案」污衊警方，其母發聲：放過我們一家人” (“The Mob Defames the Police with ‘a Case of
Adolescent Suicide;’ the Mother Speaks Out: Leave Our Family Alone”) (18 Oct. 2019), https://www.youtube.com/
watch?v=3rNza7FZN5Y.
19. “Maintaining the Integrity of Our Platforms,” Google (22 Aug. 2019), https://blog.google/outreach-initiatives/
public-policy/maintaining-integrity-our-platforms/.
20. Drew Harwell and Tony Room, “TikTok’s Beijing Roots Fuel Censorship Suspicion as it Builds a Huge U.S.
Audience,” The Washington Post (15 Sept. 2019); see also Alex Hern, “Revealed: How TikTok Censors Videos that do
Not Please Beijing,” The Guardian (25 Sept. 2019).
494
B. Facebook
On August 19, 2019, Facebook deleted seven pages, three groups and five accounts that
showed “coordinated inauthentic behavior” from a network that “originated in China and
focused on Hong Kong.”21 These pages compared the protesters to cockroaches and
terrorists, among other posts. They also posted photo montages showing armed protest-
ers, or police officers being attacked.
If the content was certainly shocking, it was removed for another reason: Facebook, in
accordance with its rules, suspended these pages, groups and accounts because of
their inauthentic and coordinated (i.e. manipulative) behavior and not because of the con-
tent they were disseminating.
An example of content removed by Facebook: “the cockroach soldiers.”
21. “Removing Coordinated Inauthentic Behavior from China,” Facebook (19 Aug. 2019), https://newsroom.
fb.com/news/2019/08/removing-cib-china/.
495
An example of content removed by Facebook: “Protesters, ISIS combatants, what’s the difference?”
An example of content removed by Facebook.
C. Twitter
The massive disinformation campaign organized by China is most visible on
Twitter. On August 19, 2019, the same day as Facebook (which had been tipped off by
Twitter), the platform made public a “significant state-backed information opera-
tion”: the company explained that it had deleted around 200,000 fake accounts implicated
in a campaign meant to “[undermine] the legitimacy and political positions of the [Hong
Kong] protest movement” and it published the archives of the 936 most active of these
accounts, all “originating from within the PRC.”22 A month later, on September 20,
Twitter published the archives of another batch of 4,301 accounts, which were also among
the most active in this operation.23
Several research teams have analyzed the data published by Twitter. In this summary we used
the analysis of the Australian Strategic Policy Institute (ASPI) and those of two indepen-
dent data scientists, one known only under his pseudonym “AirMovingDevice,” the other, Chua
22. “Information Operations Directed at Hong Kong,” Twitter (19 Aug. 2019), https://blog.twitter.com/en_us/
topics/company/2019/information_operations_directed_at_Hong_Kong.html.
23. “Disclosing New Data to Our Archive of Information Operations,” Twitter (20 Sept. 2019), https://blog.
twitter.com/en_us/topics/company/2019/info-ops-disclosure-data-september-2019.html.
496
China Hong, based in Singapore. Particularly quick to respond, the ASPI published its report,
Tweeting through the Great Firewall, on September 3, only two weeks after the data was published by
Twitter.24 The report dweals on the means used by Beijing to build its disinformation campaign
as well as its targets, but does not confine itself to Hong Kong since it gies examples of older
campaigns targeting political opponents such as Guo Wengui and Gui Minhai.
1. A Typology of the accounts involved in the disinformation campaign
First preliminary observation: this campaign was conducted in haste. The ASPI
noted that the operators behind the surge did not take the time to build credible digital
identities, unlike the Russian operators who had attacked the presidential campaign in 2016.
Of course the difference is that the calendar for the U.S. election was pre-established,
allowing the attackers to prepare their operations several years in advance and to patiently
build an “infrastructure of influence,” whereas the Hong Kong crisis was unexpected and
evolved quickly. The modus operandi of the disinformation campaign had to be adapted at
the last moment, taking “shortcuts” to buy time, such as purchasing existing accounts for
instance, which made the operation quite easy to detect.25
Two types of Twitter accounts used against Hong Kong protesters need to be
analyzed separately: on the one hand, there were “mercenaries,” which previously used
different languages (English, Indonesian, Korean, Japanese, Russian) to post content ranging
from English football to pornographic messages (sex dating websites, escort services), or
even information about Korean boy bands. Having been bought, these accounts changed
their language and cause overnight, thereafter defending Beijing and going after Hong Kong
protesters. Selling accounts is a common practice already mentioned in part three (→ p. 375).
On the other hand, there were also “dormant” accounts: among the data published by
Twitter, ASPI identified 233 accounts where there were intervals of at least a year between
two tweets. Among them, the oldest had been created in December 2007.
Their interest in Hong Kong intensified in May (one day before Carrie Lam suspended
the extradition bill) and June 2019 (with the “capture” of the Legislative Council by the pro-
testers).26 Even if the tweets referencing Hong Kong were only a small part of the sampled
data, the ASPI identified three main narratives: a rejection of the protesters and the
condemnation of their positions, support for the Hong Kong police and the “rule of law,”
and finally conspiracy theories suggesting “Westerners” were involved in the protests.
Hastily organized and lacking technical means, the ASPI describes a “blunt” influence
operation. The method was similar to those used in mainland China, where censors often
use spams to flood public opinion with certain types of content, and to armies of Weibo
accounts paid to plaster official messages under certain types of posts globally.27
Chua Chin Hong, a Singapore-based specialist on Chinese media proposed a more gran-
ulated analysis of the second batch of data published by Twitter (ten million tweets) that
refines our understanding of the campaign. These tweets, posted between May and July 2019,
represented a “campaign surge” which “could ironically be one of the key reasons
why the network set off alarm bells at Twitter and got taken down.” During this surge,
24. Tom Uren, Elise Thomas, and Jacob Wallis, Tweeting through the Great Firewall: Preliminary Analysis of PRC-linked
Information Operations against the Hong Kong Protests, ASPI, Report No. 25 (2019).
25. Ibid., 8.
26. Chua Chin Hon, “Failed Surge: Analyzing Beijing’s Disinformation Campaign Surge on Twitter,” Toward Data
Science (24 Sept. 2019).
27. Chen Na, “Guns for Hire: China’s Social Media Militia Engage on Command,” Sixth Tone (12 Mar. 2018).
497
the accounts disseminated 87,000 tweets and retweets (but only 27,000 original tweets). The
overwhelming majority of these messages were in Chinese with only 6% in English. 28
One indication of the artificial nature of these accounts was their lack of experience
using Twitter, which is banned in China. For instance, Chua highlighted that many of the
accounts unknowingly posted the welcome message on Twitter (“Just setting up my twitter
#myfirstTweet”), before they started to successively retweet the “lead” accounts. For the
Singaporean analyst, one of the main clues to the “massive and coordinated” nature of
the disinformation campaign could be found in the accounts’ dates of creation. Indeed,
as the ASPI study showed, the accounts removed in August were rather old. In the September
archive, the accounts were far more recent: out of the 4,301 accounts listed more than half
(54%) were created during the previous three months. The largest spike in creation was
on June 14 – one day before the extradition bill was suspended – but the analyst remained
prudent and, in the absence of proof of any coordination, spoke only of a coincidence.
2. Spam and dormant accounts
Another analyst, “AirMovingDevice,” underlines several other characteristics.29 First, the
tweets were synchronized with Beijing’s time zone: the time slots during which the major-
ity of the tweets were published did not correspond to the local time zone (that is to say with
the geographical zones advertised by Internet users) but to Beijing’s. Moreover, like the ASPI,
AirMovingDevice highlighted long periods of dormancy or non-activity in many
accounts (nearly 41%). He described the following model: the accounts were created before
2018, posted in different languages, and then remained inactive for more or less long periods of
time. They were reactivated around 2018-2019 and started posting in Chinese.
The analyst gave the example of 披荆斩棘 (#saydullos1d), a female user presumably based
in Colorado.30 This account, created in 2013, posted mainly in English on topics linked to hunt-
ing and fishing, without ever being retweeted or “liked.” She then progressively disappeared
from the radars before reappearing in 2018. During these five years, she had “learned Chinese”
and “developed a passion for butterflies,” for a few weeks at least. But this passion mysteriously
evaporated to make room for Chinese tabloids. AirMovingDevice then showed that she spread
content found on Weibo, even going as far as posting 1,000 messages in 2018. Then, in June
2019, she suddenly started to retweet numerous messages condemning Hong Kong protesters.
“Qujianming” was another example analyzed by Chua Chin Hong. Immediately suspect,
Qujianming boasted 28,000 subscribers for 24,000 subscriptions, but, as Chua explained, it
is rare for accounts to have that many subscriptions. Qujianming mainly tweeted in Chinese
and called for “severe punishment for the violent protesters, systematically described as
rioters (暴徒 – hooligans, thugs)” – who were accused of following their own interests only
(and not those of Hong Kong).
3. Case Study
In a complementary approach to the essentially quantitative studies published by the ASPI in
2019, and in order to better understand the methods used by Beijing to influence and manipulate
online discourse, we conducted a qualitative investigation on Twitter and Facebook.
28. Hon, “Failed Surge.”
29. AirMovingDevice on Twitter 22/08 (https://twitter.com/AirMovingDevice/status/1175689332704759810).
30. Ibid.
498
a. Methodology
We chose to start with hashtags, that are a signature for the digital narra-
tives and the narrative communities that conveyed them, to further investigate
some emblematic Twitter accounts. This approach has several advantages: first, a
hashtag is often used and shared by large groups of users at the same time, but it
nonetheless remains defined and intelligible. In the case of the Hong Kong protests,
each “side” defended its representations through clearly differentiated and divisive
hashtags. Sometimes the two sides fought over the same hashtags (#HongKongPolice
for instance), with one side defending the actions of the Hong Kong riot police, and
the other condemning their violent actions. In other words, hashtags were a point of
convergence for the different sides and a structuring force for their discourses – a
shared signature of some sort. As markers, they helped identify a group of rela-
tively homogeneous users. This also allowed us to study the users’ discourse,
which involved various channels (tweets, images, memes, allusions, wordplays, etc.).
Finally, the hashtag is also a time marker, since it allows us to date its first use (and
sometimes its decline). For all these reasons, our study analyzes of a set of hashtags
used in September and October 2019.
Starting with these hashtags (listed below), we chose representative accounts based
on three criteria that questioned their authenticity: anonymous user IDs (a Twitter
user ID made of a word, or part of a word, with many numbers); a linguistic disso-
nance (IDs using a “Western-sounding name” but tweeting only in Chinese); and their
non-linearity or non-regularity (accounts that boasted more than 5,000 subscribers
or subscriptions a few months after their creation, or accounts that had been inactive
for a long time but suddenly starting to tweet in Chinese).
These suspicious accounts were then analyzed in more details with the help of open-
source research tools (OSINT) such as “tinfoleaks” or “accountanalysis,” which gave us
the precise date of an account’s creation, the date of the first tweet, the hashtags used,
and so on. This method was relatively efficient: half the accounts quoted in our investiga-
tion have since been suspended (between the last weeks of September and mid-October
2019).
The research on these accounts focused on the period that started on September
20 and ended on October 20, 2019. The advanced research tools provided by Twitter
made it possible to target certain phrases, enabling us to concentrate on the events
that shaped this period: the protests of September 28 and October 1, and those
that followed the adoption of the law outlawing masks during demonstrations. The
choice of this one-month period allows us to propose an analysis that takes into
account the evolution of the methods and narratives used during this period, but
also to propose an original qualitative study that goes beyond a mere observation of
the virality of the messages. We formulate hypotheses regarding the intentions and
strategies that could have underlain their publication, thus complementing already
existing studies. The analysis hopes to show the technical levers used with Twitter
accounts and, therefore, their non-authentic or “unnatural” character (inherent to
their irregular uses).
499
b. Reactivated dormant accounts
• @loverealface2 (active until October 9, the day it was suspended) was registered in the
United States. Created in December 2017, it boasted 5,100 subscribers (5,800 subscrip-
tions) and only posted 56 tweets after its creation. The long period of inactivity between
August 2018 and September 17, 2019 made it suspicious.
It last tweeted in English in August 2018. And, from September 2019 on, it tweeted in
Cantonese when covering Hong Kong and, more rarely, in Mandarin when dealing with
“personal messages” (for example a message about a trip to Yunnan).
We give two examples of tweets posted by @loverealface2 here: the first (see below,
left) accused Joshua Wong, Jimmy Lai, Anson Chan, and Martin Lee of having “col-
laborated” with the “exterior anti-Chinese forces” (境外反华势力) to organize a color
revolution (颜色革命) in Hong Kong. The photo showed a “calm, clean” Hong Kong
that contrasted with the mass of umbrellas. The message meant “let Hong Kong recover
her past glow (glory).” The second picture (see below, right), exploited the theme of the
collaboration with “exterior anti-Chinese forces.” Joshua Wong was accused of wanting
to destroy the social order and to break the “One country, Two systems” compromise.
The photo was supposed to represent a U.S. diplomat having dinner with several pro-de-
mocracy figures.
• @Nirmala1244 (last consulted on October 3, and suspended around October 9) was
created in March 2019. For someone who tweeted very little (only 21 tweets), the account
boasted 5,000 subscribers and 5,400 subscriptions. This account posted in Mandarin and
Cantonese and wrote “personal” tweets (which had very little to do with the Hong Kong
protests). Furthermore, between April 8 and July 24, the account was barely active, with
only three visible messages. The peak of its activity occurred between September 21 and
October 1 – not unlike @loverealface2 – which could mean there was a rapid campaign
in the lead up to October 1. The content of the tweets was similar to that of @lovereal-
face2: patriotic images and others meant to tarnish the protesters.
500
Above are two tweets that @Nirmala1244 posted. The first (on the left) questioned
foreign influence: in this case one person presumably met with a foreign diplomat to dis-
cuss the protests. The message mentioned “meetings” between foreign agents and pro-
tests leaders. The second tweet (on the right) criticized the fact that protesters allegedly
“used” children to conduct their violent acts.
• @FunnybonesBob (consulted on October 2, suspended on October 10) had 5,400
accounts and collected 5,100 subscribers. Created in November 2017, it did not tweet
between March 2018 and September 2019 and had only published 37 tweets in total.
Most of the tweets were old and they discussed an application that supposedly “cleans”
Twitter profiles and subscriptions in particular.31 This account only became active in
September 2019 and more precisely on September 30, with only a few tweets that took
up the iconography linking protesters to the Ku Klux Klan; another tweet drew an anal-
ogy between protesters and saboteurs.
• @Lancy39681596 was also very active. Its number of subscribers grew sharply: from
1,720 on October 10 to more than 2,000 on October 19. In addition, it published more
than 12,000 tweets after its creation, five months earlier, in May 2019. Even more trou-
bling was the frequency with which this account sent messages. On Sunday, October 6,
as major protests took place in Hong Kong in reaction to the law banning the wearing
of masks, @Lancy39681596 sent nearly 1,400 tweets between 7am and 10pm. In com-
parison, on Saturday October 5, the account sent “only” 387 tweets.
31. https://play.google.com/store/apps/details?id=com.mhmdalmz.example.unfollowjet&hl=en. It is possible
that this is the application but the tweets have since been removed and we have no way of confirming this hypothesis.
501
On Sunday, October 6, @Lancy39681596 was particularly moved by the violence per-
petrated by protestors (in particular against a cab driver, stores, and individuals). It also
presented cases of looting of Chinese stores or simply of “ordinary citizens” by Hong
Kong protesters. The tweets were often bilingual (the screenshots above were taken on
October 5).
• @anyaafairy, another suspicious account, was consulted on October 14 and 18, and
suspended around October 20. This account was typical of the “improvements” made
to more basic accounts. @anyaafairy claimed to be a roleplay “NSFW” account (not safe
for work, i.e. erotic or sexual), and its profile picture was stolen from a real Thai model.32
The account had over 12,000 followers and fewer than 130 subscriptions. In other words,
it was at first glance an “authentic” account. However, in its three years of existence,
it tweeted only 234 times. Only a dozen tweets were visible at the time the account was
consulted in October 2019, with the oldest dating back to October 7. On average, two out
of three (visible) tweets were about Hong Kong and published in Mandarin.
The user deplored that, since June, more than 4,200 meters of railing, close to 400 lamp
posts and traffic lights, 2,600 square meters of asphalt and 1,700 items of city property
had been destroyed by rioters.
• @TXT_Tathyunnn (last accessed on October 18, suspended on October 30) was a
fan-made account about Kang Tae Hyun, one of the singers of the Korean K-pop group
“TXT Tomorrow-X-Together.” It was not the official account of the boys band. Created
in August 2015, it posted 8,800 tweets, with 19,000 followers and barely 50 subscriptions.
At its creation and until April 2019, it behaved “normally”: retweets of the band’s official
account (in Korean), pictures of the singers, etc. Then, between April and October 7,
2019, the account was completely inactive. In early October, the account posted a series
of photographs of members of the band but without text. Then, on October 9, the
account tweeted in traditional Mandarin to criticize the Hong Kong protesters, following
a model similar to the tweets mentioned above (infographics/photos, videos and com-
ments). The account kept tweeting about the members of the Korean K-pop group but
tweets became less frequent and interspersed with other tweets about Hong Kong. TXT
also tweeted in Cantonese and English, and seemingly focused on looting and violence.
32. Thanyarat Charoenpornkittada who can be found on https://www.facebook.com/profile.
php?id=100002074497537, Instagram: https://www.instagram.com/fearythanyarat/?hl=en, and YouTube: https://
www.youtube.com/channel/ UCT5aACooyOYKmctQsERH0Qw.
502
• Our last examples show the reactivity of these accounts’ operators, in this instance
their ability to quickly activate fallback accounts. The account @CrealDllmhg, con-
sulted on October 14, but since suspended, was a particularly active account with almost
1,100 tweets and 1,300 subscribers in its three months of existence.33 On October 17,
however, another account @DllmhgCreal, created in July 2019 with the same profile
picture, and around 1,300 subscribers, resumed its activities.34 The content was very sim-
ilar to what was already presented: highlighting the protesters’ violence was its leitmotiv.
They were often compared to terrorists. Each misdeed was scrutinized (or made up) and
the messages recycled the specificities we have observed on WeChat: shocking images
and titles that sought to prompt indignation, etc.
c. “Relay accounts”
In this study we also noted numerous accounts with a marginal audience, i.e. only a
few subscribers and subscriptions. These accounts seemed to simply retweet and share
the content disseminated by the more important accounts presented above. We named
these “relay accounts.”
There are many such fake accounts, but we could also find many “micro users” among
them who try to go undetected by Twitter moderators. For example, “Wurufan” (务如凡
@kWz33lF6nCU9Z7E – last accessed on October 19) was one account whose user-
33. See the archive we made: https://archive.fo/ozJfK.
34. The account has since been suspended, see the archive we made: https://archive.fo/62muw.
503
name seemed to have been randomly generated by computer. Not very active (123
tweets between its creation in July 2019 and October 10), the account had only five sub-
scriptions and only one subscriber. Its first message dated from September 10 and they all
followed the same principle. The account retweeted and commented on messages coming
from bigger accounts (such as those quoted previously). The first tweets sent on September
10 actually relayed messages from accounts which are now suspended.
Many accounts like this one took turns to generate content online. The account @
BeverleyMclare3, active on October 9, was suspended the next day; a similar account,
@RuthannGant9, created in September 2019 (accessed on October 9 – 2 subscriptions,
2 followers, 28 tweets), used the same codes; it subscribed to two other fake accounts @
anntaranti and @MaThao91, both created in September 2019, with one or no subscriber.
If they had been created in September, they apparently did not start tweeting until early
October, which would indicate that they were back-up or secondary accounts. Both
accounts were accessed on October 9 and suspended the next day. It is likely that these
accounts are regularly suspended and then reprogrammed. A new example is provided by
@UshaBoardman (account accessed on October 14, suspended on the 16th) created on
September 23 and almost inactive until October 14.
Another peculiarity is that most of these “new” accounts used Western names. For
instance, @GillianMonteith and @MarineMcgregor5 were two accounts that tweeted
only in Mandarin about Hong Kong-related topics. These accounts were not the most influ-
ential because their audience was after all inexistent, but they revealed the Chinese strategy
described in the ASPI reports or in other studies.
d. Anti-protester hashtags, narratives and iconographies
The following texts were collected from hashtags used both in English and in Chinese.
Generally, the tweets did not contain more than three or four keywords, among which we
always found “Hong Kong” / “香港.” Here is an overview of the main hashtags:
504
Hashtags
in Chinese
English
Translation Commentary
#香港廢青 Hong Kong wasted youth
“waste” (废 – fei has several mea-
nings: “abandoned,” “thrown away”
(for waste). This was one of the
hashtags the protestors used the
most.
#香港暴乱 Hong Kong riots One of the most common hashtags.
#(香港)暴徒 Hong Kong thugs
Another of the most frequent hash-
tags. “暴徒” (baotu), which means
“bandit,” i.e. a violent individual
without a “noble” cause.
#守護香港 Protect Hong Kong A widely-used hashtag.
支持香港警察 Supporthkpolice
Probably one of the most widely
used hashtags in both Chinese and
English.
#我支持香港警察你们可以打我
了
I support HK police you can hit me
This message appeared in late July
and early August as clashes with the
police became more frequent.
#14亿人撑香港 1,4 billion people support HK One of Diba’s slogans.
#14亿护旗手 1,4 billion protect the [PRC] flag Another of Diba’s slogans.
#饭圈女孩网络出征 The online campaign of the fan girls
One of the hashtags of Fanquan
that allowed its members to reco-
gnize themselves.
#帝吧出征 Diba’s campaign Same principle, one of Diba’s main
slogans.
#中国一点也不能少 China can’t be a bit smaller
This slogan expresses the opposi-
tion to the “secession” of Hong
Kong.
#香港是中国的香港 Hong Kong is China’s Hong Kong
Similar to the one above; absolute
refusal to lose sovereignty over
Hong Kong.
#全民撑警 The whole people support the police
A hashtag similar to the ones pre-
sented before, but sponsored by
groups endorsing the Hong Kong
(and Chinese) government, and
relayed by Guancha.a
a. “今天，香港全民撑警日” (“Today, it’s Hong Kong Police’s day”), 观察者网 (The Observer) (10 Aug. 2019).
Variations on these hashtags did exist but it was neither pertinent nor possible to include all of them here.
The few hashtags listed above were among the most recurrent but they were not the only ones.
These hashtags can be classified into two groups: those displaying patriotism or
nationalism (particularly among Diba members) and those which specifically covered
protests in Hong Kong (either by supporting the police or disparaging the protesters).
To get a better understanding of the anti-protester narratives disseminated on Twitter
one need to study the content of the tweets and images. The two main lines of attack
were the alleged violence of protesters (drawing an analogy with terrorists, highlighting
the pillaging, accusations of rape, of lynching, etc.) and the alleged manipulations of
“foreign anti-Chinese forces.” Hereafter we show several tweets and infographics, both
in English and Chinese, which are representative of these topics.
505
On image #1, the first strip shows protesters as cockroaches yelling “five demands and
not one less” (the protesters’ slogan), “I want to break stuff, hit people and stop them from
going back to work or to school.” On the strip at the bottom, under the red title (“Once
the law prohibiting the wearing of masks is enacted”), the “protester” on the right in the
foreground says to his comrade: “Come on, you go first,” the other answers: “Errrrr ok,
but I’m scared…” This image posted in late September 2019 was supposed to illustrate
the fact that protesters only dared follow through with their actions because they
felt protected by the anonymity provided by masks.
The main message was that masked protesters were violent, that they hid real weap-
onry (Image #2), and caused major damage (Image #3).
2 3 A particularly predominant leitmotiv advanced that protesters behaved more violently
than the police. A number of posters and messages highlighted the acts of violence pre-
sumably committed by protesters against the police (Image #4). For example, Image
#5, posted by Guancha, showed the following: “Neither peaceful nor reasonable,” followed
by the caption “Peaceful? Reasonable? Certainly not!” Guancha was here trying to demone-
tize the values exhibited by protesters (“we are peaceful and reasonable”) to better discredit
them.
506
54 While certain images denounced the protesters’ violence against the police, others praised
the attitude of the police in regard to the “thugs” (Images #6 and #7). This very cinematic
style could also be found among both protesters and “counter-protesters.” Once again the
role of Guancha in the dissemination of this content was noteworthy.
6 7Another angle was the “whataboutism,” a form of sophistry which was about trying to
create a diversion by invoking a different target. For example, in response to what was per-
ceived as the biased Western coverage of the Hong Kong crisis, criticizing the violence of
the police and Beijing’s support, many influential Chinese-speaking accounts turned these
criticisms back against the West using comparisons – especially to the U.S. police, as we saw
in part three (→ p. 385) and as illustrated by Images #8 and #9.
507
8 9
Không có nhận xét nào:
Đăng nhận xét